Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia

Transcription

1 Application of Tracers inaridzane Hydrology (Proceedings of the Vienna Symposium, August 1994). IAHS Pub!, no. 232, Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia J. V. TURNER CSIRO Division of Water Resources, Private Bag, PO Wembley, Western Australia 6014 M. R. ROSEN Institute of Geological and Nuclear Science Ltd., Wairakei Research Centre, Private Bag 2000, Taupo, New Zealand L. K. ÎTFTELD & G. L. ALLAN Department of Nuclear Physics, The Australian National University, GPO Box 4, Canberra, Australian Capital Territory 2601 Abstract Comparative application of the environmental tracers 36 C1, 14 C, ô 2 H and ô 18 0 in addition to chloride mass balance and hydrogeological information has been carried out in order to evaluate the groundwater recharge-discharge process in hypersaline palaeochannels of the Eastern Goldfields of Western Australia. Groundwater and chloride residence times are also estimated. The groundwater travel time through one section of palaeochannel was estimated via 14 C to be approximately years. Possible interpretations of 36 C1 concentrations are strongly dependent on the reliability of estimates of chloride and 36 C1 input into the groundwater system and the selection of an input 36 C1/C1 ratio, Rj. A slight decrease in the 36 C1/C1 ratio with distance along the flow direction of the palaeochannel was observed that provided the only suggestion of 36 C1 decay. In general however, the concentration of 36 C1 in the hypersaline groundwaters is determined by the processes of concentration and dilution. Groundwater and chloride residence times are short compared to the halflife of 36 C1. Because the accumulation time for chloride is of the order of 10 5 years, significant radioactive decay of 36 C1 is not expected. INTRODUCTION The conjunctive application of environmental isotope techniques, hydrogeological assessments and groundwater modelling is now well recognized as a powerful approach to solving water resources problems in arid regions. In this paper we focus on the isotope hydrology of a system of hypersaline, palaeochannel aquifers, located in the arid Eastern Goldfields region of Western Australia near the town of Kalgoorlie. In particular, attention is given to the interpretation of 36 C1 data in relation to the recharge and discharge process of groundwater, the groundwater residence time and the residence time of chloride in the aquifer system. Use is made of the variable distribution of hypersaline groundwater across the study region, where chloride is the dominant anion, in addition to carbon-14 and stable isotope data on groundwaters and results from hydro-

2 16 J. V. Turner et al. geological investigations. Developments in the capabilities for measuring 36 C1 at environmental levels have led to its increased use in hydrological tracer studies in recent years. In conjunction with the ubiquitous occurrence of chloride in groundwater, its highly soluble and mobile nature as well as its generally conservative geochemistry, makes 36 C1 an ideal tracer. As a result of the analytical capability provided by accelerator mass spectrometry (AMS) based on the 14UD accelerator at the Australian National University, the application of 36 C1 measurements in Australian hydrological studies has developed considerably in recent years. In the context of Australian groundwater hydrology, an isotopic tracer of the chloride ion is of particular value because much of the continental groundwater hydrogeochemistry in Australia is dominated by marine aerosol input, of which the chloride ion is a major constituent. The contributions made by 36 C1 to the understanding of salt and water movement in the saturated and unsaturated zones in Australia cover such areas as dating of groundwaters in the Great Artesian Basin (Bentley et al., 1986), the role of 36 C1 in Australian groundwater studies (Bird et al., 1989), recharge estimation (Walker et al., 1991), determination of the source of salts in the Murray Basin, (CaLfetal, 1988; Davie etal., 1989; and Tumeretal., 1991) and net marine chloride accession to the Murray Basin (Simpson & Herczeg, 1994). Several Australian case studies have been reviewed recently by Mazor (1992), however, the re-interpretations proposing groundwater age estimates based on 36 C1 decay have been challenged (Kellett etal, 1993; Phillips, 1993; Fontes & Andrews, 1993). GEOLOGY AND HYDROGEOLOGY OF THE EASTERN GOLDFIELDS The geology of the study region is well described elsewhere (Clarke, 1993; Griffin, 1990) and is described briefly here. The study region forms the eastern section of the Yilgarn block of Western Australia and is composed primarily of Archaean intrusive granitoid rocks with metamorphosed intrusive igneous assemblages known as greenstone belts. The greenstone belts trend in a north to northwesterly direction in the study region and host the important economic mineral assemblages. Much of the granitic terrain is deeply weathered forming kaolinized pallid zones that are frequently overlain by lateritic and ferricrete cappings. Palaeochannel aquifers in internal drainage basins of the eastern Yilgarn region of Western Australia were formed by the infilling of relict surface drainage channels. The drainage channels were incised in the crystalline Precambrian basement rocks during the Jurassic. Deposition of the non-marine Wollubar Sandstone, which forms the principal unconsolidated sand aquifer of this study (Kern & Commander, 1993), occurred during the Eocene and was subsequently overlain by confining clays. Salt lakes are a common feature in the present landscape and form continuous linear chains that extend over hundreds of kilometres. In many instances they form a present-day surface expression of the relict surface drainages, however the palaeochannel aquifers are not always geographically associated with salt lakes. The aquifer sands range in thickness between 5 and 20 m and are typically 400 to 800 m wide. They occur at depths of between 20 and 30 m and are overlain by confining clays. These clays also form the beds of the normally dry salt lakes. The groundwater resources of the aquifers are of crucial importance to the regional mining industry as a source of water for gold and nickel ore processing.

