CSIRO LAND and WATER. Sheet and Rill Erosion and Sediment Delivery to Streams: A Basin Wide Estimation at hillslope to Medium Catchment Scale

Transcription

1 Sheet and Rill Erosion and Sediment Delivery to Streams: A Basin Wide Estimation at hillslope to Medium Catchment Scale Report E to Project D10012 of Murray Darling Basin Commission: Basin-wide Mapping of Sediment and Nutrient Exports in Dryland Regions of the MDB Hua Lu, Chris J. Moran, Ian P. Prosser, Michael R. Raupach, Jon Olley, and Cuan Petheram CSIRO Land and Water, Canberra Technical Report 15/03, June 2003 CSIRO LAND and WATER

2 Sheet and Rill Erosion and Sediment Delivery to Streams: A Basin Wide Estimation at hillslope to Medium Catchment Scale Report E to Project D10012 of Murray Darling Basin Commission: Basin-wide Mapping of Sediment and Nutrient Exports in Dryland Regions of the MDB Hua Lu, Chris J. Moran, Ian P. Prosser, Michael R. Raupach, Jon Olley, and Cuan Petheram CSIRO Land and Water, Canberra Technical Report 15/03, June 2003

3 Copyright 2003 CSIRO and the Murray-Darling Basin Commission. This work is copyright. It may be reproduced subject to the inclusion of an acknowledgement of the source. Authors Hua Lu, Chris J. Moran, Ian P. Prosser, Michael R. Raupach, Jon Olley and Cuan Petheram CSIRO Land and Water, PO Box 1666, Canberra, 2601, Australia. Phone: For bibliographic purposes, this document may be cited as: Lu, H., Moran, C.J., Prosser, I.P., Raupach, M.R., Olley, J. and Petheram, C. (2003) Hillslope erosion and sediment delivery: A basin wide estimation at medium catchment scale, Technical Report 15/03, CSIRO Land and Water. A PDF version is available at: ISSN Important Disclaimer CSIRO Land and Water and the Murray-Darling Basin Commission advise that the information contained in this publication comprises general statements based on scientific research. The reader is advised and needs to be aware that such information may be incomplete or unable to be used in any specific situation. No reliance or actions must therefore be made on that information without seeking prior expert professional, scientific and technical advice. To the extent permitted by law, CSIRO Land and Water and the Murray-Darling Basin Commission (including their employees and consultants) excludes all liability to any person for any consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using this publication (in part or in whole) and any information or material contained in it. 2

4 Table of Contents CSIRO Land and Water... 1 Executive Summary Introduction Background Characteristics of the MDB Climate Topography Geology and Soil Land Use and Management Practices Spatial Settings and Terminology Methodology Hillslope Sheetwash and Rill Erosion Hillslope Sheet and Rill Erosion under Natural Condition Sediment Delivery Ratio (SDR) Background of SDR A New SDR Theory Statistical Analysis of Effective Rainfall duration and Intensity Estimations of Residence Time Sediment Residence Time as a Function of Particle Size Estimating Travel Time of Water Particles t h0 and t n Results Hillslope Erosion under Current Land Use Hillslope Erosion under Natural Conditions Spatial Characteristics of Effective Rainfall Duration and Intensity Spatial Distribution of Sediment Delivery Ratio Discussions and Conclusions Acknowledgments References Appendix I: Land Use Data Appendix II: Pluviograph Rainfall Data

5 List of Figures Figure 1: Major landuse groups in the MDB Figure 2: SDR vs catchment area relationships obtained from different areas around the world Figure 3: Diagram of a two storage lumped linear model of SDR at catchment scale (after Sivapalan et al. 2001, modified). See text for detail Figure 4. Comparison of SDR (%) measurements (Roehl 1962), modeled average SDR and flow response (Robinson and Sivapalan 1997). It shows that flow response represents the upper envelope of the SDR Figure 5: SDR as a function of channel residence time for different values of t er and t h (upper panel); SDR as a function of catchment area for different values of t er and t h. SDR measurements from USA catchments (Roehl 1962) are also shown as red dots (lower panel) Figure 6: Site locations of pluviograph rainfall data and their relative position to MDB Figure 7: All rainfall events characterised by their 30 intensity and duration (upper panel); Fit probability density functions to Gamma and exponential distributions for both duration and intensity (second and lower panels) Figure 8: Rainfall events which have depth equal or greater than 12.7 mm (upper panel); Fit probability density functions to Gamma and exponential distributions for both duration and intensity (second and lower panels) Figure 9: Effective rainfall events which have depth equal or greater than 12.7 mm (upper panel); Fit probability density function of effective duration to Gamma and exponential distributions (lower panel) Figure 10: Relationships between effective 30-min. rainfall intensity and the ratio between mean annual R-factor and mean annual rainfall Figure 11: Relationships between rainfall duration (t r ) to mean annual rainfall (MAR), effective 30-min intensity (MI30), MAR/MI30, and MAR 2 /R Figure 12: Relationships of effective rainfall duration and it relative errors Figure 13: Error estimations of rainfall duration. Upper Panel: Comparison between rainfall duration estimated using site specific pluviograph data and that estimated using regionalised relationships. Middle Panel: Absolute error [hrs] plotted against number of year with complete data. Lower Panel: Relative error plotted against number of year with complete data. The crosses are the sites have shorter records and relatively larger errors. They are not used in the final relationships that are applied across the MDB Figure 14: Diagram of the particle size effect on sediment travel time in relation to the travel time of water particles

6 Figure 15: Flow chart for the calculation of travel time of water particles Figure 16: Estimated annual average sheet and rill erosion rate Figure 17: Monthly distribution of total soil loss rate for the Basin Figure 18: Comparison between natural C-factor values modeled using Cubist and those extracted from C-factor map at the locations of minimum cover disturbance. There are 9916 points in total. The line of best fit and 1:1 line are shown Figure 19: Estimated annual erosion rate under natural conditions (pre-european settlement conditions) Figure 20: Estimated Ratio between erosion rate under current landuse and that under natural conditions Figure 21: Estimated effective 30-min. rainfall intensity for south-eastern Australia. The boundary of MDB is shown Figure 22: Estimated rainfall duration (left panel) and effective storm duration (right panel for south-eastern Australia. The boundary of MDB is shown Figure 23: Estimated travel time of water particles t h0 and t n0 for each sub-catchment element in MDB Figure 24: Estimated Sediment delivery ratio fro clay, silt and sand particles Figure 25: Estimated overall sediment delivery ratio from each sub-catchment elements Figure 26: Estimated specific sediment yield [t/ha/yr] for each sub-catchment element Figure AI.1: Data sources of land use used in this project

7 List of Tables Table 1: Landuse groups used to calculate sheet and rill erosion rate Table 2: Typical values of CN for some land use group Table 3: Values of Manning s n used in this study for common land use and vegetation cover groups for overland flow Table 4: channel roughness parameter a values used in this study Table 5: Three erosion groups (high, medium and low) and their relation to percentage of agricultural lands Table 6: Soil loss rate from land use categories Table AI.1: Summary of locally supplied land use data used in this study Table AII.1: Details of the pluviograph rainfall sites

8 Executive Summary This report presents a scientific and technical description of the modelling framework and main results for the long-term average hillslope erosion and sediment delivery to streams at hillslope to medium scale catchment over the Murray Darling Basin. The work was a part of project D10012 of the Murray-Darling Basin Commission (MDBC), "Basin-wide mapping of sediment and nutrient exports in dryland regions". Gully and stream bank erosion, sediment transport at larger scale with intervening deposition to flood plain, and associated nutrient exports are dealt with in separated reports (DeRose et al. 2003; Hughes and Prosser 2003). The specific objectives of this part of the work are basin-wide mapping by: (1) Quantifying the hillslope sheetwash and rill erosion under current land use condition; (2) Quantifying the inherent natural hazard of hillslope sheetwash and rill erosion; (3) Determining the amount of sediment generated by sheetwash and rill erosion delivered to the stream network from the sub-catchment elements with contributing area around km 2 ; (4) Interpreting results in terms of comparison with pre-european land use conditions. The modelling frameworks are described as follows. We undertook new assessments of hillslope sheetwash and rill erosion across the MDB, building upon our previous work for the National Land and Water Resources Audit (NLWRA) (Lu et al. 2001; Lu et al. 2003b). The mean annual hillslope sheetwash and rill erosion was modelled using the Universal Soil Loss Equation (USLE) (Wischmeier and Smith 1978; Renard et al. 1997), which is a model of surface wash and rill erosion based upon factors of rainfall erosivity, terrain, soil erodibility and vegetation cover. USLE factors were calculated from digital elevation models (DEMs), soil attribute maps, land use maps, remote sensing imagery and daily rainfall surfaces. Time series of remote sensing imagery and daily rainfall were used to incorporate the effects of seasonally varying cover and rainfall intensity. Further, we used new digital maps of soil and terrain properties. In this project, improvements to the assessment of sheetwash and rill erosion were made by compiling higher resolution land use data for the MDB from a range of sources and by incorporating a database on crop rotation, tillage and other land management practices. These new data, together with improved analysis of remote sensing data, enabled a more accurate prediction of the effect of vegetation cover and cover management on hillslope erosion. To relate hillslope erosion estimates to riverine water quality, it is necessary to estimate the properties of eroded soil that is delivered to the waterways for further transport. A theory that relates long-term averaged sediment delivery to the statistics of rainfall and catchment parameters was proposed. The derived flood frequency approach was adapted to investigate the problem of regionalization of the sediment delivery ratio (SDR) across the Basin. SDR, a measure of catchment response to the upland erosion rate, was modeled by two lumped linear stores arranged in series: hillslope transport to the nearest streams and flow routing in the channel network. The theory shows that the ratio of catchment sediment residence time (SRT) 7

