LICENTIATE T H E SIS. Critical Hydraulic Gradients in Tailings Dams Comparison to Natural Analogies. Isabel Jantzer

Transcription

1 LICENTIATE T H E SIS Critical Hydraulic Gradients in Tailings Dams Comparison to Natural Analogies Isabel Jantzer

2

3 Critical hydraulic gradients in tailings dams Comparison to natural analogies Isabel Jantzer Department of Civil and Environmental Engineering Division of Mining and Geotechnical Engineering Luleå University of Technology

4 Printed by Universitetstryckeriet, Luleå 2009 ISSN: ISBN Luleå

5 Preface This licentiate thesis was granted by the Swedish Hydropower Center SVC and was carried out at the Division of Geotechnology, Dept. of Civil, Mining and Environmental Engineering at Luleå University of Technology. The main part of this study was carried out between February 2005 and September 2008, and it was continued and finished in fall 2009 after a temporary employment abroad by the author. First of all, special thanks to my supervisor Sven Knutsson for the continuous support and encouragement, and for the patience and trust in me all the many times I was doubtful, confused or lost motivation. I would like to thank the participants in the reference group, Annika Bjelkevik at Sweco, Malin Söderman at Boliden, and Lars-Åke Lindahl at SweMin for their invaluable experience, comments and encouragement. In addition to participating in the reference group, Annika Bjelkevik always had extra time, listened, commented, helped, supported and encouraged me whenever I felt that research was a tough process, which I am deeply grateful for. Posthumous thanks to Erik Kitok, who helped me with many practical issues regarding field work. He will be remembered being a straightforward, encouraging personality with an unbelievable amount of energy to make things happen. At the department, I would like to thank Thomas Forsberg and Ulf Stenman for their invaluable experience, practical ideas and help. Both laboratory and field work can sometimes be frustrating, but they simply add a little serenity, a little more interest and commitment, making testing simpler and results clearer. I also would like to express my deep appreciation for my colleagues at the Division for the nice working environment we share. Thanks for the coffee breaks, especially when I felt overloaded and thought I did not really have the time. There is always time for coffee; I should have learned that during the years living in Sweden. Thanks to the research group, Gunnar Hellström, Hans Mattsson, and Staffan Lundström for interesting discussions and adding a new perspective. Special thanks to Gunnar who always knows about where to go for conferences, meetings and travels, which I otherwise certainly had missed. Very special thanks to Karin & Magnus and their family for their friendship and making me feel at home in Luleå. Thanks to my old friends Pia and Serge for being close despite the distance. Thanks to Lisa and John, in various ways. Thanks to Eva and all the wonderful people I shared such a good time with skiing and climbing, which gave both the physical and mental strength that I could use so well in continuing this work. And thanks to the fantastic group I started to paint with; I would not have been able to write this thesis without the creative breaks and inspiration. Finally, I do not know how I possibly could thank you, David, for your love and never ending patience, for cheering me up and never letting me down, for following me and giving me a home. May we stay Forever Young. Isabel Jantzer November, 2009 i

6 ii

7 Abstract The increasing demand of metals and minerals has made it economical for the mining industry to mine low-grade deposits, which results in immense quantities of by-products. In fact, such by-products, i.e. tailings, are the largest waste volumes produced on earth. Ore from sulfide rich geological formations result in sulfide rich tailings that can cause major environmental problems upon oxidation and acidic seepage. One way of economically and effectively preventing unwanted chemical reactions and leaching is the deposition in an impoundment surrounded by tailings dams, where tailings are allowed to settle and covered effectively by a water level against oxidation. In order to prevent environmental impacts and possible long term contamination from tailings dams, the Swedish Environmental Protection Agency (Naturvårdsverket), demands a long term stability without maintenance that refers to thousands of years or more, resulting in a design period of 1000 years for tailings dams. The stability of tailings dams in long term perspective depends, amongst others, on the prevention of internal erosion, a process that results from an exceeding seepage pressure causing particles in a dam to migrate, with possible consequences of damage and failure of the dam construction. Therefore, the main question in this thesis is: Which maximum hydraulic gradient can we allow for a tailings dam construction in order to prevent internal erosion in a long term perspective? With regard to the long term design of tailings dams, natural analogies to dam constructions are considered, i.e. formations from the last glaciation period that have fulfilled the task of damming water. Such structures are especially interesting with regard to their obvious stability against internal erosion over long time, otherwise they did not exist today. Consequently, it is assumed that a critical hydraulic gradient exists, and that the material composition and compaction reaches an optimum which allows seepage without erosion. This study provides basic knowledge on tailings dam construction and a State-of-the-Art report on current knowledge on internal erosion. A summary of natural analogies to dam constructions that have been stable dams since the last glaciation in Sweden is presented. In the context of this work, a case study was conducted at the company area of Boliden in Gällivare, northern Sweden, with the aim to study the geotechnical properties of such a natural stable embankment. This case study includes field studies, ground water monitoring and sampling. Complementing laboratory analysis covers an analysis of the materials properties with regard to composition, density, compaction and hydraulic conductivity. The formation consist of a well graded glacial till which is compacted to an optimum in situ above what could be obtained in laboratory conditions. The hydraulic conductivity of laboratory compacted samples shows a minimum of 2,60 * 10-10, which implies that the material is practically impermeable, which may be an explanation for the absence of ground water during monitoring. Critical hydraulic gradients found in literature range between 4,8 and 14 %. Current tailings dam design guidelines in Sweden relate the maximum gradient to the internal angle of friction, thus resulting in gradients of about 12 to 27 %. Gradients in long term stable natural formations are between 2 to 5 %. The calculated hydraulic gradient in the case study is 6,7 %; however, the actual gradient could not be determined due to the absence of pore pressure measurements during ground water monitoring. With regard to long term stability, possible degradation and results from comparisons to long term stable natural analogies, a modification of the design criteria for Swedish tailings dams should be considered. iii

8 iv

9 Extend of thesis The present licentiate thesis covers this embracing thesis, including a State of the art report and field studies. In addition, four conference papers are part of this thesis: o Jantzer I., Bjelkevik A. and Pousette K. (2008). Material Properties of Tailings from Swedish Mines. Nordic Geotechnical Meeting NGM. Sandefjord, Norway. September 5, o Jantzer I. and Knutsson S. (2007). Effects of Freezing and Thawing in Embankment Dams. Proceedings of International Symposium on Modern Technology of Dams the 4 th EADC Symposium. Chengdu, China. October 14, o Jantzer I. and Knutsson S. (2007). Seepage and Critical Hydraulic Gradients in Tailings Dams and Natural Formations. Proceedings of 2 nd International Conference on Porous Media and its Applications in Science and Engineering (ICPM2). Kauai, Hawaii, USA. June 19, o Jantzer I. (2006). Frost action processes in the Eastern Suorva hydropower dam. Proceedings of 17 th European Young Geotechnical Engineers Conference. Zagreb, Croatia. July 20, v

10 vi

11 TABLE OF CONTENTS Preface... i Abstract...iii Extend of thesis... v TABLE OF CONTENTS...vii 1 INTRODUCTION Background Tailings and tailings dams in general Design of tailings dams Tailings dam construction in practice Internal erosion Objective Method Outline STATE OF THE ART Introduction Classification of hydrodynamic soil deformations Suffusion Erosion Clogging Heave Geometric criteria Filter concepts Internal stability and characteristic grain size Hydraulic criteria Hydraulic conductivity Critical hydraulic gradients Design criteria for Swedish tailings dams Further aspects on hydraulic force Initiation of internal erosion Settlements Inhomogeneous material Hydraulic fracturing Statistics and risk assessments Statistical approach Risk assessment Risk analysis Discussion and conclusion vii

12 3 NATURAL ANALOGIES Natural formations in Sweden Lake Ragunda Lake Hennan in northwest Hälsingland Lake Mången at Brattforsheden Styggtjärn Skuttungesjö Summary of formations studied by SGU Case study Introduction Description of field work Results of pore pressure measurements Soil description laboratory analysis Tests on hydraulic conductivity Assessment on hydraulic conductivity in field conditions Comments on the laboratory testing technique DISCUSSION In situ density and Proctor compaction Hydraulic conductivity Ground water monitoring Critical hydraulic gradients Future work CONCLUSIONS ACKNOWLEDGEMENTS REFERENSES APPENDIX 1...I Results from hydraulic conductivity tests: Flow vs. time...i viii

13 1 INTRODUCTION 1.1 Background Tailings and tailings dams in general The term tailings refers to the solid rest product from industrial mining activity, i.e. fine crushed rock from which minerals have been extracted (Vick, 1990). During the process of extracting valuable minerals, crushed rock is grinded into fine particles. Further washing, cleaning and extraction often includes treatment with chemicals. Minerals and tailings are separated and the smaller amount of valuable minerals (about 0,4 30%) is then removed. The vast amounts of left-over tailings (about 70 99%), however, is of no value and has to be deposited. Tailings usually leave the production plant after some dewatering or thickening together with process water in a slurry form and are pumped to a tailings impoundment to be left for settlement. Process water is then clarified and either discharged in a stream, left for evaporation, or re-circulated. Due to this sequence, storage of process water in an impoundment surrounded by embankments, i.e. tailings dams, and/or natural heights is always involved to some extend during mining operation. The increasing demand of metals and minerals has made it economical to mine low-grade deposits, which in turn results in immense quantities of waste material. In fact, mine wastes are the larges waste volumes produced on earth (Mitchell & Soga, 2005; Bjelkevik, 2005a). These waste volumes comprise fine grained particles with varying characteristics dependent on the origin of the ore and different extraction processes. The physiochemical nature of the originally natural material is altered; the particles have an increased and a chemically treated surface after processing. The storage, closure and remediation of tailings storage facilities is therefore of utmost importance to inhibit environmental impact (Bjelkevik, 2005a). There are many different types of tailings dependent on the type of ore, the extraction and the processing. In Sweden, mainly ferrous ores are produced, but also ore from sulfide rich geological formations is used, resulting in sulfide rich tailings. Oxidation of such sulfide rich tailings and acidic seepage can present major environmental problems. When a mine is mined out and mining activity comes to an end, the mining company has to take measures to leave the site including the tailings storage facilities. A major concern is to minimize the environmental impact from tailings, i.e. the release of trace elements from tailings themselves, due to oxidation and weathering, or the release of acidic water due to leakage from the pond. This applies to the storage of tailings both during active mining and when a mine is closed. However, this thesis concerns the remediation of tailings deposition in a long term perspective after closure of a mine. There are basically two different ways for remediation of tailings dams: the coverage of tailings by dry cover or by wet cover. Applying the dry cover method includes the drainage of all water from the tailings storage and subsequent enclosure with suitable material, which would result in huge amounts of material for coverage. Therefore, remediation by water cover is often chosen, which does not only require less resources, but is also an effective barrier for oxygen. The wet cover method implies the coverage of tailings with a certain water level to prevent possible oxidation and weathering. This method is often chosen, because water is not only an 1

14 effective barrier for oxygen, but also because it requires less resources than the dry cover method. Tailings covered with dry material may probably become subject to external erosion and abrasion, which is not included in this thesis. Instead, this thesis refers to remediation by the wet cover method which involves water storage and a tailings dam construction, i.e. impoundments with surrounding embankments constructed partially with tailings, as such structures may become exposed to internal erosion processes in the embankments. Here, tailings dam construction becomes crucial for the long term storage Design of tailings dams Swedish tailings dams are usually constructed similar to conventional earth- and rockfill dams in order to create an impoundment for tailings storage. Yet, tailings dams are considerably different from water retention dams (WRD) in both construction material and structure. To built tailings dams as efficiently as possible, tailings themselves are often used as construction material in the dam body. An essential difference to WRDs is that tailings dams are raised in stages or continuously along with mining production. Thus, the tailings impoundment will grow in size and capacity over time. To meet increasing capacity requirements, there are basically three different methods for tailings dams construction: The upstream construction, the centerline construction, and the downstream construction. The following explanation is described according to Bjelkevik (2005a). Upstream construction Applying the upstream method, a starter dike of borrow material is constructed. The discharge of tailings is done from the crest of the dike surrounding the impoundment. Tailings settle and create a so-called beach from the crest inwards. Upon filling of the impoundment, a following dike is constructed partly on the previous dike and partly on the settled tailings as depicted in Figure 1. The upstream method is relatively cheap and simple to apply. There are several restrictions, amongst others that the storage capacity will gradually be reduced, drainage layers are difficult to apply, and the capacity of water storage is limited. Successive raise of this construction is limited by the drainage of the settling tailings and the dissipation of pore pressure. The long beach can cause problems with dust during high winds. Tailings beach Ponded water Tailings discharge Starter dike Perimeter dike Figure 1: Successive construction with the upstream method (after Vick, 1990) 2

15 Centerline construction The centerline construction is similar to the upstream construction, with a dike of borrow material as starting construction. The next raise is done partly on settled tailings, i.e. the beach, the top of the previous dike and the downstream slope, see Figure 2. The centerline method enables the application of drainage layers within the dam body to control the hydraulic gradient in the construction. Consecutive raise of the dam is not restricted as for the upstream method as the following dam section is not placed completely on tailings. Besides, the structure is less dependent on pore pressure dissipation due to application of filters. Tailings discharge Tailings beach Starter dike Internal drain Figure 2: Successive raising using the centerline construction (after Vick, 1990) Downstream construction The downstream method also applies a starter dike made of borrow material. Again, tailings are pumped as slurry together with process water from the crest of the dike along the periphery of the impoundment. Raising the embankment is done at the downstream slope of the previous dike as depicted in Figure 3, thus making it possible to integrate drainage layers to control the hydraulic gradient in the structure. This method is cost-intensive because of the large amounts of fill material needed. 3

16 Ponded water Starter dike Impervious element Internal drain Figure 3: Consecutive raise using the downstream method (after Vick, 1990) Tailings dam construction in practice Historically, remediation of tailings depositions was not regarded being important because the extend of environmental consequences from tailings were not known. Nor was the amount of tailings produced as large as today where it has become more and more profitable to mine low graded ore. Moreover, the indifference to tailings deposition was supported by the lack of economic interest; waste handling presents an additional cost to mine operation. Tailings were simply deposited without any structure or type of remediation. Yet, the increasing awareness of environmental consequences has resulted in harsh requirements on operation and closure of tailings deposition. Today, mining companies and regulators have to meet strict regulations regarding remediation and long term stability of tailings dams (Bjelkevik, 2005b). Construction methods for tailings dams were derived an developed from the construction knowledge on earth- and rock fill dams, and supplementary expertise from e.g. soil mechanics and hydrology was gradually adopted. Applying new knowledge was subject to trial and error, and as it was passed on and refined, structures of tailings dams were gradually adjusted. Tailings dams have then been built and raised continuously along with the production, sometimes lasting over three and more decades. Hence, tailings dams usually do not follow one single construction method as described above, but rather resemble a mixture of different construction principles, sometimes even with special adjustments or support structures, see Figure 4. A variety of reasons may be responsible for such variations and changes, e.g. new knowledge, environmental concern, new demands from the company and society, changing staff, or changes and developments in the mineral extraction processes (Bjelkevik, 2005a). 4

17 Upstream construction Buttress support Impoundment, i.e. tailings Downstream construction Filters Impoundment Support fill Moraine core Moraine Figure 4: Examples of Swedish tailings dam construction in practice It is common practice in Sweden to relate the design of tailings dams to basic soil mechanics and slope stability problems. The general rule given by the Swedish Environmental Protection Agency (Naturvårdsverket) assesses the safety factor for tailings dams slopes based on slope angle and internal angle of friction of the material in the slope. This results in some geometrical definition on how the tailings dam should be constructed, with a minimum seepage distance and maximum slope angle, and thereby involves basically even some kind of definition of a maximum hydraulic gradient for the dam structure (Bjelkevik, 2005a). Because tailings are partly used as construction material in the embankment structure itself, the geotechnical and environmental properties of tailings have to be identified. Knowledge about tailings density, shear strength parameters, and hydraulic conductivity is vital for tailings dams design and construction. Measurements on tailings density are relatively easy to carry out, while shear strength parameters and hydraulic conductivity are difficult to determine. Especially the friction angle is difficult to determine and varies to a large extend. Parameters also vary significantly with sampling and testing techniques, thus complicating calculations assessment of results, and construction design. Major problems in the determination of the properties of tailings arise from two different circumstances: Firstly, the nature of the hydraulic deposition makes representative sampling difficult. Tailings pumped out to the pond in a slurry settle in horizontal layers according to their grain size and specific gravity at varying distances from the outlet. The settlement in the pond is significant for the water level, the beach formation above water table, and the sloping surface below water table. Normally, the beach is steeper close to the discharge point. 5

18 Variations in grain size with distance from the discharge point have a considerable effect on the permeability in the dam construction. Secondly, the particles themselves are subject to variations in the production process. Crushing and grinding creates particular units with high angularity and grain sizes similar to clay and silt, but with different behavior than natural particles at that size. Despite these difficulties, geotechnical properties and possible variances and degradation in the material have to be considered to built safe embankments in both short and long term perspective. Regarding short term assessments, it appears that common geotechnical testing methods for determining properties of tailings are not sufficient (Jantzer et. al., 2008). Another unknown is the long term performance of the material. How are tailings and their geotechnical performance affected by degradation? Tailings dams must be stable over long time periods in terms of not having any environmental impact in the future. They are subject to an after care phase which extends from the closure of mining activity until the company actually may leave the site. In Sweden, this period is approximately 30 years. After that follows a long term phase, during which the tailings dam is supposed to be stable. In this context, the expression stability has sometimes been misunderstood, because long term stability does not only refer to slope stability in geotechnical terms. When speaking about tailings dams, the word stability is often used in terms of overall stability, often in connection with the time frame these constructions are supposed to fulfill their function, including slope stability, as well as stability against environmental harm. Stability of tailings dams therefore often also means their stability against internal erosion, i.e. the elimination of particle migration due to the hydraulic gradient from the artificial lake in long term perspective. According to the Swedish Environmental Protection Agency (Naturvårdsverket) long term stability refers to thousands of years or more, or in a philosophical sense to the next glacial period, after which we do not expect man made structures above ground to be standing. (Bjelkevik, 2005a, 2005b) Hence, internal erosion must not occur, a direct contradiction to a statement made by Lofquist in 1988: No fine-graded soil is absolutely resistant to piping failure. It is a matter of hydraulic gradient and time Internal erosion Research and discussion on internal erosion is mainly related to WRDs. Such dams are built in order to retain water for varying purposes, such as hydropower production, irrigation, or flood control. Therefore, they are exposed to a relatively high hydraulic gradient and seepage pressure, resulting in a natural seepage flow through either the porous structure of the embankment itself, the foundation, or close to abutments. This seepage is generally not regarded being a problem, as long as it occurs in a controlled manner, the leaking water is clean, and there is no particle migration involved. However, if particles in the embankment start to move due to the hydraulic gradient they are exposed to, seepage may exceed and further wash out of grains through an unprotected, i.e. open or unfiltered exit, may result. This particle migration may then become a continuous process including the formation of a pipe, i.e. an unprotected pathway for water and particles to escape. The concentrated leakage can result in serious consequences of damage or failure of a dam. Due to the formation of a pipe-shaped erosion channel internal erosion is often called piping (Terzaghi et al., 1996; Fell et al., 2005). 6

19 Because of topographical advantages and material supply found in Scandinavia, WRDs are often constructed with natural materials as earth- and rock fill dams. They may have a homogeneous structure, but are many times built as zoned embankments in order to control seepage through the dam body. A zoned embankment dam usually has an impervious core as hydraulic barrier, adjacent gradually coarser filter zones to provide seepage control and impede particle migration, i.e. internal erosion, as well as gravel and rock fill in order to provide stability for the construction, see Figure 5. Figure 5: Example of a zoned embankment dam: (1) Core, (2) filter zones, (3) rockfill, (4) rip rap (Fell et al. 2005) Knowledge on tailings dams construction has been derived from WRDs, but there are some significant differences between those two structures. One important distinction is that the tailings dam can never be removed, which is theoretically possible for a WRD. WRDs are built to their final height in one stage and are subject to maintenance and surveillance during their service live, whereas tailings dams are raised in stages and may be inspected and maintained particularly during mining operation, which is only a comparably short period when the mining company is present. Yet, the larger part of a tailings dam service period will be without maintenance and surveillance. Seepage flow through a WRD is regulated by filter zones, thereby allowing controlled drainage and pore pressure reduction in the structure. Tailings dams sometimes also comprise such filter zones (compare Figure 2 and Figure 3). However, their function is often only taken into account in short term perspective. When assessing a tailings dams long term performance, filters are usually not regarded, due to the possibility that they may become clogged, or loose their function because of other degradation processes. Internal erosion in WRDs often starts in connection with irregularities in the material of the embankment, the foundation, at the interface between embankment and foundation, or at conduits and other material intersections where grains find less resistance to migration. Regarding tailings dams, analogical initiating circumstances are encountered: inhomogeneities in the fill material leading to cracks and settlements, variations in fill materials, as well as even concrete surfaces from conduits through the embankment introduce similar problems in tailings dams as in WRDs. Contrary to WRDs, the hydraulic deposition in tailings ponds results in a higher horizontal than vertical hydraulic conductivity, a favorable situation for seepage and internal erosion. However, the continuous rise of a tailings dam compensates settlements resulting from a compressible foundation (Vick, 1990). Therefore, erosion through particle migration in a tailings dams embankment appears to be most likely. 7

20 Concerning erosion in the embankment, two different modes of erosion in a dam have been identified: backward erosion and concentrated piping. The difference between these two is the initiation of the process: backward erosion refers to a leakage on the downstream part of the dam progressing backwards to the upstream side, see Figure 6a. Concentrated piping develops in the dam by cracking or leakage directly from the water source to an exit point, where the pipe grows and the erosion increases, as shown in Figure 6b. The final breach mechanisms are profound enlargement of the pipe, unraveling of toe bank, sinkholes or settlements in the crest leading to overtopping, as well as slope instability (Fell et al., 2005). a) b) Figure 6: Backward erosion (a) and concentrated leak (b) (Fell et al. 2005) Extensive statistic research carried out by Foster et al. (2000a) recognized internal erosion as one of the major incident and failure causes for embankment dams. Almost half of all known embankment dam failures were related to piping. In the research on tailings dams operation, Bjelkevik (2005b) found that both international and Swedish data on incidents and failures is incomplete. It was shown that about 20 % of the reported events at Swedish tailings dams were related to internal erosion. These incidents did not result in failure in the past. Nevertheless is the assessment of internal erosion processes in a tailings dam crucial to its long term performance. In summary, internal erosion in tailings dams should be regarded separately as there are several important differences in both structure and construction material. The varying construction stages make it difficult to locate the internal water level and to determine the hydraulic gradient in the structure, which in turn complicates internal erosion and slope stability assessments. Moreover, incidents and failures of WRDs are more detailed reported and investigated than those of tailings dams. WRDs have been subject to extended statistical data collection and research, where especially the phenomenon of internal erosion has led to thorough discussion on filter criteria. This knowledge may not be applied to tailings dams structures without adjustments and separate research on the performance of tailings and tailings dams structures. 8

21 1.2 Objective The objective of this study is to contribute to the understanding of the long term stability of tailings dams with special regard to internal erosion. One of the most important factors with regard to long term stability is the prevention from internal erosion, i.e. particle migration initiated by seepage pressure. Internal erosion is a process not yet completely understood. It is related to the seepage rate, and the seepage rate is directly connected to the hydraulic gradient. The hydraulic gradient is therefore crucial for embankment stability and prevention of particle migration, and the basic research question covered in this thesis is: What maximum hydraulic gradient can we allow for a tailings dam construction in long term perspective in order to not to have particle migration and internal erosion? 1.3 Method Defining a maximum hydraulic gradient for a long term stable tailings dam is a complex task. In general, the design of civil engineering structures is based on experience, knowledge and research. Most materials can be tested, properties determined, and failure mechanisms be described. This is not the case for tailings dams construction, where the construction material differs significantly from natural soil, with a design and service life far longer than other civil engineering structures. Theoretic knowledge on internal erosion is collected in a literature study on the topic, which focuses on hydraulic conductivity, laboratory testing, methods for calculation and critical hydraulic gradients for various soil types. Because of the difficulties the long term perspective introduces to tailings dam design and construction, natural analogies to dam structures can give valuable information. The most recent melting of continental ice in Sweden took place about to 8000 years ago, and from this period many natural dammed lakes can be found today. Examples on such natural dammed lakes which have been stable since the last glaciation are presented and analyzed with respect to the hydraulic gradient the soil has been exposed to over a period of several thousand years. With respect to the research question, an analysis of a natural analogy to a dam construction was carried out at the company area of Bolidens mine in Gällivare. The analysis includes field studies and laboratory testing, with the aim of determining the properties of the soil. The hydraulic conductivity was tested in laboratory conditions and related to conditions in situ, as well as calculated with empirical relationships on basis of the determined properties. Results are compared to knowledge collected in the literature study, and conclusions on the hydraulic conductivity and gradient in the natural dam analogy are then related to tailings dam construction. 9

22 1.4 Outline To provide a basis for research, a State of the Art report on internal erosion is presented, covering definitions, geometric and hydraulic criteria on seepage through soil, the hydraulic conductivity and critical gradients, as well as factors initiating internal erosion, statistics and risk assessments. An overview on natural analogies based on a study provided by the Geological Survey of Sweden SGU is given in the next section. This overview provides basic information on the location, formation and age, as well as material and hydraulic gradient the structure has been exposed to. As a next step, a natural formation has been analyzed with respect to geotechnical properties. Analysis includes in situ testing and groundwater monitoring over a period of two years. Laboratory testing on material collected in test pits was carried out for determination of the soils properties, and tests on the hydraulic conductivity were conducted. On basis of laboratory testing, the hydraulic conductivity of the soil is evaluated, both by relating laboratory results to in situ conditions as well as by calculating the hydraulic conductivity with well established empirical relationships. Finally, the results are discussed and related to tailings dam construction. 10

23 2 STATE OF THE ART 2.1 Introduction Internal erosion is a problem which has been responsible for accidents and failures of dams all over the world. Wherever water is dammed by an embankment of soil, gravel and rock, the soil particles are exposed to a certain hydraulic gradient. Seepage forces and openings in the soil skeleton can create a situation where particles start to move and are washed out through an unprotected exit in the soil structure, a process which, upon progression, may eventually result in erosion holes and concentrated leakage through an embankment. An overall failure statistic carried out by Foster et al. (2000a) showed that 30,5 % of failures of large embankments, i.e. WRDs, were due to internal erosion through the embankment. Bjelkevik (2005b) analyzed incidents at Swedish tailings dams and found that 19% of the reported events were related to internal erosion. Before giving an overview on internal erosion itself, hydrodynamic soil deformations are classified in order to define various expressions related to internal erosion. Further, principle thoughts and research on internal erosion is presented in a rough timeline. Studies on internal erosion originally focused on mechanical principles: geometric criteria, i.e. particle and opening sizes in the soil matrix which allow particles to move, and hydraulic criteria, i.e. hydrodynamic forces responsible for displacement of the grains. Many geometric considerations are still based on Terzaghi s filter criteria from the early 1920 s (Terzaghi et al., 1996). Both the internal stability of a material itself as well as filtering concepts of two adjacent materials have been tested and discussed extensively by e.g. Sherard (1984, 1986), Kenney & Lau (1984, 1985), Skempton & Brogan (1994), and Fannin & Moffat (1996). In contrast, research on hydraulic criteria appears not to be as detailed as that of geometric criteria, which is due to difficulties regarding representative laboratory testing and varying methods for the calculation of the hydraulic conductivity. Critical hydraulic gradients have been summarized by Perzlmaier (2007) based on research by Bligh, Lane, Chugaev, Müller- Kirchenbauer, and Weijers & Sellmeijer. During the last few years, publications and research seem to have concentrated on phenomenological descriptions of internal erosion, i.e. how does the process start and what are the factors that lead to progression and eventually accidents and failures? Seepage paths may create from settlements and subsequent cracks in the embankment and have been described by Kjærnsli et al. (1992) and Fell et al. (2005). Other inhomogeneities such as material segregation or frost action have been identified as initiation factors by Kenney and Westland (1993), Kenney & Lau (1985), Milligan (2003), and others. In addition, the term hydraulic fracturing introduced by Sherard (1986) is explained and discussed. Finally, statistical analysis and risk assessments are presented. Thorough statistical analysis on accidents and failures in WRDs has been carried out by Foster et al. (2000a, 2000b) with the objective to make future estimations on the likelihood of piping failures from the historic frequency of accidents and failures in embankment dams. Apart from statistical analysis, the European Working Group on Internal Erosion has been working on risk assessments and risk analysis. 11

24 2.2 Classification of hydrodynamic soil deformations Hydrodynamic soil deformations include local deformations as well as global deformations such as e.g. slope stability problems arising from seepage forces. The term internal erosion is often used for particle transport in general. Many other expressions are used in the field of internal erosion, such as piping, suffusion, backward erosion. A distinct classification has been made in German-speaking countries, where Ziems was the first to define and organize hydrodynamic soil deformations in This classification has been accepted by various different German authors such as Wittmann (1980), Muckenthaler (1989) and Perzlmaier (2007). English speaking literature generally does not use similarly detailed specifications between suffusion, erosion, piping and clogging. In English literature, internal erosion is sometimes referred to as piping, suffusion may also be called internal instability, and the term contact erosion may also represent a situation of joint erosion. Therefore, an overview of the different processes and definitions are presented below. Generally, the above mentioned processes can be subdivided further according to their location, i.e. external, internal, and contact suffusion/erosion, respectively. If material transport occurs at the surface, it is external, and if it occurs inside a soil, it is internal. A special form of internal transport is contact suffusion and erosion, where the process develops at an interface between two different materials such as between fine and coarse grained soils, i.e. base soil and filter. Clogging is the inverse process to suffusion and erosion where pores are clogged by particle transport. Suffusion Internal Instability Erosion Clogging Colmation, Colmatation External S. External E. Internal S. Contact S. Internal E. Piping Contact (Joint) E. Backward Erosion / Concentrated Leak Figure 7: Classification of hydrodynamic soil deformations (Ziems, 1969) Suffusion Suffusion refers to the selective transportation and washing out of fines from the coarser soil skeleton (Figure 8), where both the coarser skeleton and the original volume of the soil initially remain (Ziems, 1969). This applies to so-called internally unstable material. Suffusion alters the grain-size distribution, increases the porosity and the hydraulic conductivity, and decreases the density and the coefficient of uniformity. The washout of fines may create an initial state for internal erosion; internally unstable soils are therefore regarded as being susceptible to erosion. Cont act suffusion describes the transportation of fines into the voids of an adjacent coarser soil in particular. Suffusion sometimes is spelled suffosion. Soils susceptible to suffusion are gap-graded and have a gravel content > 60 %, or coarse and widely graded (Wan & Fell, 2004 a, b). 12

25 Figure 8: Schematic description of suffusion after Ziems (1969) Erosion When soil particles of almost all different grain sizes are washed out from a soil to a degree that the soils volume decreases and the structure of the soil skeleton is altered, erosion takes place. The progressing loss of material may finally lead to breach, or to sinkholes and settlements of the embankment, thereby lowering the capacity of water storage. Variations in density and compaction, cohesion and stresses, and the anisotropy of soil enhance the development of internal erosion (Ziems, 1969). Erosion can be subdivided into external, internal and contact erosion. Similar to contact suffusion, contact erosion forms at an interface between different materials as shown in Figure 9. The English term contact erosion is mainly used for erosion in connection with conduits or other concrete structures through the embankment. These have been identified as increasing factor for internal erosion, since the intersection between earth fill and the concrete surface can act as a wall for a pipe, thus supporting an open channel for material transport. The German speaking countries are more specific and use contact erosion for intersections between different graded soils and joint erosion for intersections with concrete structures. However, joint erosion will be identified as contact erosion in this paper to avoid additional terms. Figure 9: Internal erosion at the interface of different materials after Ziems (1969) 13