3 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 17 REGIONAL AND LOCAL DRAINAGE BASINS The drainage systems of the southern half of Western Australia have been described by several authors. Morgan (1966) produced the first drainage map of the salt lake systems of the Yilgarn. This was followed by Bettenay & Mulcahy (1972) who mapped the major drainages and salt lake systems. Beard (1973) used vegetation distribution patterns as a means of identifying palaeodrainage patterns and Bunting et al. (1974) reported detailed investigations into palaeodrainage patterns of the Eastern Goldfields. Mann (1982) produced a map of the drainage systems of the Yilgarn Block which is reproduced in modified form in Fig. 1. The figure shows a major north-south drainage divide that partitions drainage into the westward-flowing Avon palaeodrainage and a number of eastward flowing palaeodrainages that discharge towards the Eucla basin. The study area is located entirely within the Roe palaeodrainage, named after Lake Roe, and is one of several east to southeasterly draining palaeochannel systems that discharge into the Eucla basin. Figure 1 shows the study area within the context of the regional distribution of palaeodrainage systems and provides the hydrogeological basis for analysis of the groundwater recharge and discharge processes occurring within the study area. The section of palaeochannel aquifer considered in more detail is known as the Black Flag-Gidji palaeochannel tributary. The section is approximately 55 km long and is oriented west to east. It forms the western end of the more extensive Roe palaeodrainage that discharges eastward toward the Eucla basin. The study aquifer is located 35 km north of the town of Kalgoorlie. Figure 2 shows the catchment boundaries for the Black Flag-Gidji palaeochannel and the sample locations. Prior to the commencement of groundwater resource development in the early 1980s, the direction of the groundwater hydraulic gradient was from west to east. The gradient was about in the west and levelled out to about as it enters a region of salt lakes in the east. The topographic gradient is about and this steeper gradient explains the occurrence of shallow water tables at or very close to the ground surface in the salt lake region. The western extreme of the study aquifer lies 70 km to the east of the major north-south drainage divide on the Yilgarn Block of Western Australia. GROUNDWATER SALINITY AND ITS DISTRIBUTION The groundwaters are hypersaline and attain salinities of up to mg l* 1 TDS, especially near salt lakes. The salinity of groundwaters within the palaeochannel aquifers generally increases at the regional scale from west to east, as shown in Fig. 3. In the BlackFlag-Gidjipalaeochannel, chloride concentrations increase smoothly with distance along the palaeochannel over the distance of 55 km from mg l" 1 in the west (Site KRD) to mg l" 1 in the east (Gidji Bore 17) as shown in Fig. 4. The major ion composition of the groundwaters is similar to seawater, except for slight depletions in K +, Ca ++ and SO4 relative to chloride. These depletions are due to ion adsorption and gypsum precipitation. The origin of the salts in groundwater has been identified as marine aerosol input (McArthur et al., 19 89) primarily on the basis of the predominantly marine hydrogeochemical composition and 87 Sr/ 86 Sr and <5 34 S data. Average annual pan evaporation is 2400 mm and significantly exceeds the average annual rainfall of 250 mm.

4 18 J. V. Turner et al. Major Drainage Divide Secondary Drainage Divide Fig. 1 Location of major drainages in the Yilgarn Block (after Mann, 1982). The Black Flag-Gidji palaeochannel has been studied more intensively than other palaeochannels in the study area. It has been used as a representative case study to show that the palaeochannel-salt lake systems do not form closed basins. The processes explaining the increase in salinity from west to east with increasing distance from the recharge areas at the western end of the palaeochannel aquifers have been identified (Turner et al., 1992) and are linked to the recharge-discharge cycle. At steady state, the processes of groundwater recharge, evaporative discharge from the salt lakes and subsurface groundwater outflow from the eastern end of the palaeochannels are coupled and determine the degree to which the salinity can increase along the aquifer. For the

5 Chlorine-36 in hypersaline palaeochannel groundwaters of "Western Australia 19 Black Flag palaeochannel, the twelve fold increase in chloride ( mg l" 1 to mg l" 1 ) concentration in groundwater along the flow direction in the system is explained by evaporative loss via groundwater discharge in salt lakes. This is the most important discharge process, and it is estimated that over 90% of the recharge is discharged from the aquifer by evaporation, with the remainder of the groundwater leaving the system via subsurface outflow (Turner et al., 1992). Thus the long term input of marine aerosol salt and the net loss of water by evaporation from salt lake surfaces causes the increase in salt concentration along the palaeochannels. Subsurface groundwater discharge from the end of the palaeochannel must be less than 10 % of the recharge. This small percentage of subsurface outflow (i.e. non evaporative discharge) has a very strong influence on the ratio of the salinity between the recharge and discharge ends of the system and determines the maximum salinity that can be achieved in the aquifer-salt lake systems. The small percentage of subsurface outflow indicates the degree of openness of the system. Chloride profiles taken from the unsaturated zone show that transpiration from native eucalypt vegetation prior to infiltration in the topographically elevated sections of the catchments plays a role in the primary concentration of chloride prior to recharge into the palaeochannel system (Turner et al., 1993). Concentration of chloride by trans- Fig. 2 Catchment boundaries and locations of samples.

6 20 J. V. Turner et al. 100 : INDICATES CHLORIDE CONC. IN mg I 1 X1000 «S» 400m TOPOGRAPHIC CONTOUR // PALEOCHANNEL BOUNDARIES Fig. 3 Regional distribution of chloride concentration. o GIDGIBORE# 17 I «: ' ' "J - GYPSUM SATURATION ". * J5^ WEST. s " ' ^ EAST S" / ^ 1 GIDJI BOREFIELD CD LU 50 9 cr o o 0 " ". KRA 7 B I [ KRD i a **J «/ SALT LAKE DAM BOREFIELD / PADDINGTON WEST BOREFIELD *S ROSE DAM BORES DISTANCE FROM SITE KRD (km) Fig. 4 Plot of chloride concentrations with distance along the Black Flag-Gidji palaeochannel. 60