9 to average effective rainfall duration is the most important control in the sediment delivery processes. In this study, catchment SRTs were estimated using time of the concentration for overland flow multiplied by an enlargement factor which is a function of particle size. Rainfall intensity and effective duration statistics were regionalized by using long-term measurements from 195 pluviograph sites within and around the Basin. The major findings are as follows. Hillslope Erosion under Current Land Use: Our spatial modelling of sheetwash and rill erosion estimated that a total t yr -1 of sediment moves locally across the Basin, at a mean rate of 2.1 t ha -1 yr -1. Erosion rate increases from south to north and from arid areas to temperate regions. About two-thirds of erosion occurs in the summer period. Agricultural lands have slightly higher erosion rates, at a mean rate of 2.3 t ha -1 yr -1 as most of the agricultural lands are located in the flood plains. Inherent Natural Hillslope Erosion: It was estimated that soil erosion rates are low under pre- European natural vegetation conditions. The rates are 3 to 10 times on average and up to 100 times smaller than that under current land use. Spatial Characteristics of Erosive Rainfall: Rainfall intensity increases from south to north and from west to east. Coinciding with the effect of topography, it defines the broad pattern of hillslope erosion. Rainfall duration is greater in temperate than arid regions and decreases from uplands to flat inlands. This dissipates sediment transport energy and the whole system is inefficient for sediment transport from erosion sources to basin outlet. Sediment Delivery Ratio and Sediment Yield: The averaged SDR is about 5%, which is lower than the average estimated in other countries for catchments with similar contributing area. Most sub-catchment elements have SDR smaller than 5%, suggesting inefficiency of sediment transport in the broad areas of the Basin. Larger SDRs are obtained at the eastern edge of the Basin, with the Australian Alps having the highest SDR values, followed by the central south of the Murrumbidgee and Bathurst regions. Sediment yield is low for the majority of subcatchment elements with area-specific sediment yield around 0.13 t ha -1 yr -1, which also represents the Basin average. Problem areas are located mainly in the eastern edge of the Basin. 8

10 1 Introduction Soil erosion and sediment transport are recognised as major environmental hazards in the Murray Darling Basin (MDB). It is governed by topography, climate, soil, vegetation cover, land use and management factors, through mechanisms including, particle detachment by raindrop impact, hydrology, flow hydraulics and other processes. Ability to estimate erosion rate across the whole basin is significant for three reasons. Firstly, soil erosion has a range of environmental impacts, including loss of organic matter and nutrients, reduction of crop productivity, and downstream water quality degradation (Newcombe and MacDonald 1991). The integrated impacts are often revealed and of importance at catchment or even larger scales. Secondly, effective control of soil erosion is a critical component of natural resource management when the aim is to achieve sustainable agriculture and acceptable ecosystem integrity (Pimentel et al. 1995; Rutherfurd et al. 1998). With limited resources, national scale erosion maps are useful for guiding investment prioritization in effective remediation programs. Thirdly, to aid estimations of soil erosion contributions and their impacts, the effects of changes in climatic conditions, vegetation, and land use on soil erosion rates need to be assessed at regional to continental scales (Pimentel et al. 1995). Due to the prevalence of high-value commodities in the Basin, comprehensive data on full areal extent and severity of the Basin s soil erosion and sediment delivery is of both economic and environmental importance. Information on spatially distributed sediment delivery is useful in identifying relative importance between sediment sources and the effectiveness of sediment delivery. It helps to establish strategies in effective erosion control, rehabilitation planning, and achieving long-term sustainable productivity in the Basin. In the past, there have been several attempts to estimate soil erosion rates at regional to continental scale, i.e., reviews of erosion data (Edwards 1993); synthesis of hillslope erosion rates and sediment transport (Wasson, et al. 1996), reconnaissance survey using caesium-137 (Loughran and Elliott 1996), and quantitative spatial modelling using USLE (Rosewell 1997). Variations and uncertainties exist in all the previous estimations. The major discrepancies of previous studies are largely due to lack of high quality consistent spatial data and our inability to model the complex systems which involve subsystem interactions both in time and in space. In the late 1990s, Australia launched the National Land and Water Resources Audit (NLWRA 2001) to assess the condition of its land and water resources. The continent-wide assessment of sheetwash and rill erosion was conducted as part of a broader assessment of the conditions of Australian agricultural land (NLWRA 2001). Hillslope sheetwash and rill erosion estimation in this project was building upon our work in the NLWRA project (Lu et al. 2001; Lu et al. 2003b). Improvements were made by compiling higher resolution land use data for the MDB from a range of sources and by incorporating a database on crop rotation, tillage and other land management practices. These new data, together with improved analysis of remote sensing data, enabled a more accurate prediction of the effect of vegetation cover and cover management on hillslope erosion. Only a small fraction of the soil moving on hillslopes is actually delivered to streams (Edwards 1993; Wasson et al. 1996). This implies that most of the sediment travels only a short distance (Parsons and Stromberg 1998) and is deposited before leaving the hillslope. In general, the amount of sediment deposited is intimately related to the topography, climate, soil, vegetation cover, and land use conditions, which are all closely related to the hydrological processes. The travel time for transport of sediment across a field or hillslope is 9

11 often longer than the duration of runoff-generating events so that runoff infiltrates and is not delivered to the stream, along with the sediment it carries. In some environments there is also patchy generation of runoff on impermeable areas which then infiltrates on other patches of high infiltration, often at sites with better cover. Topography can induce deposition through its influence on the capacity of overland flow to transport sediment. Reductions in gradient and the dispersion of overland flow can both cause deposition. Farm structures, such as contour banks and dams, can have similar effects, altering flow paths or trapping runoff. Deposition also results from abrupt changes to vegetation cover as runoff travels downslope. This causes deposition in backwaters and reduces the sediment transport capacity of flow. In large-scale modelling of sediment transport, this phenomenon of different sediment transport rates between hillslope and catchment scale is usually modelled using a scaling factor called the hillslope sediment delivery ratio (HSDR). This avoids the need to explicitly model patterns of deposition on hillslopes which is not possible across such large areas as the MDB. Prosser et al. (2001) developed a spatially distributed model of mean annual sediment budgets for river basins. The model, SedNet (Sediment River Network Model), used spatial modelling of the erosion, deposition, and transport processes that move sediment and nutrients within landscapes and streams to produce regional budgets for the Murray Darling Basin. The sources of sediment considered are soil erosion by surface (hillslope) processes (Lu et al. 2001), gully erosion and riverbank erosion (Hughes and Prosser 2003). These sediment sources were routed through the river network using a simple conceptual model of the primary controls on sediment export and deposition. The results demonstrate that there is a reasonable correlation between observed and predicted specific sediment yields. The hillslope delivery in SedNet is modelled by USLE-SDR approach (Lu et al. 2001). In the NLWRA project, uniform hillslope sediment delivery ratio was applied to the whole MDB by using HSDR as a calibration factor to obtain the best results (Prosser et al. 2001). This neglects the environmental variation across the MDB that would give varying sediment delivery potential. For example, short steep hillslopes experiencing long storms in the east of the Basin will have a greater delivery ratio than the long, flat hillslopes of the western regions. In this project, a new approach to model HSDR was proposed and implemented to estimate the long-term averaged spatially-distributed HSDR over the entire MDB. HSDR, a measure of catchment response to the upland erosion rate, is modeled by two lumped linear stores arranged in series: hillslope transport to the nearest streams and flow routing in the channel network. A theory developed in hydrologic scaling is adapted here to relate long-term averaged sediment delivery to the effective rainfall duration and catchment sediment residence time (SRT). Average rainfall intensity and effective duration were regionalized by using long-term measurements from 195 pluviograph sites within and around the Basin. SRT is estimated using time of the concentration for overland flow multiplied by an enhancement factor which is a function of particle size. The model was implemented across the MDB by using spatially distributed soil, vegetation, topographical and land use properties under a GIS environment. The report is organised as follows: Section 2 briefly describes some general characteristics of the MDB in terms of climate, topography and soils which have major impact on erosion and sediment transport processes. A set of terminology is given for clarity of later spatial description and interpretation of results. Section 3 presents the methods used for the modelling with emphasis on hillslope sediment delivery ratio (HSDR) and analysis of rainfall 10

12 intensity data (6-min. interval pluviograph data). The main results are presented in Section 4. Discussions and conclusions are given in Section 5. 2 Background 2.1 Characteristics of the MDB The Murray-Darling Basin covers an area of km 2 or about 14% of Australia. The Basin includes the three longest rivers in Australia. The Darling is 2,740 km long from its source in the north to its confluence with the Murray at Wentworth, the Murray is 2,530 km long from its source in the Australian Alps to its mouth on Encounter Bay in South Australia, and the Murrumbidgee is 1,690 km long. As a semi-arid country with relatively high economic dependence on agricultural revenue, the MDB is of national economic importance with rich irrigation, farming and grazing land. The Basin accounts for 40% of Australia s agricultural production, utilizing about 70% of all water used for agriculture across the nation. The 1,500,000 hectares under irrigation for crops and pastures represents 70% of the total area under irrigation in Australia. More than 80% of the divertible surface water resource is consumed in the Basin. The Basin holds a population of 2 million people, which is about 10% of the national population Climate There are a range of climatic conditions across the Basin, with cool humid conditions on the eastern uplands, and sub-tropical conditions in the northeast. The climate to the southeast is temperate, while the large western plains are semi-arid and arid areas. Annual precipitation in the Basin ranges from 185 mm to 2,500 mm. The potential evaporation rate is more than twice the precipitation rate. Mean annual evapotranspiration generally increases as rainfall decreases. Less than 10% of stream flow reaches the major rivers (Murray and Darling) and less than 5% of total rainfall is exported to the sea (Crabb 1997). The Basin has large inter-annual variability of the rainfall, mainly due to the impact of the El Nino - Southern Oscillation (ENSO) on the climate of southeastern Australia. This variability in rainfall is amplified in the annual runoff, which is more variable than runoff elsewhere in the world (except for parts of Southern Africa that experience a similar climate). These variations have profound effect on sediment delivery and transport in the basin Topography Combined with the mountains of the Australian Alps and steep hills and colluvial slopes of the Great Dividing Range, much of the Basin consists of the Murray-Murrumbidgee Riverine plain, the Darling floodplain and alluvial floodplains of other tributaries. The low relief over most of the Murray Basin occupies most of the area towards the arid west. Due to this topographic setting, the stream flow and sediment generated from the high rainfall areas are often dispersed or evaporated when the water reaches lowland floodplains. 11