26 Internal erosion is often referred to as piping, which implicates that a continuous open seepage path grows, i.e. a pipe, through which particles are carried away, see Figure 10. The development of a pipe requires that the material is able to support a roof (Fell et al., 2005). Figure 10: Piping in uniform material after Ziems (1969) A special form of internal erosion is backward erosion, where material starts to erode at a free and unfiltered exit of the seepage path. The process then continues backwards towards the upstream face of the dam. Concentrated leakage erosion may result from poor compaction, differential settlements, frost action, desiccation or hydraulic fracturing Clogging Clogging, sometimes also referred to as colmation or colmatation, is the inverse process to erosion, where fine particles are deposited in the pores and the hydraulic conductivity is reduced. The rearrangement of the particles in the soil matrix can have a positive effect on the structure as the geometric (or internal) stability is partly restored Heave Heave (or blow out ) may also be counted among hydrodynamic deformations. Here, the seepage force acts vertically upwards against gravity forces, and the pressure gradient exceeds the weight of the submerged soil, so that the effective stress between soil particles becomes zero. This is primarily a problem of balance between upward and downward acting forces; it may, however, leave an unfiltered exit and initiate backward erosion if the seepage gradient remains sufficiently high. 14

27 2.3 Geometric criteria The mutual effects between different soils and seepage forces are described in various criteria. Geometric criteria refer to the structure of soil only, where either one soil fabric or two adjacent soil layers are regarded in terms of their grain size distribution, pore volume and openings in the soil skeleton that leave a path for particles to escape Filter concepts Erosion has for a long time been regarded as a filtration problem, i.e. particle migration through a porous soil system. Focus has been on the geometry of the pores and the grain size distribution which was considered being responsible for possible particle transportation. Selffiltering or internal stability of a soil implies that the coarser particles act as filter for the finer particles, and loss of granular material through voids in the soil is impeded. Erosion and suffusion are closely related problems; for instance, erosion in a fine graded base soil may become suffusion in the adjacent filter layer. Filters have traditionally been applied to provide a base soil with the sufficient drainage to prevent from exceeding pore pressures and are therefore usually coarser than the protected base soil. On the other hand, the structure of the filter has to be such that movement of particles is impeded by sufficiently small pore spaces. Filters are often applied in the body of embankment dam structures. Filtration problems may also occur between the dam body and the foundation, in case the nature of the foundation does not meet filter criteria. Terzaghi introduced a filter concept in the early 1920 s. This criterion has been modified, refined and been adapted in different countries practises over the years. However, the traditional thought of providing a stable and filtering material remains and is still in use. The two established rules are: (1) for erosion control, the relation between filter (D) and base soil (d) should be D 15 /d 85 4, and (2) to guarantee drainage the relation should become D 15 /d The index denotes the grain size for which 15% or 85% is finer, respectively (Terzaghi et al., 1996). These relations have been discussed by many researchers, among them Sherard et al.(1984), Lafleur (1984), Kenney & Lau (1985), Lafleur et al. (1989), and Chapuis (1992). Despite all studies and filter testing, the concept of retaining particles and ensuring of drainage is still in use. Moreover, Sherard et al. (1984) stated that Terzaghis criteria employ the appropriate characteristics of the filter and the base, so that the ratio should be continued as the main criterion for judging filter acceptability. Recent laboratory studies carried out by Fannin & Moffat (2006) also approved the concept introduced by Terzaghi. Filter criteria were developed with respect to the interaction of a filter with a base soil, i.e. two adjacent soils. The self-filtration problem, i.e. internal instability in one soil itself, was not yet regarded. Kenney and Lau (1985) described the soils structure being consistent of two units, a primary fabric which transfers load and stress, and loose particles in the voids of the primary fabric. The loose particles have the ability to move through the skeleton of coarser particles. Skempton & Brogan (1994) found that internally unstable sand experienced fine grain migration, which was explained by Kenney and Lau s theory of a load carrying skeleton where the smaller particles do not transfer effective stress, see Figure

28 Figure 11: Example of stable and unstable soils. (a) internally stable material, (b) unstable gap-graded material, (c) unstable material with loose large grains (Vattenfall, 1988). Sherard (1979) defined an instability degree for suffusion, using a similar principle to that of Kenney & Lau, namely by dividing the grain size distribution curve into two parts, a coarse and a fine fraction, so that the shape of the gradation curve could be analyzed. The relation between the coarse fraction and the fine fraction of the gradation curve should meet filter criteria and be D 15(coarse) / D 85(fine) < 5. Internal stability and filtration has been recognized being not only dependent on grain size distribution, but influenced by factors such as porosity, the disturbing effects of forces from seepage and vibration, and segregation of the soil during placement (Kenney & Lau, 1985, Chapuis, 1992). High seepage forces and vibrations were considered not to occur and therefore not to be regarded in dam constructions according to Milligan (1986) Internal stability and characteristic grain size Both the internal stability and the hydraulic conductivity of a soil are related to the size of the grains. The inherent stability is ascribed to the shape of the gradation curve, i.e. the coefficient of uniformity C u, whereas the hydraulic conductivity is influenced by the size of the pore channels in the soil. However, researchers have not come to an agreement on a representative grain size that governs the size of the pore channels. Kenney et al. noted that the size of the pores primarily depends on a representative grain size, whereas the shape of the complete gradation curve was noted to have a minor effect on the pore size channels in the soil structure (Kenney et al., 1984). Hazen found in 1892 that the relationship between pore size and hydraulic conductivity was proportional, and that an increase in hydraulic conductivity is related to the square of a characteristic grain size D e (Terzaghi et al., 1996). The characteristic grain size that governed the hydraulic conductivity was set to D 10, i.e. a 10% limit of the particles of the soil. Kenney et al. (1984) stated that the permeability was primarily dependent on the finer particles with the grain size D 5. Sherard et al. (1984) found this grain size to be D 15, whereas Fell et al. (2005) agree with Hazen on D 10 being the defining factor. This discussion shows the variety of views; a consensus seems to be difficult to reach. Nevertheless, it is generally agreed on that it is the fine particles and roughly 5 to 15 % of the soil which affect the hydraulic conductivity considerably. While the hydraulic conductivity has been recognized being strongly dependent on the amount and size of soil particles between D 5 and D 15, the filtering ability of a soil has been related to the shape of the total grain size distribution. For instance, filter criteria concerning the ratio between base and filter have sometimes been extended with recommendations such as that in the USBR method: The curves of the base soil and the filter should be parallel (Fell 16

29 et al., 2005). On the contrary, Sherard et al. (1984) found that such a similar shape of gradation curves is not necessary. However, broadly graded or gap graded soils may display a lack of certain grain sizes that are needed for an optimum in structure, pore sizes and successful filtration. The coefficient of uniformity C u was not accounted for in filter criteria, and suggestions on the shape of the grading curve by means of the coefficient of uniformity C u, i.e. the ratio D 60 /D 10, have not been clearly expressed. Yet, most filter tests have been carried out using uniformly graded material with a coefficient of uniformity lower than 6. Materials for which the ratio is less than 10 are regarded as being stable, while a ratio above 20 is considered being unstable (Skempton & Brogan, 1994). Practical problems can be connected to wide graded materials because of their susceptibility for segregation during placement. Milligan (2003) stated that the instability of material tested by Kenney & Lau (1985) was closely related to material which was typically susceptible to segregation during construction works. 2.4 Hydraulic criteria Hydraulic criteria summarize the boundary conditions for critical hydraulic loads acting on a soil particle and forcing it to move. Even if the soils structure geometrically allows an opening size in the pore matrix large enough for displacement, a hydraulic load exceeding the drag force, i.e. the particles self-weight and its interaction with other grains, is needed to actually move a particle. The type of flow and the hydraulic conductivity of a soil are significant properties in internal erosion processes. In order to present limit values for the hydraulic load a soil can be exposed to without experiencing particle migration critical hydraulic gradients have been introduced Hydraulic conductivity In order to understand the process of water permeating a soil matrix it is important to look at the various forms water can exist in the matrix. A saturated soil is a two-phase system where all voids are filled with water. This system contains free pore water, i.e. water which is able to move in the voids. In the three-phase system of an unsaturated soil the voids are partly filled with water and air. Water in the pores is attracted to the particles and tends to wet the solid surface, while air strives after minimizing contact to solids and tries to form as spherical bubbles as possible. Water wetting the solid surface can be summarized as bound water. Adsorbed water is an extremely thin layer of water bound strongly to mineral surface by molecular forces so that hydrodynamic forces cannot move this layer. A thicker hydration shell of bound water around grains can be called solvate water. Density and viscosity of this water vary from normal water, but these molecules are mobile when exposed to hydrodynamic forces (Kézdi, 1974). The hydraulic conductivity of a soil is influenced by many factors. Among them are porosity, grading, fabric, its chemical composition, and the properties of the permeating fluid. However, the size and shape of the pores is essential for flow through the soil matrix. Given that smaller particles are located between larger and therefore control the void size, fine particles are much more significant for the hydraulic conductivity than the large grains (compare Figure 11). In addition, physical factors such as saturation, compaction or temperature make the hydraulic conductivity a dependent variable rather than a material constant (Mitchell & Soga, 2005). 17

30 Permeability and hydraulic conductivity When speaking about water flow through the porous media soil the soils ability to allow this water flow is generally referred to as permeability. However, this term is not correct, because permeability in general describes a materials ability to transmit a fluid. In case of water flow through soil it is therefore more precise to refer to the hydraulic conductivity k because it is dependent on the properties of water, i.e. density and viscosity, which are affected by temperature. The hydraulic conductivity k is related to the intrinsic permeability Κ, the property of fluid flow through porous media with regard to the dynamic viscosity μ, the fluid density ρ, and the acceleration of gravity g: kμ K = [m 2 ] ρ g Note that the intrinsic (or specific) permeability K is measured in units of area [m 2 ], whereas the hydraulic conductivity k is measured in units of velocity [m/s]. Darcy s Law and its validity Water flow imposes a seepage pressure on the soil mass as a result to the dimensionless hydraulic gradient i, which is defined as the ratio between the rate of head loss Δh to the flow distance ΔL: Δh i = [-] ΔL The hydraulic conductivity k (or Darcian coefficient of permeability) is a proportionality factor in Darcy s Law that states the proportionality between flow rate and gradient. The flow rate is inversely proportional to the length of the flow path. Darcy s Law in its general, onedimensional form relates the hydraulic conductivity k to the ratio between discharge velocity v and hydraulic gradient i: v k = [m/s] i This linear relationship is valid for laminar flow in porous media, i.e. the validity is dependent on the Reynolds number Re which denotes the ratio between inertia and viscous forces to describe flow transitions from laminar to turbulent. It is generally accepted that Darcy s law applies at low Reynolds numbers up to 1 and 10 (Bear, 1972; Todd & Mays, 2005). The limit value of 10 is confirmed in numerical studies carried out by Hellström, who states that inertiaeffects have to be taken into account when the Re exceeds 10; i.e. the linearity of Darcy s law cannot be applied longer. The pressure drop then increases and non-linear terms have to be added; a problem that can be solved by using Forchheimers equation: 18

31 K Δp Q Q = + b μ L A A m where K is the intrinsic permeability, μ the viscosity of the fluid, and Q the flow rate through the area A. Δp describes the pressure drop over a length L in flow direction, b denotes a property of the porous media and m is a measure of inertia. Turbulent effects occur when Re is larger than 100 (Hellström, 2007). Limit values for Darcian, Forchheimer and turbulent flow given by Perzlmaier et al. (2007) vary slightly compared to those presented by Hellström: Darcian flow occurs at Re between ,3, Forchheimer flow at Re between 5 80, and turbulent flow at Re larger than 120. These results appear to be unusually precise. For instance, Kézdi (1974) states that the it is not possible to determine the exact transition from pure laminar flow to slightly turbulent flow. Variations of these limit values may be a result of different empirical observations as well as varying definitions of the Reynolds number, which is, amongst others, defined by means of a length dimension of the flow path. The flow length can therefore be regarded on macroscopic basis being equal the length of a sample, or on a microscopic basis to account for interstitial flow and the tortuosity, as suggested by Mitchell & Soga (2005), or be defined by some representative grain size (Bear, 1972). Flow in soils finer than coarse gravel is laminar, thus obeying Darcy s law (Mitchell & Soga, 2005). Detailed values are given by Perzlmaier et al. (2007) stating that the validity of Darcy s relationship is given for effective particle diameters smaller than 1,5 mm, whereas turbulent flow starts at diameters larger than 20 mm. In addition to the deviations from Darcy s law at high Reynolds numbers, a minimum gradient below which either no or non-linear flow takes place has been discussed by some authors (Bear, 1972). Laboratory tests have shown that Darcy s law deviates in fine-grained materials and increases more than proportionally with increasing gradient. This phenomenon is most significant at low hydraulic gradients, and it is presumed being due to migrating particles that open or block passages, as well as local consolidation or swelling in the soil. The assumption that non-newtonian flow properties, i.e. a fluid with a non-linear relation between shear stress and strain rate that cannot be described by a constant viscosity value, was found not to be responsible for non-darcian flow (Mitchell & Soga, 2005). Laboratory testing The hydraulic conductivity can be measured in laboratory tests. There are various systems for available for testing, but virtually all are based on the idea to expose a soil sample to a certain gradient and measure the quantity of water percolating through a sample with a given length and cross-sectional area during a time span. A basic testing arrangement is shown in Figure

32 h Q V l Figure 12: Standard laboratory test to measure the hydraulic conductivity of a soil In case of retaining a constant water supply to the intake the hydraulic gradient will remain the same during testing and we speak of a constant-head permeameter. In a falling-head permeameter, there is no additional water supply and the total head decreases with time during testing. During a constant head test, the output flow is measured over time and related to the head that creates a constant seepage pressure. In a falling head test, the input flow is measured by using a standpipe with decreasing water level over time that indicates a negative velocity. In general, it is stated the constant-head permeameter is suited for coarse material with high to moderate hydraulic conductivity, whereas the falling-head permeameter gives better results for fine-grained material with lower conductivity (Smoltczyk, 2002; Terzaghi et al., 1996). On the other hand, Lambe & Whitman (1979) suggest to restrict the use of the falling head permeameter to pervious materials. In addition, it is stated that tests with variable head should not be carried out on unsaturated samples due to changes in saturation during testing. Testing a soils hydraulic conductivity is often difficult because of the variety of factors influencing the test. Hydraulic gradients found in nature are rarely not above one, which complicates duplication of field conditions. During laboratory testing it is usually not suitable to apply such low gradients because of the extensive and hence inappropriate time needed for testing (Mitchell & Soga, 2005). In order to receive test results in a reasonable time span fine-grained soil samples are therefore often tested at much higher gradients than those found in nature. The evident disagreement whether to use the constant-head or the falling-head permeameter may arise from varying practical experiences. A falling head test is easier to carry out on a relatively dense material because the measurement of inflow is faster and can be observed more accurate instead of trying to achieve a steady flow through the sample and measure the output. Moreover, Kézdi states that evaporation from the slow output would make constanthead permeameter test results unreliable. Permeameters can have rigid wall or a flexible wall devices. A rigid wall device has the advantage of being cheaper and are more convenient when testing compacted samples. However, the rigid wall of a cell decreases the resistance for particle migration, and so-called sidewall leakage can result in inaccurate and too large measurements of the hydraulic conductivity. Therefore, a flexible wall permeameter should be used when testing fine grained materials, because they minimize the risk for leakage along the sides of the sample. They are 20

33 also preferred when testing samples with back pressure, as the cell allows for measurement of volume change (Daniel et al., 1985) With regard to Darcy s law and its validity it is important to try to keep the flow through the sample laminar. Hence, various materials require adjustments of boundary conditions in laboratory testing in terms of keeping the seepage pressure sufficiently large to actually obtain flow through the sample, while at the same time retaining laminar flow. This is difficult to achieve when testing very fine-grained materials with low hydraulic conductivity. Such a sample would demand the application of relatively high pressures in order to receive results. An alternative is therefore to calculate the hydraulic conductivity indirectly from consolidation test (Kézdi, 1974). An additional possibility is the use of a constant-head apparatus with a cell similar a triaxial testing device, which gives the possibility to apply high pressures with an associated backpressure. The hydraulic gradient is then given by the pressure difference between top and bottom of the sample and can be controlled with suitable instrumentation. (Smoltczyk, 2002). Unsaturated samples Darcy s law applies also for unsaturated samples. However, the hydraulic conductivity is not constant, but dependent on the degree of saturation. In a non-saturated system, pores are filled with both water and air, where water covers the particle surface, and air is filling the voids center. Water will then, when applying low pressure, move first along the particles adsorbed water film. The hydraulic conductivity is therefore dependent on whether the voids are filled with water and in how far the pore water is continuous, see Figure 13. Figure 13: Unsaturated soil with continuous water film (Mitchell & Soga, 2005) The unsaturated hydraulic conductivity in a soil is, amongst others, a function of the volumetric water content, the degree of saturation, matric suction. The hydraulic conductivity can be expressed as a function of the saturation, which is related to the Kozeny-Carman equation presented in the following section: k h γ = μ k T S e S e + 3 [m/s] 21

34 where k 0 is a pore shape factor, T a tortuosity factor, S 0 the wetted surface area, and e the void ratio. As shown in the equation, the hydraulic conductivity varies as the cube of the degree of saturation. This relationship has been found adequate for degrees of saturation above 80% (Mitchell & Soga, 2005). Applying a seepage pressure, the voids first are filled with water before water can actually permeate through the system. A soil with a degree of saturation above 80% behaves as if it was saturated, while a soil with a degree of saturation below 20% appears not to be permeable (Kézdi, 1974). In the latter case, the water content is reduced so far that only water strongly attracted to the particles is left, so that the non-continuous path in the pores is impeding the flow. It will then take longer time to fill the voids with water, so that the hydraulic conductivity will increase over time (Mitchell & Soga, 2005). The flow through an unsaturated sample is not constant, but depends on the amount of pores as well as their interconnection. To determine the hydraulic conductivity in laboratory tests, the saturation conditions of the sample have to be controlled. Again, applying back pressure for saturation of the sample can be useful in order to press out air pockets. However, air pockets can not only block pores in the sample, but also other instrumental parts in the system (Knutsson, 2009). Evaluation of laboratory testing The determination of the hydraulic conductivity often involves problems and errors that are common for laboratory testing. Leakage in the interface between sample and cell wall may increase the flow and can lead to particle migration. Due to practical reasons, samples are often tested under an increased pressure gradient to reduce time and costs in the laboratory, with the result of excessive flow rates, particle migration and clogging of filters. Such an increase of pressure gradient does not only affect the flow and the application of Darcy s law, i.e. laminar, inertia dominated or turbulent (Hellström et al, 2009), but may also lead to actually testing the hydraulic conductivity of the filter rather than that of the sample in case of clogging (Stenman, 2009). The validity of Darcy s law is given when both volume and shape of the water passages remain independent from pore water pressure and time, i.e. the flow remains steady. Deviations can occur due to air bubbles or particle migration which decrease or clog percolation. On the other hand, increasing pressure can decrease air pockets in the sample, thus increase the hydraulic conductivity (Terzaghi et al. 1996). Test results can vary significantly because of the difficulty of uniform sample preparation, i.e. differences in degree of compaction or water content. Another difficulty is the saturation of a sample which cannot be controlled or checked (Fell et al, 2005), but is rather an assumption on basis of some kind of steady flow condition. Laboratory testing depends on the reliability of the sample, and the fabric of the soil can have a significant influence on testing results. The value for hydraulic conductivity tested in laboratory conditions does not take stratification or fissures in soil masses in situ into account. The horizontal hydraulic conductivity k h may be considerably higher than the vertical conductivity k v, and the ratio between horizontal to vertical conductivity k h /k v may be as high as 100. Representative samples and relations between test results and field conditions are therefore difficult to obtain (Fell et al, 2005). 22

35 Calculation of the hydraulic conductivity To calculate the hydraulic conductivity of a soil, a variety of different equations have been introduced. All equations take some kind of effective particle size and/or void ratio into account. In general, laminar flow is assumed. Hazen suggested a simple formula from his works with loose sand, where the hydraulic conductivity k was directly related to the effective grain size d 10 (in mm): k = C ( ) 2 [m/s] d 10 In order to calculate the hydraulic conductivity k in m/s, the factor C in the formula is usually set to 0,01, and the grain size is given in mm. If k is given in cm/s, C is set to 100. This equation is not recommended for application in case of clay and coarse gravel (Fell et al. 2005). Chapuis (2004) presented a method for estimation by regarding both the effective diameter d 10 (in mm) and void ratio e of the soil. The method may be applied to clean sand and gravel and is an extension of Hazen s formula, which did not take the actual porosity into account. 0, ,4622 d e k = [cm/s] 1 + e The Kozeny-Carman equation was originally proposed by Kozeny and improved by Carman and describes the permeability of porous media. In case of complete saturation, i.e. S = 1, the intrinsic permeability K can be calculated by taking a pore shape factor k 0, a tortuosity factor T, the wetted surface area S 0, the void ratio e into account (Mitchell & Soga, 2005): K = k h μ 1 = γ 2 k T S e 1+ e [m 2 ] Carrier (2003) gives detailed instructions on calculations with this formula. For water at 20º, the term μ/γ equals 9,93*10 4 1/cm s. The pore shape factor k 0 is set to 2,5, while the tortuosity T 2 is equal to 2. The wetted surface area S 0 can be estimated from the grain size distribution by S 0 = 6/D eff, and D is related to an effective diameter: 100% D eff = [cm] ( f / D ) i ave,i where f i is the fraction of particles between two sieve sizes, and D ave,i is the average particle size between two sieve sizes. 23

36 There are various formulations of the Kozeny-Carman equation; one published by Fell et al. (2005) takes the void ratio e, the specific surface area S and an angularity factor f into account: k = 3 2 e [m/s] 2 fs 1+ e The angularity factor f considers the shape of the particles and ranges from 1,1 to 1,4 for round and angular particles, respectively. In this case, the specific surface S can be calculated with regard to the maximum and minimum size particles in the soil d 1 and d 2 (in mm): S = d 6 1 / d 2 Aubertin et al. (1996) present a different version of the Kozeny-Carman equation k = μ c1g 3 2 w ρ w Dr + 1 e 2 S (1 e) [m/s] where c 1 is a material parameter, g is the gravitational acceleration, μ w is the viscosity and ρ w is the density of water, D r is the average relative density of the solid grains, and S is the specific surface. As shown here, it may appear that the Kozeny-Carman equation is slightly different in each text, because there are various parameters. However, the variation is always related to the grain size distribution and the grains specific surface, thus affecting the wetted surface area as well as the pore space and shape. Even though the parameters seem to be slightly different, the equation itself remains the same, which demonstrates the importance of pore size and the grain size distribution. Mitchell & Soga (2005) point out that the flow rate is proportional to the fourth power of the radius On basis of experimental studies, Aubertin et al. (1996) introduce a modified version of the Kozeny-Carman equation on the hydraulic conductivity of tailings. The equation includes a grain-size distribution function as well as an extended void ratio function to take tortuosity into account: γ w k = c D μ 2 10 C 1/ 3 U 3+ x e (1 + e) [m/s] The authors express some reservations in the use of this relationship: even though the results show good agreement with experimental tests, it should be considered that the angularity of tailings is not included and that the equation refers to materials which are homogenized and 24

37 saturated. The angularity of the crushed material may have a relevant influence on the hydraulic conductivity. It is therefore pointed out that the hydraulic conductivity may vary by a factor of about 1,5 2, where the value k is larger for rounded particles than for angular grains. The deficiency in the relationships presented by Hazen and Chapuis is that attention is primarily paid to particle size and void ratio. The composition, soil fabric and the saturation of the soil material are not included. However, Fell et al. (2005) do not regard the formula being more precise than Hazen s relationship, despite the attention to particle shape in Kozeny-Carman formula. In contrast to that, Aubertin et al. (1996) state that their version of the Kozeny-Carman relationship offers a relatively good fit to the experimental results for part of their experiments on hydraulic conductivity of tailings. In contrast to that, Carrier (2003) criticizes the use of Hazen s formula due to the variance of the factor C. It is stated that Hazen s formula is widely used because of its simplicity, despite its validity for a water temperature of 10º instead of the usual 20º. This discrepancy is thought to be neglected because of the formulas general inaccuracy. A reason for not using the Kozeny-Carman equation is considered to be related to the specific surface area; i.e. Carrier states that many geotechnical engineers are not used to measure the specific surface area of particles, and that factors such as surface area of tortuosity are difficult handle Critical hydraulic gradients Perzlmaier et al. give an overview of hydraulic criteria based on critical hydraulic gradients and critical flow velocities in their contribution Hydraulic Criteria for Internal Erosion in Cohesionless Soil, which was published in the intermediate report of the European Working Group of ICOLD in The following part is mainly based on this paper, if not referred to other authors. Because the flow to an unfiltered exit differs from the flow within a soil matrix, two different critical gradients are considered. For backward erosion to an unfiltered exit, the basic idea for the critical gradient is related to the relation for heave or blow-out. The critical gradient becomes ( n) ( γ s γ w ) icrit = 1 γ w The value for the critical gradient is reduced for a process occurring inside the soil matrix: i crit = { } ( 1 n ) ( γ ) s γ w 0,7 to 0,8 γ w Criteria for suffusion considered acceptable by Perzlmaier et al. (2007) are dependent on the coefficient of uniformity C u : i crit 0,3 to 0,4 = 0,2 for 0,1 for C u 10 C for C u < > 20 u 25

38 Empirically derived values for hydraulic criteria for both backward erosion and piping have been summarized by Perzlmaier and shown in Table 1. The average gradients are divided into groups according to the material. The lowest average gradients were presented by Lane in 1935, followed by the values from Bligh from 1912, and by those of Chugaev from Mueller-Kirchenbauer presented a range of critical gradients by defining upper and lower limits. Table 1: Critical hydraulic gradients to initiate backward erosion and form a pipe (Perzlmaier et al, 2007) Type of soil Gravel Coarse sand Medium sand Fine sand I crit Chugaev 0,25 0,25 0,11 0,10 I crit Bligh 0,11 0,083 0,067 I crit Lane 0,095 0,067 0,056 0,048 I crit Mueller-Kirchenbauer, lower limit 0,12 0,08 0,06 I crit Mueller-Kirchenbauer, upper limit 0,17 0,10 0,08 I crit Weijers & Sellmeijer, C u = 1,5 0,28 0,18 0,16 0,09 I crit Weijers & Sellmeijer, C u = 3 0,34 0,28 0,24 0,14 Weijers & Sellmeijer (1993) described a method for calculation of the average critical gradient in 1993, where a mathematical description by means of the geometry of the structure was presented. The difference in water head H and length of the seepage path L, as well as a certain seepage thickness layer under the dike D are related to each other. In addition, the coefficient of uniformity C u (i.e. d 60 /d 10 ) is taken into account. The study presents a solution for calculation of a critical water head and is verified by large scale tests. Yet, it is limited to geometry and material properties of the sand used in the Netherlands, and application of the results for different material is uncertain. Skempton & Brogan (1994) state that the critical gradient at which particles start to move differs from the theoretical critical hydraulic gradient that has been applied in filter studies. It is shown in tests that instable material experiences particle migration at much lower hydraulic gradients than stable material. For laboratory conditions and horizontal flow, the critical gradient of stable material is set to 70 %, whereas the critical gradient of unstable material is only 17 %. These gradients obtained in the laboratory are much higher than the gradients found in natural conditions. However, it is a well known practice to apply much higher gradients in the laboratory tests as described in chapter 2.4.1, so that conclusions from laboratory results to in situ conditions should be drawn with precaution. 26

39 2.4.3 Design criteria for Swedish tailings dams For the design and construction of tailings dams in Sweden, none of these critical values are used. Instead, the factor of safety for such constructions is related to basic soil mechanics and slope stability, which can be found in various textbooks on soil mechanics, e.g. Cernica, (1995). The slope stability is generally simplified as shown in Figure 14, where an infinite slope of constant inclination with a parallel water table and seepage flow is considered. Figure 14: Basic slope stability consideration (after Cernica, 1995) The shear strength τ f along a potential failure plane can be calculated taking the effective stress σ and the internal angle of friction φ of the material into account: τ f = σ ' tan φ' Relating the retaining and driving forces, i.e. the mobilized shear strength and the actual shear strength in the sloping mass to each other, the safety factor F for such a theoretic plane becomes τ f F = τ This equation is origin for the determination of long term stability of slopes of tailings dams applied in Sweden. The safety factor is based on the internal angle of friction, as there are currently no written criteria available, especially with regard to long term stability. A general rule is given by the Swedish Environmental Protection Agency (Naturvårdsverket), where the safety factor FS for tailings dams slopes are assessed by means of the internal angle of friction φ, the angle of the slope β, and the mass density ρ: ρ tan φ' FS = 1+ ρ tan β 27

40 A rule of thumb for practical use to determine a maximum slope angle is to apply the materials friction angle φ multiplied by the factor 0,5, and this rule is, despite the lack of scientific verification, commonly accepted for the long term stability of tailings dams slopes in Sweden. According to the Swedish dam safety guidelines RIDAS, the safety factor FS of 1,5 is regarded being sufficient (Bjelkevik, 2005a). By defining a geometric design with a maximum slope angle, a minimum length of the embankment is defined. Thus, a minimum length for the seepage path is set, and thereby also a maximum hydraulic gradient somewhat defined Further aspects on hydraulic force To develop a model on critical hydraulic head, the hydraulic shear stress has been subject for discussion. Wan and Fell made an attempt to investigate the rate of erosion of soils in embankment dams on basis of a hydraulic shear stress: the soils erosion characteristics are related to its behavior regarding the rate of erosion under a certain hydraulic shear stress. It was found that this critical shear stress was lower for coarse grained soils, which implies that they erode more rapidly than fine grained soils. By using two test methods, the hole erosion test and the slot erosion test, a critical shear stress can be determined. However, this value varies to a large degree and cannot be related to other soil properties. A calculation of the hydraulic shear stress in an open crack in the embankment is presented, but to be able to analyze this value, the geometry of the crack has to be known. The problem of non-visibility of such cracks and erosion processes inside the soil matrix sets a limit for application. The study confirms practical information about the erosion rate: it is strongly influenced by the degree of compaction and the water content of the soil. Compaction to a high dry density on the wet side of optimum water content was found to show a higher resistance against erosion (Wan & Fell, 2004a). The critical hydraulic gradient is the average value for the hydraulic force acting on a soil volume. Perzlmaier et al. (2007) consider the average value not sufficient for description of the force acting on a soil particle, which is why the hydraulic load should be regarded on micro scale. Critical pore velocity derived from the particle sink velocity, taking the friction forces on the particle at rest into account, are considered to be more precise. The critical flow velocity may be developed from the particle sink velocity, from the drag force on the resting particles, or from sediment transport. The particles sink velocity in relation to the flow relative to a particle, forming an equilibrium between forces is thought to be suitable to describe the flow to an unfiltered exit. A critical pore velocity may be calculated based on the materials friction angle resulting in a friction force. Concepts of deriving a critical pore velocity for sediment transport from shear stress and adhesion require quantification of the ratio between mean pore velocity and shear velocity at the particles boundary layer. In general, critical pore velocities seem to involve extensive calculations based on a variety of variables, such as a drag force coefficient based on the Reynolds number, or the tortuosity, i.e. the ratio between the shortest distance of two points and the effective length of the winding flow path through pores. Even though critical pore velocities would be more precise than average critical gradients, field conditions do not seem to allow such specified values at present, especially in relation to the continuously ongoing construction and the nonhomogeneous structure of a tailings dam, but may become more interesting in the future considering the increasing possibility to use computational tools. 28