7 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 21 piration results in chloride concentrations of up to mg l" 1 in the unsaturated zone. The distribution of brine concentration therefore reflects an approximately steady state condition that has been established over the long term (tens of thousands of years) between marine aerosol input, solute concentration by evaporative discharge of groundwater in salt lakes and groundwater outflow to the east. The degree of hydrological openness accounts for the absence of significant abundances of evaporite minerals such as gypsum and halite in the system, although minor gypsum is found in the beds of many salt lakes. CHLORIDE AND 36 C1 MASS BALANCES AND ACCUMULATION TIME Issues related to 36 CI systematics in groundwater Before any model can be developed and used to estimate the residence time of chloride in the palaeochannel aquifer system based on radioactive decay of 36 C1, several factors other than radioactive decay alone that could result in changes to the 36 C1/C1 ratios in the brines must be assessed. These factors include consideration of whether the groundwater system is closed or open with respect to cycling of 36 C1 and CI. Specifically, whether 36 C1 is gained from the rock matrix by in situ production and whether the aquifer is closed with respect to outflow of chloride. In the case of the palaeochannel aquifers, in situ production refers to epigene or hypogene production, i.e. whether 36 C1 is gained from production of 36 C1 in the soil or rock matrix by cosmic ray interaction with targets at the ground surface or by capture of neutrons by 35 C1 from radioactive decay of uranium or thorium series elements in the aquifer matrix. Hypogene production is likely to be irrelevant because (a) a posteriori, groundwater in the system has been found to have relatively short residence time in relation to likely production rates of 36 C1 by this process; (b) the potential source rocks are highly weathered and oxidized and the hypersaline groundwaters reside in aquifers within the weathered and oxidized sediments; and (c) groundwaters are found at relatively shallow depths ( < 60 m) and circulation within the basement granitoid rocks is limited. Other factors include the uncertainty on the input value for R D the 36 C1/C1 ratio of chloride input to the system and whether it should be based on theoretical literature values or field measurements. In addition, there is the question of whether modern measurements of chloride fallout are valid over the much longer time scale of aerosol accumulation in a groundwater system. Chloride deposition from aerosol input also influences the 36 Cl/Clratio of input, decreasing it if the CI input increases and increasing it if the CI input decreases. Modern Australian measurements of 36 C1 fallout are potentially affected by high Rj values caused by thermonuclear 36 C1 fallout. Mazor (1992) has proposed input ratios of 157 and 44 x 10" 15 respectively for local intake of 36 C1 for the Lachlan Fan and Murray- Mallee of eastern Australia. Walker etal. (1991) show 36 C1/CI ratios in the unsaturated zone of about 200 x 10" 15 in the Murray Darling Basin, which means that input ratios are significantly higher than calculated by Davie et al. (1988) and indicate thermonuclear 36 C1 fallout. In past climatic regimes the chloride fallout could have been different and therefore the fallout of 36 C1, expressed as the ratio 36 C1/C1, could have been different. Thus, there

8 22 J. V. Turner et al. is a possibility that Rj values in groundwaters are less than in input, not due radioactive decay, but because of a higher chloride content in the input at the time of recharge. Input and accumulation of meteoric chloride The water balance and groundwater residence times can be evaluated by estimating the solute balance for the catchment. The following solute balance calculations were made for the northern catchment area, as shown in Fig. 2. The catchment area was estimated as 4040 km 2. The chloride fallout in rainfall was measured during 1989 to 1991 in rainfall collected in static funnel rainfall collectors in the region at Kalgoorlie and at Southern Cross, 200 km to the west of Kalgoorlie (Farrington et al., 1993). The average annual chloride concentration in rainfall at Kalgoorlie was found to be 1.2, 1.8 and 2.9 mg l" 1 respectively and the corresponding rainfall amounts were 216, 258 and 178 mm. The corresponding chloride fallout amounts are 0.26, 0.46 and 0.52 g m" 2 year" 1. At Southern Cross, the average annual chloride concentration was 2.2, 3.1 and 3.6 mg l" 1 and the annual rainfall amounts were 310, 261 and 285 mm, corresponding to chloride fallouts of 0.68, 0.81 and 1.03 g m" 2 year" 1. A rainwater sample collected from an underground domestic rainwater storage tank at Gindalbie Station, 50 km north of Kalgoorlie in February 1992 had a CI concentration of 1.5 mg l" 1 (Table 1). The rainwater sampled from this source integrates an unknown number of years of rainfall in the region. The chloride fallout estimated from this sample of rainfall is 0.38 g m" 2 year" 1 based on the annual average rainfall of 250 mm and agrees well with the static rainfall collector measurements at Kalgoorlie. The chloride fallout in rainfall at Gindalbie Station of 0.38 g m" 2 year" 1 falls within the range Table 1 Input data for CI and 36 C1 for the study region. Rainfall chloride input Date [CI] mg l" 1 CI fallout (g m" 2 year" 1 ) Static collectors Gindalbie station 1992 (Integrated) Collector transect Rainfall 36 C1 input Atoms nr 2 s" 1 36 C1/C1 x 10" 15 Calculated at 31 S Gindalbie (measured) Calculated at 27 S S transect (measured) 'Farrington et al. (1993). 2 Keywood et al. (1994). 3 Following Andrews & Fontes (1991).

9 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 23 measured by Farrington et al. (1993). Thus the best estimate of chloride accession from rainfall in the study region is between 0.26 and 0.52 g m" 2 year 4. Using an average rainfall of 250 mm year 1 and this range of annual chloride accession amounts gives an annual net input of chloride into the catchment of about 1.5 ± 0.5 x 10 6 kg. The chloride storage within the aquifer and overlying clays was estimated using figures presented by Commander et al. (1992) except that the catchment area was taken as 4040 km 2 as above. The chloride storage is calculated as x 10 6 kg. This corresponds to an accumulation time of between and years at the present rate of chloride fallout in rainfall. The mass of chloride exported from the catchment by groundwater outflow at the eastern end of the palaeochannel is found by combining the estimates of groundwater outflow based on modelling of x 10 6 m 3 year" 1 (Turner et al., 199'4) with its chloride concentration of 80 kg m" 3. The estimate of chloride export is 0.24 x 10 6 kg year 1. This compares closely with the estimated input rate of chloride and indicates an approximate steady state between accession and discharge of chloride. The mass of chloride lost from the catchment via deflation of halite efflorescences on salt lake surfaces and airborne dust is not known, but could be an important mechanism for salt transport in the system. For example, Table 2 shows a 36 C1 analysis from Reedy Dam, an artificial open water body designed to impound local surface runoff. Its chloride concentration is low, but its 36 C1/C1 ratio is consistent with that of halite and is indicative of chloride recycling in the landscape. These water and solute balance results are summarized in Table 3. Meteoric J6 C1 input Estimates of the meteoric 36 C1 input to the region can be made both on a theoretical basis and from experimental data. Natural 36 C1 production and distribution in for the Murray Basin in Australia has been estimated by Davie et al. (1988) following Bentley et al. (1986). This approach has been used to estimate the 36 C1 fallout for the Eastern Goldfields region as the ratio 36 C1/C1 by combining the measurement of chloride fallout in rainfall with estimates of the deposition rates of cosmogenic 36 C1 fallout. The deposition rate of 36 C1 as a function of latitude has been estimated by Bentley et al. (1986). The latitude of Kalgoorlie is between 30 and 31 S and according to Bentley et al. corresponds to a 36 C1 fallout of about 21 atoms m" 2 s" 1. Recently however Jiang et al. (1990) and Andrews & Fontes (1991) have shown that the estimates of cosmogenically produced 36 C1 estimated by Bentley et al. (1986) should be reduced by a factor of 11/16. Taking this into account gives an estimate of the 36 C1 fallout in the Eastern Goldfields study area of about 14 atoms nr 2 s" 1. Combining this estimate with the data on chloride fallout described above gives a 36 C1/C1 ratio of between 100 and 50 x 10" 15 respectively. Gindalbie rainfall has a 36 C1/C1 ratio of 153 ± 7 x 10" 15 and the corresponding 36 C1 fallout calculated from it is 31 atoms m" 2 s" 1. This measured 36 C1/C1 ratio is significantly higher than the fallout estimate determined from the production of 36 C1 as a function of geomagnetic latitude. Further evidence that the actual input 36 C1/C1 ratio is higher than the estimated value is apparent from the measured ratios in the two dilute groundwaters at Credo Well and Bellvue which have ratios of 102 and 129 x 10" 15 respectively. Both of these are greater than the calculated fallout ratio. The 36 C1 content of bulk depositional precipitation in Western Australia has been