13 2.1.3 Geology and Soil The spatial distribution and the properties of soils reflect the effect of climate, topography, flora and fauna acting on parent material over time. Organic soils are found at high altitude in alpine areas. Soils on steep mountain slopes, upper valleys and their terraces reflect the sequence of periodic erosion-deposition driven by tectonic activity and/or climate change (Butler et al. 1983). Soil thickness and horizon development are a function of the age of the deposits; buried soils are widespread (Rowe et al. 1978). Gradational soils of various levels of differentiation reflect the age of their parent material. Red and yellow duplex soils are widespread on rounded spurs, ridges, and hills, and dissected colluvial deposits (Rowe et al. 1978). On the alluvial deposits of the Riverine Plain, sediment texture and drainage control soil profile colour and development. A characteristic soil catena is associated with leveefloodplain transects of prior streams (Butler et al. 1983). Red-brown earths are found on levees, sandy on the crest of the levee and loamy on the backslope. Grey, brown, and red clays are extensive on the floodplains, with gilgai and soluble salt content increasing downstream and with distance transverse from the levee (Butler et al. 1983). Wind-blown parna mantles much of the Riverine Plain but is most prominent on foothills and hilly inliers (Butler et al. 1983). In the Darling alluvial plain, grey self-mulching clay soils derived from basalt are extensive (Butler and Hubble 1978) Land Use and Management Practices The major land use types in the Basin are dryland grazing (native pasture), cropping dominated by winter cereals, improved pasture, open forest, and agroforestry. The native vegetation is diverse with grassland, open woodland, woodland and shrubland environments and a very small area of dense vegetation growth in the eastern part of the Basin. The predominant land use is grazing but due to the economic benefits there is a shift towards cropping. As the high resolution land use and land management data sets currently only cover selected dryland areas as part of the Landmark project, 1-km resolution snap-shot land use data derived from 1996 NOAA LAC remote sensed images by Bureau of Rural Sciences (BRS 2001) was used for the other areas (see Figure AI.1 and Table AI.1 in Appendix I for detail). Figure 1 shows spatially distributed current land use classes in the Basin. In terms of management practices, burning was a widespread practice in the early 1970s throughout the Basin. In the 1990s, general observations suggest that stubble retention is much more common north of about Parkes, relative to southern areas. In the northern part of the NSW south-western slopes, a survey carried out by Vanclay (1997) suggested that about 50% of farmers burn stubble, and that burning tends to be associated with cultivation rather than direct drilling. In northern NSW, a survey by Martin et al. (1988) showed that the incidence of stubble burning ranged from 0% in dry years to 28% in years with disease problems. Burning generally occurred soon after harvest it removes approximately 90% of the stubble cover. The trend in tillage practice is for greater retention of crop residues and fewer tillage operations, especially in the northern part of the Basin. This helped not only to reduce erosion but also increased profits of the farmers. Those features of tillage and crop rotation were considered in our hillslope sheetwash and rill erosion modelling. 12

14 Typical pasture dry matter production in the Basin is around 8 10 t/ha DM in a good season down to 3 t/ha DM in a poor season. In severe droughts, dry matter production is negligible. Stocking rates tend to be in the range DSE/ha. In marginal western areas, suitable pasture systems for rotation with cereals have not yet been developed. In the Cobar-Nyngan- Walgett region, where most of the new development has occurred since 1970, both forest and open grassland have been converted to crop and pasture production (Swift and Skjemstad 2002). Figure 1: Major landuse groups in the MDB. 2.2 Spatial Settings and Terminology For modelling purposes, SedNet (Prosser et al. 2001; DeRose et al. 2003) spatially divided the MDB into around 10,000 sub-areas according to its topography using ESRI ArcInfo software (ESRI 2003) and 9 digital elevation model (DEM) derived by the Australian National University (Hutchinson et al. 2001). The sub-areas, which are constituted by many grid cells, are the basic constituent elements used to compute hillslope sheet and rill erosion, hillslope sediment delivery ratio, gully erosion, and bank erosion. Those sub-areas are called subcatchment elements and have contributing area around km 2. Grid cells are the basic constituent element for hillslope sheetwash and rill erosion modelling and the results are presented as the same raster GIS formant. For HSDR modelling, grid cells remain the basic constituent element but the results are presented at the sub-catchment element level. For clarity, in this report, we use the following terminology: Contribution point: This refers to the river export point for evaluation of suspended sediment contribution. Sometimes, it is called sediment control location. Sub-catchment element: It is the basic constituent element of SetNet model. Normally, each has a contributing area around km 2. There are nearly 10,000 subcatchment elements covering the MDB. 13

15 Sub-catchment: A group of sub-catchment elements. These equate to tributary rivers of catchments. E.g., the Cotter sub-catchment is a tributary of the Murrumbidgee catchment. Catchment: Refers to the major upland catchment areas and associated rivers, e.g., the Murrumbidgee Catchment. Basin: Refers to the Murray Darling Basin as a whole. SDR: The mean value of sediment delivery ratio from a sub-catchment element. It is also called HSDR in the other reports produced by this project. Hillslope Erosion: Refers to hillslope sheetwash and rill erosion only. 3 Methodology 3.1 Hillslope Sheetwash and Rill Erosion Mean annual soil erosion under current land use was predicted using the Universal Soil Loss Equation (USLE), a model of surface wash and rill erosion based upon factors of rainfall erosivity, terrain, soil erodibility, and vegetation cover. We mapped USLE factors from digital elevation models (DEMs), soil property maps, remotely sensed images and climate surfaces. Innovations were made in obtaining high resolution terrain properties from coarse resolution DEMs (Gallant 2001), and seasonal vegetation cover mapping from 14 yr of imagery (Lu et al. 2003c), and seasonal rainfall erosivity estimation using 20 yr of daily rainfall (Lu and Yu 2002). Technique details of the modelling process can be found in Lu et al. (2001). Further improvement on the estimation of erodibility (USLE K-factor), cover and management factor (USLE C-factor) and model validation can be found in Lu et al. (2003b). Table 1: Landuse groups used to calculate sheet and rill erosion rate. Groups Land use Descriptions 10 Built-up area 11 Perennial watercourse and lake, Mangrove, Reservoir, Saline, Coastal flat, Swamp 12 Non-perennial watercourse and lake 21 Closed forest 22 Open forest 23 Woodland 24 Commercial native forest production, Plantation fruit, Agroforestry, Apples, Citrus, Grapes, Stone fruit, Pears, Plantation 25 National Park 31 Cereals excluding rice 32 Legumes 33 Other non-cereal crops 34 Oilseeds 35 Non-cereal forage crops 36 Rice 37 Cotton 38 Potatoes 39 Sugar cane 14

16 40 Other vegetables 41 Nuts 42 Improved Pastures 51 Residual/Native Pasture This report provides an update on the previous assessment by Lu et al. (2001) by including improved vegetation cover estimation (Lu et al. 2003c) and high resolution land use data supplied by local agencies. Vegetation cover and land use data were used to calculate USLE C-factor. Compared with the snap-shot 1 km resolution land use coverage produced by the Bureau of Rural Sciences (BRS), we found that the locally supplied land use data are more consistent with long-term averaged vegetation cover estimated from remote sensing images. As locally supplied land use data only covers part of the Basin, the BRS 1 km land use map (BRS 2001) was used for the rest of the areas. Details of land use data sources are given in the Appendix I. All the landuse data are re-classified into 22 classes given in the Table 1 for sheetwash and rill erosion rate calculations. 3.2 Hillslope Sheet and Rill Erosion under Natural Condition To better understand the relative impact of land use and management practices on hillslope erosion, the predicted sheetwash and rill erosion needs to be put in the context of erosion under natural vegetation cover. We predicted natural erosion using a similar procedure as for modelling soil erosion under current land use conditions with a cover factor for native vegetation, keeping the other factors as for the present day. An empirical modelling framework to predict the pre-european settlement (undisturbed) USLE C-factor was implemented. Instead of directly using the results from NLWRA sediment delivery and transport project Theme 5.4b, the modelling work was redone for this project using an updated current C factor and an improved technique for sampling remnant native vegetation. There are two basic assumptions of this modelling framework: 1) climate, soil type, geology and terrain conditions remain unchanged since European settlement; 2) the natural vegetation and soil surface conditions remains similar to pre-settlement condition for those areas with limited disturbance. Based on those two assumptions, we sampled the C-factor from those areas with limited disturbance, built statistical models using climate, soil, geological and terrain variables as predictor variables and used the models to extrapolate to those areas with substantial disturbance by human intervention, especially agricultural activities such as cropping, grazing and tree clearing. Statistical models were constructed using the Cubist data mining tool (Rulequest Research 2001) in a similar way as we used for the predictions of hillslope length and slope (Lu et al. 2003b). In this study, we reserved a proportion of the sample set to test the model, calculating statistics of model performance for both the model-built data and test data sets. 50% of the total sampling points were used for model building and the other 50% of points for model testing. The sampling and modelling were carried out at 0.05 degree resolution. A stepwise model building approach was used. For the first step, each predictive variable was used independently and the best variable was identified on the basis of correlation coefficient and relative error. This variable was then combined with every other variable, to find the second variable that most improved the model. This procedure was repeated until all variables were included. Final selection of the model was based on the statistical diagnostics, and visual comparisons of predicted and measured maps. 15

17 The predictive variables for modelling the C-factor under pre-european conditions using Cubist were selected to represent the major factors presumed to determine vegetation cover and soil distribution across the continent. They can be broadly grouped into four categories: natural vegetation; soil parent material; climate; and geomorphology. Specifically, the following nineteen predictive variables were selected: (1) Australia - Natural Vegetation (Carnahan, J.A. and AUSLIG (1989) 1:5 M scale); (2) aggregated geology classifications derived from the 1:2.5M scale geology map of Australia; (3) the Australian Soil Classification derived from the Atlas of Australian Soils; (4) mean annual temperature, mean diurnal change, isothermality, temperature seasonality and diurnal temperature range; (5) mean annual rainfall, rainfall seasonality index, annual moisture index and moisture index seasonality; (6) mean annual radiation and radiation seasonality; and (7) 9 DEM, averaged slope and slope length derived from 9 DEM and relief, and their scaled estimations (Gallant 2001). 3.3 Sediment Delivery Ratio (SDR) Background of SDR Soil erosion models, such as the Universal Soil Loss Equation (USLE) (Wischmeier and Smith 1978) estimate gross soil erosion rate at plot-scale. Erosion rates estimated by USLE are often higher than those measured at catchment outlets. Sediment delivery ratio (SDR) is used to correct for this reduction effect. It is defined as the fraction of gross erosion that is transported from a given area in a given time interval and it is a measure of sediment transport efficiency which accounts for the amount of sediment that is actually transported from the eroding sources to a measurement point or catchment outlet compared to the total amount of soil that is detached over the same area above that point. Mathematically, it is expressed as Y SDR = (3.1) E where Y is average annual sediment yield per unit area and E is average annual erosion over that same area. It compensates for areas of sediment deposition that become increasingly important with increasing catchment area, and therefore, determines the relative significance of sediment sources and their delivery. Factors influence SDR including hydrological inputs (mainly rainfall), landscape properties (e.g., vegetation, topography, and soil properties) and their complex interactions at the land surface. The multitude of such interactions makes it difficult to identify the dominant controls on catchment sediment response and on catchment-to-catchment variability. In reality, erosion is not normally measured directly. It is measured as sediment yield at a small scale, such as a hillslope plot. Thus, SDR is a scaling factor used to accommodate differences in arealaveraged sediment yields between measurement scales. Physically, it stands as a mechanism for compensating for areas of sediment deposition that becoming increasingly important with increasing catchment area. Therefore, transport and storage lie in the heart of SDR. At regional scale, the most widely used method to estimate SDR is through a SDR-area power function: SDR = α A β (3.2) 16