41 2.5 Initiation of internal erosion Some general principles about the initiation of internal erosion have been summarized by Fell et al. (2005). Four factors have to exist for internal erosion to occur: a water source and seepage path, erodible material in the seepage path, an exit where the eroded material can escape, and soil material that is capable to form and maintain a so-called pipe, i.e. an open channel for the eroded particles. Seepage paths may emerge from settlements and inhomogeneities as described in the following. Erosion can occur in the dam body itself, in the foundation, or from the dam body into the foundation. Further, the process of internal erosion is divided into four stages: initiation, continuation, progression, and finally breach or failure. The initiation stage is characterized by the formation of leakage. This leakage continues until progression, where the erosion path expands and a pipe is formed, which may finally lead to an accident or failure of the dam construction Settlements Dams are subjected to a variety of deformations and displacements because of diverse forms of compression, extension, or shear distortion. The state of pore pressure and effective stress changes during proceeding construction works. Horizontal and vertical stress increases with the imposed load, thus leading to consolidation and settlements in the dam body. Steep abutments, varying materials and properties, a compressible foundation or settlements of the embankment itself are major reasons for differential settlements and can result in fissures and cracks in the dam body. Conduits through the embankment are generally recognized giving rise to variations in stress distribution because of the interface between different materials. Compaction usually is difficult adjacent to the stiff walls, thus creating weak zones susceptible to initiate piping. Differential settlements in the dam body have been recognized being responsible for so-called arching and the development of cracks, which in turn may lead to potential weak zones as starting point for internal erosion. Displacements and settlements have been described by Fell et al. (2005). In addition, Kjærnsli et al. (1992) described several cases of dams which have experienced settlements and displacements. These observations and descriptions were mainly based on experiences from zoned water retention dams, where different materials cause varying settlements and cracks develop at interfaces. It was considered that settlements in the core are larger than that in the fill so that the core material is hanging on the shoulders of the coarser fill material. Fissures or cracks in the embankment may be found along the dam crest or transversal to the crest as shown in Figure 15, the latter introducing serious problems as they create a path for seepage. Settlements resulting from a compressible foundation and leading to a lowering of the crest are not considered being a problem for tailings dams, as such settlements can be adjusted by continuous raising of the structure (Vick, 1990). However, open cracks may remain also in a tailings dam. 29

42 Figure 15: Transverse differential settlements (Fell et al., 2005) Inhomogeneous material A majority of incidents and failures in embankment dam engineering are due to some kind of inhomogeneities in either the embankment material of the foundation. Inhomogeneities are either given, e.g. fractures in the foundation, or develop during or after construction, e.g. segregation of fill material during placement, variations in water content and compaction, conduits and walls in contact with fill material, or variations in weather conditions such as rain and freezing temperatures. Generally, most of such variations in the embankment or foundation can create a potential weakness zone susceptible to increased seepage and erosion, either by cracks, hydraulic fracturing, settlements, sinkholes, or various forms of piping, which is why these all have been summarized under the term inhomogeneities. The segregation of granular materials has been studied in the laboratory by Kenney & Westland (1993), where it was found that all dry sands and gravels segregate and that the patterns of segregation in such materials can repeatedly be tested, thus suggesting that many core and filter materials will separate into finer and coarser layers during construction placement. The use of broadly or gap-graded internally unstable material as discussed by Sherard (1979), Kenney & Lau (1985) and Milligan (2003) may also be regarded as contribution to inhomogeneities. For instance, the failure of the Teton dam in Idaho 1976 is considered to be due to a fractured foundation in connection with easily erodible core material, so that a combination of cracks, settlements at steep abutments, as well as internally unstable material led to failure (Sowers, 1993). Variable weather conditions, especially cold weather, may affect embankment construction significantly in countries in North America, northern Asia, or Scandinavia. Milligan (2003) describes several embankment dam constructions which have suffered from frost action. Freezing and thawing changes a soils hydraulic conductivity, thus reducing the function of a hydraulic barrier, increasing the flow and eventually resulting in internal erosion (Viklander, 1997). Annual freezing and thawing of frost susceptible fill material in embankment dams has been reported to result in fissures and cracks (Jantzer, 2006; Jantzer & Knutsson, 2007). 30

43 2.5.3 Hydraulic fracturing The term hydraulic fracturing still is discussed in Sweden, though Sherard has raised this subject almost four decades ago (Kenney, 1974). In 1986, an article on Hydraulic fracturing in embankment dams was published, where Sherard stated that concentrated leaks are common in embankment dams. Reservoir water pressure is considered to have the capability to open and enter cracks when the hydrostatic pressure exceeds the effective stress. Leakage, often initially caused by differential settlements and cracking, is thought to be kept open by acting water pressure. Cracking by means of differential settlements in connection with reservoir filling is considered to create a horizontal plane in the core material where water is able to enter, i.e. hydraulic fracture, in a thin channel, thus forming a wet seam. The initial crack is regarded being too small for erosion to take place, because the flow velocity is not sufficient. However, water pressure in the crack is higher than the pore water pressure in adjacent layers, which creates seepage and can leave the crack open. Increasing water pressure during further reservoir filling may be responsible for further opening of cracks. As a result, a wet seam either softens, collapses, and eroded material is carried away, or the flow velocity is so low that no significant erosion takes place. Sherard (1986) considers the development of wet seams in embankment dams being common. It appears that the term hydraulic fracture has led to misunderstanding. Whether or not water has the ability to jack open cracks has been questioned, as the total stress condition has not been regarded being likely to become equal to the hydraulic pressure, thus reducing the effective stress to zero regarding non-cohesive soils. Yet, the term hydraulic fracturing is also used in case of water entering already existing cracks, which seems to be more likely to occur. Lofquist (1988), as well as Mesri & Ali (1988), stated in a following discussion that hydraulic fracturing itself is not the cause for damage and failure, but rather inhomogeneities in the material arising either from the layered placement or from settlements in the embankment. Differential settlements and movement of material in different zones relatively to each other may cause zones with increased porosity. Circumstances during construction, such as poor compaction and separation of the material during placement may contribute to the development of such weak zones. Influences on the likelihood of hydraulic fracturing have been summarized by Foster and Fell: Hydraulic fracture is more likely to occur in case of deep and narrow valleys, where the abutment slopes are steep. Changes and irregularities, differential settlements in the foundation, a narrow core with different stiffness compared to shell layers, as well as irregularities during construction may contribute to hydraulic fractures (Fell et al., 2005). 2.6 Statistics and risk assessments Research on internal erosion and piping has been extended by the fields of statistics, risk assessments and risk analysis during approximately the last two decades. Statistical analysis of incidents, deteriorations and failures have been carried out by the International Commission on Large Dams ICOLD, showing that overtopping and internal erosion are the main causes to failure (Fell et al., 2005). Risk assessment has traditionally focused on a dams structural elements, i.e. the construction materials and the foundation conditions, where monitoring and surveillance was a central instrument for assessments. Today, risk assessment is based on quantitative and qualitative methods, where both the relative importance of different structural factors are linked together with uncertainties and unknown factors. 31

44 2.6.1 Statistical approach A detailed statistical analysis of embankment dam failure has been carried out by Foster et al. (2000a), where the relative likelihood of failure is estimated from the frequency of incidents related to dam characteristics. In their following study based on the obtained statistical data, it was assumed that is reasonable to make estimates of the relative likelihood of failure from the historic frequency of failures. Besides core soil types, filters and zoning of a dam, which have been taken account for earlier, characteristics such as the age of a dam, the foundation geology, compaction and the dam s performance, monitoring and surveillance were regarded. Weighing factors were developed by means of structural elements and their potential likelihood of influencing initiation of piping of failure, which are multiplied by an average historical frequency of failure based on a statistical analysis of historic cases. This method is intended to serve as primary assessment in order to develop a priority range for dams which need further detailed studies (Foster et al., 2000b). For Swedish tailings dams, Bjelkevik (2005b) developed statistics on incidents and failures. It was found that data, especially in historical perspective, is incomplete, and difficult to gather in international perspective due to different levels in reporting. However, it could be concluded that documentation and accessibility of data on incidents and failures provides valuable information, and that internal erosion is one of the main causes leading to incidents and failures. In general, it appears that qualitative analysis of internal erosion processes is mainly applied on WRDs and should be implemented also for tailings dams Risk assessment In their latest report the European Working Group on Internal Erosion has summarized 8 steps to assess the risks of internal erosion in an embankment as shown in Figure 16 (Fry, 2007). These steps are based on the initiation and progression principles presented by Fell et al. (2005). The initiation results from a load condition which may be hydrostatic (water level or flood), seismic, or environmental (e.g. freezing and thawing) that has an impact on the structure at a certain location, i.e. in the embankment, the foundation, or between both. Step 3 is the actual initiation of internal erosion may arise from either a concentrated leakage, backward or contact erosion, or suffusion. Continuation, also referred to as filtration, is the following fourth phase where filter criteria apply. Dependent on the grain size distributions of base and filter material, particle transport occurs respectively, so that four different levels from no erosion, some, excessive or continuing erosion can be distinguished. Phase 5, progression, is dependent on the hydraulic or mechanical condition. Steps 6 and 7 are detection and intervention, including the ability for detection and stopping the process. Finally, breach prediction includes different failure modes such as enlargement of the pipe, general or local instability, or overtopping resulting from crest settlement. 32

45 Embankment Foundation Embankment to Foundation No Erosion Some Erosion Excessive Erosion Continuing Erosion 1 Load Condition 2 Location 3 Initiation 4 Filtration 5 Progression 6 Detection 7 Intervention 8 Breach Prediction & Population Protection Hydrostatic Seismic Environmental Backward Erosion Concentrated Leak Contact Erosion Suffusion Mechanical Condition Hydraulic Condition Figure 16: Steps to assess the risks of internal erosion (after Fry, 2007) Risk analysis The work of the European Working Group on Internal Erosion includes practical tools for risk analysis and management. As part of this work, a framework for risk analysis has been presented by Brown (2007), where the process of risk management was broken down into four basic questions: 1. What are the failure modes? 2. What are the consequences of progressing erosion? 3. How large is the probability of a failure mode? 4. What are the consequences in case of a reservoir failure? Brown (2007) states that risk assessment is the process of deciding whether a risk is sufficiently significant to require additional control measures. Risk assessment should use quantitative methods to pay respect to the relative importance of various factors, whereas qualitative methods provide an understanding for risks. Failure modes need to be studied in a systematic way in order to become valuable for the risk analysis. Significant effects on the risk analysis may arise from unknown factors as well as uncertainties in known issues and factors, so that these have to be considered in particular. Due to the fact that the understanding of internal erosion still is limited, case histories, theoretical analysis as well as field investigations and laboratory testing should be taken into account. Basic techniques for risk control include detection by instrumentation, i.e. monitoring and visual observation, i.e. surveillance, emergency planning, and physical upgrades of the dam structure for intervention (Brown, 2007). 33

46 2.7 Discussion and conclusion Knowledge on internal erosion is often related to embankment dams constructed to retain water. Internal erosion in tailings dams is only regarded in a limited extent, which can be noticed in this literature study. Research in the field of internal erosion has changed over the years; from the original thought of creating criteria for filters and hydraulic conditions to statistical analysis, attempts to describe the mechanism, and based on that risk assessments. Geometric criteria are designed for natural materials, and the same applies for the assessment of the hydraulic conductivity by empirical relationships. The uniformity of tailings and their angularity are usually not taken into account in these considerations; both the filtering capacity in relation to adjacent fill material of dikes, as well as flow patterns in tailings themselves are generally not studied as well as those of natural, rounded material in literature. The hydraulic conductivity of a soil is difficult to determine in laboratory conditions. Besides many problems and possibilities of error in the testing procedures, it is difficult to draw conclusions from laboratory results to the complex conditions in situ. In addition, there are numerous equations to calculate the hydraulic conductivity, as well as a great variation in the form of an equation, as can be seen from the different presented versions of the Kozeny- Carman equation. The basic principle is the same, but there are many variations including different factors for particle shape, pore shape, effective grain size and others. All these different versions show the importance of an effective grain size, the pore size and shape, and the particle shape for the hydraulic conductivity. Nevertheless, the chemical nature of tailings, their angularity and specific surface area may affect the flow pattern and hydraulic conductivity considerably. Despite all various factors for particle shape, pore space, etc., can the specific characteristics of tailings not be taken into account, thus making it difficult to calculate reasonable values. Due to the presented difficulties in laboratory testing should empirically derived values for the critical hydraulic gradient to initiate erosion be regarded with attention to their verification. The design criteria for Swedish tailings dams is completely independent from values for the critical gradient, which may be a result of deviating values for different materials and in situ conditions. Following geometric and hydraulic criteria does not eliminate the risk for internal erosion, because the boundary conditions appear to vary to a relatively large extent. Besides, researchers have not come to conclusions on hydraulic and geometric criteria, which is shown by the continuous and ongoing discussion in both fields. In addition, structural variations and construction practices cannot be calculated or accounted for in construction, which is why statistical analysis and risk assessments have to complete structural shortcomings. In general, this study shows that most research has been directed to water retention dams and natural materials which may be a historic effect of economic interests regarding water retention dams. The knowledge on and geotechnical testing of tailings and their behavior is not sufficiently developed, especially with regard to the hydraulic conductivity. 34

47 3 NATURAL ANALOGIES Tailings dams are supposed to be long term stable without supervision and maintenance. As described in the introduction, the uncontrolled release of trace elements from tailings can result in unwanted effects on the environment. The remediation of tailings dams has to be regarded with respect to providing a safe structure that is environmentally vindicable for future generations. In order to prevent environmental impacts and possible long term contamination, the physical stability of a tailings dam, e.g. slope stability, stability against erosion, or settlements, is essential. In Sweden, the design period for this long term stability is normally regarded being about 1000 years. Experience on such extended design periods is relatively limited. There are few examples on long term stable constructions that can be found in ancient structures, such as e.g. pyramids. Additional examples are e.g. stable slopes of man-made earthen mounds in China that have been stable over several thousands of years. These mounds were studied with respect to stable slope angles; their inclination varies from 16º to 28º. There are additional examples of manmade earth mounds, burial and entombment sites all over the world that are up to years old. (Bjelkevik 2005a). In Sweden, the current long term design on tailings dams in Sweden is not related to ancient constructions, but instead based upon knowledge on natural examples; i.e. moraine formations which have been stable in a long-term perspective (Bjelkevik, 2005b). Geological formations which have and partly still do fulfill the function of damming water have been examined by Agrell at the Geological Survey of Sweden SGU (2002). The landscape in Scandinavia, especially northern Sweden, is characterized by its numerous lakes in various dimensions. These are a result from an inland ice cover that was responsible for sedimentation and erosion. The most recent melting of continental ice in took place about to years ago, so that the landscape is relatively young with regard to its geological age. However, such landforms are considered to give valuable information on the long-term performance of manmade dam constructions. Agrell states that there is an infinite number of natural landforms similar to dam constructions creating long-term stable hydraulic barriers, and that it is most probable that a barrier of soil can withstand aging and deterioration and retain water without experiencing significant erosion processes. The study provided by the SGU described several examples of natural dammed lakes that can be dated back to the last glaciation in Sweden. It is shown that natural barriers are capable to operate effectively as a hydraulic barrier in long-time perspective. However, information on the geotechnical properties is for the most part missing, which is why it is necessary to study such analogies in their geotechnical composition. Generally, the design of civil engineering structures is based on well known and laboratory-tested materials with quite accurately defined failure mechanisms. This is usually not the case in geotechnical engineering; e.g. a slope stability problem cannot be tested under laboratory conditions. It is therefore important to study analogies in nature in their geotechnical formation. Such a study can give valuable information for the long-term design of man-made dams based on the geotechnical properties of natural long-term stable structures. In order to be able to evaluate information from natural analogies for tailings dam construction, the material and properties of the natural dam have to be known. Natural dammed lakes are especially interesting with regard to their obvious stability against internal erosion. Consequently, it is assumed that a critical hydraulic gradient exists, and that the 35

48 material is compacted to some kind of optimum which allows a hydraulic conductivity without migrating particles. Analysis in a case study included in this work covers, besides monitoring of ground water conditions in the field, grain size distribution, density, compaction and hydraulic conductivity in the laboratory. 3.1 Natural formations in Sweden In the study conducted by SGU, Agrell listed several geological formations that fulfil a function similar to tailings dams. The primary focus in this study is the long-term stability of a geological structure with regard to the mean hydraulic gradient it is exposed to. In addition, information about the geological structure, the material, as well as to how a formation can be related to a dam construction is given. Unfortunately, the given information about the landforms acting as hydraulic barrier is rather general from a geotechnical point of view; the main aspect in the study is the hydraulic gradient, i.e. width of the landform acting as dam, and differences in water table on the upstream and downstream side. Three of the given examples are presented more detailed: lake Ragunda, lake Hennan, and lake Mången at Brattforsheden. Additional examples are Styggtjärn and Skuttungesjö. Locations of examples of natural analogies, as well as the location of the case study, are shown in Figure 17. Lake Ragunda has been subject to a doctoral thesis in geology. Geotechnical investigations have been carried out at lake Hennan, because the structure of the landform is similar to a dam construction with different zones. Lake Mången has been subject to examination due to its relatively high hydraulic conductivity despite its damming capacity; there are numerous springs in the area. Apart from the existing doctoral thesis on lake Ragunda, further information beyond Agrells report has not been available, which is why additionally selected examples can only be roughly described, but not presented more detailed. In order to study a natural dammed lake with respect to geotechnical properties, a case study has been carried out in south of Gällivare in northern Sweden. Figure 17: Location of natural analogies 36

49 3.1.1 Lake Ragunda The lake Ragunda (Ragundasjö in Swedish) was a 25 km long lake at Indalsälvens valley in northwest Sweden (see Figure 17). The lakes original southern part formed two narrow bays. At the eastern outlet the mean water level was regulated by the waterfall Storforsen, and the western bay Sandviken formed a natural dam, see Figure 18. Sandvikens bay was about 600 m long and 200 m wide, and the formation of the southern barrier which dammed the lake was dated back to approximately 6500 B.C. Sandvikens eastern and western border consisted of rock, while the southern barrier was formed by clay. The clay ridge consisted of 60m glacial clay in layered deposits of alluvium, i.e. sediments with low hydraulic conductivity deposited by a stream or other running water. In the doctoral thesis, Ahlman et al. (1924) stated that the difference in height between the water levels upstream and downstream of the ridge was 40 m over a distance of 1000 m, thus corresponding to a mean hydraulic gradient of 4 %. The lake is well known because of historical events and has been documented comprehensively. It is presumed that the natural dam at Sandviken would still exist if not a human being had destroyed it: by digging a channel to drive timber, the dam broke and the lake was drained about 200 years ago. Figure 18: Lake Ragunda before and after Sandviken bay broke (modified after Ahlman et al, 1924) Lake Hennan in northwest Hälsingland Lake Hennan is located upstream of the Lake Storsjön in central Sweden in the county of Gävleborg. The water level of lake Hennan ranges between 206,9 to 208,8 masl, while Storsjöns water level is only 186,8 masl. This results in a water level difference of about 20 m with the headland between these two lakes only being about one kilometer wide, resulting in a mean hydraulic gradient of 2 %, see Figure

50 Figure 19: Map of Hennan The village of Hennan is located on the ridge between the lakes. The responsible geologist for this area strongly believes the ridge to consist of moraine and well compacted material from an earlier glaciation. Drilling carried out in the headland showed that bedrock was located at a depth of 47 m, with an overlying sediment layer from a glacial stream which is about 32 m thick. Above that, a moraine layer with a thickness of 8 m was found. The lake is therefore considered to be dammed by a core of fine stream deposit surrounded by moraine, which is similar to an earthfill dam construction with central core and filter layers (Agrell, 2002) Lake Mången at Brattforsheden Brattforsheden is an area in Värmland, western Sweden. Several lakes can be found in this area, among them lake Alstern and lake Mången. Alstern is the largest lake and is 13 km long,1 km wide, has a maximum depth of 54 m, and a mean water table at about masl. Mången is a smaller lake southwest of Alstern at a level of 166,5 masl, see Figure 20. Both lakes are dammed at Brattforsheden by a southern border of glacio-fluvial deposits (i.e. moved and deposited material by running water associated with glaciers), a combination of rubble, gravel, sand, fine-sand and silty loam reaching a height of 170 to 180 masl. It is noted that lake Mången has been dammed by this formation during the past years, despite the relatively high hydraulic conductivity southwest of Brattforsheden. Several springs can be found in this area, and ground water observations have shown that the difference in water tables between lake Mången and the southern springs (east of Lindfors, see Figure 20) is 16,5 m over a distance of 1,4 km, thus resulting in a hydraulic gradient of 1,2 %. Agrell considers this formation being the best example of a natural dammed lake in Sweden. Figure 20: Lake Mången and lake Alstern at Brattforsheden 38

51 3.1.4 Styggtjärn Styggtjärn is located at Rogen natural preserve in Härjedalen in the county of Jämtland, close to the Norwegian border. This landscape is, as many other places in Sweden, characterized by hilly terrain with many small lakes, which is a result from dead ice from the last glaciation. A plateau consisting of glacial till serves as natural dam for the eastern lake Styggtjärn with a water table at 779,6 masl. On the downstream side to the west of the ridge lake Kärringsjön is located, which has a water level of 777,5 masl. The difference between water tables is 2,1. The natural dam between is about 500 m long and 100 m wide, so that the formation is exposed to a hydraulic gradient of 2 %. The plateau is marked in Figure 21. Figure 21: Styggtjärn and Kärringsjön at Rogen nature preserve Skuttungesjö Lake Skuttunge (Skuttungesjö in Swedish) was a lake that existed between 1500 B.C. to 500 B.C. in the Uppsala region. The lake was shallow, and Agrell (2002) states that it likely that climate changes increased the water level so that the natural dam at the southern border was overtopped. This natural dam was 10 m high and 200 m long, and the mean hydraulic gradient had a maximum of as much as 5 % Summary of formations studied by SGU The natural formations presented withstand hydraulic gradients between 1.2 % to 5 % and have been stable up to the past years. A summary is presented in Table 2. Table 2: Natural analogies to dam constructions in Sweden (modified after Bjelkevik, 2005a) Place Difference in height Δh [m] Distance Δl [m] Hydraulic gradient Δh/Δl [%] Material Ragundasjö ,0 Hennan ,0 Mången 16, ,2 glacial clay in layered deposits of alluvium glacial till over well compacted sediments from glacial stream from earlier glaciation rubble, gravel, sand, fine-sand and silty loam Styggtjärn 2, ,1 glacial till Skuttungesjö ,0 not known 39

52 3.2 Case study Introduction In order to analyze a natural analogy in its geotechnical properties, a natural dammed lake in northern Sweden was chosen, see map in Figure 22. Several small lakes at varying water levels can be found at the company area of Bolidens mine Aitik in Gällivare, about 260 km northwest of Luleå. These lakes have not been studied with respect to their geological age, but referring to the study conducted by SGU (Agrell, 2002) it is assumed that such small lakes are typical landforms resulting from dead ice from the last glaciation in northern Sweden. The selection for studying the formation located south of the mining area was of practical and economical nature: because of the development of mining activity, construction works were carried out close to the three lakes which were noticed to have different water levels with short distances in between. In connection with these construction works it was possible to carry out field investigations. Figure 22: Location of field studies of a natural dammed lake southwest of Gällivare The following aerial view in Figure 23 shows the location of three lakes at Kiilavaara road south east of Gällivare. The central u -shaped lake has a water level of approximately 459,5 masl. Originally, it was thought to analyze the soil structure between the central lake and the lake located north at a distance of 260 m. The water level of this lake is 441,5 masl, which results in a relatively high hydraulic gradient being 6,9 %. Unfortunately, the companies construction works were extended over this area, so it was not possible to carry out field research here. Instead, the formation between the central and the southern lake, which has a mean water table of 456,8 masl, was chosen. The difference between water tables results in 2,7 m over a distance of 40 m, which gives a mean hydraulic gradient of approximately 6,75 %. 40

53 Figure 23: Layout of the three lakes at Kiilavaara, Gällivare. Despite the relatively high hydraulic gradient the soil is exposed to when comparing to hydraulic gradients in natural formations (see Table 2), the soil seems to be well compacted and the hydraulic conductivity sufficiently low, thus resulting in a capacity to withstand exceeding seepage, particle migration, and internal erosion in a time perspective that probably may not be long-term in the sense of thousands of years, but at least extending over a time span of a few thousand years. If the soil was loosely compacted and had a high hydraulic conductivity, the lakes would not exist the way they do. However, even though the material well consolidated, the water tables are expected to correspond with each other, so that not only material properties, but also ground water conditions need to be analyzed. In order to be able to analyze the soil mechanical properties of the natural dam, field investigations were carried out. These included the excavation of two test pits where soil samples were collected at a depth of 2 m below ground surface. In addition, the materials compaction in situ was studied at the bottom of each test pit using the balloon test principle, see chapter To be able to study the ground water condition, pore pressure gauges were installed. Further laboratory analysis was carried out to study the grain size distribution, the water content, the optimum dry density by using the Proctor compaction method, and to determine the density of the solid particles. With these values, the degree of saturation and porosity could then be calculated. Finally, tests on the hydraulic conductivity were carried out applying different gradients on samples with varying porosity. The results from laboratory tests were then compared to those from field studies for evaluation. Studies of the water balance in the area have not been carried out, but should be considered in future research. 41

54 3.2.2 Description of field work Field investigations were carried out in September 2007; a schematic illustration of the analyzed profile can be found in Figure 31. The field work was carried out together with lic. tech. Håkan Åkerlund, working at the company WSP during that time. On both sides of the road, a test pit was excavated to a depth of approximately 2 m from ground surface using an excavator. The location of the test pits is shown in Figure 23. The depth was measured using a GPS, see Figure 24. At the bottom of each pit the soils density was measured by using the principle of a rubber-balloon test. Soil samples were taken for further laboratory analysis. For mapping, surveying was carried out using a GPS system as shown in Figure 24. To be able to monitor the ground water conditions, several pore pressure gauges were placed in the profile at varying depths. Figure 24: Excavation of test pits and surveying using a GPS with amplification In-situ density and collection of samples Determination of the density in situ can be done using various methods; in this case, the principle of the rubberballoon method was applied with slight modification. The rubber-balloon test is described in textbooks on soil mechanics (e.g. Cernica, 1995) or in Swedish Standard SS There are different instruments for this test, but the principle is the same: A volume of soil from a small test pit with a diameter of about cm is removed and collected for water content determination in the laboratory. The apparatus, see Figure 25, is then put above the resulting pit, which is supplied by a tightly fixed rubber-balloon on the bottom. The pit is then filled by a water volume in the balloon. The amount of water needed to level out the pit can be measured, so that the volume of the removed soil can be determined. Figure 25: Rubber-balloon test apparatus 42

55 One problem using the rubber-balloon test apparatus is that the soil sample is relatively small and easily disturbed by stones of small sizes. It is difficult to obtain a test pit in a well graded glacial till because of the large variety of particle sizes. The pit then becomes irregular and cannot be filled with the balloon properly. As can be seen in Figure 26, there is a significant amount of gravel and boulders included in the material. In order to increase the reliability of the sample it was therefore chosen to apply the same principle, but a larger test scale. Instead of a rubber-balloon, the test pit, which had a diameter of about 0,5 m, was provided with a watertight film and filled with water from containers which contained a certain amount of water, see Figure 26 left. In doing so, the volume of the test pit could be determined. Figure 26: In-situ density measurement using the principle of a balloon density apparatus In order to be able to obtain the density in situ, the water content of the removed sample from the pit has to be determined in the laboratory, which is described in chapter Two such density tests were carried out, one in each test pit, respectively. The test result was modified by means of adjustment of the water volume in subsequent evaluation of the test. As can be seen in Figure 26 to the left, the plastic film used in the pit was quite stiff, so that the pit was difficult to fill accurately. The plastic film was pulled away, so that the water used to fill the pit ran directly into the pit, see Figure 26, right. It is visible that the water did not level out the pit completely. Therefore it was considered suitable to add an additional liter of water to the measured quantity in both tests. The results are shown in Table 3. Table 3: Results from in situ density measurements Test Weight soil sample [kg] Quantity of water [l] Water content w [%] Density ρ Dry density ρ d Degree of saturation [%] Porosity n [-] Void ratio e [-] 1 28,5 10,57 (+1,0) 2 24,9 8,74 (+1,0) 7,8 2,46 2, ,185 0,227 8,8 2,56 2, ,164 0,196 The calculation of both the degree of saturation and the porosity in Table 3 includes the density of the solid particles ρ s, which was analyzed in laboratory tests and found to be 2,81 t/m 3, see chapter

56 Installation of pore pressure gauges For ground water monitoring and study of variations in the hydraulic gradient in the soil between the two lakes, an overall number of six pore pressure gauges at three different places in the profile were installed, see Figure 31. These pore pressure gauges were expected to give valuable information on as to how far both water tables communicate with each other and how water permeates through the soil, which is an important part for comparison and assessment to dam construction. For placement of the gauges, the section was organized geometrically as shown in the schematic diagram in Figure 27. As a first step, the ground water table was assumed to be a straight line between both water tables. Thereafter, the distance between both lakes was divided into thirds. The first pore pressure gauges were placed in the middle of the section, and additional gauges one third of the distance to each side, so that a distance of one sixth from the gauges to the left and the right to each water table remained. Figure 27: Schematic diagram on the placement of pore pressure gauges The difference in height of the two water levels was also divided schematically. The difference in height between the water levels of the lakes was calculated and divided into four parts. Pore pressure gauges were then placed according to the geometric scheme. With this arrangement, it was assumed that those gauges located deepest were placed at the same level as the lower lakes water table. The difference in water tables was measured by GPS and set to 1,70 m. For instance, in accordance to the measured height difference of 1,70 m, pore pressure gauges in the middle were supposed to be located at the assumed water table, about 0,40 m and about 0,85 m below the presumed ground water level. It was chosen not to include actual heights in this schematic diagram in order to present the basic way of thinking. During the progress of the study it was found out that mapping with the GPS did not work reliable and various values for heights of water tables and ground surface were measured. Several measurements were compared and evaluated, and a final measurement was carried out with the help of the pore pressure measurement device. To be able to measure the height of a water column, a cylinder was placed on ground surface and filled with water, see Figure 28, left. This water level was measured so that the distance from cylinder to ground surface was known. It was then possible to measure relative heights to several points in the section with the help of pore pressure measurement equipment, which shows the height of the water column in mh 2 O, see Figure 28, right. With these values, a new evaluation on actual location of ground surface and depth of pore pressure gauges could be made, which is presented in Figure