10 24 J. V. Turner et al. Table 2 3S C1 data on chloride from regional groundwaters, surface water, rainwater, and surficial efflorescence halite from salt lakes. Data under the headings of Lake Ballard and GSWA Kalgoorlie regional groundwaters are from Commander et al. (1994). The surface halite data are from Chivas et al. (1987 and 1988). Index numbers 1 to 24 are used to identify the samples in Figs 6 and 7. Sample Sampling date CI (mg l 1 ) 36 C1 X 10 6 Atoms H 36 C1/C1 X 10" 15 Rainfall (10) Gindalbie rain 22 Feb ± ± ± 7 Black Flag-Gidji palaeochannel groundwaters (1) Credo Well (2) Bellvue Bore (3) KRA 7 (4) PW 15 (5) Pad Seep 4 (6) Pad Seep 1 (7) AL 37 (8) BH 4 24 Jan Sep Jan Dec Dec Dec Dec Jan ± 13 ± 12 ± 389 ±446 ± 696 ± 993 ± 1002 ± ± 90 ± 160 ± 3600 ± 2000 ± 2900 ±4100 ±4300 ± ±5 ± ± 3 ±3 ±3 ±3 ±5 (9) Gidji Bore Jan ± ± ±5 (11) Reedy Dam 22 Feb ± ± ±4 Other regional groundwaters (12) Kaltails A (13) Kaltails R (14) KRC5 (15)JP 1 (16) NB 2 Lake Ballard (17) LBC-8C (18) LBC-8B 9 Dec Dec June Dec Dec ± 1102 ± 862 ± 601 ±405 ± 924 ± 110 ± ±5000 ± 3100 ± 3100 ±2600 ± 3900 ± 3000 ± ±3 ±3 ±4 ±5 ± ± 3 (19) LBC-8A GSWA Kalgoorlie regional groundwaters ± ± ±3 (20) KRE-4 (21) KRL ±294 ± ±2000 ± ±6 ±5 (22) KRM-2 Surface halite ± ±6 (23) Lake Ballard (24) Lake Lefroy ± ± ± ±7 ±5

11 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 25 Table 3 Solute balance results for the Black Flag-Gidji palaeochannel. Catchment area 4040 km 2 Catchment CI input Catchment CI storage Chloride accumulation time 1.5 ± 0.5 X 10 s kg year" X kg X 10 3 years determined by Key wood et al. (1994) from a west-east array of rainfall collectors trending inland from the coast at a latitude of 27 S for a distance of 1200 km. Precipitation samples were collected every three months for two years during 1992 and This transect of rainfall sampling locations passes approximately 300 km to the north of the study region. The 36 C1/C1 ratios increase steadily with increasing distance from the coast from about 15 X 10" 15 at the coast to between 210 and 260 X 10" to 800 km inland. This range of distances inland from the west coast brackets the distance inland of the study area, which is 600 km. The increase in 36 C1/C1 ratio with distance inland reflects the decreasing marine influence and decreasing chloride content of rainfall further inland. The average chloride fallout measured between 400 and 800 km inland was 0.12 g m" 2 year" 1 and, as expected, is lower than at Kalgoorlie because it is further inland. The estimated fallout of 36 C1 at 27 S latitude corrected according to Jiang et al. (1990) and Andrews & Fontes (1991) is about 12 atoms m~ 2 s" 1. Combining this with the chloride fallout of 0.12 g m" 2 year" 1 gives a 36 C1/C1 ratio of 186 x 10" 15. While this ratio is higher than that at Kalgoorlie, it is still lower than the ratios measured by Keywood et al. (1994) (Table 1). There are two possible explanations for the discrepancy between theoretical and observed meteoric 36 C1 input. The first is that there is a component of thermonuclear derived 36 C1 in the input. Comparisons of observed (Gindalbie rainfall) and theoretical 36 C1 fallout at 31 S (Table 1) suggests that as much as 50% (31-14)/31 x 100%) of the observed fallout could be derived from thermonuclear sources. At 27 S the discrepancy is not as significant, amounting to about 20 % of the observed fallout. The second explanation is the effect of inadvertently collecting recycled terrestrial chloride in field collectors, including Gindalbie rainfall. This would lead to a lowering of the measured 36 C1/C1 ratio in the collected chloride, and consequently could not explain a discrepancy where the measured 36 C1/C1 ratio is greater than the theoreticval. This point is based on the 36 C1/C1 ratio of the Reedy Dam sample of 35 ±4 X 10~ 15 (Table 2) mentioned above which is considered to be characteristic of the 36 C1/C1 ratio of terrestrial recycled chloride. Were the theoretical value greater than the observed, then it could be argued that inadvertently collected terrestrial recycled chloride was the cause of such a discrepancy. GROUNDWATER RESIDENCE TIMES DETERMINED FROM 36 C1: COMPARISON WITH CARBON-14 AND MODELLING Analysis of the 36 C1 data is based firstly on a comparison with 14 C data from the Black Flag-Gidji palaeochannel, and secondly on an analysis of the entire regional data set.