18 where A is the catchment area (in km 2 ), the constant α and a scaling exponent β are empirical parameters (Maner 1958; Roehl 1962). Field measurements suggest that β is in the range to (Walling 1983; Richards 1993), which means that SDR decreases with increasing catchment area. The scaling exponent β contains key physical information about catchment sediment transport processes and its close linkage to rainfall-runoff processes. It seems that β decreases with increasing aridity (Richards 1993). Lower value of β (up to 0.7) were found in the Sicilian region and in former USSR catchments (Ferro and Minacapilli 1995). Field data (Figure 2) show that the relationships between SDR and drainage area changes considerably between different catchments over the world. Extrapolation of those empirical relationships can be misleading and results in SDR exceeding 100%. Figure 2: SDR vs catchment area relationships obtained from different areas around the world. For catchments with similar area, field data show the values of α and β in equation (3.2) are also different in different regions (Walling 1983; Roehl 1962). It is because the SDR-area relationship does not take into account local descriptors, such as rainfall, topography, vegetation, land use and soil characteristics. There are other empirical relationships which show that SDR varies with various physiographic attributes but the data that went into these relationships are few and of only local extent (Khanbilvardi and Rogowski 1984). This limits the usefulness of such a lumped empirical approach. Williams (1977) developed a procedure for determining SDR based on runoff models for small catchments. Recent development in this direction is towards the spatially distributed modelling using GIS techniques (Ferro 1995). There are other methods to predict sediment delivery and deposition through calculation of sediment transport capacity, avoiding the need for a lumped SDR (Morgan et al. 1998; Van Rompaey et al. 2001). Although those methods were based on improved physical understanding of sediment transport processes, they require high resolution DEMs to route the flow and sediment. They also rely on detailed sediment transport or runoff data to calibrate parameters, such as the sediment transport capacity coefficient. However, such methods often require many parameters which are generally too expensive or even impossible to determine reliably and the input data such as hydraulic resistance, infiltration rate, and soil properties including particle size distributions are not commonly available over large spatial extents. The traditional SDR methods are often data-driven. They depend on the existence of long periods of sediment yield records at the stream gauging stations and a sensible measure or estimation of hillslope erosion rate. However, there are few consistent long periods of 17

19 sediment yield data available in the MDB to allow such an analysis to be carried out. In addition, approaches based on analyzing sediment yield records cannot identify the separate effects of changing climate, land use and management practices on sediment delivery as catchment response to change is often longer than the record length. It is known that there are some limitations of SDR methods (Walling 1983; Richards 1993). One is that SDR methods cannot explicitly predict the locations and rates of sediment deposition in the lowland phases, and another is the problem of temporal and spatial lumping and lack of physical basis. However, SDR is a very useful concept to model regional scale sediment delivery processes. It avoids the need to explicitly model patterns of deposition on hillslopes which is not possible across such large areas as the MDB. There is little quantitative SDR information available within the Basin for the scale we are interested here. Existing measurements are either at much smaller or larger catchment scales. Studies based on sediment budgets carried out in forest areas of south-eastern New South Wales (NSW) and the East Gippsland show that the values of SDR are in the range of 10% - 45% for catchment areas of about 2 km 2 (Croke et al. 1999) and range from 2% - 95% for those sub-catchments (with areas of around 100 km 2 ) within the Bega Catchment (Fryirs and Brierley 2001). A SDR of 70% was found for the Upper Wolumlar Creek (area = 18 km 2 ), located in the South Coast of New South Wales (Brierley and Fryirs 1998). For the catchment area in which we are interested, most of the measurements and studies were carried out in humid areas outside of the Basin. Little has been done for the arid and semiarid regions. In summary, SDR is the result of numerous complex interactions among hydrological inputs (mainly rainfall) and landscape properties (e.g., vegetation, topography, and soil properties) through a number of hydrological processes at the land surface. The multitude of such interactions makes it difficult to identify the dominant controls on catchment sediment response and on catchment-to-catchment variability within the MDB. In addition, field measurement of SDR is severely limited. Therefore, it is difficult to model spatially distributed SDR accurately A New SDR Theory One important aim of this study is to develop a SDR model that incorporates the key elements of the catchment storm response and sediment delivery process. Sivapalan et al. (2001) showed that the interactions between time scales, namely between rainfall duration and catchment response lay at the heart of the regional flood frequency estimations. The way that catchment response time varies with catchment area depends on the relative dominance of hillslope response, channel hydraulic response, and network geomorphology. A simple linear model of catchment response (Sivapalan et al. 2001) is used in this study. Instead of using the model for studying catchment response of flood, we use the same concept to model SDR. The model consists of two independent components: sediment transport on hillslopes and sediment routing in the channel network. As shown in Figure 3, these are represented through two linear stores, arranged in series. The hillslope store is supplied with sediment by soil eroison at a rate e [mass/area/time] over an effective storm duration t er (erosion only occurs during this time period). The hillslope stores part of the eroded sediment and delivers the rest to the channel network store, located downstream of it, at a rate y h [mass/area/time]. y h is assumed to be a linear function of the mass of sediment stored in the hillslope per unit area, denoted by S h [mass/area]. The area specific sediment yield from the 18

20 network store, y [mass/area/time], which is the same as the area specific sediment yield from the catchment outlet, is assumed to be a linear function of the sediment stored in the channel network, denoted by S n [mass/area]. The continuity equation of sediment for the two stores can be expressed as: dsh() t = e( t) yh( t) dt yh() t = Sh()/ t th dsn() t = yh( t) y( t) dt yt () = S()/ t t n n (3.3) where t h is the mean hillslope residence time and t n is the mean channel residence time. e(t) S h (t) Hillslope Storage S () () h t yh t = t h S n (t) Channel Storage S () () n t yt = t n Figure 3: Diagram of a two storage lumped linear model of SDR at catchment scale (after Sivapalan et al. 2001, modified). See text for detail. For simplicity, we assume that the upland erosion rate e is constant during t er. Equation (3.3) can then be solved analytically. The final expressions for the ratio between the peak of the resulting sedigraph, denoted by y p [mass/area/time] (which is equal to max(y)), and upland erosion rate e can be written as follows: y p t n t er = 1 exp e tn th tn t h t er 1 exp t 0 t t tn th th 2 3 y p 1 t 1 er t er =... t n = t e 2 tn 3 tn h n h h (3.4) On an event basis, we assume SDR = y p / e. The peak sediment yield Y p [mass/time] can be estimated by multiplying area specific sediment yield y p [mass/area/time] by the catchment area A. Equations were firstly derived by Sivapalan et al. (2001) for studying the scaling effects on regional flood frequency under different rainfall and catchment conditions. 19

21 Sivapalan et al. (2001) showed that equation (3.4) is capable of explaining the power law relationship between flow response and catchment area and changing value of the scaling exponent which is caused by a change of hydrological processes. Similarly, equation (3.4) can be used to explain the obtained SDR vs area relationships. As shown in Figure 4, SDR measurements gathered by Roehl (1962) in several American catchments including Blackland Prairies, the Red Hills of Texas and Oklahoma, the Missouri Basin Loess Hills, the Mississippi Sand Clay Hill, and the South-eastern Piedmont (shown in dots) suggested that, in general, SDR decreases with catchment area. The solid line, which is the average flow response (the scaling factor of mean flood discharge defined as the ratio between average rainfall input rate and runoff at the catchment outlet during flood events) calculated using the equation (28a) of Robinson and Sivapalan (1997), represents the upper envelope of SDR. The averaged modeled SDR estimated by equation (3.4) is shown as the dashed line. The reason that SDR is often smaller than flow response is due to the settling velocity of soil particles (compared with water particles) and other effects such as sediment transport capacity. For a given catchment area, the variations in SDR measurements (by up to two orders of magnitude) are due to heterogeneity in catchment properties (e.g. rainfall, catchment slope and curvature, soil texture, etc). The combination of the above physical properties results in differences in the time variables t er, t n and t h in eq. (2). Therefore, eq. (2) can be used to model spatially distributed SDR if the time variables t er, t n and t h can be spatially differentiated SDR (Roehl 1962) SDR (modelled) Flow Response 100 SDR Area (km 2 ) Figure 4. Comparison of SDR (%) measurements (Roehl 1962), modeled average SDR and flow response (Robinson and Sivapalan 1997). It shows that flow response represents the upper envelope of the SDR. Equations (3.3) and (3.4) can be used to compute the magnitudes of the SDR for different values of the timescales t er, t h and t n. The results are presented in the Figure 5 (upper panel) as families of curves relating SDR to t n for different values of t er and t h. They show that the SDR remains constant for small values of t n, while for larger values of t n they decrease linearly with increasing values of t n (scaling exponent is -1). The effect of t h is to smooth and reduce the magnitude of SDR, without changing the scaling exponents with respect to t n. 20

22 1 0.1 SDR t n (hrs) t er = 1, t = 50 t = 1, t = 5 t = 1, t = 0 h er h er h t er = 5, t = 50 t = 5, t = 5 t = 5, t = 0 h er h er h t = 10, t = 50 t = 10, t = 5 t = 10, t = 0 er h er h er h 1 Measurements 0.1 SDR E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09 A (km 2 ) Figure 5: SDR as a function of channel residence time for different values of t er and t h (upper panel); SDR as a function of catchment area for different values of t er and t h. SDR measurements from USA catchments (Roehl 1962) are also shown as red dots (lower panel). Equations (3.3) and (3.4) can also be used to explain the SDR and area relationship observed from measurements. Assume that t h is independent of catchment area for the scale of catchment shown in Figure 5 and t n can be expressed as a function of catchment size A of the form: tn = α A β (3.5) where t n is in hours, A is in km 2. Assuming that parameters α = 0.76 and β = 0.38 (ARR 1987) Figure 5 (lower panel) shows SDR versus catchment area relationships. It shows that for a given catchment area, SDR values vary by three orders of magnitude due to the differences in t er and t h. While the SDR, in the sense of equations (3.3) and its solution (3.4) remains linear for all catchment sizes, the observed change in scaling exponent is cause by a change of hydrological 21