57 Figure 28: Checking heights in the profile with the help of a water reservoir. Measurement of water column from one point on ground surface to location of water table of lake. The new mapping made it possible to present a section with locations of pore pressure gauges which appears relatively exact to the author with respect to problems with the GPS. The section with heights of ground surface, water tables and placement of pore pressure gauges is presented in Figure 31. From the middle of the section (about 13,5 m to the left) two pore pressure gauges were installed (A and B) at a distance of 6,75 m from the upper lake; i.e. the lake with a water table at about 459,5 masl. These two were set at two different levels: A at a depth of 2,2 m from ground surface, and B at a depth of 3,0 m from ground surface. In the middle, three piezometers were set at three different levels, C, D1 and E at depths of 3,1 m, 3,6 m, and 4,1 m, respectively. At a small distance, a fourth pore pressure gauge (D2) was set at the same depth as D1, namely 3,1 m, in order to be able to compare and check the reliability of the following measurements. In the right section, piezometer F was installed at a distance 13 m from the middle section at a depth of 3,6 m, which corresponded to the water table of the lower lake. The system installed is BAT piezometer, which is a proven system for pore pressure measurements in soils with low hydraulic conductivity in both saturated and unsaturated conditions (Åkerlund, 2007) Drilling and placement of the piezometers was done by personnel from WSP. The filter tip with a flexible membrane was saturated before placement, see Figure 29, left. It was then set on the pipe, Figure 29, right, to be lowered into the drilled hole. 45

58 Figure 29: BAT Piezometer To measure the pore pressure, the membrane can be penetrated by lowering an injection needle connected to a pressure sensor as shown in Figure 30. The pore pressure is the directly converted and displayed in water column [mh 2 O] on the instrument. In order to cover possible seasonal variations, pore pressures were checked once a month during the first year. The piezometers need some time to adjust to the ground conditions (Åkerlund, 2007), so that the first measurement was carried out in late November 2007, two months after installation. The section displayed in Figure 31 also includes two additional piezometers that were installed in February 2009, because it was found that most of the pore pressure gauges in place did not provide valuable information, which is described in detail in the following chapter. Those two devices, numbered H and G, were set at a depth of 5,4 m below ground surface, about 1,2 m below the lower lakes water table. Another adjustment at device B was made when reading the latest pore pressure measurement. The gauge was driven down another 0,8 m into the ground in order to check if an additional measurement at a deeper depth could be carried out. Figure 30: BAT piezometer (modified after Geometric.se, 2009) 46

59 Figure 31: Profile of the analyzed structure and location of pore pressure gauges

60 3.2.3 Results of pore pressure measurements The results of the pore pressure measurements are shown in Figure 32. The diagram shows the location of each piezometer in meters above sea-level [masl] as a starting point on the y- axis. The measured pressure can then be described by a water column with a height according to the measurement [mh 2 O] and can be displayed in [masl] by adding the measured water column to the piezometers level. Note that the diagram starts at September 2007, even though the first measurement took place in November. This first level at Sept. 07 was included to illustrate the location of the piezometer as a starting point. Measurements were carried out with approximately one month intervals during the first year for piezometer A F. After that, the two new piezometers G and H were installed. In order to be able to depict their variations their original location is displayed according to all other piezometers on the y-axis denoted Sept. 07, even though they were set in February, Devices G and H were first checked in May 2009, and finally all piezometers were read for the last time in October ,5 Pore pressure level with respect to location of pore pressure gauges [masl] 459,0 458,5 458,0 457,5 457,0 456,5 456,0 455,5 455,0 Sep. 07 Feb. 08 Jul. 08 Dez. 08 Mai. 09 Okt. 09 A C B D1 D2 E F G H Figure 32: Results of pore pressure measurements When measuring the pore pressure, the sensor is lowered into the pipe and the needle penetrates the membrane, which was explained in the previous section. Before lowering the needle into the membrane, it should be left hanging just above, so that the instrument can be set to zero. Otherwise, the reading is not correct, because the zero value fluctuates. After setting the instrument to zero, the sensor can be lowered into the membrane and the pressure is measured. After that, the sensor is lift up just above the membrane again in order to check if zero still is zero, i.e. if the zero value for the reading has been constant or fluctuating. If another value than zero is shown, this value is added to the reading, e.g. if the pore pressure is 48

61 0,50 m and the subsequent zero-reading is -0,05, the final value for the pore pressure becomes 0,55 m. The procedure of finding a zero value is especially difficult to carry out during winter time, because the instrument is sensitive to temperature (Forsberg & Stenman; 2007, 2009). The zero value is therefore almost always shifting. The instrument has to be handled with accuracy, because even slight changes in the pressure chamber can cause deflection. It can be seen from Figure 32 that most readings were discrete, with relatively insignificant pore pressure changes around zero. Analysis of readings of pore pressure gauge A lead to the conclusion that the gauge was located above the ground water level, because all measurements were zero. The same applied to pore pressure gauge F. Readings B, C, D1, D2 and E show an amplitude during late winter Table 4 shows the exact measurements, where measurements above 0,10 mh 2 O were marked red. It can be seen that most of the increases in pore pressure seen in the diagram are variations below 0,30 mh 2 O. Such small values are difficult to interpret, because variations of displayed values can cover a range of about +0,67 mh 2 O to -0,67 mh 2 O, which is especially the case during wintertime, when the sensor is more sensitive to temperature. It was noted that measurements during winter were difficult to carry out due to shifting readings and reference of zero. Measurements in October 2008 showed negative pore pressure in point C and D1, which implies that the gauge is located above the ground water table. Pore pressure gauges D1 and D2 were set at the same distance from both lakes and same depth in order to check the reliability of measurements. In spring 2008, both gauges showed an increase in measured pore pressure between 0,15 mh 2 O and 0,30 mh 2 O. However, no conclusion could be drawn from the measurements; both gauges vary differently in ranges of values between 0,0 mh 2 O and 0,3 mh 2 O, which are found to be too small values for interpretation with respect to shifting the zero reference. Two additional pore pressure gauges G and H were set at a significant larger depth of 5,4 m below ground surface in order to find out if the ground water level was located deeper than originally hypothesized. The first measurement of G and H in May 2009 showed 0,59 mh 2 O and 0,27 mh 2 O, respectively. However, gauge G was not readable at the last occasion, while the pore pressure at gauge H had increased to 0,53 mh 2 O. With Pore pressure gauge G not operating, even these values were difficult to evaluate. Finally, pore pressure gauge B was driven down 0,8 m deeper in the ground and has a new location at 3,8 m below ground surface. A measurement carried out directly after movement showed a value of 0,25 mh 2 O, a value that also was fluctuating continuously with respect to the zero value. Conclusions of the pore water pressure measurements With most of the measurements in the region of zero and some readings only slightly above zero, it is concluded that the pore pressure gauges are either located above ground water table or that the ground water flows orientation is perpendicular to the drawn section. Even those measurements above 0,20 m appear to be unreliable, since the zero value was shifting permanently, as described above. On several occasions, the instrument displayed negative values, which were set to zero when they occurred in gauges that were located above others that displayed zero or just above zero. Such negative values sometimes occur when the pressure chamber is slightly deflected by some air bubble that can occur in the instrument, or when the injection needle is bent and has to be replaced. 49

62 In addition to the gauges being located above the ground surface it may even be possible that individual piezometers do not operate correctly. The flexible membrane in top of the filter tip is supposed to seal the filter tip and is penetrated by a needle for measurement. It has been hypothesized that this membrane does not seal the tip, and that it can only withstand a limited amount of penetrations. Table 4: Results of pore pressure measurements Gauge Level A 458,8 0,01 0,01 0,01 0,05 0,02 0,07 0,00 0,00 0,25 B 458,0 0,08 0,02 0,00 0,30 0,02 0,07 0,00 0,00 0,00 C 457,9 0,11 0,02 0,02 0,24 0,02 0,03 0,02-0,37 0,11 D1 457,4 0,10 0,41 0,01 0,20 0,30 0,27 0,25-0,14 0,03 D2 457,4 0,10 0,10 0,01 0,15 0,37 0,16 0,06 0,03 0,00 E 456,9 0,09 0,14 0,42 0,11 0,14 0,03 0,02 0,05 0,05 F 456,8 0,07 0,02 0,08 0,19 0,00 0,02 0,02 0,02 0,00 G 455,4 0,59 0,00 H 455,4 0,27 0,53 50

63 3.2.4 Soil description laboratory analysis Soil samples were taken from the test pits described in chapter for further laboratory analysis. At first, the water content and the grain size distribution of the till was determined on samples that resulted from the pit excavated for the rubber balloon test; the determination of the water content in the laboratory is needed for the final result of the in-situ density. The grain size distribution also was determined using samples from the in-situ density test pit. In addition, analysis of the density of solid particles, the optimum compaction according to the Standard Proctor test, as well as tests on the hydraulic conductivity was carried out. The samples analyzed in these tests were collected at the bottom of the test pit, i.e. at various places at a depth of 2,0 m below ground surface. Grain size distribution The grain size distribution was determined according to the Swedish Standard SS on samples being used by the in-situ density measurement. Sieving was carried out on two samples per in-situ density test, see Figure 33. The particle size of finer grains was determined by sedimentation, i.e. the pipette method. Sedimentation was carried out on one sample per in-situ density test, see Figure 33. Figure 33: Explanation on sieving, sedimentation, and development of grain size distribution curves The two sieving curves then were completed with the corresponding sedimentation curve from the in-situ density test. Because the rubber balloon test was carried out in both test pits, two soil samples were available, so that the above described procedure was carried out two times, resulting in four grain size distribution curves, see Figure 34. The result shows that the material can be described as well graded fine grained till (in Swedish gr sa si morän) with approximately 7 % clay, 19 % silt, and 45 % sand, and the uniformity coefficient C u is

64 100 Clay Silt Sand Gravel percent finer by weight < d [%] ,002 0, ,001 0,01 0, particle size d [mm] Figure 34: Grain size distribution of samples taken in field tests Regarding water retention dam construction, such a material has a sufficiently low hydraulic conductivity and would be adequate as hydraulic barrier if compacted at optimum water content or up to 3 % above optimum water content, with a degree of saturation of %, according to the Swedish guidelines for dam safety RIDAS (2004). Mass density of solid particles The density of the solid particles ρ s is defined by their mass divided by the volume of the mass and is determined as described in Swedish Standard SS A soil sample is placed in a pycnometer, a glass bottle with close fitting lock with known mass and volume. The soil sample is placed in the pycnometer and filled with completely with distilled water, so that no air remains in the pycnometer. From the measured masses of the sample, the pycnometer and glass plate and the density of water, the density of solid particles can then be determined. This test was carried out four times to verify the relatively high density of ρ s = 2,81 g/cm 3, see Table 5. Table 5: Results of grain density test Test No Mean value ρ s 2,81 2,81 2,83 2,81 2,81 52

65 Proctor compaction test In order to find out the optimum dry density for the till, compaction tests were carried out according to the Standard Proctor test procedure described in Swedish Standard SS This test is described in textbooks on soil mechanics (e.g. Cernica, 1995) and is used to test a soils optimum dry density at a certain water content. In order to standardise the volume of soil a special mold is used. The soil is then tamped in the mold with a standardized hammer at defined falling height to obtain a constant energy impact. To determine the optimum dry density ρ d, several soil samples with varying water contents are compacted and the result can then be plotted in a diagram. In relation to the other samples, sample 4 was best compacted to an optimum dry density of 2,22 g/cm 3 at an optimum water content of 7,6 %. The dry density of samples 1, 2, and 3 is relatively similar with 2,15 g/cm 3, 2,13 g/cm 3, and 2,16 g/cm 3, respectively. However, sample 1 has the least water content with 5,1 %, while samples 2 and 3 were compacted almost equally with a water content of 8,6 % and 8,3 %. Sample 5 was least compacted with an optimum dry density of 2,08 g/cm 3, at a water content close to saturation, 12 %. The result of the test is shown in Figure 35 below, and discussed further in chapter 4. 2,24 No 4 Dry density [g/cm 3 ] 2,20 2,16 2,12 2,08 No 1 No 3 No 2 No 5 2,04 5,0 7,0 9,0 11,0 13,0 Water content [%] Figure 35: Result of Proctor compaction test The compacted samples remained in the mold for further tests on the hydraulic conductivity described in chapter Therefore, the samples water content in Proctor compaction could not be determined on the samples by removing soil from the mold and drying it. Instead, it was chosen to determine the water contend on the relatively small amount of soil that was left from each particular test. This small portion of leftover soil may have been affected by drying out, which may give some variation in the test result. In order to compare and check the samples density, the porosity of the sample was determined after conductivity tests. This procedure is shown in Figure

66 Figure 36: Procedure of determination of the soil samples properties before and after testing the hydraulic conductivity By determining the density of the solid particles and determining the dry density of the sample, the porosity can be calculated. Both the porosity and the dry density of the samples after performing hydraulic conductivity tests are displayed in Figure 34. Comparing the samples dry density before and after hydraulic conductivity tests it can be seen that the dry density after testing is fractionally lower. The relation between the samples compaction is nevertheless the same. For instance, sample 4 was best compacted and sample 5 least, while samples 1, 2 and 3 differ between 2,11 g/cm 3 and 2,14 g/cm 3. Comparing the porosity of the samples to the dry density, it is shown that the higher the dry density, the lower the porosity. Hence, the values for porosity and dry density are mirror-inverted as shown in Figure 37. Porosity 0,26 0,26 0,25 0,25 0,24 0,24 0,23 0,23 2,20 2,18 2,16 2,14 2,12 2,10 Dry density ρ d 0, Sample No 2,08 Porosity Dry density Figure 37: Porosity of the compacted samples after testing the hydraulic conductivity From this comparison, it is concluded that the Proctor compaction test is representative, despite the fact that the water content of the samples could only be determined on a small amount of soil from the sample compacted as described above. 54

67 Apart from porosity and dry density, the values of the bulk density ρ, water content w and degree of saturation S r before and after hydraulic conductivity tests were determined and are presented in Table 6. During hydraulic conductivity tests, the samples are exposed to a seepage pressure and allowed for saturation. Therefore, the degree of saturation in the samples after conducting tests on the hydraulic conductivity is be expected to be 100 %, with a corresponding water content at approximately 12 %. However, the degree of saturation in samples 1, 3, 4, and 5 is found to be considerably lower, i.e. between 55 and 75 %. Only sample 2 has a water content of 12 % and is 100 % saturated, which appears to be appropriate after allowing the sample to saturate during conductivity testing. Well graded glacial till such as the present contains a low pore volume due to the large amount of fines in pore spaces. Upon hydraulic conductivity testing, air can remain confined in few and small pores, so that water cannot fill the total pore space. Therefore, the sample does not attain complete saturation. A value somewhat below 100 % saturation could therefore regarded as acceptable, with the explanation of confined air bubbles. However, the discrepancy in saturation found is too large for such an effect. It is therefore considered being due to not drying the samples completely after carrying out tests on the hydraulic conductivity. The amount of soil to be dried in the oven is about 2,2 kg per sample. The used drying time is based upon a much smaller soil sample, normally less than 100 g. Thus, the drying time should have been increased significantly. The water content determined is therefore considered to be too low, resulting also in a too low degree of saturation. Nevertheless, the difference between dry density ρ d before and after testing the hydraulic conductivity is considered relatively small, so that the Proctor compaction is regarded being representative for dry density, water content and degree of saturation of the samples. Table 6: Results from Proctor compaction compared with values after testing the samples on hydraulic conductivity Sample no ρ [g/cm 3 ] 2,26 2,32 2,34 2,39 2,33 Proctor compaction ρ d [g/cm 3 ] 2,15 2,13 2,16 2,22 2,08 w [%] 5,1 8,6 8,3 7,6 12,0 S r [%] e 0,309 0,317 0,303 0,268 0,353 n 0,236 0,241 0,233 0,211 0,261 ρ [g/cm 3 ] 2,31 2,36 2,30 2,35 2,23 after conductivity test ρ d [g/cm 3 ] 2,13 2,11 2,14 2,18 2,09 w [%] 8,5 12,0 7,8 7,6 6,8 S r [%] e 0,317 0,333 0,314 0,287 0,346 n 0,241 0,250 0,239 0,223 0,257 55

68 3.2.5 Tests on hydraulic conductivity The hydraulic conductivity in field conditions can be determined in pumping tests in situ. Such large scale tests are both cost and time demanding. To determine the hydraulic conductivity of the natural dam, it was therefore chosen to carry out laboratory tests, which are presented in this section. The hydraulic conductivity k varies with void ratio e and porosity n in the soil. In order to be able to assess the hydraulic conductivity k in the natural dam with a certain compaction and porosity in situ, laboratory tests were carried out on samples compacted to different dry densities and porosities using the Standard Proctor method. These laboratory tests provide a basis for the development of a relation between compaction and hydraulic conductivity in laboratory conditions and in situ, with the final aim to estimate the hydraulic conductivity in field conditions. In order to be able to relate compaction, density and flow through the samples to the situation in nature, each sample was tested under three different hydraulic gradients: i = 3, i = 5, and i = 7,3. An additional testing gradient of i = 0,3 was applied on one sample for practical reasons, which are explained in the following. The hydraulic conductivity was tested under constant head conditions, i.e. with a constant water supply level, as described in chapter The hydraulic gradient stands for the relation between the rate of head loss Δh to the flow distance ΔL through the sample. For the samples with the height of 11,6 cm, the hydraulic gradient was adjusted by the distance to the water supply, i.e. 34,8 cm, 58 cm and 84,7 cm, respectively. For the gradient 0,3, the distance was 2,5 cm. The testing principle is shown in Figure 38. The order of the applied gradient was not appointed, but random. Due to the compaction method, the samples had to be allowed for saturation. Figure 38: Testing principle for hydraulic conductivity tests 56

69 Preparation of test series After compaction in the Proctor cylinder, the bottom of the cylinder was removed and sample was closed by rubber forms on both sides with double o-rings and double filter layers, as shown in Figure 39. The stamp shown in Figure 38 was used to cut filter sheets so that they fit accurately into the bottom of the rubber form, which corresponds to the samples diameter. Figure 39: Testing equipment for hydraulic conductivity tests The samples were closed tightly with the rubber lock and placed in the testing assembly shown in Figure 40. The samples were arranged on metal frame and fixed tightly by metal screws on top of a spacer piece which allowed for the discharge pipe to fixed to the top lock of the sample. The frame with the sample in place could then be moved up and down along a metal bar for adjusting the hydraulic gradient, i.e. the samples height in relation to the water supply level. Unfortunately, the water supply level is not included in Figure 40. The flow through the sample was directed from bottom to top as suggested by Forsberg (2007). By directing the flow upwards, particle migration was considered to be reduced. The discharge was collected in bottles that were covered by plastic sheets in order to prevent evaporation. It was possible to test several samples in the experiment arrangement at the same time. 57

70 Figure 40: Experiment arrangement for testing the hydraulic conductivity 58

71 TEST RESULTS The test results of the different tests with varying gradients were collected for setting up one diagram for each sample. In each diagram, the hydraulic conductivity k is plotted versus time. Even though the hydraulic conductivity k is measured in m/sec, it was decided to present time on the x-axis in hours to make it easier for the reader to see how long each test lasted. On samples no 1, 2, 3 and 4 three tests were carried out, i.e. one test for gradient 3, 5, and 7,3, respectively; sample 5 was tested twice under a gradient of 0,3, because it appeared that the sample was much more pervious than the previous samples. All tests with the corresponding evaluation of k are described below. Due to the proctor compaction method used for preparation of the samples they had to be allowed for saturation. The saturation phase can be identified in all diagrams presented in the following test description. It is characterized by an increase of the hydraulic conductivity during the first hours of testing. The length of the saturation phase varies between 2,5 to 65 hours; however, most samples reach saturation during the first day, approximately. The hydraulic conductivity then remains at the attained level or decreases. A decrease implies that grains in the soil structure reorganize due to the seepage pressure they are exposed to; migrating particles open and clog pathways for percolating water. In an ideal case, the hydraulic conductivity reaches a steady flow through the sample at an ideal soil structure where all particles remain in place. The hydraulic conductivity can then be evaluated from this steady state condition. This was not the case in all tests; the corresponding evaluation of results is given in each section on the particular sample tested. Sample 1 Sample 1 was tested first under a gradient of 7,3. Saturation of the sample took place during the first three to four hours of the test. Plotting the hydraulic conductivity k in m/sec versus time showed a relatively straight and nearly horizontal curve, so that the value for the hydraulic conductivity was easy to evaluate; k = 8, m/sec. The following test was carried out with a gradient of i = 3. The hydraulic conductivity decreased steadily during the nearly two-week long test, but did not reach an equilibrium. It is assumed that the curve would have reached some kind of balance after the last reading, so that the last measured value of 5, m/sec was regarded being representative. The final test on sample 1 carried out with a gradient of i = 5 showed that the hydraulic conductivity first increased slightly, decreased afterwards and finally evened out at a medium value of 2, m/sec. The results are shown in Figure

72 3,5E-08 Hydraulic conductivity (m/sec) 3,0E-08 2,5E-08 2,0E-08 1,5E-08 1,0E-08 5,0E-09 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 Figure 41: Test results for sample 1 Sample 2 Sample 2 was first tested under a gradient of i = 3. After the saturation phase the hydraulic conductivity reached a maximum and then reduced gradually to a value of 4, m/sec. The subsequent test carried out with a gradient of i = 5 showed a steady increase in flow, but appeared to even out after the last reading. The last two values measured were considered relevant, i.e. 2, m/sec. The sample was then tested under a gradient of 7,3, and the hydraulic conductivity showed to be relatively constant at 2, m/sec. However, the final two measurements appear to include a mistake in reading, because the curve shows a crack upwards. This section has therefore not been included in the evaluation. The test results are shown in Figure 42. 3,0E-09 Hydraulic conductivity (m/sec) 2,5E-09 2,0E-09 1,5E-09 1,0E-09 5,0E-10 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 Figure 42: Test results for sample 2 60

73 Sample 3 Sample 3 was first tested under a gradient of 7,3. During the first 20 hours, the sample saturated and k reached an equilibrium at 1, m/sec. During the following test with i = 3 the hydraulic conductivity decreased to a value of 3, m/sec. Due to the small scale in the diagram, k appears to be relatively steady. However, it was gradually decreasing and the curve had not evened out after 14 days of testing, which is why the final measurement was chosen as representative. During the last test under the gradient of 5 the hydraulic conductivity was relatively consistent at 1, m/sec. The test results are shown in Figure 43. 2,0E-09 Hydraulic conductivity (m/sec) 1,8E-09 1,6E-09 1,4E-09 1,2E-09 1,0E-09 8,0E-10 6,0E-10 4,0E-10 2,0E-10 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 Figure 43 : Test results for sample 3 Sample 4 The first test on Sample 4 was carried out with a gradient of 7,3. The flow increased during 3 days until a maximum was reached; this was the longest saturation phase. The flow then decreased steadily and did not reach an equilibrium after 8 days. The last measured value of 4, m/sec is considered being representative for this test. The following test was carried out under a gradient of 3, with a result similar to the first test. The flow gradually decreased and the final reading of 3, m/sec was regarded as representative. The final test under a gradient of 5 showed an almost constant flow through the sample which resulted in a mean value of 2, m/sec. The test results are shown in Figure

74 8,0E-10 Hydraulic conductivity (m/sec) 7,0E-10 6,0E-10 5,0E-10 4,0E-10 3,0E-10 2,0E-10 1,0E-10 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 Figure 44: Test results for sample 4 Sample 5 Sample 5 was compacted at a water content close to complete saturation. It was therefore chosen to test the sample first with application of i = 3. Shortly after starting the test, water together with fine particles was pressed out of the cylinder at both top and bottom lock. Since this seepage pressure obviously was too high for the sample, a gradient of i = 0,3 was applied, which corresponds to a height difference of only 2,50 cm from water supply level to sample. The first test carried out with a lower gradient resulted in k = 4, m/sec, see Figure 45. Because the sample was relatively pervious, the discharge collector had to be emptied after each measurement, it would otherwise overflow. This test lasted only for 37 hours. 4,5E-05 Hydraulic conductivity (m/sec) 4,0E-05 3,5E-05 3,0E-05 2,5E-05 2,0E-05 1,5E-05 1,0E-05 5,0E-06 0,0E Time steps (hrs) Figure 45: Result of the first test on sample 5 62

75 It was then decided to test the sample again under the same conditions, which resulted in a decrease of the hydraulic conductivity by two orders of magnitude: k = 4, m/sec, see Figure 46. Hydraulic conductivity (m/sec) 5,0E-09 4,5E-09 4,0E-09 3,5E-09 3,0E-09 2,5E-09 2,0E-09 1,5E-09 1,0E-09 5,0E-10 0,0E Time steps (hrs) Figure 46: Result of the second test on sample 5 Despite the fact that sample 5 was leaking and fine particles had been washed out by seepage pressure when trying to test it under a gradient of i = 3, the hydraulic conductivity measured in the second test reduced by two orders of magnitude. This reduction in hydraulic conductivity might be due to some rearrangement of particles in the sample. However, it is considered that the fine particles migrated and were captured by the filter layers. The filters then become plugged by particles, and as a result it is rather the hydraulic conductivity of the filter that is measured instead of that of the samples actual hydraulic conductivity. Because of this uncertainty, it was decided not to regard the test results of sample 5 in following considerations. The order of applied gradients was 7,3 3 5, except for sample 2, which was tested in the order of 3 5 7,3. This exception was not planned. It was assumed that starting testing samples under the highest gradient would increase the density due to reorganization of particles under a larger seepage pressure, so that k would decrease in the following tests. However, such an effect could not be concluded from test results. A summary of all test results is given in Table 7. It can be noted that measurements of k in sample 1 for i = 5 is larger than at other gradients and differs by one order of magnitude. The difference is less in sample 2 and 3, but the hydraulic conductivity is largest for the intermediate gradient of 5. In general, sample 4 has the lowest hydraulic conductivity compared to other samples, except sample 5. This can be related to the degree of compaction, which was highest for this sample. In contrast to the above described samples, the hydraulic conductivity measured under gradient i = 5 in sample 4 is lowest. 63

76 Table 7: Summary of measured values for the hydraulic conductivity Gradient Sample 1 ρ d = 2,15 t/m 3 w = 5,1 % n = 0,236 [m/s] Sample 2 ρ d = 2,13 t/m 3 w = 8,6 % n = 0,241 [m/s] Sample 3 ρ d = 2,16 t/m 3 w = 8,3 % n = 0,233 [m/s] Sample 4 ρ d = 2,22 t/m 3 w = 7,6 % n = 0,211 [m/s] Sample 5 ρ d = 2,08 t/m 3 w = 12,0 % n = 0,261 i = 0,3 [m/s] i = 3 5, , , , Test 1: 4, i = 5 2, , , , Test 2: 4, i = 7,3 8, , , , Assessment on hydraulic conductivity in field conditions Assessments on the hydraulic conductivity in field conditions were carried out in two different ways. First, the hydraulic conductivity measured in laboratory tests was related to field conditions as described below. The conductivity was then calculated using well known equations presented in chapter In order to be able to assess the hydraulic conductivity of the natural dam in field conditions, the results from the laboratory test series are collected and related to the samples porosity. The hydraulic conductivity varies, among other factors, with the porosity, as described in the Sate of the Art report in chapter two. By developing a relation between porosity, gradient and hydraulic conductivity in test conditions and applying the measured value for the porosity in situ in this relation, conclusions on the hydraulic conductivity in situ can be drawn. Figure 46 is based on information given in Table 7, where all test results are summarized. In addition, this table includes information on the samples porosity which was determined in the Standard Proctor compaction. The hydraulic conductivity measured under the three different gradients was plotted versus the porosity n in samples 1 4. In this diagram, all obtained values for k were included without review, even though k in sample 1 under gradient i = 5 and i = 7,3 is by one order of magnitude lower than all other measured values for k. Assessing the hydraulic conductivity in situ on basis of laboratory testing To be able to calculate the hydraulic conductivity for the porosity in situ, a mathematical relation between the test results has to be established. In reference to the section on mathematical calculations of the hydraulic conductivity presented in the State of the Art report, a power function was chosen. The corresponding equation for each trend line is displayed in the diagram shown in Figure

77 3,00E-08 Hydraulic conductivity k [m/sec] 2,50E-08 2,00E-08 1,50E-08 1,00E-08 5,00E-09 y = 2E-08x 2,4832 y = 7,836E+06x 2,427E+01 y = 14,038x 15,501 0,00E+00 0,21 0,215 0,22 0,225 0,23 0,235 0,24 0,245 Porosity n i = 3 i = 5 i = 7,3 Power (i = 5) Power (i = 7,3) Power (i = 3) Figure 47: Measured values for the hydraulic conductivity in laboratory conditions versus the samples porosity, all values included The applied porosity in situ is a result from the field density test described in chapter Table 3 presents the results from both rubber balloon tests, where the porosity was determined: n Test 1 = 0,185, and n Test 2 = 0,164. Further calculations on the hydraulic conductivity including an effective grain size and a mean value for the porosity will be presented in the next section. With respect to these following calculations a mean value for the porosity is also used in this section: n m =0,175. Inserting this mean porosity measured in field conditions into the equations evaluating the relationship between hydraulic conductivity and porosity in laboratory conditions results in a hydraulic conductivity between 2, m/s and k = 3, m/s in the natural dam, see Table 8. Table 8: Calculation of k in field conditions using relationships obtained in laboratory Gradient Equation (see Figure 46) Result [m/s] i = 3 k = n 2,4832 k = 2, i = 5 k = 7, n 24,27 k = 3, i = 7,3 k = 14,038 n 15,501 k = 2, Because the measured hydraulic conductivity in sample 1 under gradient 5 and 7,3 was by one order of magnitude higher than that measured in other samples (sample 5 excluded), an additional relation between hydraulic conductivity and porosity excluding those two values was established. The trend curve then appears to fit measured values of k better, see Figure

78 3,50E-09 Hydraulic conductivity k [m/sec] 3,00E-09 2,50E-09 2,00E-09 1,50E-09 1,00E-09 5,00E-10 y = 2E-08x 2,4832 y = 0,0234x 11,43 y = 4,035E+02x 1,799E+01 0,00E+00 0,21 0,215 0,22 0,225 0,23 0,235 0,24 0,245 Porosity n i = 3 i = 5 i = 7,3 Power (i = 5) Power (i = 7,3) Power (i = 3) Figure 48: Measured values for the hydraulic conductivity in laboratory conditions versus the samples porosity, two values from sample 1 excluded Again, the hydraulic conductivity was calculated using the mean porosity in field conditions, n m = 0,175, as done previously in calculations presented in Table 8. The result is a hydraulic conductivity between 2, m/s and k = 9, m/s in the natural dam, see Table 9. Table 9: Calculation of k in field conditions using relationships obtained in laboratory, excluding values for k in sample 1, i = 5 and i = 7,3 Gradient Equation (see Figure 47) Result [m/s] i = 3 k = n 2,4832 k = 2, i = 5 k = 403,5 n 17,99 k = 9, i = 7,3 k = 0,0234 n 11,43 k = 5, Comparing results from Table 8 and Table 9, the calculated hydraulic conductivity increases slightly for gradients 5 and 7,3 when excluding laboratory values that deviate by one order of magnitude. The lower value of k = 2, m/s remains the same, because the trend curve for gradient 3 is not affected by excluding values. However, all calculated results are in the range of to 10-12, which implies that the soil in the natural dam is practically impermeable, according to Mitchell & Soga (2005). Regarding the pore pressure measurements presented in chapter 3.2.3, which did not provide information on the ground water conditions, this result shows that it may be possible that the ground water table is significantly lower than assumed due to relatively impermeable material. 66