12 26 /. V. Turner et al. Comparison of groundwater residence times for the Black Flag-Gidjipalaeochannel can be based on 36 C1 and 14 C data and on hydraulic considerations. Estimation of groundwater flow rates in the palaeochannel based on carbon-14 data and results from groundwater flow modelling have been given in Turner et al. (1994). The difference in 14 C activity of dissolved inorganic carbon between samples along the palaeochannel was used to estimate travel times. A steady decline in the 14 C activity of groundwater was observed with distance along the flow direction. Thus, an approach of considering relative differences in 14 C activity in groundwater along the flow system avoided the necessity of using correction schemes to determine initial 14 C activities. A groundwater travel time calculated from the usual equation t = lna/a 0 of years was estimated over a distance of about 15 km. The 14 C analyses selected for this calculation were chosen to be consistent with the domain of the model calculations mentioned below. Within the uncertainty of the determinations, this gave a maximum groundwater flow rate of approximately 1 m year" 1. This flow rate was used to estimate a steady state groundwater flux through the aquifer of about 2000 m 3 year" J. Numerical modelling was also used to evaluate the steady groundwater flow through the aquifer and indicated a steady groundwater flux of 9000 m 3 year" 1 in good agreement with the estimate based on 14 C (Turner et ah, 1994). Sample locations such as KRA 7 and Gidji Bore 15, where both 14 C and 36 C1 data are available, are selected for this calculation because they are the most widely separated 36 C1 analyses available for the Black Flag-Gidji palaeochannel. The difference in 14 C activity between KRA 7 and Gidji Bore 15 is = 25.3 pmc. Once again, under the condition outlined above, this translates into a travel time calculated from the equation t = 8033 lna/a 0 of approximately years. To enable direct comparison of the estimates of groundwater travel times based on 14 C, we proceed to interpret the 36 C1 data from the same section of palaeochannel. As with the differential approach to interpreting the 14 C data, the 36 C1 data are also interpreted differentially. Table 2 shows 36 C1 and CI data on seven samples (Index numbers 3 to 9) that were obtained sequentially from west to east along the Black Flag-Gidji palaeochannel. Figure 2 shows the location of these samples. Samples from Credo Well and Bellvue Bore characterize the low salinity groundwaters in the topographically elevated areas of the catchment. These groundwaters are in fractured rocks that flank the palaeochannel aquifer and are not within the palaeochannel system itself. Sample KRA 7 is at the western extreme end of the palaeochannel and has a relatively low salinity. The linear distance along the aquifer from KRA 7 to Gidji Bore 15 is approximately 55 km. Sample PW 15 is to the east of KRA 7 and within the palaeochannel aquifer. Samples Pad Seep 4 and 1 further east again, are groundwater seepages from drill holes in an open mine pit wall adjacent to the palaeochannel, along gradient from PW 15. These seepages are considered to be derived from palaeochannel aquifer groundwater. Samples AL 37 and BH 4 are from the palaeochannel aquifer itself and the overlying, saturated clays respectively. Sample Gidji Bore 15 is from a production bore into the palaeochannel aquifer and is at the eastward extreme end of the sampled section of palaeochannel. From the data in Table 2 it can be seen that there is a significant increase of 4 to 5 fold in the chloride concentration along the palaeochannel system from the upper, recharge, end in the west at KRA, to the furthest eastward sampling of the palaeochannel at Gidji Bore 15. At another bore transect at the western end of the palaeochannel (Transect KRD, not sampled for 36 C1) the chloride concentration is

13 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia mg l" 1, thus the ratio of chloride at the outflow to the palaeochannel to that at the recharge end is as high as 12. Figure 5 shows the 36 C1/C1 ratios plotted against CI concentration for all the Kalgoorlie regional data. Data on groundwaters in the Black Flag-Gidji palaeochannel (Index numbers 3 to 9) are shown connected by a broken line. The dilute end member sample (Credo Well) and the concentrated end member (halite) are important to interpretation of the system. The data of Chivas et al. (1987,1988) on salt lake halite represents a concentrated end member for the system. From the two 36 C1/C1 ratios for halite in salt lakes (Table 2) and under the assumption that the halite is associated with concentrated brines at saturation with respect to halite (about 240 g l" 1 chloride), the 36 C1 concentration of the brines is approximately 140 x 10 9 atoms l" 1. These points are plotted in Figs 5 and 6. Given the explanation for the increase in chloride concentration along the palaeochannel aquifer described above, it is expected that the 36 C1 concentration will increase along the aquifer with the increase in total chloride concentration as shown in Fig. 4. However, the 36 C1/C1 ratio of groundwater will remain constant when the total chloride concentration increases due to evaporative loss. Thus the data can be examined for evidence of radioactive decay, since it is the only process other than in situ production, that can affect the 36 C1/C1 ratio. The range of 36 C1/C1 ratios along the palaeochannel is narrow and values are not significantly different from each other. However, Fig. 5 shows a slight decrease in 36 C1/C1 ratio from 49 ± 7 x 10" 15 at KRA 7 to 34 ± 5 X 10" 15 at Gidji Bore 15. If radioactive decay of 36 C1 along the palaeochannel is the Gindalbie Rainfall -T- " 2 Bellvue Bore 120,. s 1 > 36 O T ><_ CI/CI Credo Well -pi 15 _[20 Jl4 Halite H I Reedy Dam 113 ^FÏ Chloride (g I' 1 ) Fig. 5 Plot of 36 C1/C1 ratios vs CI. The broken line links samples from west to east along the Black Flag-Gidji palaeochannel.