23 processes. In small catchments, effective rainfall duration is long compared to the catchment residence time, and consequently the sediment transport reaches steady state, with the whole catchment area contributing to the sediment yield. As long as this remains true, the sediment yield increases linearly with catchment area with the exponent at unity. On the other hand, in large catchments, effective storm duration is smaller than the catchment residence time. The fraction of the catchment area contributing to the sediment yield is proportional to the ratio of effective storm duration to catchment residence time. This ratio decreases at the rate of A -β with an increase of catchment area A. Thus the partial area contributing to sediment yield increases only at the rate of A 1-β, with exponent 1-β less than unity. The above analysis was based on a single storm. The derived flood frequency method can be used to deal with multiple storms. In this study, for simplicity, we treat effective storm duration t er as a random variable and calculate t h and t n as catchment averaged values. According to equation (3.4), SDR becomes a random variable. By knowing the probability distribution of t er, we can derive a probability distribution of SDR. According to standard statistical procedures, we can calculate the mean values of SDR. To apply the model to the whole MDB, we need to estimate two groups of input variables: 1) statistical properties of effective rainfall duration and intensity using pluviograph rainfall data; and 2) hillslope and channel residence time t h and t n as a function of particle size and other catchment properties. The following sections describe the procedures for estimating the two groups of input variables. 3.4 Statistical Analysis of Effective Rainfall duration and Intensity Rainfall varies both in space and time. Both rainfall duration and intensity are important and interrelated, and have significant impacts on sediment generation at the point scale (source) and sediment transport at catchment scale. The main aim of analyzing high temporal resolution rainfall data is to discover the possible controls on the spatial variability of sediment delivery due to temporal variability of rainfall intensity. Adelaide # # Brisbane # Sydney # Canberra # Melbourne Figure 6: Site locations of pluviograph rainfall data and their relative position to MDB. 22

24 Pluviograph rainfall data with 6-min sampling interval was collected for 195 sites from the Bureau of Meteorology (BoM). The site selection is based on two conditions: 1) that sites are within or nearby the MDB; and 2) they have at least 10 years of rainfall record. Among those 195 sites, four sites which have a large number of missing values were eliminated. The locations of the gauges and their relative position to the MDB are shown Figure 6. More detailed site information can be found in Table AII.1 of Appendix II. The analyses of rainfall data is divided in two parts: 1) a statistically analysis of rainfall data at a single site to search for a suitable probability distribution function; and 2) regionalization of the parameters of the suitable probability distribution function. Search for a suitable probability distribution function The storm events, characterized by their intensity and duration, are estimated by statistical analyzing pluviograph rainfall data, with temporal resolution of 6 minute interval. In this study, the storm events are defined as rain periods separated by dry periods of at least 6 hours or longer. Once a storm is defined, the 30 minute rainfall intensity and storm duration can be estimated. The events which have total rainfall depth equal or larger than 12.7 mm are considered to be potentially harmful in terms of erosion and sediment transport. Therefore, only those events are included in the calculation. The value of 12.7 mm is chosen here to be consistent with that used in the Universal Soil Loss Equation (USLE) (Wischmeier and Smith 1978). In addition, an event which has total rainfall depth equal or larger than 12.7 mm does not necessarily cause erosion during its whole rainfall duration. For those events with low intensity but long duration, the effective duration in terms of causing erosion is shorter than the rainfall duration. To consider this effect, we calculate the effective duration for an event for a given site using the following equation R t = min t t R j j= 1 i er, i N ri,, ri, (3.6) where R i is the R factor for i-th event, N j= 1 R j is the sum of R factor for all the events, and t r,i is the duration for i-th event. Equation (3.6) simply states that the effective duration is shorter for events with smaller erosivity compared to events with larger erosivity, and the effective duration cannot exceed the actual rainfall duration. The computer program of Yu and Rosewell (1998) was modified for the calculations of rainfall duration and intensity. Figure 7 to Figure 9 show the event 30-intensity and duration for the Wagga Wagga site (upper panels). Figure 7 shows all the rainfall events. It shows that many events have small 30-min intensity and duration. Those events have little effect on sediment generation and transport and can be excluded from the analysis. This exclusion is done by only including the events which have total rainfall depth equal or greater than 12.7 mm as suggested by USLE. Figure 8 shows the events after excluding those small events. The events in which the effective duration is calculated using equation (3.6) are shown in Figure 9. 23

25 Figure 7: All rainfall events characterised by their 30 intensity and duration (upper panel); Fit probability density functions to Gamma and exponential distributions for both duration and intensity (second and lower panels). 24

26 Figure 8: Rainfall events which have depth equal or greater than 12.7 mm (upper panel); Fit probability density functions to Gamma and exponential distributions for both duration and intensity (second and lower panels). 25

27 Figure 9: Effective rainfall events which have depth equal or greater than 12.7 mm (upper panel); Fit probability density function of effective duration to Gamma and exponential distributions (lower panel). Figure 7 to Figure 9 show that the Gamma distribution fits the data better than the exponential distribution for both intensity and duration (lower two panels). However, due to the awkwardness of operating with the Gamma distribution in analytical form and the reasonable fit of the exponential distribution for effective rainfall for larger events, we decided to use the exponential distribution in this study. The exponential distribution density function is written as 1 x f( x) = exp λ λ (3.7) where x is the random variable, λ is the mean value of the random variable x (in this study, x can be 30-min rainfall intensity, rainfall duration or effective rainfall duration). For each rainfall site, the mean value λ is obtained by averaging values across all the events considered. To apply the statistical estimation of SDR using the exponential distribution (3.7) to the whole Basin, we need to regionalize the mean value λ for both intensity and duration. Regionalization of the Mean Values of Intensity and Duration To apply the SDR model cross the Basin, we need to regionalise the mean values of rainfall intensity and duration by linking them to the existing climatic surfaces. The regionalization is done as follows. 26

28 As shown in Figure 10, a regression relationship with r 2 = 0.98 is obtained between MI30 and the ratio between mean annual R factor (R) and mean annual rainfall (MAR). Good relationships are obtained for the mean values of rainfall duration (t r ) and the effective duration (t er ) in relation to the ratio between mean annual rainfall (MAR) and mean 30- min maximum intensity (MI30) (Figure 11 and Figure 12, respectively). As we have already regionalized R and MAR, those relationships shown in Figures 10 to 12 are used to regionalise t r, t er and MI30. Note that only the events with rainfall depth equal or greater than 12.7 mm are considered in the calculation. Figure 10: Relationships between effective 30-min. rainfall intensity and the ratio between mean annual R-factor and mean annual rainfall. Figure 11: Relationships between rainfall duration (t r ) to mean annual rainfall (MAR), effective 30-min intensity (MI30), MAR/MI30, and MAR 2 /R. 27

29 t (hr) er er t (hr) MAR (mm) MI30 (mm) y = x (r = 0.84) Relative Error MAR/MI t er (hr) Figure 12: Relationships of effective rainfall duration and it relative errors. Figure 13 shows the errors between rainfall duration and rainfall duration estimated using regionalisation equations derived in this study for all the rainfall sites. It suggests that the prediction accuracy increases as the number of years with complete records increases. This is consistent with the USLE which suggests at least 22 years pluviograph records are appropriate for long term erosion estimation. Relative larger errors occur mainly at the sites which have shorter rainfall record. Larger errors are observed at some sites within the high rainfall regions (Australian Alps), which might suggest more complex and non-linear rainfall patterns in those regions. 28

30 40 t (Simulated) r t (Calculated from Pluviograph data) r Absolute Error (hr) Number of Year with Complete Data 1 Relative Error Number of Year with Complete Data Figure 13: Error estimations of rainfall duration. Upper Panel: Comparison between rainfall duration estimated using site specific pluviograph data and that estimated using regionalised relationships. Middle Panel: Absolute error [hrs] plotted against number of year with complete data. Lower Panel: Relative error plotted against number of year with complete data. The crosses are the sites have shorter records and relatively larger errors. They are not used in the final relationships that are applied across the MDB. 29

31 3.5 Estimations of Residence Time Sediment Residence Time as a Function of Particle Size The residence time of sediment can be estimated as a function of sediment particle size and the travel time of water particles. Trajectory of Water Particle Fine Particle Suspension Silt Particle Suspension and Saltation Sand Saltation Figure 14: Diagram of the particle size effect on sediment travel time in relation to the travel time of water particles. Suppose we can estimate the travel time of water particles as a function of local slope, roughness, rainfall intensity, etc. The effect of sediment particle size can be reflected as shown in the diagram in Figure 14. Very small clay particles, which are characterized by their slower settling velocity, remain suspended in the water most of the time and their trajectories of travel differ little to that of water particles. Silt particles with faster settling velocity travel with water particles during high velocity flows and settle to the soil bed during low flows. Large sand particles saltate near the soil bed with slow overall velocity. The travel time for different size particles within flowing water was modelled as follows: t ( d) = t F ( d) h h0 h t ( d) = t F ( d) n n0 n (3.8) where t h (d) and t n (d) are the hillslope and channel residence time for particles with diameter d, respectively, and t h0 and t n0 are the hillslope and channel travel time of water particles, respectively. F h (d) and F n (d) are the enlargement functions describing the influence of particle size d. The mathematical forms of F h (d) and F n (d) were modelled as: F F ( γ w d ) ( γ w d ) = exp ( ) h h t = exp ( ) n n t (3.9) where w t (d) is the settling velocity for particles with diameter equal to d, and γ h and γ n are the parameters inversely relating to water depth. In general, γ h is larger than γ n as the typical water depth in overland flow is of order of millimetres and the water depth in small channels is of order of centimetres. The settling velocity was calculated as: 30