79 Calculations of the hydraulic conductivity using empirical equations A second comparison can be made by calculating the hydraulic conductivity on basis of the grain size distribution and mean porosity in situ, using equations presented in the State of the Art report in chapter 2. Three equations were chosen for comparison: Hazen, Chapuis, and Kozeny-Carman. To calculate the hydraulic conductivity with these equations, an effective diameter based on the grain size distribution of the glacial till is used. The effective diameter d 10 refers to the grain size for which 10 % is finer. The analysis of the grain size distribution is presented in chapter 3.2.4, and the mean value from all four sieving curves results in an effective diameter of d 10 = 0,0172 mm. The porosity used for calculations is the mean value n m = 0,175 as described above, so that the mean void ratio is e m = 0,212. The Kozeny-Carman equation includes several factors that are explained in detail in the State of the Art report, see chapter The wetted surface area S 0 is related to an effective 100% diameter and has to be calculated using the relation D eff =, which results in ( f / D ) D eff = 0,0187 mm for the given material. i ave,i The results of these calculations shown in Table 10 are considerably higher than those obtained from laboratory results and differ between one up to six orders of magnitude. The calculated values are in the range of those proposed by Todd & Mays (2005) for glacial till, and can be regarded as low or very low according to Mitchell & Soga (2005). Table 10: Results from calculations of the hydraulic conductivity with well-established equations Equation Result [m/s] Hazen k = C ( ) 2 2, d 10 Chapuis 0, ,4622 d e k = 9, e Kozeny- Carman K = k h μ 1 = γ 2 k T S e 1+ e 1, The values obtained in laboratory tests are considerably lower compared to those calculated by well established empirical relationships. The range for the hydraulic conductivity of glacial till presented by Todd & Mays (2005) is between 10-6 to 10-10, so that the laboratory tests themselves may have acceptable results, whereas results on the hydraulic conductivity in field conditions from laboratory relations appear to be remarkably low. The material practically would be impermeable, which probably could be applied in case of a different material, i.e. clay instead of well graded till. 67

80 3.2.7 Comments on the laboratory testing technique As Hellström et al. (2009) state, the flow through a soil sample in geotechnical laboratory testing is not always clear whether the flow is truly laminar, if inertia effects occur or the flow in the sample is turbulent. As described in the State of the Art report, Darcy s law states that the hydraulic conductivity is independent from the gradient, i.e. when plotting the relationship conductivity versus gradient, the results should be located on a horizontal line. Thus, to experimentally derive if the flow is laminar, the various applied hydraulic gradients are related to the hydraulic conductivity k. A dependency of k to the pressure gradient implies non-darcian flow, i.e. inertia dominated or turbulent flow. The measured hydraulic conductivity in present laboratory tests is therefore related to the applied gradient. The hydraulic conductivity in sample 1 differs by as much as one order of magnitude, and the dependency between k and the applied pressure gradient becomes clearly visible, see Figure 49. 3,0E-08 Hydraulic conductivity k [m/sec] 2,5E-08 2,0E-08 1,5E-08 1,0E-08 5,0E-09 0,0E Gradient i [-] Sample 1 Sample 2 Sample 3 Sample 4 Figure 49: Results for hydraulic conductivity versus gradient The dependency between hydraulic conductivity and gradient is striking. Tests on sample 1 appear to be representative, as discussed in the previous section. As done in the assessment on the conductivity in situ, results from sample 1 are removed from the diagram displayed in Figure 48 to obtain a better view on the relation between conductivity k and gradient i in samples 2, 3, and 4, see Figure

81 3,0E-09 Hydraulic conductivity k [m/sec] 2,5E-09 2,0E-09 1,5E-09 1,0E-09 5,0E-10 0,0E Gradient i [-] Sample 2 Sample 3 Sample 4 Figure 50: Results for hydraulic conductivity versus gradient, excluding Sample 1 Both diagrams show that the flow through the samples during laboratory testing may not have been laminar. Other possibilities for non-darcian flow behavior can be of practical nature and be due to redistribution and migration of particles, as well as leakage or other irregularities in the measuring cell (Hellström et al, 2009). Hence, the presented deviation may imply nonlinear flow through the sample, but could also be due to migrating particles or leakage along the rigid cell walls which could not be observed due to the non-transparency of the proctor cylinders used. It should therefore be considered to test the material in a closed system with a transparent cell where it is possible to apply backpressure and control flow conditions through the sample much more accurate than in a constant head permeameter. 69

82 4 DISCUSSION The experience of dam constructions and their function in long term perspective is limited. One way to overcome this lack of information is to study long term stable natural formations from the last glaciation. Such structures can give important information on the performance of water retaining embankments over long time periods. This time period may be up to thousands of years. A study provided by the Swedish Geological Survey SGU (2002) shows that many formations similar to dam constructions exist in nature, and such structures are stable over time periods extending up to 8000 to years. In this context, the author carried out field studies on a natural analogy located at the company area of Boliden in Gällivare, northern Sweden. These studies were complemented by laboratory tests and analysis, with the aim to provide knowledge of the geotechnical properties of such a long term stable formation. Field work included the excavation of test pits, measurement of in situ density, sampling, and installation of pore pressure gauges for ground water monitoring. The laboratory analysis ranges from grain size distribution, water content, the degree of compaction and the determination of the density of the solid grains. Further, a test series on the hydraulic conductivity of differently compacted samples were carried out. The results are discussed in the following. 4.1 In situ density and Proctor compaction The material analyzed is a well graded glacial till, with a clay content of 7 %, and a silt content of 19 %, see chapter For hydraulic conductivity tests, 5 samples were compacted following the Standard Proctor method, which enables compaction of different samples with varying water content under a defined energy input. The obtained dry density from Proctor compaction ranges from 2,08 g/cm 3 at a water content of 12 % to 2,22 g/cm 3 at a water content of 7,6 %. However, the two in situ density measurements resulted in 2,29 g/cm 3 at a water content of 7,8 % and 2,35 g/cm 3 at a water content of 8,8 %. Because of the comparatively higher dry density in situ it is assumed that a possible maximum for the dry density in laboratory conditions was not reached. A possible modified result of the Proctor compaction is therefore considered as shown in Figure 51. Figure 51: Possible modification of the Standard Proctor test results 70

83 The higher degree of compaction in situ compared to what was obtained in laboratory may be due to an inaccuracy in both field and laboratory testing methods. It may possibly even display how well nature can compact a material over time. In any case, the measured void ratio and porosity of the soil in situ is lower than that obtained by manual compaction in the laboratory, which plays an important role when assessing the hydraulic conductivity of the natural dam. 4.2 Hydraulic conductivity The hydraulic conductivity was tested on 5 samples with porosities ranging between 0,21 to 0,26. The latter value is that of sample 5, which was not included in further analysis due to leakage in the testing apparatus. Therefore, this discussion covers the remaining samples 1 to 4 with a porosity between 0,21 to 0,24, which corresponds to a void ratio of 0,27 to 0,32, see chapter The samples were tested under three different hydraulic gradients, and the hydraulic conductivity measured is between 2, m/s to 2, m/s, see Table 7, chapter Most measurements are in the range of 10-9 m/s to m/s. Such low values imply that the till is practically impervious (Kézdi, 1974; Mitchell & Soga, 2005; Todd & Mays, 2005). With the results from conductivity tests, a mathematical relation between porosity n and hydraulic conductivity k is established. The hydraulic conductivity in the analyzed natural formation can then be determined by applying the porosity in situ. The porosity in situ is a mean value from both in situ density tests, n m = 0,175, which corresponds to a void ratio in situ of e m = 0,212, compare Table 3, chapter Because the compaction of the material in nature is denser and thus both the corresponding porosity and the void ratio lower, the hydraulic conductivity in situ results in a lower value compared to laboratory measurements. Performing an evaluation according to the established relations the hydraulic conductivity in field conditions is determined to range between 2, m/s to 3, m/s, see Table 9, chapter The hydraulic conductivity appears to be too low for a glacial till by several orders of magnitude. The range given by Todd & Mays (2005) for glacial till is between 10-6 m/s and m/s, whereas Kézdi (1974) considers conductivities between 10-3 m/s to 10-7 m/s reasonable for fine sand, sandy silt or silt, and values below 10-7 impervious and typical for clay. Applying empirical relationships presented in chapter for the calculation of the hydraulic conductivity predicts a significantly higher hydraulic conductivity between 2, m/s to 1, m/s. This is a difference up to five orders of magnitude compared to results related to laboratory tests. The hydraulic conductivity in situ often differs significantly from that determined in the laboratory, since the natural soil mass is affected by stratification and inhomogeneities which cannot be simulated in laboratory tests (Kézdi, 1974). It may be possible that the test result from the in situ density test may be too low due to irregularities in the testing procedure, i.e. errors in the determination of the volume of the pit as described in chapter However, the strikingly low results from laboratory tests compared to those calculated with empirical relationships remain, thus leading to the presumption that the testing technique, i.e. the application of a constant head permeameter, may not have been suitable for the material. To check the results, testing in a triaxial cell under application of backpressure should be considered. Such tests would also allow the change of flow direction, reduce the risk for clogging filters and thus most likely give better results. 71

84 4.3 Ground water monitoring The low hydraulic conductivity could be an explanation for the absence of ground water during pore pressure monitoring, which is described in Results and conclusions of the pore pressure measurements were presented in chapter The originally assumed hydraulic gradient was 6,7 %, which results in an inclination of approximately 4º, see Figure 52. If the material actually has such a low hydraulic conductivity, it could be possible that the inclination of the groundwater table, i.e. the hydraulic gradient, is significantly higher than assumed in the first place. The water table would then have an inclination of 30º, which corresponds to a gradient of as much as 58 %, see Figure 52. Figure 52: Assumption of ground water table In this case, all pore pressure gauges would be located above the ground water table, which could explain the deficiency of pore pressure measurements. Another possibility for not being able to locate the phreatic surface may be a different water balance in the area, i.e. an absence of ground water communication between both lakes. The hydraulic gradient could correspond to the lower lake at a distance of 260 m with a water table of 441,5 masl, which was originally considered for analysis, see Figure 53. Figure 53: Possible alternative for ground water conditions in the area 72

85 Several pore pressure gauges may not function correctly, but it appears unlikely that all of them are malfunctioning, so that this source of error is not considered responsible for the complete absence of ground water. 4.4 Critical hydraulic gradients A summary of critical hydraulic gradients according to different soil types (gravel, coarse, medium and fine sand) was presented in Table 1, chapter The critical hydraulic gradient varies with grain size and hydraulic conductivity of a material. Further, it increases with increasing particle size. The natural analogies to dam structures studied by the Geological Survey of Sweden SGU mainly consisted of glacial till, clay, fine sand or silty loam, see Table 2, chapter The material of analyzed structure presented in the authors case study was well graded glacial till with a significant amount of fines, see chapter For comparison and discussion of critical hydraulic gradients in such material, the part referring to the least particle size, i.e. fine sand, presented in Table 1 is extracted and shown in Table 11. Table 11: Excerpt of critical hydraulic gradients from Table 1 (after Perzlmaier et al., 2007) Type of soil Fine sand I crit Chugaev 0,10 I crit Bligh 0,067 I crit Lane 0,048 I crit Mueller-Kirchenbauer, lower limit 0,06 I crit Mueller-Kirchenbauer, upper limit 0,08 I crit Weijers & Sellmeijer, C u = 1,5 0,09 I crit Weijers & Sellmeijer, C u = 3 0,14 The gradient in the formation analyzed in the case study is 0,067, which is in the range of values given in Table 11. However, the calculated gradient of 0,067 appears relatively high with respect to the gradients found in natural analogies, which were lower and in a range between 0,012 and 0,05, see Table 2, chapter The hydraulic gradient of 0,012 applying for lake Mången should be disregarded in this context because of the considerably coarser material and its notable large hydraulic conductivity. The range for natural formations presented by SGU is therefore reduced to 0,02 to 0,05, which is lower compared to the given range of 0,048 to 0,14 for fine sand, see Table 1. The decrease for the critical hydraulic gradient in natural formations could probably be related to the lesser particle size in the material, i.e. glacial till with a significant amount fines and a lower hydraulic conductivity compared to fine sand. Due to the lack of reliable ground water measurements in the case study it is difficult to determine the hydraulic gradient that actually exists in this structure. Because of the large 73

86 amount of fines in the soil, the calculated hydraulic gradient of 6,7 % may be reduced with regard to suggestions on critical gradients made in Table 1. In contrast to that, the hydraulic gradient might also be considerably higher and be as much as 58 %, as discussed above. A well graded glacial till may also be exposed to larger gradients due to its increased internal geometric stability, as discussed in the literature study, see chapter 2.3. Critical values for well graded material with smaller fractions than fine sand are not available. Reference values in Table 1 can therefore only be applied limited to the analyzed structure, thus limiting the comparison to the given critical gradients. In addition, the origin of the critical hydraulic gradients in Table 1 varies, they are partly developed in laboratory studies, or calculated with regard to the geometry of a soil structure, thus making further conclusions difficult. Considering a gradient of 5 % for a stable natural formation, the water table would have an inclination of about 3º. The design criteria for Swedish tailings dams presented in chapter relates the safety factor to the internal angle of friction, and a rule of thumb is to apply the materials friction angle φ multiplied by the factor 0,5 under the assumption that the water table and the slope surface coincide. As presented in Paper A, a major problem for this assessment is the determination of tailings friction angle, as it can vary between 18º up to 45º. For a material with a friction angle of 20º, a slope inclination of 7º and a gradient of 12 % could be accepted, and for a material with a friction angle of 40º, a slope inclination of 15º and a gradient of 27 % would meet requirements for the safety factor. Hence, the design criteria for tailings dams results in larger inclinations for the required slope of an embankment than in natural examples. However, the slope angle does not necessarily correspond to the inclination of the phreatic surface and the resulting hydraulic gradient in a tailings dam. In addition, it is not know how degradation affects the internal angle of friction on which the design criteria is based upon. With regard to long term perspective, it is likely that the material properties of tailings will change. The originally angular material will probably become more round, which reduces the friction angle, thus reducing the allowed inclination for the embankment. The degradation and its effects on the material properties and design criteria are beyond the scope of this work. However, further research on the material properties of tailings is needed to provide a basis for comparison and conclusions on the hydraulic gradient for the long term stability of tailings dams. 4.5 Future work Due to the low hydraulic conductivity obtained in laboratory testing it should be considered to carry out further tests on the material for comparison and evaluation. These tests should be carried out in a closed system with the possibility to apply backpressure. The water balance in the area should be considered for further studies in order to locate the ground water table for clarification of the flow conditions and the hydraulic gradient the natural dam is exposed to. 74

87 5 CONCLUSIONS Natural analogies to dam constructions are considered giving valuable information on the long term stability of tailings dams. On basis of information provided by the Geological Survey of Sweden SGU, natural formations being exposed to seepage and a hydraulic gradient over long time periods have been presented and analyzed. These examples show that it is basically possible for an embankment to withstand a certain hydraulic gradient over long time periods; in this case since the last glaciation, i.e. thousands of years. The objective of this study was to contribute to tailings dams design in long term perspective, in respect to critical hydraulic gradients. Information on critical hydraulic gradients given in literature was collected, and long term stable natural formations were presented. In addition, a case study on a natural dammed lake was carried out. From these studies, the following conclusions can be drawn: o Critical hydraulic gradients for fine sand given in the literature range from 4,8 % to 14 %. o The design criteria for Swedish tailings dams results in theoretic hydraulic gradients between 12 % to 27 % based on a friction angle ranging from 20º to 40º. It should be stressed that using a criteria of this type excludes the degradation of the material over time. The design should not be based upon properties obtained from laboratory testing of fresh materials only. Upon degradation, the angularity of tailings is expected to decrease and the particles will become more round, thus reducing the internal angle of friction. With regard to both degradation and the comparison to gradients found in natural formations, the design criteria should be modified to reduce the currently allowed gradient. o The hydraulic gradients in natural formations presented on basis of the Study conducted by SGU are between 2 % to 5 %. o The calculated hydraulic gradient in the case study is 6,7 % and therewith higher than that of other natural formations presented. In case of a ground water table located below pore water gauges, the hydraulic gradient would increase to a value of 58 %. However, the actual hydraulic gradient remains unclear due to the absence of pore pressure readings. o It is considered that natural deposits can be compacted to an optimum over long time periods, thus resulting in considerable low hydraulic conductivities. o Whether the results from laboratory tests on the hydraulic conductivity are reasonable or not remains questionable; calculations with empirical relationships result in a hydraulic conductivity up to almost five orders of magnitude higher. o The low hydraulic conductivity in the natural dam could explain the absence of ground water during pore pressure monitoring. Another possibility for not being able to locate the ground water level is a different water balance in the area. 75

88 ACKNOWLEDGEMENTS The research presented in this thesis was carried out as a part of "Swedish Hydropower Centre - SVC". SVC has been established by the Swedish Energy Agency, Elforsk and Svenska Kraftnät together with Luleå University of Technology, The Royal Institute of Technology, Chalmers University of Technology and Uppsala University. Participating hydro power companies are: Andritz Hydro Inepar Sweden, Andritz Waplans, E.ON Vattenkraft Sverige, Fortum Generation, Holmen Energi, Jämtkraft, Karlstads Energi, Linde Energi, Mälarenergi, Skellefteå Kraft, Sollefteåforsens, Statkraft Sverige, Statoil Lubricants, Sweco Infrastructure, Sweco Energuide, SveMin, Umeå Energi, Vattenfall Research and Development, Vattenfall Vattenkraft, VG Power and WSP. 76

89 REFERENSES Agrell, H. (2002). Naturligt dämda sjöar - analogier av dammkonstruktioner. Uppdragsrapport för Svarliden Guld AB. Geological Survey of Sweden SGU, Uppsala, Sweden. (In Swedish) Ahlmann, H., Caldenius, C., and Sandegren, R. (1924). Ragundasjön. En geomorfologisk, växtgeografisk undersökning. Doctoral Thesis. Geological Survey of Sweden SGU, Stockholm. (In Swedish) Åkerlund, H. (2007). Personal communication. Aubertin, M., Bussière, B., and Chapuis, R. P. (1996). Hydraulic conductivity of homogenized tailings from hard rock mines. Canadian Geotechnical Journal, Vol. 33, Nr. 3, pp Bear, J. (1972). Dynamics of fluids in porous media. American Elsevier Publishing Company, Inc., New York. ISBN X. Bjelkevik, A. (2005a). Water Cover Closure Design for Tailings Dams. State of the Art Report. Research Report, Luleå University of Technology, Luleå, Sweden. ISSN Bjelkevik, A. (2005b). Stability of Tailings Dams. Focus on Water Cover Closure. Licensiate Thesis, Luleå University of Technology, Luleå, Sweden. ISSN Brown, A. J. (2007) A framework for the Management of Risk from Internal Erosion. In: Assessment of the Risk of Internal Erosion of Water Retaining Structures: Dams, Dykes and Levees. Intermediate Report of the European Working Group of ICOLD. Contributions to the Symposium in Freising, Germany, September ISBN Carrier, W. D. (2003). Goodbye, Hazen; Hello, Kozeny-Carman. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 129, No. 11, pp Chapuis, R. P. (1992). Similarity of internal stability criteria for granular soils. Canadian Geotechnical Journal, Vol. 29, pp Chapuis, R. P. (2004). Predicting the saturated hydraulic conductivity of sand and gravel using the effective diameter and void ratio. Canadian Geotechnical Journal, Vol. 41, pp Cernica, J. N. (1995) Geotechnical engineering: soil mechanics. John Wiley and Sons, Inc. ISBN Daniel, D. E., Anderson, D. C., and Boynton, S. S. (1985). Fixed-Wall Versus Flexible-Wall Permeameters. In: Hydraulic Barriers in Soil and Rock. ASTM STP 874. Johnson, A. I., Frobel, R. K., Cavalli, N. J., and Pettersson, C. B., Eds. American Society for Testing Materials, Philadelphia, 1985, pp Fannin, R. J., and Moffat, R. (2006). A large permeameter for study of internal stability in cohesionless soils. Geotechnical Testing Journal, Vol. 29, No. 4, pp Fell, R., MacGregor, P., Stapledon, D., and Bell, G. (2005). Geotechnical Engineering of Dams. A.A. Balkema, Leiden. ISBN

90 Forsberg, T. (2007, 2009). Personal discussion. Foster, M., Fell, R., and Spannagle, M. (2000a). The statistics of embankment dam failures and accidents. Canadian Geotechnical Journal, Vol. 37, pp Foster, M., Fell, R., and Spannagle, M. (2000b). A method for assessing the relative likelihood of failure of embankment dams by piping. Canadian Geotechnical Journal, Vol. 37, pp Fry, J. J. (2007). Context and Framework of the Report of the European Working Group on Internal Erosion. In: Assessment of the Risk of Internal Erosion of Water Retaining Structures: Dams, Dykes and Levees. Intermediate Report of the European Working Group of ICOLD. Contributions to the Symposium in Freising, Germany, September ISBN Geometric (2009) Produktblad om tryckgivare. Stockholm, Sweden. (October 2009) Hellström, J. G. I. (2007). Parallel Computing of Fluid Flow Through Porous Media. Licentiate Thesis, Luleå University of Technology, Luleå, Sweden. 2007:06. ISSN Hellström, J. G. I., Lundström, T. S., Jantzer, I. and Knutsson, S. (2009). Theoretical limits of the pressure gradient applied to drive the flow during geotechnical experiments. In: Internal erosion in embankment dams Fluid flow through and deformation of porous media. Hellström, J. G. I. Doctoral Thesis, Luleå University of Technology, Luleå, Sweden. ISSN Kézdi, Á. (1974). Handbook of Soil Mechanics. Vol. 1. Soil Physics. Elsevier Scientific Publishing Company, New York. ISBN X. Kenney, T. C. (1974). Embankment-dam engineering: Book Review. Canadian Geotechnical Journal, Vol. 11, pp Kenney, T. C., and Lau, D. (1985). Internal stability of granular filters. Canadian Geotechnical Journal, Vol. 22, pp Kenney, T. C., Lau, D., and Ofoegbu, G. I. (1984). Permeability of compacted granular materials. Canadian Geotechnical Journal, Vol. 21, pp Kenney, T. C., and Westland, J. (1993). Laboratory study of segregation of granular filter materials. In: Filters in geotechnical and hydraulic engineering. Edited by Brauns, J., Heibaum, M., and Schuler, U. A.A. Balkema, Rotterdam. pp ISBN Kjærnsli, B., Valstad, T., and Höeg, K. (1992). Rockfill Dams - Design and Construction. Hydropower Development, Vol. 10. Published by Norwegian Institute of Technology, Division of Hydraulic Engineering, Trondheim. ISBN Knutsson, S. (2009). Personal discussion. Lafleur, J. (1984). Filter testing of broadly graded cohesionless tills. Canadian Geotechnical Journal, Vol. 21, pp Lafleur, J., Mlynarek, J., and Rollin, A. L. (1989). Filtration of broadly graded cohesionless soils. Journal of Geotechnical Engineering, Vol. 115, No. 12, pp Lofquist, B. (1988). Hydraulic Fracturing in Embankment Dams. Discussion. Journal of Geotechnical Engineering, Vol. 114, No. 6, pp , June

91 Lambe, T. W. and Whitman, R. V. (1979). Soil Mechanics, SI Version. John Wiley and Sons, Inc., Delhi. ISBN Mesri, G., and Ali, S. (1988). Discussion of Hydraulic Fracturing in Embankment Dams. Journal of Geotechnical Engineering, Vol. 114, No.6, pp Milligan, V. (1986). Internal stability of granular filters: Discussion. Canadian Geotechnical Journal, Vol. 23, pp Milligan, V. (2003). Some uncertainties in embankment dam engineering. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 129, No. 9, pp Mitchell, J. K. and Soga, K. (2005). Fundamentals of soil behavior. John Wiley and Sons, Inc., Hoboken, New Jersey. ISBN Muckenthaler, P. (1989). Hydraulische Sicherheit von Staudämmen. Doctoral Thesis, Technische Universität München, Germany. (In German) Perzlmaier, S. (2007). Verteilte Filtergeschwindigkeitsmessung in Staudämmen. Doctoral Thesis, Technische Universität München, Germany. (In German) Perzlmaier, S., Muckenthaler, P. and Koelewijn, A. R. (2007). Hydraulic Criteria for Internal Erosion in Cohesionless Soil. In: Assessment of the Risk of Internal Erosion of Water Retaining Structures: Dams, Dykes and Levees. Intermediate Report of the European Working Group of ICOLD. Contributions to the Symposium in Freising, Germany, September ISBN RIDAS (2004) Kraftföretagens riktlinjer för damsäkerhet. (In Swedish.) Sherard, J. L. (1979). Sinkholes in Dams of Coarse, Broadly graded Soils. 13 th ICOLD Congress, New Delhi, India, Vol. II, pp Sherard, J. L., Dunnigan, L. P., and Talbot, J. R. (1984). Basic properties of sand and gravel filters. Journal of Geotechnical Engineering, Vol. 110, Nr. 6, pp Sherard, J. L. (1986). Hydraulic Fracturing in Embankment Dams. Journal of Geotechnical Engineering, Vol. 112, No.10, pp Skempton, A. W., and Brogan, J. M Experiments on piping in sandy gravels. Géotechnique, Vol. 44, No. 3, pp Smoltczyk, U. (Editor) (2002). Geotechnical Engineering Handbook. Vol. 1: Fundamentals. Ernst & Sohn Verlag, Berlin. ISBN Sowers, G. F. (1993). Human factors in civil and geotechnical engineering failures. Journal of Geotechnical Engineering, Vol. 119, No. 2, pp Stenman U. (2007, 2009). Personal discussion. Swedish Standard SS Geotechnical tests Compaction tests in field, edition 4. SIS Standardiseringskommissionen i Sverige and BST Byggstandardiseringen. Mars (In Swedish) Swedish Standard SS Geotechnical tests Grain size distribution - Sieving, edition 2. SIS Standardiseringskommissionen i Sverige and BST Byggstandardiseringen. Mars (In Swedish) Swedish Standard SS Geotechnical tests Compaction properties Laboratory compaction, edition 4. SIS Standardiseringskommissionen i Sverige and BST Byggstandardiseringen. Mars (In Swedish) 79

92 Swedish Standard SS Geotechnical tests Grain density and bulk density, edition 3. SIS Standardiseringskommissionen i Sverige and BST Byggstandardiseringen. November In Swedish. Terzaghi, K., Peck, R. B., and Mesri, G. (1996). Soil mechanics in engineering practice. John Wiley & Sons Inc., New York. ISBN Todd, D., and Mays, L. (2005). Groundwater Hydrology. John Wiley and Sons, Inc. ISBN Vattenfall (1988). Jord- och Stenfyllningsdammar. Vattenfall, Stockholm. ISBN (In Swedish). Vick, S. G. (1990). Planning, Design, and Analysis of Tailings Dams. BiTech Publishers Ltd. Vancouver, B.C., Canada. ISBN Viklander, P. (1997) Compaction and Thaw Deformation of Frozen Soil. Permeability and Structural Effects due to Freezing and Thawing. Doctoral Thesis, Luleå University of Technology, Luleå, Sweden. 1997:22. ISSN Wan, C. F., and Fell, R. (2004a). Investigation of rate of erosion of soils in embankment dams. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 130, No. 4, pp Wan, C. F., and Fell, R. (2004b). Experimental investigation of internal instability of soils in embankment dams and their foundations. UNICIV Report No. R 429. University of New South Wales, Sydney, Australia. ISBN Weijers, J., and Sellmeijer, J. (1993). A new model to deal with the piping mechanism. In: Filters in geotechnical and hydraulic engineering. Edited by Brauns, J., Heibaum, M., and Schuler, U. A.A. Balkema, Rotterdam. pp ISBN Wittmann, L. (1980). Filtrations- und Transportphänomene in porösen Medien. Doctoral Thesis at University of Karlsuhe (TH), Germany. (In German) Ziems, J. (1969). Beitrag zur Kontakterosion nichtbindiger Erdstoffe. Doctoral Thesis at Technische Universität Dresden, Germany. (In German) 80

93 APPENDIX 1 Results from hydraulic conductivity tests: Flow vs. time 1,4E-02 Flow Sample 1 1,2E-02 Flow (ml/sec) 1,0E-02 8,0E-03 6,0E-03 4,0E-03 2,0E-03 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 1,4E-03 Flow Sample 2 1,2E-03 1,0E-03 Flow (ml/sec) 8,0E-04 6,0E-04 4,0E-04 2,0E-04 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 I

94 1,0E-03 9,0E-04 Flow Sample 3 8,0E-04 Flow (ml/sec) 7,0E-04 6,0E-04 5,0E-04 4,0E-04 3,0E-04 2,0E-04 1,0E-04 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 5,0E-04 Flow Sample 4 4,5E-04 4,0E-04 Flow (ml/sec) 3,5E-04 3,0E-04 2,5E-04 2,0E-04 1,5E-04 1,0E-04 5,0E-05 0,0E Time steps (hrs) i = 3 i = 5 i = 7,3 II

95 3,5E-02 Flow Sample 5 Test 1 3,0E-02 2,5E-02 Flow (ml/sec) 2,0E-02 1,5E-02 1,0E-02 5,0E-03 0,0E Time steps (hours) 1,5E-04 Flow Sample 5 Test 2 1,3E-04 Flow (ml/sec) 1,0E-04 7,5E-05 5,0E-05 2,5E-05 0,0E Time steps (hours) III

96

97 Conference paper #1 Jantzer, I., Bjelkevik, A., and Pousette, K. (2008). Material Properties of Tailings from Swedish Mines. Nordic Geotechnical Meeting NGM. Sandefjord, Norway. Sept. 5, 2008.

98

99 Material properties of Tailings from Swedish mines I. Jantzer Luleå University of Technology, Sweden, A. Bjelkevik Sweco AB, Sweden, K. Pousette Luleå University of Technology, Sweden, Abstract: Tailings impoundments are designed and constructed for disposal of mine waste, i.e. tailings, environmentally safe in a long-term perspective. Tailings dams are raised along with a mines production. Important for tailings dam design and construction are the material properties of tailings, with shear parameters and hydraulic conductivity in focus. Studies on these properties have been carried out and are presented. Results show a large variation of properties, thus suggesting that common geotechnical test methods may not be applicable for tailings. 1 INTRODUCTION Tailings impoundments are constructed to store the fine residual from mining activities. Ore is crushed and milled to a fine sand in the plant to enable extraction of material of interest, which normally are metals. About 70-99% of the ore is leftover, non-valuable crushed and milled material, i.e. tailings, which is often pumped with process water as slurry from the plant to a tailings impoundment. At the tailings impoundment the rest material is allowed to settle, and thus the water is cleared and can be reused in the plant. The tailings impoundment is normally surrounded by tailings dams (embankments) and, if the topography allows for it, natural heights. In order to build the tailings dams as efficiently as possible, tailings themselves are often used as construction material. When tailings are used for dam construction the geotechnical (and environmental) properties of the material become important to know in order to build safe embankments in both short and long term perspective. To design and construct tailings dams, geotechnical properties of tailings, in particular density, shear strength parameters and hydraulic conductivity, normally have to be known. While the density is relatively easy to measure, shear strength parameters and hydraulic conductivity vary significantly with sampling and testing technique and are therefore difficult to determine and assess. This has an impact on calculation procedures and obtained results. A major problem is that the production process creates angular particles at a size similar to clay, slit and fine sand, which do not behave in the same way as natural geological material. This paper presents laboratory test results from tailings in a short term perspective. Parameters are important to know in a long-term perspective. They will be essential when assessing the mechanical and environmental behavior of embankment structures in the future, but are not discussed in this paper.