14 28 J. V. Turner et al Chloride (g r') Fig. 6 Plot of 36 C1 vs CI for regional groundwaters. The broken line links samples from west to east along the Black Flag-Gidji palaeochannel. cause of the decrease in 36 C1/C1 ratio then a travel time can be calculated. In order to do this, as above with the 14 C based estimates of groundwater travel times, the equation t = 4.35 x 10 5 In AIA 0 can be used to estimate the travel time. This calculation yields x 10 3 years. The uncertainties in estimated travel time are large because of the uncertainty in the 36 C1/C1 ratio, and only the lower end of the range of uncertainty (i.e years) approaches the travel time estimated by 14 C. The rather large uncertainties but nevertheless broadly consistent conclusions with regard to travel times are reached with the 36 C1 data because of the short groundwater travel times in this system, compared with the 36 C1 half life of years. An alternative approach to analysing the regional 36 C1 data is to establish an intake line on Fig. 6. Several alternatives are available to establish an intake line. The ideal intake is that measured in rainfall and for this, a dilution-concentration line based on Gindalbie rainfall is shown in Fig. 6 with a slope of x 10" 15. Alternatively, the 36 C1 content of dilute end member samples (Credo Well and Bellvue Bore) can be regarded as the best available measure of the 36 C1 content in recharge because dilute groundwater samples have the advantage that they integrate the input over time. In order to establish groundwater residence times from the regional 36 C1 data, it is first necessary to account for the effect of evaporation (or dilution) on the concentration of both CI and 36 C1. As is shown in Fig. 6, the effect of concentration or dilution is that values fall along a line of positive slope. The numerical value of this slope is determined by the 36 C1/C1 ratio of the low salinity end member. The effect of decay of 36 C1 is a vertically

15 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 29 downward displacement from the concentration-dilution slope. To interpret the field data in Fig. 6 in terms of groundwater residence times, it is necessary to determine as accurately as possible the 36 C1/C1 ratio of the intake so that the slope of the concentration-dilution line can be established. Using only the Credo Well sample and the origin to establish this slope, a value of 102 x 10~ 15 is obtained. This line is plotted on Fig. 6 and has an uncertainty of ±5 x 10" 15 (Table 2). However, assigning such considerable weight to one analysis in determining such a critical parameter as the slope of the concentration-dilution line is unwise. Consequently, the slope shown in Fig. 6 was established using the five additional low salinity samples: Bellvue Bore, LBC-8C, KRA 7, KRE 4 and PW-15. The slope of the concentration-dilution line for intake is shown in Fig. 6 as 61.1 X 10" 15. Because of experimental errors in both the 36 C1 and CI analyses, a corresponding uncertainty is unavoidable in the slope of the concentration-dilution line. Most of the uncertainty in calculating the slope is due to uncertainty in the 36 C1 analysis. The curved and dashed lines shown are the 95 % confidence limits for the concentration-dilution line. As expected, the form of the 95 % confidence boundaries shows that the minimum uncertainty occurs near the mean value of the points used to determine the slope. The uncertainty increases at higher concentrations. The 95% confidence limits mean that points that lie within its boundary cannot be definitely distinguished from having evolved from low salinity end members by the concentration-dilution process. Figure 6 shows that most higher salinity samples, except LBC-8A, Gidji Bore 15 and the halite samples fall within these confidence limits and therefore their 36 C1 and CI concentrations can only be interpreted as being due to concentration-dilution and mixing, but not due to radioactive decay. To illustrate the effect of radioactive decay, the year isochron has been calculated and is shown in Fig. 6. The isochron was calculated using the value of 61.1 X 10" 15 for the initial slope of the concentration-dilution line. The time taken for any point on the concentration-dilution line to decay to a lower 36 C1 activity can be readily determined by constructing a line vertically downward from the line to the point of interest and reading off the corresponding isochron. It can be seen from Fig. 6 that the 0 to year isochron overlaps with the 95 % confidence interval boundary. The consequence of this is that an upper limit of years is imposed on the residence time of chloride within the palaeochannel aquifers, but more importantly it shows that given the uncertainties in the fundamental measurements there is little evidence for radioactive decay of 36 C1. The chloride mass balance and accumulation time considerations for the northern catchment area given above support this conclusion. Samples such as LBC-8A (19) and halite (23, 24) possibly indicate chloride with the longest residence times. The sample Kaltails R (13) is an outlier in this analysis and may be somewhat older than other samples. Thus we conclude from the 36 C1 data that the residence time of chloride in the palaeochannel systems is much less than one 36 C1 half life. A further observation is that the 36 C1 data provide additional confirmation that the source of chloride is predominantly from concentration of marine aerosol inputs (McArthur et al., 1989). The chloride cannot have been derived from relict sequestered seawater from Miocene marine transgressions, in which case the 36 C1 data would fall along a low slope of probably less than 10 X 10~ 15 representing secular equilibrium sustained mainly by in situ production of 36 C1 from nuclear reactions within the rock matrix and subsequent leaching.

16 30 J. V. Turner et al. STABLE ISOTOPE DATA A plot of ô 2 H and ô 18 0 in rainfall and groundwater from the Black Flag-Gidji palaeochannel is given in Fig. 7. A more extensive data set of stable isotope compositions on regional groundwaters is given in Turner et al. (1993). The ô 2 H and ô 18 0 of the groundwaters show a trend of increasing stable isotope enrichment with increasing salinity. The Kalgoorlie meteoric water line (MWL) was derived from the amount-weighted mean isotopic composition of monthly rainfall samples collected at Kalgoorlie during It has a lower slope and intercept than, for example, the Perth MWL, due to the influence of light rainfall events that are partially evaporated during descent. The partial evaporation causes enrichment in the residual rainwater and such rainfall tends to show isotopic enrichment that lowers the slope of the local MWL. Low salinity groundwaters from fractured rocks plot close to the lower end of the MWL as expected for groundwaters derived from rainfall-recharge of larger rainfall events. Low-salinity, isotopically depleted groundwaters were sampled in the topographically elevated recharge areas of the catchment. These water types are the precursors to isotopically enriched, high salinity groundwaters in the palaeochannels that have undergone substantial evaporative concentration in salt lakes. Because the groundwaters plot along a line between these end-member groups and show the characteristic signature of evaporation, the high salinity groundwaters in 30 Kalgoorlie Rainfall E3 Black Flag - Gidgi Groundwaters 20 - Fractured Rock Groundwaters SEPT. '90 MAR. ' Kalgoorlie Meteoric Water Line 8 2 H= OCT. '90 X CM -10 MAR. '92,. AUG. '91 % SEPT. '91,--' OCT. '91,.- $ FEB. '92-30 JUNE'91 APRIL'91 Increasing salinity JULY '91 I -2 s18, Ô 1B 0(%o) Fig. 7 Stable isotope data from rainfall and groundwaters from the Black Flag-Gidji palaeochannel. The isotopically enriched groundwater values are from the eastern end of the palaeochannel.