32 4ρ pgd wt ( d) = 3 ρcd(re p) 1/ 2 (3.10) where ρ p is the particle density, ρ is the water density, g is acceleration due to gravity, Re p = w t d/ν is the particle Reynolds number at the settling velocity, and C D is the drag coefficient modelled as a function of the particle Reynolds number Re p : (Durst et al. 1984) C D(Re p) = ( Re p ) (3.11) Re p Finally, SDR was calculated for each particle size group and then weighted by the particle size distribution to get an overall SDR as follows: SDR = N i= 1 i N i= 1 w = 1 w SDR i i (3.12) where N is the total number of particle groups, w i and SDR i are the mass percentage and SDR for particle group i, respectively. Three particle size groups are considered in this study. These are: d 2 µm (clay), 2 d 20 µm (silt), and 20 d 1000 µm (sand). Particles with diameter larger than 1000 µm are considered too large to be transport far away from their source areas. The mass percentage of each particle size group was estimated using the Australian Soil Resource Information System (ASRIS) product (Carlile et al. 2001) Estimating Travel Time of Water Particles t h0 and t n0 Novotny and Olem (1994) pointed out that land cover and slope are the key factors in affecting sediment delivery rates. Additionally, they stated the importance of factors specific to storm events, such as rainfall intensity, infiltration, ponding, and overland flow energy. However, because this research used average annual erosion rate, consideration of detailed infiltration, ponding and storm specific factors was not feasible. The travel time of water particles is calculated separately for overland flow and stream flow. The travel time is inversely related to flow velocity. During a storm event when overland flow occurs, the flow carries sediment from surface runoff until it reaches a stream. In the stream component, the runoff water is influenced by a different set of factors affecting travel-time compared to that of the overland component. To capture this, the travel time of channel flow is calculated from each cell in the catchment to the outlet by aggregating stream segments. Along each path, travel-time is calculated by aggregating time taken within each cell using procedures described below. 31

33 Overland flow component: For the hillslope cells, the overland flow velocity is estimated by combining a kinematic wave approximation with Manning s equation. The depth of flow at equilibrium (m) is given by (Overton and Meadows 1976): 0.6 e 0.5 ni L y = s (3.13) where L is the travel distance along the flow path (m), n is Manning s roughness coefficient, i e is the rainfall excess rate (mm/s), and s is the decimal slope. By substituting the depth of flow at equilibrium in Manning s equation, the velocity of overland flow (m/s) can be calculated as: ( iel) s Vo = (3.14) 0.6 n The travel time (s) through each cell can be estimated as: D t o = (3.15) V o where D is the distance travelled through that cell (m). For orthogonal flow, the flow distance is the cell width, while for diagonal flow, it is equal to 2 D. To calculate travel time by implementing the above procedure, four input parameters are needed in the overland component: rainfall excess rate i e, Manning s coefficient n, flow travel length L, and slope s. Estimations of those input parameters are made as follows. Estimation of Excess Rainfall Rate i e : Excess rainfall generated in a catchment is known to vary spatially. The variation in excess rainfall follows that of land use, land cover, and soil type. Typically, the way to account for this variation is to divide the catchment into smaller areas of uniform land use, land cover, and soil type combinations. An average curve number (CN) for the whole catchment determined using the area weighting method is then given by: CN A+ CNA CN A m m CN = m A i= 1 i (3.16) where CN i is the curve number of the sub-area i (with area equal to A i ). m is the total number of sub-areas. This procedure is the standard procedure used in the USDA SCS rainfall-runoff relationship (SCS 1983). It gives an average excess rainfall depth for the entire catchment, P e that corresponds to an average rainfall depth, P. The equations used to calculate P e are: P e ( P 0.2 S) = P + 0.8S 2 (3.17) where S is the storage term (in mm) which can be obtained using the formula: 32

34 254 S = (3.18) CN where CN is the curve number that can be obtained from standard tables for different combinations of land use and land cover, soil hydrologic group, treatment, and conditions. The hydrologic soil group reflects soil permeability and surface runoff potential. Following is a description of the four different hydrologic soil groups: Group A are soils with low total surface runoff potential due to their high infiltration rates. They consist mainly of excessively drained sands and gravels. Group B are soils with low to moderate surface runoff potential. They have moderate infiltration rates and moderately fine to moderately coarse texture. Group C are soils with moderate to high surface runoff potential. They have slow infiltration rates and moderately fine to fine textures. Group D are soils with high surface runoff potential. They have very slow infiltration rates and consist chiefly of clay soils. Typical values of CN for certain land use groups are given in Table 2. Table 2: Typical values of CN for some land use group. Sources of CN: SCS (1983; 1986), Novotny and Olem (1994) 33

35 Once the spatially-distributed CN map is developed, the total storage can be obtained by equation (3.18). The excess rainfall equation (3.20) gives the accumulated depth of excess rainfall from the start of the storm to the current time. For an unsteady rainfall/flow event, the incremental value of excess rainfall of a time interval t, i e can be calculated as the difference between the accumulated excess rainfall at the end of that time interval and the accumulated excess rainfall at the beginning of the that same interval as follows: ie ( t) = Pe ( t) Pe ( t 1) (3.19) In this study, we assume steady-state rainfall, we calculate i e as: i e Q = (3.20) t r where t r is the rainfall event duration and Q = P e is the total excess rainfall for the event. The procedure of estimating t r has been given in Section 3.4. Estimation of Manning s Roughness Coefficient, n: For simplicity, Manning s n roughness coefficient is estimated using available land use and landcover data. Table 3 shows estimated typical values of n for overland flow. Table 3: Values of Manning s n used in this study for common land use and vegetation cover groups for overland flow. Land use Veg. cover (c v ) c v 30% 30% < c v 70% c v > 70% Annual (not managed) Pasture Sow (improved) Pasture Crop Forest Built-up areas Wetland and ponds Estimation of Travel Length and Slope These parameters can be extracted from a digital elevation model (DEM) by using a geographic information system (GIS). As the 9 DEM is used in this study, the resolution of the DEM is not capable of capturing the overland flow path in detail. In the implementation of equations (3.14) and (3.15), flow length L and the distance of travel D are approximately equal to hillslope length, which is a product of NLWRA sediment transport and delivery project (Gallant 2001). Like hillslope length, slope grid s is also a product of NLWRA sediment transport and delivery project. Both grids were statistically derived using higher resolution DEM, 9 DEM and other climatic, geology, and soil attributes (Gallant 2001). 34

36 Channel component: The travel time in the channels can be calculated based on the SCS flow velocity equation (Haan et al. 1994) Vch 1/ 2 = as (3.21) where V ch is flow velocity [m/s], s is the slope [m/m], and a is a coefficient relating to stream roughness condition. Landuse Soil hydrologic Group Representative Rainfall Event DEM Manning s n Curve Number Slope Flow direction Rainfall Excess Volume Flow Accumulation Rainfall excess intensity Delineate Channel Network Flow length Flow Velocity Overland component Channel Component Calculate travel time for each cell by dividing the travel distance by the flow velocity Calculate the cumulative travel time Figure 15: Flow chart for the calculation of travel time of water particles. Similar to overland flow, the travel time (t c ) through each channel cell can be estimated as: D t c = (3.22) V ch where D is the distance travelled through that cell (equal to horizontal, vertical or diagonal distance across a cell flow direction). For a given cell i, the cumulative travel time was estimated by summing the travel time along its flow path. More specifically, if a sediment particle in cell i travels through m o cells overland and m c cells in the stream to reach the catchment outlet, equations (3.14) and (3.15) were used in each of the m o upland cells to calculate the concentrated shallow flow travel time and equation (3.22) was used in each of the m c stream cells and aggregated to estimated total 35

37 stream flow time (T ic ). Figure 15 shows the overall procedure for calculating travel-time t h0 and t n0. Two input parameters are needed for the channel component: slope s and channel roughness parameter a. The channel roughness parameter a is parameterised as in Table 4. Table 4: channel roughness parameter a values used in this study. Channel Section Upstream Area (ha) A Concentrated shallow flow Intermittent stream (grass waterway) Permanent Stream (little cover) and up Results 4.1 Hillslope Erosion under Current Land Use Figure 16 shows the predicted sheet and rill erosion across the Basin. In general, it is predicted that erosion rate increases from south to north and from west to east. The major source areas are: Brigalow Belt, New South Wales South West Slopes and north part of Darling Riverine Plains. It was predicted that about tones of soil is moved annually on hillslopes over the Basin. The average erosion rate across the basin is 2.1 t ha -1 yr -1. If we denote that a pixel with soil loss rate below 0.5 t ha -1 yr -1 as low erosion, larger than 10 t ha -1 yr -1 as high erosion, and in between as medium, it is estimated that about 40% of the Basin experiences low erosion, 4% faces high erosion and 56% of the Basin experiences medium hillslope erosion. Agricultural lands in steep and higher rainfall intensity areas experience higher erosion rate than other land use groups, showing the potential to target erosion control. Table 5 shows the percentage erosion for those three groups and in relation to percentage agricultural lands in each group. Table 6 divides hillslope erosion into land use classes. In general, the average erosion rate is higher for agricultural lands compared with other land use groups though many of agricultural lands are located in floodplains where the slope is low. The rates are higher compared with surrounding non-cropping areas where other conditions are similar. This confirms that land use and management practices have a major impact on soil erosion. 36

38 Figure 16: Estimated annual average sheet and rill erosion rate. Table 5: Three erosion groups (high, medium and low) and their relation to percentage of agricultural lands. High Erosion Rate (> 10 t/ha/year) Percentage in Basin Area (%) Area (10 3 km 2 ) 4 42 Percentage of Agr.Lands (%) Area (10 3 km 2 ) Medium Erosion Rate ( t/ha/year) Percentage in Basin Area (%) Area (10 3 km 2 ) Percentage of Agr.Lands (%) Area (10 3 km 2 ) Low Erosion Rate ( < 0.5 t/ha/year) Percentage in Basin Area (%) Area (10 3 km 2 ) Percentage of Agr.Lands (%) Area (10 3 km 2 ) *Total Basin area: 108 million ha. Lakes and reservoirs are not included for erosion statistics. Agricultural lands in the Basin: 17.3 million ha (around 16% of the Basin area). Figure 17 shows the monthly distribution of total soil loss. It is found that over 75% of the erosion occurs in the summer period, especially in the north part of the Basin. However, the high erosion zone detected within the east part of the Basin, shows weaker summer dominance due to its temperate climate condition. 37