100 2 TAILINGS DAM CONSTRUCTION Tailings dam construction is normally divided into three main construction methodologies; upstream, centre line, and downstream construction, as depicted in Figure 1, respectively. Many Swedish tailings dams are originally constructed more or less like conventional earth and rock fill water retention dams. One main difference between tailings dams and water retention dams is that tailings dams are raised in stages or continuously as the impoundment is filled up with tailings along with mining production. a) b) c) 1 Starter wall, 2 deposited tailings, 3 support fill Figure 1. The three main construction methods for tailings dam construction. Upstream (a), centre line (b) and downstream (c) construction. A common trend involves more and more tailings being used for construction of tailings dams and application of the upstream method. Applying the upstream method, a starter dike is constructed with borrow material, because tailings are not produced at this time. Tailings are then discharged from the crest of the dam along the impoundments periphery and left for sedimentation, thus creating a tailings beach from the dam crest towards the impoundment. When the impoundment is full, or rather before that, a second dike is constructed on the settled and consolidated tailings beach. This process continues as the tailings dam increases in height [3]. Due to many reasons, such as new knowledge, new demands from both the company s and society s point of view, extended environmental concern, improved extraction process, changing staff, etc. the construction methods at Swedish dams have changed over the years.

101 Figure 2 shows some example of how Swedish tailings dams may look in practice. Tailings dam construction does not follow one construction principle, but often becomes a mixture of different construction practices and special adjustments. Upstream construction Impoundment, i.e tailings Buttress support Downstream construction Filters Impoundment Support fill Moraine core Moraine Dyke Beach Impoundment Moraine Support fill Figure 2. Tailings dam construction in practise for three different Swedish tailing dams (after [1]). Advantages of the upstream construction method are [1]: Low cost and simplicity. The man made dikes may be constructed of sand and this process is simple and ongoing. The method results in a low hydraulic gradient due to long beaches and gradually coarser fractions of tailings closer to the dam crest as a result from the hydraulic deposition. A low hydraulic gradient is favorable for long term conditions [1]. The outer slope can be remediated during operation as the crest moves inwards.

102 Disadvantages or factors that are a constraint to the application of the upstream construction include [1]: control of the hydraulic gradient as filter layers are difficult to apply gradually reduced space for deposition of tailings and water because of inward construction susceptibility to seismic liquefaction due to ongoing construction on deposited tailings lower raise rate because the deposited material has to settle and consolidate before further construction dust control at high winds In order to construct a safe tailings dam using tailings and the upstream construction method it is important to understand and know the: principles for upstream construction geotechnical properties of the tailings behavior of the tailings material water balance This article does not discuss all four issues, but the geotechnical properties of the tailings. 3 MATERIAL PROPERTIES OF TAILINGS Tailings are particles of crushed rock with particle sizes ranging from clay to sand. Normally, the grain size varies from 0,01 mm to 1,0 mm, but up to 20 % clay-sized particles, i.e. 0,002 mm, can be found. Such variations occur dependent on sedimentation, site and processing methods. Tailings could be regarded as natural materials as they are basically crushed rock, but the fine material contains chemicals and metals, which may be environmentally harmful when released in combination with water and air. They can be described in soil mechanical terms and the geotechnical properties can partly be compared to natural materials. Yet, tailings characteristics can diverge in their characteristics due to the variations in origin and processing of the ore, as well as deposition methods. The origin affects the size and the gradation of the grains, the internal friction angle and the particle density; whereas the deposition method is responsible for bulk density, void ratio and porosity, and the hydraulic conductivity. Tailings generally have high water content and porosity, a low to moderate hydraulic conductivity and a low plasticity when compared to soil. The shear strength was rated low to moderate by ICOLD, but found to be moderate to high in relation to the grain size compared to natural material [2]. 3.1 Shear strength Direct shear tests and triaxial tests were carried out on tailings samples from Boliden s mine Aitik in northern Sweden. Samples were taken at different points from the outlet of the pumping station at different depths. A test program for a variety of undisturbed samples according to CPT-results was carried out. Samples with low CPT resistance are called loose, whereas samples taken from depths with higher resistance are called consolidated. Samples from beach are taken close to the discharge point at the dam crest. Fabricated

103 samples are tailings samples fabricated in the laboratory. Both loose and consolidated samples from different depths in the impoundment were tested. Loose samples were taken at depths ranging between 3-17m, whereas consolidated samples were taken at 12-20m depth. It is not only the depth, but grain size and settlement that control the depositions influence on the samples compaction and drainage. Direct shear tests where carried out on 18 samples, where 15 samples where taken from boreholes and 3 samples where fabricated in the laboratory. Consolidation was carried out at 20 kpa, 150 kpa, and 300 kpa, respectively. Results are shown in Table 1. Table 1. Results from direct shear tests (drained) on tailings [4] Sample Friction angle, respectively Inclination of stress path Loose 24 / 22 / 20 c = 3,9 kpa φ = 19,4 Loose, fabricated sample 33 / 22 / 19 c = 8,6 kpa φ = 18,2 Consolidated 46* / 32 / 28 c = 13,7 kpa φ = 26,7 Consolidated (from beach) 34 / 31 / 18 c = 9,4 kpa φ = 16,0 Consolidated (from beach) 45 / 34 / 26 c = 8,9 kpa φ = 23,9 * peak value The angle of friction varies between 18 and 46 regarding single values, and it is even lower when regarding the inclination of the failure envelope. The failure shear strength is obtained from the deformation at 0,15 rad, a peak value was only obtained in one case. The friction angle decreases with increasing consolidation stress. The evaluation of test results as shown in Figure 3 contains both values for the angle of friction for each of the three points in the graph, i.e. the inclination of the connecting line to the origin, and the angle of the envelope, which is an approximation based upon linear regression from the failure shear strength at different consolidation stresses. Even though tailings are considered non-cohesive, the stress path does not intercept the origin on extension, thus implying a value for cohesion. τ c σn Figure 3. Principle for evaluation of results from direct shear tests [4] The test program for triaxial tests included 12 tests, whereof 10 samples where taken from boreholes and two samples where fabricated in the laboratory. Active isotrope consolidation was carried out to in-situ stress. Pore pressure was measured in undrained conditions. Results from the tests are shown in Table 2.

104 Table 2. Results from triaxial tests on tailings [4] Sample Test conditions Cons. stress Evaluation of φ Deviatoric stress (maximum value) Evaluation φ Residual stress Evaluation φ 15% deformation Consolidated drained Consolidated drained Consolidated drained Consolidated undrained Loose drained 40 *42 *41 Loose drained Loose drained Loose undrained Loose undrained Loose undrained Loose, fab. drained Loose, fab. drained Maximum deviatoric stresses have been evaluated in most of the tests. Those samples which did not have a peak value were evaluated using a 15% deformation limit. In one case, marked with *, a maximum value was reached after a 15% deformation. Results from triaxial tests do not vary as much as results from direct shear tests, but friction angles are generally higher compared to results from direct shear tests. Despite the larger variation of consolidation stress in triaxial tests, the friction angle has a comparatively narrow range from Hydraulic conductivity Tailings generally have a relatively low uniformity coefficient C u between 3 and 8 [2]. The value is not constant, as it varies with the distance to the discharge point because of the sorting effect of sedimentation. The hydraulic deposition affects the properties of tailings, so that the parameters change with distance from the outlet. The slurry s flow, velocity and solid content have additional effect on the sedimentation. With increasing distance from the outlet, the bulk and grain density decrease, thus resulting in varying void ratio, porosity and dry density. The dry density also decreases at larger distance, whereas both the void ratio and porosity are relatively smaller closer to the discharge point [2]. Results from the study on tailings from different Swedish mines carried out by [2] are shown in Table 3. The vertical hydraulic conductivity k v has been measured on undisturbed samples and compared with calculated values. The hydraulic conductivity can be calculated in numerous ways; well known empirical relationships are those from Hazen, Kozeny-Carman, and Chapuis, which take grain size, porosity and particle shape differently into account. Measured and calculated values have been compared to each other by calculation of a ratio. Maximum and minimum values of measured and calculated values, as well as maximum and minimum ratios are printed bold. Results vary significantly, so that the calculation of an actual value by empirical relationships does not appear adequate. The study conducted by [2] shows that observed values for void ratio and hydraulic conductivity do not correspond to each other.

105 Table 3. Hydraulic conductivity of tailings [2] Ratio Ratio Ratio Measured Calculation Measurement Calculation Measurement Calculation Measurement k v Hazen / Kozeny- / Chapuis / Calculation Carman Calculation Calculation Site [10-6 m/s] [10-6 m/s] [%] [10-6 m/s] [%] [10-6 m/s] [%] Kiruna 14,7 36,00 40,8 12,20 120,5 64,34 22,8 Svappavara 6,08 49,00 12,4 51,56 11,8 263,99 2,3 5,67 39,96 14,3 23,99 23,6 124,62 4,5 Malmberget 16,3 56,25 29,0 19,25 84,7 92,12 17,7 18,7 39,69 47,1 18,08 103,4 93,35 20,0 2,54 1,00 254,0 0,50 510,1 5,73 44,3 Aitik 1,41 2,56 55,1 1,58 89,2 14,91 9,5 1,01 2,56 39,5 3,24 31,1 32,11 3,1 Boliden 2,56 1,44 177,8 1,67 153,2 18,58 13,8 2,78 0,36 772,2 0,47 588,2 7,19 38,7 Garpenberg 2,68 1,21 221,5 0,79 338,6 8,81 30,4 1,70 0,36 472,2 0,52 328,9 7,94 21,4 Zinkgruvan 5,41 4,00 135,3 2,92 185,0 25,22 21,4 4 CONCLUSIONS Studies on shear parameters show that the variation of results is generally large, and that results from direct shear tests vary more than those from triaxial tests. The angularity of tailings may be one explanation for diverging results; the stress concentration at edges during shear probably leads to further crushing and higher resistance to shear. Crushing and grinding creates particles that are much more angular than natural grains. While natural material has been tested and described extensively, tailings properties raise new questions. Porosity, shear strength, and hydraulic conductivity are usually well-described properties for natural round and well-graded material. Because of saturation and the lack of cohesion, the placement of an undisturbed tailings sample in a triaxial test device is difficult to carry out, thus making representative laboratory testing on tailings difficult. Measurement and calculation of the hydraulic conductivity of tailings appears to be difficult as the study shows that obtained values have a large variation and that increased void ratio does not correspond to increased hydraulic conductivity. The nature of the hydraulic deposition seems to play an important role. In addition, questions arise whether the flow pattern in tailings is the same as in natural materials; shape, chemical and mineralogical properties seem to affect the flow significantly.

106 5 ACKNOWLEDGEMENTS This research study is part of the work within Swedish Hydropower Center SVC, which has been established by the Swedish Energy Agency, Elforsk and Svenska Kraftnät together with Luleå University of Technology, The Royal Institute of Technology, Chalmers University of Technology and Uppsala University. The study is carried out in cooperation with the Swedish mining industry and dam safety committee. REFERENSES 1. Bjelkevik, A. Water Cover Closure Design for Tailings Dams-State of the Art Report. Research Report. Luleå University of Technology. 2005:19, Bjelkevik, A. Stability of tailings dams-focus on water cover closure. Licentiate thesis. Luleå University of Technology. 2005:85, Vick, S.G. Planning, Design and Analysis of Tailings Dams. BiTech Publishers Ltd. Vancouver, Canada ISBN Pousette, K. Laboratorieförsök på anrikningssand från Aitik. Ödometerförsök, skjuvförsök, triaxialförsök. Internal working document, Luleå University of Technology

107 Conference paper #2 Jantzer, I. and Knutsson, S. (2007) Effects of Freezing and Thawing in Embankment Dams. Proceedings of International Symposium on Modern Technology of Dams the 4 th EADC Symposium. Chengdu, China. Oct. 14, 2007.

108

109 EFFECTS OF FREEZING AND THAWING IN EMBANKMENT DAMS Isabel Jantzer Sven Knutsson Dept of Civil, Mining and Env. Eng. Dept of Civil, Mining and Env. Eng. Luleå University of Technology Luleå University of Technology Luleå, Sweden Luleå, Sweden Abstract: Embankment dams located in cold climate are subjected to freezing condition, which is usually not taken into account in the design. Frost susceptible fill material may be exposed to frost action, a process known for ice lens formation which increases water content and soil volume. Upon thawing, rearrangement of particles takes place and the soil structure is altered. Fine grained soils may therefore exhibit weak zones with reduced shear strength and increased hydraulic conductivity. Internal erosion may start in such weak zones, thus causing problems for the dam construction. Therefore, climate should be considered in order to find temperature distribution and pore pressures in the embankment. A study of freezing effects has been carried out at the Suorva hydropower embankment dam in northern Sweden. The study included field investigations, i.e. test pit excavation, where weak zones have been observed. The comparison of results to those from temperature calculations with a commercial finite element program showed that the core is exposed to frost action, thus causing seepage problems. Keywords Dam design, frost action, internal erosion. 1 Introduction Construction in cold climate areas involves special engineering problems and designs in order to manage the impact of freezing and thawing on the soil material. Such special problems generally refer to unwanted movements and stress increases and it normally affects constructions like roads, foundations, water pipes etc., but also dams are affected. The effects are most obvious in tailings dams but also hydropower dams might be affected. This paper deals with a study of a frost exposed hydropower dam in northern Sweden, where the length of winter varies from two months in the south to 8 months in the north. The cold climate has significant influence on construction periods and routines. Most of the embankment dams in Sweden have been constructed during the 1950s and 1960s. A common fill material used for the hydraulic barrier is glacial till, which has a complex grain size distribution, ranging from clay to boulder. The relatively high amount of fines results in highly frostsusceptible material. The constructions have been exposed to repeated freezing and thawing since construction. Cold climate can have significant influence on embankment dams both during the construction period and the service life after construction. Effects of temperature and precipitation are not always taken into account when making judgments on stresses, deformations or seepage. One generally accepted definition of the term cold regions is that of areas with a minimum seasonal frost penetration of 300 mm. Cold regions may be subdivided into permafrost regions, both continuous or discontinuous, or seasonally frozen ground. Such regions are found in northern America, Canada, parts of Europe and Asia.

110 2 Frost action and its effects on earth fill material Civil engineering structures may suffer from climate impact in many different ways. Freezing conditions in soil may result in frost heave and reduced bearing capacity upon thawing. Soil exposed to repeated freezing and thawing processes may experience changes in its volume and structure, thus altering properties such as strength, compressibility or hydraulic conductivity (Viklander, 1997). 2.1 Frost-susceptibillity and freezing index Three conditions must exist for frost action to occur in a soil: Sufficiently cold temperatures for freezing, the presence of a frost-susceptible material, and presence of water. The frost-susceptibility of a soil is measured in terms of its behavior during frost heave and at thaw. The classification of a soil regarding its frost-susceptibility can be done by laboratory frost heave tests. An approximate classification of the frost-susceptibility of soils can be done by screening or based on the Unified Soil Classification System [Andersland, 1994 #5]. However, many different methods have been suggested for categorization, which implies that there is no generally accepted criterion. In Sweden, four different frost-susceptibility classes are used for categorization: negligibly, low, medium and highly frost-susceptible soils. The underlying criteria for classification are the percentage of fine-grained material, the hydraulic conductivity and the capillary rise. Glacial till, which is commonly used as hydraulic barrier in embankment dams in Sweden, has a complex grain-size distribution ranging from clay to boulder. In order to fulfill its function as impermeable barrier, the material has to meet several specifications: The content of fines is less than 40 % of the material passing the 20 mm sieve. Out of these maximum 40 %, the fine-grain content < 0,06 mm is 15 to 40 %.The hydraulic conductivity should be 3 x 10-7 m/s to 3 x 10-9 m/s (RIDAS, 2004). With these regulations, the material is supposed to be easily handled, with adequate bearing capacity and limited separation of grains during placement. However, with this composition the material is frostsusceptible: it is sufficiently fine-grained and permeable enough to support water transport to the freezing front for ice lens formation. Further, it is dense enough for not draining away water very easily during thaw and can therefore exhibit thaw weakening phenomena. 2.2 Frost heave and thaw weakening Frost heave and ice lens formation are well-described processes, but is not very much studied in relation to dams and dam design. The accumulation of ice in an embankment can be responsible for regions with excess pore pressure and weak zones with increased seepage. This may start a process of internal erosion and therefore frost action has to be considered in dam engineering. This paper highlights the risk of internal erosion. Ice lenses are formed perpendicular to the heat flow and therefore normally parallel to the ground surface when the frost line advances into the soil and heat is extracted. A soil sample with an already frozen upper part, an active growing ice lens, partly frozen pores and underlying unfrozen soil is shown in Figure 1. Water in the pores of the soil changes its state from liquid to solid due to the subfreezing temperatures. However, not all water in the pores is frozen. Dependent on the capillarity effect and surface absorption, some water remains unfrozen. The adsorbed water provides an unfrozen water film creates a gradient which enables water transport along the temperature gradient. Thus, water from lower parts in the soil profile can be sucked to the freezing plane. Below the active growing ice lens, an area with partly frozen and liquid pore water grows. This so-called frozen fringe impedes the water flow to the freezing plane. Water reaching the active growing ice lens at the frost front accumulates and freezes.

111 T ice Temperatures Frozen soil with ice lenses Active growing ice lens Area with partly frozen pores: frozen fringe Heave Original surface Unfrozen soil Pore water drawn to freezing plane Figure 1. Frost heave process (Andersland and Ladanyi, 1994) The rate at which ice lenses are created depends on the rate of heat extraction and water flow. The ice lens will grow as long as the conditions of freezing temperatures and water supply in the plane prevail. Once the ice pressure becomes larger than the overburden pressure, the soil skeleton will be lifted up. When the freezing temperatures advance deeper in the ground, a new ice lens will grow at a lower level. During ice lens formation, the pore pressures in the soil follow a cyclic pattern which has been described by Eigenbrod et al. (1996). Results of freezing tests showed that the pore-water pressure in the soil increases along with temperature increase. The repeating process is initiated with ice lens growth and expansion into unfrozen soil, hence increasing the pressure on unfrozen pore water. The increased pore water pressure results in an increased suction force, followed by pore pressure decrease. The effective stress becomes larger and compression of the soil occurs. Again, the pore pressure increases and the effective stress decreases, thus accelerating the ice lens formation. This process is summarized in Figure 2. T f T water (-) 0 (+) Effective stress decreases Pore water pressure increases Active growing ice lens Expands into unfrozen soil creates pressure on unfrozen pore water Effective stress increases Compression of soil Pore water pressure increases Suction force increases Pore water pressure decreases Figure 2. Cyclic pattern of pore-water pressures during ice lens formation Thawing in the ground proceeds normal to the ground surface, thus creating melting water in the pores. Here, the temperature is lower deeper down in the ground, so that lower parts in the soil remain frozen and the downward drainage path is blocked. Water has to move laterally or upwards. Due to

112 the water-ice accumulation during the frost heave process, excess pore pressures and reduced shear strength and bearing capacity results. 2.3 Structural changes Freezing and thawing causes water migration, ice lens formation, excess pore pressures and displacements in the soils fabric. Heaving forces due to ice pressure and dissipating melt water from thawing are responsible for movement of soil particles. When freezing and thawing occur repeatedly, the rearrangement of the grains leads to structural changes. Structural changes have been summarized by Viklander (1997). These are, among others, changes in the pore water distribution, volume, strength, compressibility, and Atterberg limits. Significant changes for embankment dam engineering are the formation of cracks, particle movements and increasing hydraulic conductivity. Structural effects are dependent on the grain-size distribution, specific surface of particles, and the amount of adsorbed water which govern the amount of unfrozen water and thereby suction force and water migration during freezing. Due to the cohesion of a fine-grained soil, the cracks which developed during freezing are not closed. The lower hydraulic conductivity of fine-grained soils does not have a sufficient drainage capacity during thawing. Water remains in the cracks, thus leaving paths open for water flow. Coarse grained soils exhibit self-healing and drainage capacity as particles can move and fill the cracks when melting water is drained. Both the initial void ratio and degree of compaction are essential for the effects on hydraulic conductivity (Viklander, 1997). An originally loose, poorly compacted soil may become denser during cyclic freezing and thawing. It will increase its volume during freezing, but become better compacted when water dissipates after thawing. The thawing consolidation reduces the void ratio and the hydraulic conductivity. On the other hand, originally dense and compacted material may be subject to increasing hydraulic conductivity and void ratio after freeze/thaw. The repeated freezing and thawing loosens the structure of the soil. Studies on fine grained till carried out by Viklander showed that the hydraulic conductivity increased at least 10 times when exposed to freeze-thaw cycles. Loose soil Unfrozen Frozen Thawed Dense soil Increased void ratio and hydraulic conductivity Unfrozen Frozen Thawed Figure 3. Effects of freezing and thawing on fine-grained till (Viklander, 1997) 3 Internal erosion Failure statistics carried out by Foster et al. (2000) showed that internal erosion through the embankment is the second largest failure mode after overtopping for large embankment dams. It was found that a large portion of erosion accidents occurs in earth-rock fill dams with a central core of broadly graded material with glacial origin, and that glacial core materials have a larger tendency to initiate piping than materials of other origin. Terms such as piping and hydraulic fracturing are commonly used in connection with internal erosion. Piping generally refers to a process of backward or concentrated erosion such as shown in Figure 4, where a pipe develops in the core from an exit seepage point. Two different modes of erosion have been identified: backward erosion or piping, starting at the downstream part of the core and regressing to the upstream side (Figure 4a), and concentrated leakage, where piping develops directly from the water source inside the core to an exit point (Figure 4b).

113 a) b) Figure 4. Principles of erosion (Fell, 2005) Hydraulic fracturing is commonly used to explain the cracking of the core due to settlements from reservoir filling, creating a condition of reduced effective stress which allows water to enter and open a path for leakage. Hydraulic fracturing has been questioned in terms of the condition of zero effective stress, allowing water to penetrate and open a pipe. Whether the stress condition in the embankment can support hydraulic fracturing may be questioned, since it appears unlikely that the total stress condition is equal to the hydraulic pressure. However, zones with cracks or increased porosity may not only be caused by differential settlements or other movements. These loose parts of the structure can very well be caused by the impact of freezing and thawing which will be showed here. Piping and internal erosion occurs when the following conditions are met: A water source, seepage flow, and erodible material in the seepage path, an exit where the eroded material can escape, and the presence of a material which can support a so-called roof for the pipe (Fell, 2005). Cracks or wet seams in earthfill cores have been reported to be responsible for concentrated leaks. Such phenomena may be observed during the first filling of a reservoir, in connection with inadequate filter layers, or settlements. Sherard (1985) reports experiences from homogeneous dams where wet areas with completely saturated soil have been observed. 4 Freezing effects in embankment dams Few cases of dam constructions experiencing frost action have been reported so far. The Waterloo Lake Dam in Saskatchewan, Canada, was completed in 1961 as zoned earth-rockfill structure. Deterioration due to frost action was identified from the first winter. Longitudinal cracks in the crest and the downstream face have been subscribed to ice lens formation in the silt core. Test bore holes showed that the frost penetrated 2,7-3,0 m into the hydraulic barrier, and that water was drawn from the reservoir to the freezing plane. Thawing and consolidation of the permafrost foundation was made responsible for the cracks in the upstream face (Solymar, 1983). Paré et al. (1982) report on observations on frost penetration into glacial till. Even though damage due to frost was not regarded being common for embankment dams, longitudinal fissures and ice lens formation has been recorded at the Whitehorse dam in the Yukon, Canada. A study on frost depth penetration in glacial till was carried out, showing that maximum frost depth of 12 ft., i.e. 4 m was reached in the dyke crest of Lac Jacques. An overview of some uncertainties in embankment dam engineering in terms of cold climate has been given by Milligan (2003). These uncertainties refer to cold climate conditions both during and after construction. 5 Freezing effects in the Eastern Suorva hydropower dam The Suorva dams are three embankment dams, the Eastern Suorva dam, Sågviks dam, and the Western Suorva dam, regulating Sweden s second largest reservoir and most important hydropower source. All

114 three dams are rockfill dams with a central core of glacial till. A sketch of the construction is shown in Figure 5. Figure 5. Basic cross-section of the Eastern Suorva dam (Jantzer, 2006) Suorva is located north of the artic circle in a mountain region which is known for cold winters and severe wind conditions. The mountain valley is known for wind channeling effects and measurements showed that Sourva is Sweden s windiest inland location. The Swedish Meteorological and Hydrological Institute SMHI measured a mean annual temperature of 0,4 C in this area over a 30- year period from 1961 to The Suorva dams were constructed during 1966 to 1972 and have since then been exposed to annual freezing and thawing cycles. 5.1 Field investigations Tests pit excatations was carried out at the Eastern Suorva dam during the summer The original dam crest was removed and the condition of the material in the core examined. Four test pits in different sections of the dam were excavated. The pits had a side length of 2 4 m and were excavated by the help of an excavator. Disturbed samples were collected in each test pit at five levels: 0,0 m, 0,5 m, 1,0 m, 1,5 m, and 2,0 m. Visual inspections as well as collection of samples from pits were expected to provide information of whether the core had been exposed to freeze/thaw cycles or not. If freezing had occurred, the inspections were estimated to reveal in which way such frost action had influenced the fill material. In case of freezing, the material was expected to be layered and to show visible water accumulations; differences in the grain-size distribution, water content or varying dry density were hypothesized to be detected. For documentation of the condition of the moraine core, pictures were taken. In addition, the temperatures were measured when the soil felt unusually cold during sampling. A few selected pictures will be presented here to give a general overview of the observed aspects. Results of field investigations During field investigations, the overall impression was that the soil was comparatively cold in comparison with the air temperatures, and that the soil had a layered structure with a considerable higher water content in upper regions at a depth around 0,30 0,80 m. Three of the four test pits showed such similar features. In case of one test pit, no water accumulation or low temperature was measured. The material was homogeneous and did not give information about freezing processes. The samples taken from the Eastern Suorva dam section 750 are significant for observations of temperature and water content and will therefore be discussed in detail.

115 Field investigations in section 750 were carried out on July 5 th, 2004, on a sunny day with temperatures around 25 C. The material on top of the core was removed right before excavation of the test pit, which included a layer of boulders and an insulation layer. Remains from measures against leakage in the embankment dam were found, such as standpipes which had been used for bentonite grouting, see Figure 6. (Jantzer, 2005) standpipe boulder Figure 6. Surface of test pit in section 750. Note leftover standpipes from grouting as well as stones and boulder. (Jantzer, 2005) While the top layer of fill material was dry and well-graded, it appeared to have an increased moisture content already at an excavation level of 0,3 m, see Figure 7. The soil seemed to be completely saturated at this point and water was expelled when pressure was imposed by setting a foot on it. The consistency varied from very soft to an almost liquid state. The water content was found to be extremely high and the temperatures of the soil comparatively low, even though the weather conditions warm and dry, see Figure 7. The water content ranged from a minimum of 6,6 % at level 2,0 m to a maximum of 13,4 % at 0,5 m. (Jantzer, 2005)

116 Temp [ C] 3,3 0,5 W [%] 11,4 0,3 0,8 m 1,8 13,9 1,0 2,0 8,4 1,5 1,6 10,4 Crack with high water content, Temperature measurement see Figures 5,6,7 1,3 2,0 6,6 Depth [m] Figure 7. Overall view of the trench and summary of values for water content and temperatures measured of the excavated test pit. (Jantzer, 2005) At a depth of 1,0 m the moisture content changed considerably to a much lower value: 8,4 %. A horizontal fracture was found at 0,5 m depth and when excavation reached this level a long gap was left open, see Figure 8. The surrounding area was wet and was found to have temperatures of about 1,7 C to 1,9 C in comparison to 3,3 C in the upper layer and 2,0 C in the lower layer. It was obvious that the moraine core had been frozen at this level, see Figures 9, and 10. (Jantzer, 2005) Cracks Figure 8. Fractured moraine core at level 0,4-0,5 m. (Jantzer, 2005)

117 wet zone Figure 9. Wet area in the fracture after removing loose material. (Jantzer, 2005) Figure 10. Temperature measurement showing 1,7 C in the area of moisture accumulation. (Jantzer, 2005) The grain-size distributions of samples from the test pit in section 750 are shown in Figure 11. It was noted that the particle-size distribution from level 2,0 m was coarser compared with the other four samples, and that samples taken between 0,0 m and 1,5 m were more fine-grained. The difference was associated with the comparatively high water content at upper levels, thus implicating that fine-grained material is more frost susceptible, and that water is collected preferably in such locations. (Jantzer, 2005) Clay & silt Sand Gravel ,01 0,06 0, ,00m 0,50m 1,00m 1,50m 2,00m W atercontent: 0,00 m: 11,4 % 0,50 m: 13,9 % 1,00 m: 8,4 % 1,50 m: 10,4 % 2,00 m: 6,6 % Figure 11. Grain-size distribution and water content section 750. (Jantzer, 2005)

118 Discussion In order to be able to draw conclusions about impact of frost action, pictures, analysis of water content and temperature measurement were much more significant than the comparison of grain-size distributions. As the water content was exceptionally high in some places, and the temperatures at the same time low compared to the air temperatures, it was likely that the soil had been frozen. A summary of these values is given in Table 2. Table 2. Comparison water content values and temperature Test pit section Level w [%] w [%] Temp [ C] w [%] Temp [ C] w [%] 0,0 m 6,9 7,6 4,4-3,8 11,4 3,3 6,5 0,5 m 6,5 7,7-13,9 1,9-1,7 7,1 1,0 m 6,8 6,9 2,3-2,1 8,4 2,0 6,8 1,5 m 7,3 8,1 1,9 10,4 1,6 7,6 2,0 m 7,7 7,2 1,6 6,6 1,3 6,6 Mean value Ø 7,04 7,50 10,14 6,92 It was assumed that ice lens formation had been taking place in section 750, where the soil was found to be saturated and temperatures between 1,3 to 3,3 C were measured. Other observations such as cracks or a layered structure were made in almost all test pits, mostly at depths between approximately 0,3 m to 0,8 m. Nevertheless, evidence for effects of frost action, such as changes in structure, density, volume, permeability or particle movements is not given. 5.2 Temperature calculations Temperature calculations were performed by a FEM program in order to illustrate the penetration of the frost line during winter. One simplified sketch of the model is shown in Figure 12. The vertical and horizontal scales differ. The result of this calculation is displayed in Figure 12. It can be noticed that the freezing plane had advanced into the moraine core approximately two meters and is marked by 0 C isotherm. Again, the water saturated part of the dam body is comparatively warmer than the rest of the dam body because of seepage. Figure 12. Calculated temperatures in the beginning of May, 2004.