17 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 31 sections of palaeochannel associated with salt lakes and the dilute end-member recharge groundwaters, can be linked in the recharge-discharge process. Because of the widespread isotopic enrichment observed in the palaeochannel groundwaters (including many that are several kilometres from the nearest salt lake but not discussed in this paper), it is evident that at some time in their history they have undergone isotopic enrichment by evaporation within a salt lake environment. DISCUSSION AND CONCLUSION The goal of this work was to evaluate the application of the 36 C1 method to the estimation of groundwater and chloride residence time estimation in hypersaline groundwaters in the palaeochannel aquifers of the Eastern Goldfields. The results demonstrate that various approaches and prior understanding were necessary in order to fully assess the meaning of the 36 C1 data. Despite the excellent information available on input of 36 C1 and CI to the regional aquifers it has proven difficult to extract more than semiquantitative information from the 36 C1 data. The primary reasons for this are: (a) The measured input 36 C1/C1 ratios from recharge waters and an extensive set of rainfall data (Keywood et al., 1994) show an excess fallout of 36 C1 in the region compared to that expected for 36 C1 fallout determined from cosmogenic 36 C1 production, combined with data on measured chloride fallout. It could not be discounted that the rainfall and low salinity recharge waters are free of thermonuclear 36 C1. (b) It was difficult to interpret the 36 C1 in this case to estimate groundwater ages because (i) groundwater residence and CI residence times were short compared to the 36 C1 half life, and (ii) even though we have comparatively excellent data on e.g. chloride fallout and 36 C1 input, the variability in these measurements results in large uncertainties in calculated residence times. The decline in 36 C1/C1 ratio along the Black Flag-Gidji palaeochannel was broadly consistent with residence time estimates based on 14 C, but the most favourable combination of extremes in uncertainties in 36 C1/C1 ratios needed to be used, and even then the 36 C1 method indicates residence times in excess of those from 14 C. Although the CI contents in rainfall were measured quite extensively, their uncertainty contributed to the uncertainty in the use of 36 C1 to estimate groundwater ages. In the palaeochannel system where chloride concentrations typically increase by four or five orders of magnitude from rainfall to the most concentrated brines as a result of evaporation, determination of the slope of a concentration-dilution regression in 36 C1/C1 space is essential and uncertainty in its value has a strong bearing on whether groundwater 36 C1 contents can be distinguished from the effects of concentration and dilution. The estimated accumulation time for chloride is probably an underestimate because we assume a closed system with respect to chloride discharge. Nevertheless the accumulation time is of the order of 10 5 years, and thus significant radioactive decay of 36 C1 is not expected, and neither was it encountered. The most likely evidence of thermonuclear derived 36 C1 can be seen in the high 36 C1/C1 ratio in the rainfall sample from Gindalbie Station. The much lower 36 C1/C1 ratios in palaeochannel groundwaters indicate that thermonuclear derived 36 C1 has not entered the majority of the aquifer system. The very low salinity groundwaters in recharge areas also appear to contain some thermonuclear derived 36 C1.

18 32 /. V. Turner et al. Despite some of these problems however, no gross inconsistencies in interpretation emerged in relation to residence time estimations based on 14 C, 36 C1, or CI accumulation times or hydraulic considerations, and this emphasizes once again the necessity of applying a number of isotopic and non-isotopic techniques to the study of groundwater systems. Acknowledgements Project funding was provided by the Minerals and Energy Research Institute of Western Australia (MERIWA) and six mining industry sponsors: Pancontinental Gold Mining Areas, Kaltails Mining Services, Western Mining Corporation, Newcrest Pty Ltd, Kalgoorlie Consolidated Gold Mines and Peko Gold. Funding was also received from the CSIRO Institute of Minerals, Energy and Construction under the Priority Research Funds programme. Support in kind was received from the Geological Survey of Western Australia. The project was coordinated by the Australian Mineral Industries Research Association (AMIRA). Mr Ken Wright, Mr Robert Woodbury, Mr Vit Gailitis and Mr Gerald Watson of the CSIRO Division of Water Resources are thanked for their contributions to the field and laboratory work. REFERENCES Andrews, J.N. & Fontes, J. Ch. (1991) Importance olin-situ production of chlorine-36,argon-36 and carbon-14 in hydrology andhydrogeochemistry.in:/sorqpe Techniques in Water Resources Development, IAEA, Vienna. Beard, J. S. (1973) The elucidation of paleodrainagepatterns in Western Australia through vegetation mapping. Occasional Pap. no. 1, Vegmap publications, Applecross, Western Australia. Bentley, H. W., Phillips, F. M., & Davis, S. N. (1986) ChIorine-36in the terrestrial environment. In: Handbookoj'Environmental Isotope Geochemistry, vol. 2, The Terrestrial Environment (ed. by P Fritz & J. C. Fontes), Elsevier, Amsterdam. Bettenay, E. &. Mulcahy, M. J. (1972) Soil and landscape studies in Western Australia. II: Valley form and surface features of the south-west drainage division./. Geol. Soc. Austral. 18, Bird, J. R., Calf, G. E., Davie, R. F., Fifield, L. K., Ophel, T. R., Evans, W. R., Keilet, J. R. & Habermehl, M. A. (1989) The role of 36 C1 and 14 C measurements in Australian groundwater studies. Radiocarbon, 31(3). Bunting, J. A., van de Graaf & Jackson, M. J. (1974) Palaeodrainagesand Cainozoicpalaeogeography of the Eastern Goldfields, Gibson Desert and Great Victoria Desert. Annual Report ofthe Geological Survey ofwesternaustraliafor 1973, Calf, G. E., Bird, J. R., Kellett, J. R., & Evans, W. R. (1988) Origins of chloride variation in the Murray Basin using environmental chlorine-36. In: Murray Basin 88, Abstracts, Chivas, A. R., Fifield, L. K., Davie, R., Bird, J. R., Ophel, T. R. & Kiss, E. (1987) Chlorine-36 investigations of Australian salt lakes. Australian National University, SLEADS Workshop, p. 23. Chivas, A. R., Fifield, L. K., Davie, R., Bird, J. R., Ophel, T. R. & Kiss, E. (1988) Chlorine-36 investigations of Australian salt lakes. SLÉADS Conference 1988, Salt Lakes in Arid Australia. Abstracts volume. Clarke, J. D. A. (1993) Stratigraphy ofthelefroy and Cowan palaeodrainages, Western Australia./. Roy. Soc. Western Austral. 76, Commander, D. P., Kern, A. M. & Smith, R. A. (1992) Hydrogeology of the Tertiary paleochannels in the Kalgoorlie region (Roe Paleodrainage). Geological Survey of Western Australia Hydrogeology Report, Record 1991/10. Commander, D. P., Fifield, L. K., Thorpe, P. M., Davie, R. F., Bird, J. R. & Turner, J. V. (1994) Chlorine-36 and carbon-14 measurements on hypersaline groundwater in Tertiary paiaeochannelsnear Kalgoorlie, Western Australia. Geol. Survey of Western Austral. Prof. Pap., Report ZT, Davie, R. F., Fifield, L. K., Bird, J. R., Ophel, T. R. & Calf, G. E. (1988) The application of 36 C1 measurements to groundwater modelling. In: Murray Basin 88, Abstracts, Davie, R. F., Kellett, J. R., Fifield, L. K., Evans, W. R., Calf, G. E., Bird, J. R., Topham, S. & Ophel, T. R. (1989) Chlorine-36 measurements in the Murray Basin: preliminary results from the Victorian and South Australian Mallee region. BMR J. Austral. Geol. Geophys. 11, Farrington, P., Salama, R. B., Bartle, G. A., & Watson, G. D. (1993) Accession of major ions in rainfall in the south western region of Western Australia. CSIRO Division of Water Resources, Divisional Report 93/1.