39 Table 6: Soil loss rate from land use categories. Landuse Approx. Total Area Total Erosion Ave. Erosion Rate Rate of acceleration Group (km2 * 10^3) (Mt yr-1) (t ha-1 yr-1) ratio of current and natural rates National park Woodland Plantation Forest Residual/Native Pastures Improved Pastures/Legumes Cereals excluding Rice Other agricultural lands Total Soil Loss (Mt/month) Jan Feb March April May June July Aug Sep Oct Nov Dec Month Figure 17: Monthly distribution of total soil loss rate for the Basin. 4.2 Hillslope Erosion under Natural Conditions The best predicting variable is the mean annual soil moisture index, which explains 72% of the variance of the sampled data, followed by the annual mean radiation explaining 59% of the variance of the sampled data. The correlation to polygon based data, such soil, and geology are lower (around 9% to 20%) and they improve little to the overall needed correlation coefficient or spatial patterns of predicted maps. The final predicting variables are clim1, clim2, clim3, clim4, clim7, clim12, clim15, clim20, clim23, clim28, clim31 and austdem. Figure 18 shows the comparison between samples of C values extracted from the current C map for those undisturbed (or minimum disturbed) points and modelled C values using Cubist for the testing data from the final model. 38

40

41

42

43

44

45 5 Discussions and Conclusions Hillslope sheetwash and rill erosion: There are three dominant forms of water-borne erosion in the Murray-Darling Basin. These are sheetwash and rill erosion (sometimes termed hillslope erosion in the reports produced by this project), the formation and erosion of gullies, and the erosion of riverbanks. In this study, new assessments of hillslope erosion across the MDB were reported, building upon our previous work for the National Land and Water Resources Audit (NLWRA) (Lu et al. 2001; Lu et al. 2003b). Improvements to the assessment of sheetwash and rill erosion were made by compiling higher resolution land use data for the MDB from a range of sources and by incorporating a database on crop rotation, tillage and other land management practices. These new data, together with improved analysis of remote sensing data, enabled a more accurate prediction of the effect of vegetation cover and cover management on hillslope erosion. It is estimated that tonnes of sediment were moved in the MDB annually as hillslope erosion at a mean rate of 2.1 t ha -1 yr -1. Erosion rate increases from south to north and from arid areas to temperate regions with most of the erosion generated from the east and north part of the MDB. Under any given rainfall regime, the reduction of protective ground cover increases the risk of high soil losses. About two-thirds of erosion occurs in the summer period. Agricultural lands have relatively high erosion rates and higher increment of soil erosion rates. Very low soil erosion rates are estimated under pre-european natural vegetation conditions. The rates are 3 10 times on average and up to 100 times smaller than that under current land use. A New Theory for Modelling Spatially Distributed Sediment Delivery Ratio: In this report, we have proposed a theory for sediment delivery ratio and implemented the theory across the MDB in a spatially distributed manner. Spatially, sediments are produced from different sources distributed throughout the Basin. Each source is characterized by its sediment detachment, transport and storage. The SDR model argues that sediment delivery can be closely linked to temporal hydrological control. For each source area, SDR is characterized by two important time variables, namely, its travel time, i.e., the time that particles eroded from the source area and transported through the hillslope conveyance system take to arrive at the channel network and eventually to the catchment outlet, and the typical rainfall duration, which is the primary driving force of sediment transport. For instance, for the same rainfall event, we expect that a source area with a shorter travel time would have a higher SDR. Alternatively, for the same source area, a rainfall event with shorter duration would have a lower SDR as less eroded particles would make their way to the catchment outlet. Those particles will be stored (or deposited) somewhere in the system. These interactions between rainfall attributes (including intensity, duration and intermittency) and catchment characteristics are important factors for understanding spatially distributed sediment delivery. For the arid part of the Basin, rainfall events are often smaller in size spatially with shorter duration but more intense than in humid temperate climates. For a given slope steepness and slope length, local erosion rate in the arid areas is relatively high due to insufficient vegetation cover and relatively more intensive rainfall. However, the sediment delivery follows a different pattern. The shorter rainfall duration and larger variations in interannual rainfall also cause a greater variation in sediment transport. The sediment yield differs from one catchment to another depending upon whether the storm duration is larger or smaller than the sediment residence time (SRT) of the catchments. The SDR model proposed 44

46 in this study is able to differentiate the catchments for which storms usually last longer than the SRT or for those for which residence time is seldom met. The SDR model allows quantitative estimates of the non-linear effects on sediment delivery due to changes in climate and land use. It expresses the spatial variability of catchmentaveraged SDR in terms of the statistical time variables and particle size distributions. It relies on rainfall intensity (6-min interval) and daily rainfall records (which cover larger areas) instead of stream flow records. It offers a means to understand the dominate processes which control sediment delivery. The model has a simple analytical form which can be implemented in a GIS environment. Applying the model to the MDB, we found: 1) sediment delivery ratio and sediment yield are low for most parts of the Basin except some upland areas in the east and north part of the Basin; 2) the sediment transport can be very effective at sub-catchment level, especially in the areas of the Australia Alps, South West Slopes, Brigalow Belt South, and Darling Downs; 3) only about 5% of sheet and rill erosion are transported from sub-catchment elements in to the streams. The average area specific sediment yield at sub-catchment element level is around 0.1 t ha -1 yr -1. About 14 million tones in total of sediment generated from sheet and rill erosion is delivered from the sub-catchment elements to the major streams. The quantitative, spatially distributed estimations of SDR have important implications not only for the study of off-site environment impact due to exported sediment but also to on-site erosion control. It has been demonstrated that there is economic advantage from identifying the areas that have a higher potential to deliver sediment and prioritizing control implementation in those areas (Dickinson et al. 1990). The spatially distributed SDR map contributes to the development of cost-effective strategies for erosion control (Lu et al. 2003a). SDR and Sediment Yield due to Hillslope Erosion: In summary, it is found that sediment delivery ratio and sediment yield are low for most part of the Basin except some sloping land in the eastern part of the Basin. Estimated at Basin outlet, spatial patterns of topography, rainfall intensity and rainfall duration suggests the system is not effective in terms of sediment transport. However, the sediment transport can be very effective at sub-catchment scale, such as in the areas of South West Slopes, Brigalow Belt South, and Darling Down regions. Average area specific sediment yield from subcatchment is 0.13 t ha -1 yr -1. On average, about 5% of sheet and rill erosion is transported from sub-catchment elements in to the streams. In total, around 14 million tonnes per year of sediment generated from hillslope sheet and rill erosion is delivered from the sub-catchment elements to the major streams. The Australian Alps have relatively high sediment delivery ratio though the local erosion rate is medium. Caution is recommended for any vegetation clearance in those ranges. Acknowledgments We acknowledge with appreciation the financial support for this work from the Murray Darling Basin Commission, and the personal support from the MDBC Office particularly from Ms. Lisa Robins. The project steering committee led by Dr. Pat Feehan are thanked for their efforts and time to oversee and to guide the progress of the work. 45

47 We thank Bofu Yu for supplying the source code of RECS, which was modified to calculate 30-min rainfall intensity and duration for each site with rainfall record. We also thank the local agencies who kindly supplied us the land use data. The interactions and discussions with colleagues within and outside the team are gratefully acknowledged. Individuals include Elisabeth Bui, Greg Cannon, Francis Chiew, Barry Croke, Mick Fleming, John Gallant, Tony Jakeman, Russell Mein, Neil McKenzie, David Simon, David Smiles, and Bill Young. 46

48 References ARR (1987) Australian Rainfall and Runoff. A Guide to Flood Estimation Volume 1, D.H. Pilgrim (ed.), Institution of Engineers, Australia. Brierley, G.J. and Fryirs, K. (1998) A fluvial sediment budget for upper Wolumla Creek, South Coast, New South Wales, Australia, Australian Geographer, 29, BRS (2001) Land use mapping at catchment scale: Principles, procedures and definitions. Bureau of rural Sciences, Canberra. Butler, B.E., Blackburn, G. and Hubble, G.D. (1983) Murray-Darling Plains (VII). In: (Eds.), Soils: An Australian viewpoint. pp Butler, B.E. and Hubble, G.D. (1978) The general distribution and character of the soils in the Murray-Darling River system. Proceedings of the Royal Society of Victoria. 90, Carlile P, Bui, E., Moran, C., Minasny, B., and McBratney A.B. (2001) Estimating soil particle size distributions and percent sand, silt and clay for six texture classes using the Australian Soil Resource Information System point database. Technical Report 29/01, CSIRO Land and Water, Canberra. Carnahan, J.A. and Australian Surveying and Land Information Group (1989) Australia - Natural Vegetation. Crabb, P. (1997) Murray-Darling Basin Resources. Murray-Darling Basin Commission, Canberra. Croke, J., Wallbrink, P., Fogarty, P., Hairsine, P., Mockler, S., McCormack, B. and Brophy, J. (1999) Managing Sediment Sources and Movement in Forests: The Forest Industry and Water Quality, Cooperative Research Centre for Catchment Hydrology Industry Report 99/11. DeRose R.C., Prosser I.P., Weisse M., and Hughes A.O. (2003) Summary of sediment and nutrient budgets for the Murray-Darling Basin. Technical Report K to Murray-Darling Basin Commission, Basin-wide mapping of sediment and nutrient exports in dryland regions of the MDB (Project D10012). Dickinson, W.T., Rudra, R.P. and Wall, G.J. (1990) Targeting remedial measures to control nonpoint source pollution. Water Resour. Bull. 26, Durst, F., Milojevic, D. and Schönung, B. (1984) Eulerian and Lagrangian predictions of particulate two-phase flows: a numerical study. Appl. Math. Modelling, 8, Eagleson, P.S. (1972) Dynamics of flood frequency, Water Resour. Res. 8(4), Edwards. K. (1987) Runoff and soil loss studies in New South Wales. Technical Handbook No. 10, Soil Conservation Service of NSW, Sydney, NSW. Edwards, K. (1988) How much soil loss is acceptable? Search 19, Edwards, K. (1993) Soil erosion and conservation in Australia. In: World soil erosion and conservation, Ed: Pimentel D, Cambridge, ESRI (2003) ESRI GIS software, World Wide Web ESRI, Redlands, California. Ferro, V. and Minacapilli, M. (1995) Sediment delivery processes at basin scale. Hydrological Sciences Journal 40, Fryirs, K. and Brierley, G.J. (2001) Variability in sediment delivery and storage along river courses in Bega catchment, NSW, Australia: implications for geomorphic river recovery, Geomorphology 38, Freebairn, D.M. and Wockner, G.H. (1986). A study of soil erosion on vertisols of the eastern Darling Downs, Queensland. 1. Effect of surface conditions on soil movement within contour bays. Australian Journal of Soil Research 24, Gallant, J.C. (2001) Topographic scaling for the NLWRA sediment project. Technical Report 27/01, CSIRO Land and Water, Canberra. Haan, C.T., Barfield, B.J. and Hays. J.C. (1994) Design Hydrology and Sedimentology for Small Catchments. Academic Press, New York. Hughes, A.O., and Prosser, I.P. (2003) Gully and Riverbank Erosion Mapping for the Murray-Darling Basin. CSIRO Technical Report 3/03, CSIRO Land and Water, Canberra. 20pp. Hutchinson, M.F., Stein, J.A. and Stein, J.L. (2001) Upgrade of the 9 Second Australian Digital Elevation Model. Khanbilvardi, R.M. and Rogowski, A.S. (1984) Quantitative evaluation of sediment delivery ratios. Water Resour. Bull. 20, Loughran, R.J. and Elliott, G.L. (1996) Rates of soil erosion in Australia determined by the caesium-137 technique: a national reconnaissance survey. In: Erosion and Sediment Yield: Global and Regional Perspectives. IAHS Publication. 236,