119 6 Discussion The field investigations showed that the core of the dam actually is exposed to freezing; a frost line at a depth of about 0,3 0,8 m was found. Layered structures, water content and temperatures of the moraine core give evidence for frost actions. The soil displayed different formations: it appeared partly to be homogeneous, well-graded and dry, while other tests showed that the water content was comparatively high and the soils temperature pointed at being close to the freezing point. A comparison of water content and temperatures gives evidence for freezing of the core. With regard to wind conditions, which have not been included in the analysis, it is assumed that the freezing plane advances even deeper down in the central core. The water table, which tends to be lower during winter time, reduces the possibility to warm up the inner part of the dam because of seepage, as water passing through the dam during winter has a comparatively warmer temperature than the air. Because of wind snow usually not cover the embankment dam. Snow can act as insulation for ground in general. This factor will also increase the frost front advance. Thermal modeling with the finite element program gives additional evidence that the hydraulic barrier of the dam is exposed to freezing and thawing. Using an annual mean temperature of -0,4 C, the results show that the frost front advances approximately 2 m into the core. The calculations were carried out several times with varying values for the mean annual temperature, water table and construction of the dam to be able to asses the variation of frost line depth. Freezing does even occur in filter zones but these calculations show that frost action has to be considered. 7 Conclusions During field investigations it was found that the soil had a layered structure, a comparatively high water content, and low temperatures. The features were typical for soil structures affected by frost. The frost depth was found to range between 0,3 m to 0,8 m from the top of the moraine core. Calculations carried out with a commercial finite element program confirmed the observations. The calculations, which were based on the actual temperatures measured in the area, showed that the frost front could advance to a depth of 2 m down to 5 m into the core. This supported the idea that the dam had been exposed to freezing and thawing. However, it is not possible to draw conclusions about in which way the freezing actions affect the core and the filter zones. Detailed information about porosity, void ratio and permeability is needed in order to be able to compare the condition of the core today with the condition right after construction. A thorough analysis of these properties would be necessary. However, it is highly probable that freezing and thawing has caused structural changes of the core material, thus affecting critical parameters like hydraulic conductivity. This is believed to be able to be the start of an internal erosion process. Acknowledgements Acknowledgements is given to Vattenfall AB, Sweden for providing me with information about the dam and giving me the chance to perform the field tests. SVC (Swedish Hydropower Center) is also acknowledged for financial support like Luleå University och Technology. References O.B. Andersland and B. Ladanyi. An Introduction to Frozen Ground Engineering. Chapman & Hall, New York, USA K.D. Eigenbrod, S. Knutsson, and D. Sheng. Pore-water pressures in freezing and thawing finegrained soils. Journal of Cold regions Engineering, Volume 10, number 2, pages 77-92, 1996.

120 M. Foster, R.Fell and M. Spannagle. The statistics of embankment dam failures and accidents. Canadian Geotechnical Journal, Volume 37, pages , Jantzer, I.. A documentation of the Eastern Suorva dam core. Field investigations and thermal modeling regarding frost action. Master s thesis 2005:076, Luleå University of Technology. ISSN , 2005 V. Milligan. Some Uncertainties in Embankment Dam Engineering. Journal of Geotechnical and Geoenvironmental Engineering, Volume 129, number 9, pages , D.K.J. Noonan, L.H. Smith, and V. Milligan. The use of insulation to provide frost protection for the Waterloo Dam in northern Saskatchewan. Proc., 35 th Canadian Geotechnical Conf., Volume 2, J.J.Paré, J.G. Lavallée and P. Rosenberg. Frost penetration studies in glacial till on the James Bay hydroelectric complex.. Canadian Geotechnical Journal, Volume 15, pages , P. Viklander. Compaction and Thaw Deformation of Frozen Soil. Permeability and Structural Effects due to Freezing and Thawing. Doctoral thesis. Luleå University of Technology. 1997:22. RIDAS. Kraftföretagens Riktlinjer för Damm Säkerhet (Swedish HydropowerIindrustry Dam Safety Guidelines) J.L. Sherard. Hydraulic Fracturing in Embankment Dams. In: Seepage and Leakage from Dams and Impoundments. R.L. Volpe and W.E. Kelly (eds). ASCE, New York Z.V. Solymar and J.O.H. Nunn. Frost sensitivity of core materials. Case histories. Canadian Geotechnical Journal, Volume 20, pages , 1983.

121 Conference paper #3 Jantzer, I. and Knutsson, S. (2007) Seepage and Critical Hydraulic Gradients in Tailings Dams and Natural Formations. Proceedings of 2 nd International Conference on Porous Media and its Applications in Science and Engineering ICPM2. Kauai, Hawaii, USA. June 19, 2007.

122

123 Proceedings of the 2nd International Conference on Porous Media and its Applications in Science and Engineering ICPM2 June 17-22, 2007, Kauai, Hawaii, USA SEEPAGE AND CRITICAL HYDRAULIC GRADIENTS IN TAILINGS DAMS AND NATURAL FORMATIONS Isabel Jantzer and Sven Knutsson Department of Civil Engineering, Luleå University of Technology, Luleå, Sweden ABSTRACT Embankment dams, both water Retention Dams (WRD) and tailings dams, are constructed of granular material. Natural seepage flow through porous media, i.e. soil, rock, and tailings, occurs in all types of dams. During this process, particles in the porous media are exposed to hydraulic gradients. Under certain circumstances seepage can initiate internal erosion which can seriously damage the construction, eventually resulting in serious accidents or failure. The process starts when a critical hydraulic gradient is reached, which may vary with the way of construction, materials used and its properties and compaction. This is of special importance for dams constructed of mine waste, i.e. tailings dams, where the stability and function has to be guaranteed over very long time periods (>1000 years). Here, seepage, internal erosion and the corresponding critical gradient are fundamental parameters. It is not clear under which circumstances the process of internal erosion begins, i.e. when one particle in the system starts to move, thus creating a further process of particle transportation in the system. A common way to control seepage, filter layers are included in the construction. However, erosion, piping or sinkholes are still observed. Different models on the hydraulic conductivity and hydraulic gradient are presented and discussed. Properties of tailings in comparison with natural geological materials are identified. Observations from several natural geological formations in Sweden, which have fulfilled the function of a dam since the last glaciation, are presented and the critical hydraulic gradients with respect to vary long time periods are compared. INTRODUCTION Dams are designed and constructed to retain water for different purposes, such as the production of hydropower, irrigation, and flood control, or in special cases for the storage of mining waste. Natural seepage flow through porous media, i.e. soil, rock, and tailings, occurs in all kinds of dams, either through the dam body itself, through the foundation of the dam or close to the abutments. Seepage through dams is generally not considered being a problem, as long as the amount of seeping water through the structure can be controlled and no particle migration is involved. However, if particle transport occurs and seepage is exceeding, internal erosion may result, with serious consequences of damage or breach of the dam. The phenomenon of internal erosion is commonly discussed in reference to two groups of dams: embankment dams, i.e. water retention dams (WRD) and tailings dams, where water flow through the dam body occurs. In Scandinavia, WRD are often made of natural materials such as soil and rock and known as earth- and rockfill dams. They can be homogeneous or zoned; the latter referring to a structure with some impervious core, filter, and rockfill, shown in Figure 1. Figure 1: Example of zoned embankment dam(fell et al. 2005): (1) Core, (2) Filter zones, (3) rock fill, (4) rip rap Tailings dams are constructed with by-products from mining activities. Their purpose is often to store both mining waste and process water. Tailings are a result from the process of crushing rock and extracting ore. This refining process leaves material in a typical grain size corresponding to silt and sand, and the particles may be toxic if released to the environment. A main difference between WRD and tailings dams besides the maximum hydraulic gradient is their life

124 time, and with that surveillance and maintenance. WRD can theoretically be removed when they are not longer needed, i.e. their design life is finite. In addition, they are subject to regular monitoring and surveillance. On the contrary, tailings dams need to be stable over long time spans, even after the closure of mining activities. Tailings storage facilities are designed to fulfil their purpose under an infinite period; they have to be free of maintenance to guarantee the safe storage over a time period of thousands of years, i.e. to the next glaciations. In order to control seepage, homogeneous embankment dams as well as tailings dams are normally exposed to low hydraulic gradients. Zoned earth and rock dams are designed and built to withstand higher hydraulic gradients. Each zone has a particular purpose giving a structure with a material combination of acceptable permeability and stability. Filter zones between the core and rockfill comprise intermediate grain size distributions. Filters are designed to fulfil two basic functions: the proper drainage for the reduction of pore pressure, and the prevention of particle transport. It is not clear under which circumstances the process of internal erosion begins and when a particle starts to move. The study of internal erosion has included filter criteria, filtering and internal stability characteristics, critical hydraulic gradients, and even statistical approaches. However, even though filters have been provided and materials have been tested, erosion and sinkholes are still observed. Because of the increasing demand of metals and minerals the mining industry is prospering. It has become economic to mine low-grade deposits, which in turn leads to the production of large quantities of fine grained waste material that has to be stored. Former small mines producing a relatively small amount of tailings have to meet challenging requirements on tailings treatment. Their tailings storage facilities have been constructed in stages and developed in stages, sometimes not according to recent knowledge and design standards. Nevertheless, the stability has to be guaranteed. Considerations of the critical hydraulic gradient therefore become vital for such structures. In Sweden, it has been observed that several natural geological formations in Sweden have fulfilled the function of a dam since the last glaciations, i.e. approximately years, although they do not comprise any filtering or stabilizing layer. These evidently successful structures lead to the assumption that we can learn from nature. An analysis of material properties and critical hydraulic gradient is carried out and compared with conventional information and models. The study is limited to non-cohesive materials to meet the conditions of coarse grained structures in WRD and tailings dams. NOMENCLATURE WRD = Water retention dam v = Flow velocity k = Hydraulic conductivity i = Hydraulic gradient D α = Grain size of filter material with α% mass of smaller particles d α = Grain size of base material α% mass of smaller particles D e = Characteristic grain size C u = Coefficient of uniformity k h = hydraulic conductivity in horizontal direction k v = hydraulic conductivity in vertical direction τ f = shear strength in failure plane φ = friction angle of material β = slope angle γ = buoyant unit weight 1 Seepage and internal erosion Natural seepage occurs in all kinds of dams, either through the foundation of the dam, the abutments or through the dam body itself. This is basically not a problem, as long as the hydraulic gradient is sufficiently low, so all particles stay in place and the construction remains stable. When hydraulic forces exceed and particles start to move, a zone develops where further migration of grains is possible, which may result in internal erosion or piping. The problem with internal erosion is that it is a complex, non-visible process, which makes it difficult to describe and understand. Most often internal erosion is observed when the process already has progressed so far that it is detected by measurements, or visible on the sides of the dam. At this point, it is too late to determine the initial action. Some generally accepted basic principles have, however, been explained by Fell et al. (2005). Four conditions have been recognized being essential for internal erosion or piping to occur: (1) the existence of a water flow, (2) erodible material in this path which is carried away by the seepage flow, (3) an open exit for the eroded material, (4) material which is capable to create and support a pipe. Two different modes of erosion in a dam have been distinguished: backward erosion and concentrated piping. The difference between these two is the initiation of the process: backward erosion (Figure 2a) refers to a leakage on the downstream part of the dam progressing backwards to the upstream side. Concentrated piping (Figure 2b) develops in the dam by cracking or leakage directly from the water source to an exit point, where the pipe grows and the erosion increases.

125 a) b) Figure 2: Principles of erosion (Fell et al. 2005) Other terms used in this context are suffosion and hydraulic fracturing. Suffosion is the washing out of fine particles in a soil and implies internal instability, i.e. filtering incapability (Fell et al. 2005). Hydraulic fracturing has been discussed by Sherard, who states that differential settlements in a dam can cause stress redistribution and zones with zero or tensile stress, where water can penetrate and result in leakage (Sherard 1986). Whether or not the stress condition in the dam body may enhance the development of hydraulic fracturing could be questioned, as the total stress condition does not seem likely to be equal to the hydraulic pressure, thus reducing the effective stress to zero regarding non-cohesive soils. However, differential settlements and movement of material from different zones relatively to each other may cause zones with increased porosity. Circumstances during construction, such as poor compaction and separation of the material during placement may contribute to the development of such weak zones. Extensive statistic studies recognized internal erosion as being one of the major incidents and failure causes, both regarding WRD and tailings dams. Foster et al. published a study on the statistics of dam failures and accidents where it was stated that almost half of all known failures were related to piping (Foster et al. 2000). Bjelkevik analyzed Swedish tailings dams and found that more than 30% of incidents reported from Swedish tailings dams were related to internal erosion. Failures and incidents were mainly associated with earth- and rock fill material (Bjelkevik 2005b). 2 Theory The interaction between soil and seeping water is a compound process involving the soils characteristics and stability when exposed to a hydraulic gradient. Successful filtration depends on the composition of the coarse grained materials, the hydraulic gradient and flow velocity. When no grains are washed out the material is regarded being stable. Instability and particle transport, i.e. internal erosion, implicates increased hydraulic conductivity and deformation. The multiple natures of such an event and the lacking possibility to observation make explanations difficult. 2.1 Hydraulic conductivity The hydraulic conductivity k is a basic property dependent on the structure of the soil, i.e. particle size distribution, fabric and mineralogical composition, void ratio and degree of saturation. Besides that, the type of flow, the nature of the fluid and temperature has influence. In geotechnical engineering, the hydraulic conductivity is often simply referred to as coefficient of permeability. Using the term permeability may, however, be confusing, as it does not take the viscosity of the fluid into account. The hydraulic conductivity of a soil may exhibit a large variety of values as a function of void ratio. Darcy s law is the empirical relationship between the flow velocity v, the hydraulic gradient i and the hydraulic conductivity k of a soil: v = k i (1) To measure the hydraulic conductivity of a soil, laboratory tests are performed; the principle is shown in Figure 3. Special attention has to be given to the preparation of the soil sample, as the saturation and presence of air bubbles may become sources of error. Q V Figure 3: Standard laboratory test to measure the hydraulic conductivity of a soil Unfortunately, such laboratory testing depends on the reliability of the sample. It is difficult to obtain a representative, i.e. more or less undisturbed sample under laboratory conditions. Testing clay samples almost always involves some kind of erosion processes in the testing device, where the walls of the test mould h l

126 serve as part of a pipe. Such dense material often needs to be tested under large hydraulic gradients to obtain results. The value of the hydraulic conductivity of a soil tested in laboratory conditions is, besides that, a value for the vertical conductivity k v. However, the horizontal conductivity k h and the relation between horizontal and vertical conductivity is an important factor for embankment dam engineering and may vary significantly from the test result. In-situ testing of undisturbed soil may therefore regarded being more reliable. Such in-situ tests are well-pumping or borehole tests. These methods may give results taking the horizontal hydraulic conductivity into account. The water flow measured comprises both a vertical and horizontal component. Yet, also field testing implies factors of disturbance and unknown sources of error, so that additional laboratory tests are usually needed. To calculate the hydraulic conductivity of a soil, different equations have been introduced by Kozeny and Carman, Hazen, and recently by Chapuis. In general, laminar flow is assumed. The relationship proposed by Kozeny and improved by Carman takes the void ratio, the specific surface area and specific weight of the soil into account. In addition, the viscosity and a factor for pore shape and ratio for actual flow length path are considered. Hazen, on the other hand, suggested a simple formula from his works with sand, where the hydraulic conductivity was directly related to the effective grain size D 10. The deficiency in both these relationships is that only particle size and void ration are regarded. Hazen s formula only refers to the particle size. The composition, soil fabric and saturation are not included (Lambe and Whitman 1969). Chapuis presented a method for estimation of the hydraulic conductivity as recent as in 2004, regarding both the effective diameter and void ratio of the soil. The method may be applied to clean sand and gravel and is an extension of Hazen s formula, which did not take the actual porosity into account (Chapuis 2004). 2.2 Filters and filter criteria The concept of filter criteria for a base soil filter system is related to the idea of internal stability or selffiltering of a material. Filtering or stability implies that the structure of the material prevents the transport and loss of particles through voids. A filter that has the task of protecting a base soil is commonly designed in respect to the gradation of the base soil, which usually has a larger content of fines than the filter. Thus, the grain size distribution of the filter has to be adjusted so that the voids are sufficiently small. The same applies basically in case of only one material; the soil would be considered self filtering or internally stable if its gradation is such that no particles are lost. Traditionally, the ratio between grain sizes of base soil and filter has been applied to prevent erosion. The geometrical criteria present relatively easy applicable design and construction rules, since they are independent from the hydraulic condition. Terzagi introduced a basic concept for designing filters in the early 1920 s by establishing two rules (Terzaghi, 1996): To control erosion the relation between filter (D) and base (d) material D 15 /d 85 4 should be met, to guarantee permeability and drainage D 15 /d The index indicates the grain size for which 15% or 85% is finer, respectively. These two basic rules have been modified, refined, and adapted to different countries materials and practices by a number of researchers. Extensive test programs have been carried out by Sherard et al. (1984), Lafleur (1984), Lafleur et al. (1989), and Kenney and Lau (1985), with the aim to account for filtration and filter criteria. Despite all studies and testing, the concept of ensuring drainage and retaining particles by using filter criteria is still in use. Sherard et al. (1984) considered that they employ the appropriate characteristics of the filter and the base, so that the ratio should be continued as the main criterion for judging filter acceptability. Åberg (1993) stated that the criteria were logical because of the defining grain size D 15 for pore size and permeability. Moreover, Terzaghis rules were approved by Fannin and Moffat through laboratory testing in Both the hydraulic conductivity and the internal stability of a soil are related to the size of the grains. The inherent stability against particle loss has been ascribed to the shape of the gradation curve, i.e. the coefficient of uniformity, which will be explained in the next section. The hydraulic conductivity of a soil is influenced by the size of the pore channels in the soil matrix. Kenney et al. noted that the size of the pores primarily depends on a representative grain size, whereas the shape of the complete gradation curve was noted to have a minor effect on the pore size channels in the soil structure (Kenney 1984). Researchers have, on the other hand, not come to an agreement on the actual grain size that governs this property. Hazen found in 1892 that the relationship between pore size and hydraulic conductivity was proportional, and that an increase in hydraulic conductivity is related to the square of a characteristic grain size D e (Terzaghi 1996). The characteristic grain size that governed the hydraulic conductivity was set to D 10, i.e. a 10% limit of the particles of the soil. Kenney et al. (1984) stated that the permeability was primarily dependent on the finer particles with the grain size D 5. Sherard et al. (1984) found this grain size to be D 15, whereas Fell et al. (205) agree with Hazen on D 10 being the defining factor. This discussion shows the variety of views; a consensus seems to be difficult to reach. Nevertheless, it is generally agreed on that it is the fine particles and roughly 5 to 15 % of the soil which affect the hydraulic conductivity considerably. The reason for transport and washout of particles is the inability of the porous material to filtrate and provide internal stability. The grading stability of a soil is primarily based on its grain size distribution. Again, the size of the pores and loose grains inside the pores play a major role. The soils structure was described by Kenney and Lau (1985) being consistent of two units, a primary fabric which transfers load and stress, and loose particles in the voids of the primary fabric. The loose particles have the ability to move through the skeleton of

127 coarser particles. Skempton and Brogan (1994) found that internally unstable sand experienced fine grain migration. This was explained by Kenneys theory of a load carrying skeleton where the smaller particles do not transfer effective stress, see Figure 4. Figure 4: Example of stable and unstable soils (a) internally stable material, (b) unstable gap-graded material, (c) unstable material with loose large grains. While the hydraulic conductivity has been recognized being strongly dependent on the amount and size of soil particles between D 5 and D 15, the filtering ability of a soil has been related to the shape of the total grain size distribution. For instance, filter criteria concerning the ratio between base and filter have sometimes been extended with recommendations such as that in the USBR method: The curves of the base soil and the filter should be parallel (Fell et al. 2005). On the contrary, Sherard (1984) found that such a similar shape of gradation curves is not necessary. However, broadly graded or gap graded soils may display a lack of certain grain sizes that are needed for an optimum in structure, pore sizes and successful filtration. Suggestions on the shape of the grading curve by means of the coefficient of uniformity C u, i.e. the ratio D 60 /D 10, have not been clearly expressed. Yet, most filter tests have been carried out using uniformly graded material with a coefficient of uniformity lower than 6. Materials for which the ratio is less than 10 are regarded as being stable, while a ratio above 20 is considered being unstable (Skempton 1994). Practical problems can be connected to wide graded materials because of their susceptibility for segregation during placement. Milligan stated that the instability of material tested by Kenney and Lau was closely related to material which was typically susceptible to segregation during construction works (Milligan 2003). 3 Tailings dams Tailings are one result from mining activity. The process of mining ore and extracting metals includes the crushing and milling of the excavated rock into small particles. The valuable metal, the smaller portion of about 0,4-30%, is removed, and the by far larger portion of left over waste material, i.e. tailings, have to be stored. Tailings could be regarded as natural material as they are basically crushed rock. However, the fine waste material contains chemicals and metals, which may be harmful to the environment when released in combination with water and air. Tailings are commonly discharged as slurry in impoundments, together with the process water from the extraction of metals. The slurry is pumped directly from the plant to the impoundment, where sedimentation of the tailings takes place. An impoundment may be created by natural heights and/or tailings. The tailings settle, and clarified water is either re-circulated to processing, released into a river or stream, or is left for evaporation in arid climate. When the mining activities come to an end and the facility is closed down, mining structures and office buildings are subject to decommission. The tailings storage facility, i.e. the tailings dam and the impoundment, however, will be left. These facilities have to be taken care of in order to provide security against containment. The safety and stability has to be guaranteed over a long time period, which has been defined covering > 1000 years (Bjelkevik 2005b). Two methods can be applied for the remediation of tailings dams: the dry method or the wet method. The dry method implies the complete drainage of the material and coverage against weathering and oxidation. Enclosure of tailings by an artificial lake is referred to as wet method, and this technique is subject to this report. The disposal of tailings in a dam construction and coverage with water is the most cost effective and practical way of handling mine waste. Using tailings in the embankment construction reduces the amount of natural material, which makes this solution relatively inexpensive. Yet, it is difficult to ensure the safety and stability of the pond and the surrounding embankment. The water balance has to be maintained even during extreme events, such as drought or flood. Focus for the long term stability lies on the embankment dams, with the hydraulic gradient and pore pressure as critical properties in question. 3.1 Construction of tailings dams Various fundamental methods for the construction of tailings dams are described by Bjelkevik (2005a): the upstream method, the downstream method, the centreline method, and others. The tailings disposal is in that way used as a part of the embankment construction. Especially for long term stability a low hydraulic gradient is essential for the structure. The location of the seepage surface differs in particular when considering the upstream and the downstream method, as the structure in the downstream method is supplied with filter layers that control the hydraulic gradient. Therefore, these two methods are explained more detailed. Applying the upstream method, a starter dike is constructed and the tailings slurry is then discharged to the downstream side from the crest of this dike, forming a tailings beach, see Figure 5 (a). As the tailings reach the height of the dike, the construction is raised by a next perimeter dike (b), located somewhat more on the downstream side of the first dike. The raise of the embankment will then follow these sequences (c).

128 a) b) c) Tailings beach Tailings discharge Perimeter dike Figure 5: Principle of upstream method Starter dike The downstream method also is set up with a starter dike and raised along with the tailings level. However, in this case the dam is increased on the downstream side, which makes it possible to include filter zones in the construction, see Figure 6. Both ways of construction include the utilization of tailings as a main part of the structure. Ponded water Starter dike Impervious element Internal drain Figure 6: Principle of downstream method Tailings leave the production plant together with process water. The slurry is pumped through pipelines to the deposition place. The actual deposition at site is completed by single point discharge, i.e. from an open end of a pipeline, by spigotting, or cycloning. Spigotting refers to a technique where smaller pipes derive from a main pipeline. Several discharge points are supposed to provide better control over the deposition to create a more uniform beach. Due to the variation in grain sizes, the coarser particles will settle closer to the discharge point than the smaller particles. Reduction of pore pressures and lowering the hydraulic gradient makes it necessary to control the fines settlement. To obtain a uniform structure where the fines are considerably far from the actual dam construction, the points of discharge have to be moved. Cycloning is an alternative to separate the coarse from the fine fraction, so that the coarse fraction can be utilized for the construction of the dam. The tailings are sorted by centrifugal action, the sand fraction is used for embankment construction, and the fines are discharged in the impoundment. A significant feature of tailings dams is that they are raised along with the mines production, so that the construction is an ongoing process. This implies that tailings dams face problems such as changing design and construction, as well as the change of staff on site and development in knowledge. The design of both the impoundment and the tailings disposal may be subject to more or less unpredictable adjustments during the mines production period. The difficulty is to create an adaptable, safe and stable construction under any circumstance, which is of utmost importance. Environmental awareness and concern about dam safety has increased, leading to the mining industry facing problems arising from their past handling and deposition of tailings. Design and construction of the first tailings dams were based on knowledge from water retention dams, as well as trial and error. Knowledge and experience to construct and maintain safe tailings dams is available today. Yet, the problem is that original structures are still a part of today s tailings dams, which can not be removed. A major question for stability and safety is the pore pressure and the hydraulic gradient; both closely related to the deposition of tailings. The following sections will therefore explain these main principles. 3.2 Material properties of tailings Tailings are particles of crushed rock at a size of 0,01mm to 0,1 mm. This corresponds to a natural material classified as fine sand or silty sand. Tailings can be described in soil mechanics terms; the geotechnical properties can partly be compared to natural materials. Yet, tailings characteristics can diverge in their characteristics due to the variations in origin and processing of the ore, and deposition methods. The origin affects the size and the gradation of the grains, the internal friction angle and the density; whereas the deposition method is responsible for bulk density, void ratio and porosity, and the hydraulic conductivity. Tailings generally have high water content and porosity, a low to moderate hydraulic conductivity, and a low plasticity when compared to soil. The shear strength was rated low to moderate by ICOLD, but found to be moderate to high in relation to the grain size compared to natural material (Bjelkevik 2005a). The grinding creates particles that are much more angular than natural grains. The angularity of the tailings involves some difficulties for description of material parameters. While natural material has been tested and described extensively (see 2.2 Filters and filter criteria), tailings properties rise new questions. Porosity, shear strength, and hydraulic conductivity are usually well described properties for natural round and well graded material. In contrast, representative laboratory testing on tailings is difficult. Figure 7 shows the principle structure of both tailings and natural material. While geological material is often more round and has a smoother surface, tailings are angular, which may be an explanation for varying results in shear tests, where stress concentration at edges of tailings can lead to further crushing of the particles and higher resistance to shearing. The friction angle of tailings is relatively high with values up to 45º. The structure of tailings may also raise the question whether the flow pattern is the same as in geological materials. The porosity maybe the same in

129 both materials; yet, the shape of the grains may affect the seepage. therefore be considered more stable in long-term thinking with regard to the lower seepage line (a). Hydraulic gradient a) b) a) Figure 7: Comparison between a) tailings and b) natural material Tailings generally have a relatively low uniformity coefficient; Bjelkevik and Knutsson (2005b) noted a C u between 3 and 8. The value is not constant, as it varies with the distance to the discharge point because of the sorting effect of sedimentation. The hydraulic deposition does affect the properties of tailings; dependent on the distance from the outlet the parameters change. In addition, flow, velocity, and solid content in the slurry affect the sedimentation. For instance, the hydraulic conductivity of tailings in place varies by a factor of more than 10 when comparing the horizontal conductivity k h and vertical conductivity k v. The horizontal layering leads to a much higher value for k h. A calculation of an actual value by empirical relationships is therefore not accurate. Both the bulk and grain density are also controlled by the distance of the outlet, both decreasing with increasing distance. This results in varying void ratio, porosity and dry density. The dry density also decreases at larger distance, whereas both the void ratio and porosity are relatively smaller closer to the discharge point. In the study conducted by Bjelkevik and Knutsson the observed values for void ratio and hydraulic conductivity do not correspond to each other; a higher void ratio would imply that the hydraulic conductivity also becomes larger. This may also be due to the nature of the hydraulic deposition. 3.3 Pore pressures and hydraulic gradient Understanding the principles of water flow through tailings embankments is basic for stability considerations. Important for the tailings dams stability is the pore pressure inside the construction. The pore water pressure affects the effective stress condition; increased pore water pressure reduces the effective stress, leading to decreased stability and resistance to sliding. The hydraulic gradient emerges from different water levels and induces seepage flow. This seepage flow may induce a stress condition on particles and force them to move, resulting in erosion processes. A basic rule is to increase the hydraulic conductivity in the direction of the flow to achieve an as low phreatic surface as possible near the embankment face. Comparing the hydraulic gradient in the upstream and downstream construction, it can be noted that the gradient is higher in the downstream structure (b), where a sealing element of moraine reduces the permeability, see Figure 8. The upstream method may b) Hydraulic gradient Figure 8: Comparison of the hydraulic gradient The hydraulic gradient in the tailings dam is determined by the water level in the pond, the relative proportion of the foundations hydraulic conductivity to that of the tailings, and variations of the tailings hydraulic conductivity due to grain size segregation, see Figure 9. Low water level in pond a) Effect of location of water level in pond Pervious foundation Impervious foundation b) Effect of the foundations permeability Low beach segregation High beach segregation c) Effect of grain size and segregation High water level in pond Figure 9: Factors influencing the location of the phreatic surface A lower water level in the pond (a) and a permeable foundation (b) will draw down the seepage line in the tailings dam construction. Segregation of grains will take place because of the difference in size; the smaller the grains, the longer they will be transported and sediment from the outlet. High beach segregation, i.e. separation of large and small particles, will lead to a major part of large particles with a higher permeability and a lower location of the phreatic surface (c). Especially the deposition and segregation of particles has been a source of problems; fine particles in the tailings dam construction raise the pore pressure and hydraulic gradient. The stability of the dam construction is directly related to the friction angle, density, and cohesion of the tailings (Bjelkevik 2005a).

130 3.4 Hydraulic conductivity of tailings To be able to understand seepage and pore pressure conditions in tailings dams we need to understand the flow conditions and, with that, the hydraulic conductivity of the material. As mentioned before, the grains themselves vary in shape, composition, and density regarding their origin. In addition, the composition of tailings, the method of discharge and segregation is a factor which makes it difficult to assess the hydraulic conductivity of the material and the seepage conditions in the structure. The hydraulic conductivity as explained in section 2.1 has so far been described for natural geological materials. Tailings, however, are angular, have a different chemical composition and vary in their structure and porosity because of the deposition method. The horizontal layering results in an increased horizontal hydraulic conductivity. Thus, flow through tailings may vary considerably compared to flow through natural materials. A study on material properties carried out by Bjelkevik and Knutsson (Bjelkevik 2005b) showed that the hydraulic conductivity differed to a large extend when compared to conventional materials. It was found that the hydraulic conductivity decreased with increasing distance from the discharge point, whereas the porosity increased. Since the porosity is directly related to the value of k, it would be expected that the porosity also decreased. Laboratory testing on samples was compared with field measurements, and again, the results did not coincide. Values for the horizontal and vertical conductivity differ up to one order of magnitude. Finally, attempts on calculating the hydraulic conductivity on basis of the models named above showed that measured and calculated values did not correspond. The variety in results is shown in Table 1. It can be noted that the ratio between measured and calculated values vary to a large extend, not only for different calculation models, but also for samples from the same site. The measured values range from 1,01 to 18,7 x 10-6 m/s. Calculations with Hazen s and Kozeny-Carman s formula result in a variation between 0,36 to 56,25 x 10-6 m/s and 0,47 to 51,56 x 10-6 m/s, respectively. Results from calculations on basis of Chapius method are even more extreme and lie between 5,73 to 263,99 x 10-6 m/s. This shows that the properties of tailings have so far not been studied sufficiently to provide a basis for seepage estimations and stability analysis. Nevertheless, design criteria for tailings dams exist, and embankment stability has been discussed. Table 1: Measured and calculated values of the hydraulic conductivity at different tailings dams in Sweden Measured Calculation Permeability (Hazen 1991) Ratio Measurement / Calculation Calculation Permeability (Kozeny- Carmen) Ratio Measurement / Calculation Calculation Permeability (Chapius) Ratio Measurement / Calculation [10-6 m/s] [10-6 m/s] [%] [10-6 m/s] [%] [10-6 m/s] [%] Kiruna 14,70 36,00 40,8 12,20 120,5 64,34 22,8 Svappavaara 6,08 5,67 49,00 39,69 12,4 14,3 51,56 23,99 11,8 23,6 263,99 124,62 2,3 4,5 Malmberget 16,30 18,70 56,25 39,69 29,0 47,1 19,25 18,08 84,7 103,4 92,12 93,35 17,7 20,0 2,54 1,00 254,0 0,50 510,1 5,73 44,3 Aitik 1,41 2,56 55,1 1,58 89,2 14,91 9,5 1,01 2,56 39,5 3,24 31,1 32,11 3,1 Boliden 2,56 2,78 1,44 0,36 177,8 772,2 1,67 0,47 153,2 588,2 18,58 7,19 13,8 38,7 Garpenberg 2,68 1,70 1,21 0,36 221,5 472,2 0,79 0,52 338,6 328,9 8,81 7,94 30,4 21,4 Zinkgruvan 5,41 4,00 135,3 2,92 185,0 25,22 21,4

131 4 Design and stability of tailings dams Geometrical criteria for filtering are related to soil geometry where constrictions are too small for particles to migrate. The hydraulic criteria, on the other hand, refer to the retaining and driving forces that result from seeping water. Seepage through the embankments is commonly considered by applying a flow net with stream lines and equipotentials to approximate the flow situation. The top flow line, i.e. the phreatic surface, has to be located inside the embankment at each point, so that the discharge point always lies within the structure. This method of controlling the pore pressure and hydraulic gradient is essential for both the slope stability and internal erosion. The hydraulic gradient and flow velocity has to be kept sufficiently low to avoid displacement of particles (Bjelkevik 2005a). 4.1 Design criteria for tailings dams The slope stability for an embankment exposed to seepage flow is generally simplified as shown in Figure 10, where an infinite slope of constant inclination with a parallel water table and seepage flow is considered. β Flow net τ f σ Figure 10: Basic slope stability consideration The shear strength τ f along a potential failure plane can be calculated taking the effective stress σ and the internal angle of friction φ of the material into account: τ f = σ ' tan φ' (2) Relating the retaining and driving forces, i.e. the mobilized shear strength and the actual shear strength in the sloping mass to each other, the safety factor F for such a theoretic plane becomes τ f F = (3) τ This equation is origin for the determination of long term stability of slopes of tailings dams applied in Sweden. The factor of safety for tailings dam slopes is currently assessed by means of the friction angle of the τ material φ, the angle of the slope β, and the mass density ρ: ρ tan φ' FS = (4) 1+ ρ tan β This design method may be questionable, since a real embankment slope is not infinite, and both the water table and the slope surface are not parallel to the failure plane. However, a more correct practical solution to the problem is currently not available, so that this principle is commonly used (Bjelkevik 2005b). 4.2 Other models on stability and erosion analysis To develop a model on critical hydraulic head, the hydraulic shear stress has been subject for discussion. Wan and Fell made an attempt to investigate the rate of erosion of soils in embankment dams on basis of a hydraulic shear stress (Wan 2004). The soils erosion characteristics were related to its behaviour regarding the rate of erosion under a certain hydraulic shear stress. It was found that this critical shear stress was lower for coarse grained soils, which implies that they erode more rapidly than fine grained soils. By using two test methods, the hole erosion test and the slot erosion test, a critical shear stress was determined. However, this value varied to a large degree and could not be related to other soil properties. A calculation of the hydraulic shear stress in an open crack in the embankment was presented, but to be able to analyse this value, the geometry of the crack has to be known. Again, the problem of non-visibility of such cracks and erosion processes inside the embankment set a limit for application. Nevertheless, useful practical information of the study was that the erosion rate is strongly influenced by the degree of compaction and the water content of the soil. Compaction to a high dry density on the wet side of optimum water content was found to show a higher resistance against erosion. Arulanandan and Perry examined piping failures in dams by means of the critical shear stress (1983). Yet, this study was mainly related to cohesive soils. Sellmeijer and Weijers published a study on piping mechanism which presented a model used as basis for the design of dikes in the Netherlands. A mathematical description was presented by means geometry of the structure. The difference in water head H and length of the seepage path L, as well as a certain seepage thickness layer under the dike D were related to each other. The study presented a solution for calculation of a critical water head and verified by large scale tests. Yet, it was limited to geometry and material properties of the sand used in the Netherlands, and application of the results for variant material is uncertain (Weijers 1993). Skempton stated that the critical gradient at which particles start to move differs from the theoretical critical hydraulic gradient that had been applied in filter studies. Skempton and Brogan showed in tests that instable material experiences particle migration at much lower hydraulic gradients than stable material. For horizontal flow, the critical gradient of stable material was set to 70 %, whereas the critical gradient

132 of unstable material was only 17 %. (Skempton 1994). These gradients are much higher than those gradients found in nature, and the values show that laboratory tests often require extremely high gradients which are not comparable to field conditions. 5 Observations of natural formations Little knowledge or experience is available regarding the long term stability of man-made embankments. Therefore, valuable information may possibly be found in geological formations which have fulfilled a similar function. It is obvious that a natural barrier of soil can withstand ageing or deterioration, and retain water over long time periods without experiencing erosion processes. A study by the Geological Survey of Sweden (SGU) covered several landforms, giving information about the geological structure as well as the hydraulic gradient the formation was, or still is, exposed to. These structures are dated back to the last glacial period, i.e to 8000 years ago (Agrell 2002). A well-known Swedish example of a natural dam is Ragundasjön, which was located at the Indalälvens valley the middle of the country. The lake was 25 km long, 6-10 m deep and had a medium water level of 139 masl. The southern part of the lake formed two narrow bays: one where the lakes mean water level was regulated by a waterfall, and Sandviken, with a border formed as natural dam. The lakes history is comprehensively documented because it was accidentally emptied in After a ditch was burrowed to create a channel by erosion, the barrier broke and the lake was drained. Sandvikens southern barrier dammed the lake that was formed about 6500 B.C. until the accident and it is presumed that it would still exist if not a human being had destroyed it. The eastern and western border consisted of rock, while the southern barrier was formed by layers of clay. Parts of this clay ridge are left, the western plateau at a height of 155 masl, and the eastern plateau at 160 masl. The ridge consisted of 60 m glacial clay in layered deposits with alluvium with a low hydraulic conductivity. Agrell stated in his report that the difference in height between the lakes highest water level and the water level downstream was 40 m over a distance of about 1000 m, which corresponds to a hydraulic gradient of 4 %. Another example is the lake Hennan, in the county of Gävleborg in central Sweden, south of Ragundasjön. At this site, two lakes are located right beside each other. The lake Hennan upstream of the lake Storsjön has a water level of about 206,9 to 208,8 masl, while the lake Storsjön has a water level of 186,8 masl, i.e. a difference of 22 m. The headland between these two lakes is about one kilometre wide, which results in a hydraulic gradient of 2 %. The ridge between the two lakes has an elevation of 240 masl. It is strongly believed that the delta consists of a barrier of moraine and well compacted material from an earlier glaciation. Drilling showed that bedrock was located at a depth of 47 m, with an overlying layer of sediment from a glacial stream which is at least 32 m thick. The upper layer of moraine has a thickness of 8 m. The lake is therefore considered to be dammed by a core of stream deposit surrounded by moraine, which is similar to an earthfill dam with a core and filter layer. Another formation is located close to the mining area Aitik in northern Sweden, which will be interesting for further examination in connection with ongoing work on tailings dams construction. Here, two lakes at a distance of 250m show a difference of 19m in water level, which results in a hydraulic gradient of about 7,5%. Such examples from nature show that it is obviously possible to retain water over long time periods. Unfortunately, we meet the same problem here as we did when considering the hydraulic gradient. Tailings differ from natural materials to a large extend, and our engineering models to deal with the material may not be sufficient in case of tailings. Hence, we need to learn more about tailings and their properties. Nevertheless, regarding the life time a tailings dam has to be stable, geological formations are the only structures we are able to refer to. Even though there are limitations in the application of geotechnical models on tailings, we need to learn from nature. CONCLUSIONS Tailings dams differ significantly from embankment dams in their structure, the material they are constructed of, the hydraulic gradient they are exposed to, and their design life. Because of the mineral composition of tailings as a rest product of mining activity, tailings have to be stored so that chemicals and metals are not released to the environment. This is often done by wet cover, i.e. by creating an artificial lake where tailings are used for construction of the surrounding embankments. A major problem is to guarantee the stability of a tailings dam over a time period of more than 1000 years. Many man-made embankments have, however, no comparable life time and have suffered from accidents and failures because of seepage and internal erosion. To be able to design a tailings dam, information about the materials friction angle and shear strength is taken into account. Nevertheless, the maximum hydraulic gradient for such a structure is not defined. Yet, this maximum gradient is crucial for the stability of the dam. Geological formations that act as natural analogies of dams have been identified. Such geological formations are the only structures that can be related to the subject when regarding the design life of a tailings dam. It has been shown that the gradients natural dams are exposed to range from 2 % to 7,5 %. A further study and comparison of material properties of both the geological material and tailings will give additional information on the critical hydraulic gradient for the design of tailings dams.

133 ACKNOWLEDGEMENTS This research study is part of the work within Swedish Hydropower Center SVC, which has been established by the Swedish Energy Agency, Elforsk and Svenska Kraftnät together with Luleå University of Technology, The Royal Institute of Technology, Chalmers University of Technology and Uppsala University. The study is carried out in cooperation with the Swedish mining industry and dam safety committee. REFERENCES [1] Åberg B (1993) Washout of grains from filtered sand and gravel materials. Journal of Geotechnical Engineering. 119 (1): [2] Agrell H (2002) Naturlig dämda sjöar. Analogier av damkonstruktioner.geological Survey of Sweden (SGU). Project report for Svartliden Guld AB (in Swedish). [3] Arulanandan K, Perry EB (1983) Erosion in relation to filter design criteria in earth dams. Journal of Geotechnical Engineering. 129(5): [4] Bjelkevik A (2005a) Water cover closure design for tailings dams- State of the art report. Research report, Luleå University of Technology. 2005:19. [5] Bjelkevik A (2005b) Stability of tailings dams- Focus on water cover closure. Licentiate thesis. Luleå University of Technology. 2005:85. [6] Chapuis, RP (2004) Predicting the saturated hydraulic conductivity of sand and gravel using the effective diameter and void ratio. Canadian Geotechnical Journal. 41: [7] Fannin R, Moffat R (2006) Observations on internal stability of cohesionless soils. Géotechnique. 56(7): [8] Fell R, MacGregor P, Bell G (2005) Geotechnical engineering of dams. A.A. Balkema Publishers, London. [9] Foster M, Fell R, Spannagle M (2000) The statistics of embankment dam failures and accidents. Canadian Geotechnical Journal. 37: [10] Kenney TC, Lau D (1985) Internal stability of granular filters. Canadian Geotechnical Journal. 22: [11] Kenney TC, Lau D, Ofoegbu GI (1984) Permeability of compacted materials. Canadian Geotechnical Journal. 21: [12] Lafleur J (1984) Filter testing of broadly graded cohesionless tills. Canadian Geotechnical Journal. 21: [13] Lafleur J, Mlynarek J, Rollin AL (1989) Filtration of broadly graded cohesionless soils. Journal of Geotechnical Engineering. 115(12): [14] Lambe WT, Whitman, RV (1969) Soil engineering. John Wiley & Sons, Inc., New York. [15] Milligan V (2003) Some uncertainties in embankment dam engineering. Journal of Geotechnical and Geoenvironmental Engineering. 129(9): [16] Sherard JL (1986) Hydraulic fracturing in embankment dams. Journal of Geotechnical Engineering. 112(10): [17] Sherard JL, Dunnigan LP, Talbot JR (1984) Basic properties of sand and gravel filters. Journal of Geotechnical Engineering. 110(6): [18] Skempton AW, Brogan JM (1994) Experiments on piping in sandy gravels. Géotechnique. 44 (3): [19] Terzaghi K, Peck RB, Mesri G (1996) Soil mechanics in engineering practice. John Wiley & Sons. New York. [20] Wan CF, Fell R (2004) Investigation of rate of erosion of soils in embankment dams. Journal of Geotechnical and Geoenvironmental Engineering. 130(4): [21] Weijers JBA, Sellmeijer JB (1993) A new model to deal with the piping mechanism. In: Filters in geotechnical and hydraulic engineering. Brauns, Heibaum & Schuler (eds). Balkema, Rotterdam.

134

135 Conference paper #4 Jantzer, I. (2006). Frost action processes in the Eastern Suorva hydropower dam. 17 th European Young Geotechnical Engineers Conference EYGEC. Zagreb, Croatia. July 20, 2006.

136

137 Frost action processes in the Eastern Suorva hydropower dam core I. Jantzer Dept. of Civil, Mining and Environmental Engineering, Luleå University of Technology, Sweden ABSTRACT: The hydropower embankment dam in Suorva, Sweden, is located in north of the Arctic Circle. The region is known for cold winters and severe wind conditions. The construction was completed in 1972 and has since then been exposed to a wide temperature range and low annual mean temperatures, leading to the presumption that the fine-grained frost susceptible moraine core of the embankment was exposed to cyclic freezing and thawing. The aim of this project was to find indications for freezing processes that affect the structure of the soil. The project included field investigations and a theoretical thermal analysis carried out with a commercial finite element program. Results from this analysis showed that the freezing plane advances to a maximum depth of 5 m from the top of the core. Hence, it may be possible that frost action affects the construction in its function as hydraulic barrier. 1 INTRODUCTION The hydropower embankment dam in Suorva, Sweden, is situated north of the Arctic circle in a mountain valley which is known for its low annual mean temperature and channelling effects on wind. The dam was built during the years and is one of the most important hydro power sources in Sweden as it regulates second largest reservoir with a capacity of m³ water. The mountain region is a part of cold regions with deep seasonally frozen ground. It is exposed to severe wind conditions and annual mean temperatures around -1 C to -0,4 C. Hence, the core of the embankment dam may possibly be affected by repeated freezing and thawing. As a part of the Swedish dam safety program the dam crest was removed and the moraine core was raised. The original material in the core was analyzed to ensure the quality of the material. In combination with these works, the possibility of investigations regarding freezing and thawing on the core was given. (Jantzer, 2005) Cyclic freezing and thawing influences the structure of soil; studies of fine-grained soils exposed to freezing reported changes in volume and structure, as well as significant increases in permeability. (Viklander, 1997) Such changes may have an impact on seepage flow and reduce the function of the core is significantly. Temperature and climate in this northern region provide conditions for repeated freezing and thawing actions. In order to find out if, and in which way, the central core may had been affected of the repeated cycles of freezing and thawing an in-situ study was carried out. In addition, a theoretical part was carried out to illustrate the possible impact of climate on the dam with the help of the finite element method, Temp/W.

138 2 BACKGROUND 2.1 Structure of the embankment The Suorva dams are rockfill dams and consist of three different dams, East Suorva dam, Sågvik dam, and West Suorva dam. They were constructed during the years 1966 until The total length of the crest is 1370 m; the length of the Eastern dam which is subject to this work is 780 m. All three dams have an impervious central core of moraine. They are founded on rock with a grout curtain beneath the central core. A schematic cross-section of the East Suorva dam is shown in Figure 1, where the different zones are numbered: 1 rock fill, 2 moraine core, and 3 filter zone. Figure 1. Basic cross-section of the Eastern Suorva dam (Vattenfall) 2.2 Location and impact of climate Frozen ground is defined as soil or rock with a temperature below 0 C. This definition is only dependent on temperature and gives no information about the content of water or ice. If temperature in the ground remains below freezing point throughout two subsequent winters and the intermediate summer, the term permafrost is applied. If the ground is only frozen during the winter period, it is referred to as seasonally frozen ground. Cold regions are characterized with respect to air temperatures, depth of ground freezing, snow depth, or ice cover on lakes. Such areas are assumed to have a minimum frost penetration of 0,3 m occurring at least once during a period of 10 years. Cold regions are divided into permafrost or seasonally frozen ground. Continuous and discontinuous permafrost regions may be found in Polar Regions, such as Alaska, Canada or Siberia. Scandinavia, despite high latitude, belongs mainly to regions with seasonally frozen ground, together with parts of the United States and vast areas of Asia. Nevertheless, there are some areas found in Scandinavia which belong to the category of discontinuous permafrost. The embankment dam at Suorva is located north of the Arctic Circle in the upper reaches of Lule River. The reservoir regulated by the Suorva dam is about a 60 km long. It originates from five lakes. The area is marked with a circle in Figure 2, which shows the mean annual temperatures based on records by Swedish Meteorological and Hydrological Institute. The map illustrates the mean annual temperatures over a 30-year period from 1931 until These temperatures range from -2 C to +8 C. For comparison, the registered mean annual temperature in Stockholm is +8 C, in Luleå +2,5 C, and in Suorva -1 C. (Taesler, 1972) The following 30- year period until 1990 showed a slightly increased mean annual temperature. The increase ranges from 0 C to 0,6 C in different regions of Sweden. For the location of the embankment dam, an increase of the annual mean temperature of 0,4 C was noted. (Alexandersson, 1991) The reservoir is surrounded by a mountain valley with a main direction from north-west to south-east, which is known for its channelling effects on wind. Wind conditions with an annual

139 average mean wind speed of 7,85 m/s were measured on a 35 m high tower situated between the Eastern dam and Sågviks dam. This number is considered remarkably high for an inland location. For comparison, average wind speeds at coastal sites were measured to approximately 7,0 m/s. Luleå Stkhlm Figure 2. Map of Sweden with annual mean temperature zones (SMHI, 1991) 2.3 Properties of the core material The core of the dam, which is subject to this study, is made up of earth fill material. Its function is to limit the seepage through the dam and to ensure a reasonable low hydraulic conductivity. In Sweden, the use of moraine, a glacial till with a complex grain-size distribution ranging from fine-grained material such as clay and silt up to gravel, stones and boulder, is most common. For this purpose, hydraulic barriers of moraine are predominant since large deposits of the material can be found in mountain areas within economic hauling distance. Requirements suggested by Vattenfall (1988) for moraine used as hydraulic barrier in Sweden are following: - The content of fines is less than 40 % of the material passing a 20 mm sieve. - A minimum amount of 70 % of particles passing a 64 mm sieve is smaller than 20 mm. - A maximum amount of 85 % of particles passing a 64 mm sieve is larger than 2 mm. - The hydraulic conductivity should range between 3 x 10 7 m/s to 3 x 10 9 m/s. - Large boulders may occur, but a maximum size of two thirds of the placement layer thickness may not be exceeded. The maximum diameter allowed is mm.

140 2.4 Frost action The term frost action refers in general to three features: frost heave, thaw weakening and thaw settlement. Apart from the fact that soil in its frozen state may introduce positive properties such as strength, increased bearing capacity and reduced hydraulic conductivity, does the process of frost action also involve negative effects to engineering constructions. Frost action occurs in frost-susceptible soils, when temperatures are sufficiently low and water for freezing and ice lens formation is available. Frost heave results from freezing of the soil. Water in the ground undergoes a phase change from fluid to solid state, thereby increasing its volume. Yet, it is not only water freezing in situ in the voids that is responsible for the expansion, but also unfrozen water migrating and turning to ice at the freezing front. The accumulation of ice results in increased water content when the soil thaws. Consequently, the bearing capacity is reduced. (Andersland/Anderson, 1978) Effects of frost action are changes in volume and water content, as well as loss of shear strength and bearing capacity. Cyclic freezing and thawing influences the structure and permeability. Both may have significance for fill material in the core of an embankment dam. The process of freezing causes water to migrate, freeze, and increase the volume of the soil, and finally the soil skeleton to move. Pressure created by ice breaks up bonds between particles and lifts them up. In contrast to that, thawing generates melting water which is pressed out of the voids. Thaw settlement and consolidation, causing movement of particles and rearrangement in the skeleton, are the result. It has been observed that fine-grained soils exposed to freezing and thawing undergo structural changes. (Viklander, 1997) These changes are not similar for fine-grained and coarse-grained soils, but differ because of cohesion, specific surface of particles and the amount of bound water. A clayey soil may have a different structure, where cracks filled with ice are located close to the freezing plane and shrinkage occurs because of negative pore pressure. Cracks, ice lenses or fissures are alternating layers in the soil with varying thickness. Such spaces will not be completely closed in finegrained soils during thawing because of cohesive forces of the fines in the soil. After thawing, water filled cracks create new paths for water to flow through the soil, thus increasing its hydraulic conductivity. A coarse-grained soil, on the other hand, has a self-healing capacity because of the lack of cohesion and will have a similar structure before and after a freeze/thaw cycle. Effects on hydraulic conductivity and volume depend on the initial void ratio and the degree of compaction. An originally loose soil with a small degree of compaction shows some volume increase during freezing and after that an extended decrease of volume upon thawing. In this case, the void ratio and hydraulic conductivity finally will be decreased, since particles move closer together after each freezing cycle. After a number of repeated freeze/thaw processes, a residual void ratio is reached and the hydraulic conductivity has a lower value than before. Tests with 10 cycles have been reported, and a decrease in hydraulic conductivity was noted to a magnitude of approximately 50 times. In contrast, a dense and more compacted soil showed a void ratio and permeability increase up to 11 times. Repeated freeze/thaw loosens the structure of the soil, and again, the void ration reaches a residual value (Viklander, 1997). 3 FIELD STUDIES 3.1 Description of practical work For field investigation, four test pits in different sections of the dam were excavated. The pits had a side length of 2 4 m and were excavated by the help of an excavator. Disturbed samples were collected in each test pit at five levels: 0,0 m, 0,5 m, 1,0 m, 1,5 m, and 2,0 m. Date, location and depth were registered. The samples were stored in plastic bags and in containers. Visual inspections as well as collection of samples from pits were expected to provide information of whether the core had been exposed to freeze/thaw cycles or not. If freezing had occurred, the inspections were estimated to reveal in which way such frost action had influenced the fill

141 material. In case of freezing, the material was expected to be layered and to show visible water accumulations; differences in the grain-size distribution, water content or varying dry density were hypothesized to be detected. For documentation of the condition of the moraine core, pictures were taken. In addition, the temperatures were measured when the soil felt unusually cold during sampling. A few selected pictures will be presented here to give a general overview of the observed aspects. 3.2 Results of field investigations During field investigations, the overall impression was that the soil was comparatively cold in comparison with the air temperatures, and that the soil had a layered structure with a considerable higher water content in upper regions at a depth around 0,30 0,80 m. Three of the four test pits showed such similar features. In case of one test pit, no water accumulation or low temperature was measured. The material was homogeneous and did not give information about freezing processes. The samples taken from the Eastern Suorva dam section 750 are significant for observations of temperature and water content and will therefore be discussed in detail. Field investigations in section 750 were carried out on July 5 th, 2004, on a sunny day with temperatures around 25 C. The material on top of the core was removed right before excavation of the test pit, which included a layer of boulders and an insulation layer. Remains from measures against leakage in the embankment dam were found, such as standpipes which had been used for bentonite grouting, see Figure 3. (Jantzer, 2005) standpipe boulder Figure 3. Surface of test pit in section 750. Note leftover standpipes from grouting as well as stones and boulder. (Jantzer, 2005) While the top layer of fill material was dry and well-graded, it appeared to have an increased moisture content already at an excavation level of 0,3 m, see Figure 4. The soil seemed to be completely saturated at this point and water was expelled when pressure was imposed by setting a foot on it. The consistency varied from very soft to an almost liquid state. The water content was found to be extremely high and the temperatures of the soil comparatively low, even though the weather conditions warm and dry, see Figure 4. The water content ranged from a minimum of 6,6 % at level 2,0 m to a maximum of 13,4 % at 0,5 m. (Jantzer, 2005)

142 0,3 0,8 m Temp [ C] 3,3 0,5 1,8 1,0 W [%] 11,4 13,9 2,0 8,4 1,5 1,6 10,4 Crack with high water content, Temperature measurement see Figures 5,6,7 1,3 2,0 6,6 Depth [m] Figure 4. Overall view of the trench and summary of values for water content and temperatures measured of the excavated test pit. (Jantzer, 2005) At a depth of 1,0 m the moisture content changed considerably to a much lower value: 8,4 %. A horizontal fracture was found at 0,5 m depth and when excavation reached this level a long gap was left open, see Figure 5. The surrounding area was wet and was found to have temperatures of about 1,7 C to 1,9 C in comparison to 3,3 C in the upper layer and 2,0 C in the lower layer. It was obvious that the moraine core had been frozen at this level, see Figures 6, and 7. (Jantzer, 2005) Cracks Figure 5. Fractured moraine core at level 0,4-0,5 m. (Jantzer, 2005)

143 wet zone Figure 6. Wet area in the fracture after removing loose material. (Jantzer, 2005) Figure 7. Temperature measurement showing 1,7 C in the area of moisture accumulation. (Jantzer, 2005) The grain-size distributions of samples from the test pit in section 750 are shown in Figure 8. It was noted that the particle-size distribution from level 2,0 m was coarser compared with the other four samples, and that samples taken between 0,0 m and 1,5 m were more fine-grained. The difference was associated with the comparatively high water content at upper levels, thus implicating that fine-grained material is more frost susceptible, and that water is collected preferably in such locations. (Jantzer, 2005) Clay & silt Sand Gravel ,01 0,06 0, ,00m 0,50m 1,00m 1,50m 2,00m W atercontent: 0,00 m: 11,4 % 0,50 m: 13,9 % 1,00 m: 8,4 % 1,50 m: 10,4 % 2,00 m: 6,6 % Figure 8. Grain-size distribution and water content section 750. (Jantzer, 2005)

144 3.3 Discussion The field investigations were supposed to show if the core of the dam had been exposed to freezing and thawing, thus influencing the structure of the core and its hydraulic conductivity. The grain-size distribution was analyzed in order to find out if freeze/thaw cycles had influenced on the soils structure and if segregation had taken place. Therefore, archive data and grain-size distributions from the moraine pit from the time of construction works were viewed and compared with the fractions of the material found in the test pits, see Table 1. Upper and lower limits of fines, sand and gravel were identified, as well as their range of variation max min. It was found that the grain-size distributions generally varied to a larger extent than expected. For example, test pit 750 had a variation of fines around 20 %, and the grain-size distribution found in archive records from section 737 showed a variation of 23 %. The same applies to the fraction of sand, where test pit 750 had a variation of 16 % between the upper and lower limits of this fraction, while archive records showed 23 %. Hence, statements about variations in grain-size distributions or rearrangement of particles could not be made. Table 1. Comparison of upper and lower limits in grain-size distribution Test pit section ,5 archive 737 archive Borrow pit Fraction [%] [%] [%] [%] [%] [%] [%] < 0,06 mm min 29,4 26,3 24,4 32, < 0,06 mm max 35,4 34,6 44,4 36, max - min 6,0 8,3 20,0 4, < 2,00 mm min 74,7 62,6 65,1 73, < 2,00 mm max 84,1 74,8 81,1 79, max - min 9,4 12,2 16,0 6, < 20,0 mm min < 20,0 mm max max - min 9,0 14,0 17,0 9, In order to be able to draw conclusions about impact of frost action, pictures, analysis of water content and temperature measurement were much more significant than the comparison of grain-size distributions. As the water content was exceptionally high in some places, and the temperatures at the same time low compared to the air temperatures, it was likely that the soil had been frozen. A summary of these values is given in Table 2. Table 2. Comparison water content values and temperature Test pit section Level w [%] w [%] Temp [ C] w [%] Temp [ C] w [%] 0,0 m 6,9 7,6 4,4-3,8 11,4 3,3 6,5 0,5 m 6,5 7,7-13,9 1,9-1,7 7,1 1,0 m 6,8 6,9 2,3-2,1 8,4 2,0 6,8 1,5 m 7,3 8,1 1,9 10,4 1,6 7,6 2,0 m 7,7 7,2 1,6 6,6 1,3 6,6 Mean value Ø 7,04 7,50 10,14 6,92 It was assumed that ice lens formation had been taking place in section 750, where the soil was found to be saturated and temperatures between 1,3 to 3,3 C were measured. Other observations such as cracks or a layered structure were made in almost all test pits, mostly at depths between approximately 0,3 m to 0,8 m. Nevertheless, evidence for effects of frost action, such as changes in structure, density, volume, permeability or particle movements is not given.

145 4 THERMAL ANALYSIS 4.1 Introduction and definition of the problem The question whether the core of the embankment dam had been subjected to freezing cannot be answered by the results of the field investigations only. In order to get a more detailed picture of the temperature distribution in the construction, a thermal analysis was carried out. Modeling and calculation of thermal changes was carried out with they finite element program Temp/W. The Swedish Meteorological Institute SMHI provided temperature records measured at the meteorological station in Stora Sjöfallet, 10 km east from the Suorva dam. To be able to compare the results of the thermal analysis with observations from field investigations, the temperatures records from 2002, 2003 and 2004 where regarded as relevant for computation. The problem is drawn and defined by a CAD-similar function. For drawing a cross-section of the embankment dam a drawing from archive records was used as basis. The original drawing from the archive is shown in Figure 9. In this drawing of section 400, the construction is 44 m high and has a maximum width of 120 m. The core is sloped at water side and placed between filter zones. Figure 9. Archive drawing used to sketch the problem. (Vattenfall) Since the water level is varying by time and the mesh cannot be changed during the computation of the problem, a medium water level was used. This is essential for the following definition of material properties in different zones and unsaturated and saturated condition, as the thermal properties are substantially dependent on the water content. During the period 2002 to 2004 the mean water level was set to m. The seepage level throughout the whole construction should have been calculated in order to get a more exact border line for material properties. However, as the actual seepage condition is only a presumption based on a medium value, and is mainly supposed to show the influence of the water temperature in general, seepage was simplified and approximated by straight lines, see Figure 10. To generate finite elements, a mesh was drawn, consisting of 2214 elements and 2384 nodes. The elements have different side length, varying from 0,5 m in the upper zone up to m above sea level, 1,0 m between m to 443 m and 2,0 m down to the foundation. Since the area of the cross section of the embankment is quite large, the element size was increased at lower levels to avoid a too large number of nodes and mistakes in the verification and sorting of the elements.

146 Figure 10. Basic drawing of the problem for computation 4.2 Steady-state analysis The steady-state analysis has to be carried out to create a basic condition of temperature distribution. This basic condition is an essential part of the calculation as it influences considerably the following transient analysis. The obtained temperature distribution governs the amount of heat energy stored in the body of the embankment dam. It was therefore chosen to base the initial calculation upon the mean annual temperature, which was assessed according to SMHIs temperature records for a thirty-year mean value. The value of about -1 C to -0,4 C was chosen; the results for calculations based on the mean annual temperature for the last 30 years, -0,4 C, are presented below. The temperature distribution inside the body of the embankment dam is also dependent on the water table and the seepage through the construction. With regard to the special property of water having its highest density at +4,0 C, a boundary condition of +4,0 C was used at the adjacent nodes. Figure 11 shows the result of the calculation of an initial temperature condition on basis of an annual mean temperature of -0,4 C. The freezing plane advances about 3 m from the top of the dam crest, but does not reach the central moraine core. The influence of the water can be seen in the right corner of the drawing, where the boundary condition of +4,0 C was set. In this water saturated area more thermal energy is stored; the temperature ranges between +4,0 to +2,0 C. Approx. 3 m Figure 11. Result of calculation of initial condition based on a mean value of 0,4 C.