49 Lu, H., Gallant, J., Prosser, I.P., Moran, C., and Priestley, G. (2001) Prediction of Sheet and Rill Erosion Over the Australian Continent, Incorporating Monthly Soil Loss Distribution. Technical Report 13/01, CSIRO Land and Water, Canberra, Australia. Lu, H., Moran, C. J., DeRose, R. and Cannon, G. (2003a) Spatially Distributed Investment Prioritization for Sediment Control in the Murray Darling Basin. Technical Report, (In press)/03. Lu, H., Prosser, I.P., Moran, C.J., Gallant, J. Priestley, G. and Stevenson, J.G (2003b) Predicting sheetwash and rill erosion over the Australian continent, Australian Journal of Soil Research. (revised) Lu, H., Raupach, M.R., McVicar, T.R. Barrett, D.J. (2003c) Decomposition of Vegetation Cover into Woody and Herbaceous Components Using AVHRR NDVI Time Series. Remote Sensing of Environment 5850, Lu, H. and Yu, B. (2002) Spatial and seasonal distribution of rainfall erosivity in Australia, Australian Journal of Soil Research 40, Maner, S.B. (1958) Factors affecting sediment delivery rates in the Red Hills physiographic area. Trans. Am. Geophys. 39, Martin, R.J., MCMillan, M.G. and Cook, J.B. (1988) Survey of farm management practices of the northern wheat belt of New South Wales. Australian Journal of Experimental Agriculture 28, Morgan, R.P.C., Quinton, J. N., Smith, R.E., Govers, G., Poesen, J.W.A., Auerswald, K., Chisci, G., Torri., D. and Styczen, M. E. (1998) The European soil erosion model (EUROSEM): A dynamic approach for predicting sediment transport form fields and small catchments. Earth surf. Process. Landforms 23, Newcombe, C.P. and MacDonald D.D. (1991) Effects of suspended sediments on aquatic ecosystems. North American Journal of Fisheries Management 11, NLWRA (2001) Australian Agriculture Assessment. National Land and Water Resources Audit, Canberra. Novotny, V. and Olem, H. (1994) Water Quality-Prevention, Identification, and Management of Diffuse Pollution. Van Nostrand Reinhold, New York. Overton, D.E. and Meadows, M.E. (1976) Stormwater modeling, Academic Press, New York. Parsons, A.J. and Stromberg, S.G.L. (1998) Experimental analysis of size and distance of travel of unconstrained particles in interrill flow Water Resour. Res. 34, Pimentel, D., Harvey, C., Resosudarmo, P., Sinclair, K., Kurz, D., McNair, M., Crist, S., Shpritz, L., Fitton, L., Saffouri, R. and Blair, R. (1995) Environmental and economic costs of soil erosion and conservation benefits. Science 267, Prosser, I.P., Rustomji P., Young, W.J., Moran, C.J. and Hughes, A.O. (2001) Constructing river basin sediment budgets for the National Land and Water Resources Audit. CSIRO Land and Water Technical Report 15/01, CSIRO Land and Water, Canberra, Australia. Quinlan, J.R. (1993) C4.5: Programs for Machine Learning, Morgan Kaufmann Publishers, San Mateo, California.. For further information see: Renard, K.G., Foster, G.A., Weesies, D.K., McCool, D.K., and Yoder, D.C. (1997) Predicting soil erosion by water: A guide to conservation planning with the Revised Universal Soil Loss Equation. Agriculture Handbook 703, United States Department of Agriculture, Washington DC. Richards, K. (1993) Sediment delivery and the drainage network, In: Channel Network Hydrology, Eds. Beven, K. and Kirkby, M.J., Robinson, J.S. and Sivapalan, M. (1997) An investigation into the physical causes of scaling and heterogeneity in regional flood frequency, Water Resour. Res., 33, Roehl, J.E. (1962) Sediment source areas, and delivery ratios influencing morphological factors. Int. Assoc. Hydro. Sci. 59, Rosewell, C.J. (1997) Potential source of sediments and nutrients: Sheet and rill erosion and phosphorus sources. Australia: State of the Environment Technical Paper Series (Inland Waters), Department of the Environment, Sport and Territories, Canberra. Rowe, R.K., Crouch, R.J. and van Dyke, D.C. (1978) Soils in the upper valleys of the Murray River basin. Proceedings of the Royal Society of Victoria 90, Rulequest Research (2001) Rulequest Research Data Mining Tools. World Wide Web Rutherfurd, I.D., Ladson, A., Tilleard, J., Stewardson, M., Ewing, S., Brierley, G. and Fryirs, K. (1998) Research and development needs for river restoration in Australia. LWRRDC Occasional Paper No. 15/98, NSW. SCS (1983) TR-20 Computer Program for Project Formulation Hydrology, U. S. Department of Agriculture, Soil Conservation Service, Washington, D. C., May, SCS (1986) Urban Hydrology for Small Watersheds, Technical Release No. 55, U. S.Department of Agriculture, Soil Conservation Service, Washington, D. C., June,1986. Sivapalan, M., Jothityangkoon, C. and Menabde, M. (2001) Linearity and non-linearity of basin response as a function of scale: Discussion of alternative definitions. Water Resour. Res. 24,

50 Swift, R. and Skjemstad, J. (2002) Agricultural land use and management information. National Carbon Accounting System, Technical Report No. 13. Vanclay, F. (1997) Land degradation and land management in central NSW. Charles Sturt University, Wagga Wagga). Van Rompaey, A.J.J., Verstraeten, G., Van Oost, K., Govers, G. and Poesen, J. (2001) Modelling mean annual sediment yield using a distributed approach, Earth surf. Process. Landforms 26, Walling, D.E. (1983) The sediment delivery problem, Journal of Hydrology 65, Walling, D.E. (1988) Erosion and sediment yield research - some recent perspectives. Journal of Hydrology 100, Wasson, R.J., Olive, L.J. and Rosewell, C. (1996) Rates of Erosion and Sediment Transport in Australia. In: Erosion and Sediment Yield: Global and Regional Perspectives. D.E. Walling and R.Webb (eds) IAHS Publ. Williams. J. R. (1977). Sediment delivery ratios determined with sediment and runof models. In: Erosion and solid matter transport in in land waters. IAHS-AISH publication, 122: Wischmeier, W.H. and Smith, D.D. (1978) Predicting rainfall erosion losses: A guide to conservation planning. US Department of Agriculture. Agriculture Handbook No (US Government Printing Office: Washington, DC). Yu, B. and Rosewell, C.J. (1998) A program to calculate the R-factor for the USLE/RUSLE using BoM/AWS pluviograph data. ENS Working Paper 8/98. 49

51 Appendix I: Land Use Data Table AI.1 shows the details of the land use data supplied status and contact details of the agencies. Figure AI.1 shows the land use extent used in this study. Areas shown in blue have land use supplied by local agencies. Table AI.1: Summary of locally supplied land use data used in this study. Region Data Supply Status address Data supplied by a CD with Rob Brownbill 1: map sheets of the rbrownbill@dlwc.nsw.gov.au Murubidgee region. CD includes Arcview shape files, Sally Keane metadata and readme Resource Officer (GIS) documents. (02) Murrumbidgee (New South Wales) Barwon region (New South Wales) Murray region (New South Wales) South-east South Australia (New South Wales) Golbourn region (New South Wales) Condamine (Queensland) Upper Billabong * (New South Wales) Bendigo Region (Victoria) A CD of Landuse mapping was supplied. It contains 1: landuse mapping of Barwon region done during late 1980s, Showing timber, pasture, and cropping lands. It also contains 1: landuse mapping of eastern part of Walgett shire and all of Moree shire ( ). Data supplied for Upper Murray and Billabong regions. A CD with landuse data was supplied. A CD was supplied by MDBC. It contains land use data for Golbourn, Condamine and Upper Billabong regions. Products of MDBC Project D2006 Land mark Task 6a. See above. See above. Current landuse mapping for Victoria is as same as BRS 1: landuse. Just starting a landuse project for NC CMA but won t be finished until next financial year. Angela McCormack amccormack@dlwc.nsw.gov.au Stuart Lucas slucas@dlwc.nsw.gov.au David Tonkin Tonkin.David@saugov.sa.gov.au Michael Htun GPO Box 409 Canberra, ACT 2601 Ph: (02) Fax: (02) See above. See above. Maree Platt Maree.Platt@nre.vic.gov.au Christian Writte No data supplied. David Burton 50

52 (Queensland) GIS/Drafting Officer(Graphics Unit) Department of Natural Resources and Mines Far West of New South Wales Central West Region of New South Wales We were told that detailed landuse data does exist for the region but they were a bit hesitant to give it to us because the information was quite sensitive. Despite assuring that the data would be used in the strictest of confidence, no data supplied to us. No data supplied. Aaron Colbran acolbran@dlwc.nsw.gov.au Michael Casey mcascey@dlwc.nsw.gov.au * The Murrumbidgee data includes 90% of the Billabong data. The Murrumbidgee landuse data is used for the overlapped part. Figure AI.1: Data sources of land use used in this project. 51