19 Chlorine-36 in hypersaline palaeochannel groundwaters of Western Australia 33 Fontes, J. -C. & Andrews, J. N. (1993) Comment on "Reintcrpretationof 36 CI data: physical processes, hydraulic interconnections and age estimations in groundwater systems" by E. Mazor. Appl. Geochem. 8, Griffin, T. J. (1990) Eastern Goldfields Province. In: Geology and Mineral Resources of Western Australia: Australian Geological Survey, Memoir 3, Western Jiang, J. J. et al. (1990) Measurements of the 36Ar(n.p.)36Cl cross section at thermal energies using the AMS technique. Proc. 5th Internat. Conf. on AMS, Paris (ed. by F. Yiou& G. M. Raisbeck), Nuclear Inst. Methods Phys. Res. B52. Kellett, J. R., Evans, W. R., Allan, G. L. & Fifield, L. K. (1993) Reinterpretation of 36 C1 data: physical processes, hydraulic interconnections and age estimations in groundwater systems discussion. Appl. Geochem. 8, Kern, A. M. & Commander, D. P. (1993) Cainozoic stratigraphy in the Roe palaeodrainage of the Kalgoorlie Region, Western Australia. Geol. Surv. Western Austral. Prof. Pap. Reportai, Keywood, M. D., Chivas, A. R., Fifield, L. K., Allan, G. R. & Cresswell, R. (1994) Chlorine-36 in Australian Rainfall. In: Fifth Australian Conference on Isotopes in the Environment, Abstracts, Mann, A. W. (1982)Physical characteristicsof thedrainagcsof the Yilgarn Block, South Western Australia. CSIRO Institute of Energy and Earth Resources, Division of Mineralogy Report no. FP 25. Mazoi, E. (1992) Reinterpretationof 36 C1 data: physical processes, hydraulic interconnections and age estimation in groundwater systems. Appl. Geochem. 7, McArthur, J. M., Turner, J. V., Lyons, W. B. &Thirlwall, M. F. Salt sources and water-rock interaction on the Yilgarn Block: Isotopic and major element evidence, (1989). Appl. Geochem. 4, Morgan, K. H. (1966) Hydrogeology of the East Muchison and North Coolgardie Goldfields. Geol. Surv. Western Austral. Annual Report 1965, Phillips, F. M. (1993) Comment on "Reintcrpretationof 36 C1 data: physical processes, hydraulic interconnections and age estimations in groundwater systems" by E. Mazor. Appl. Geochem. 8, Simpson, H. J. & Herczeg, A. L. (1994) Delivery of marine chloride in precipitation and removal by rivers in the Murray- Darling Basin, Australia. /. Hydrol. 154, Turner, J. V., Bradd, J. M., &. Waite, T. D. (1991) The conjunctive use of isotopic techniques to elucidate solute concentration and flow processes in dryland salinized catchments. In: Proc. Internat. Symp. on the Use of Isotope Techniques in Water Resources Development (Vienna, Austria, March 1991), IAEA-SM-319-1, IAEA, Vienna. Turner, J. V., Townlcy, L. R., Rosen, M. R. & Sklash, M. K. (1992) Coupling the spatial distribution of solute concentration and stable isotope enrichments to hydrological processes in hypersaline palcochannel aquifers. In: 7ih International Symposium on Water-Rock Interaction (Park City, Utah, USA, July 1992), Turner, J. V., Rosen, M. R., Milligan, N., Sklash, M. K. & Townley, L. R. (1993) Groundwater recharge studies in the Kalgoorlie Region. Minerals and Energy Research Institute ofwestern Australia Report no. 98. Project M146/P321. Turner, J. V., Townley, L. R., Rosen, M. R, & Milligan N. (1994) Groundwater recharge to palcochannel aquifers in the Eastern Goldficldsof Western Australia. In: Water Down Under (Proc. 25th Congressof the International Association of Hydrogeologists, Adelaide), Vol. 2B, IAH/IE Australia. Walker, G. R., Jolly,I. D., Stadter,M. H., Leaney,F. W., Davie, R. F., Fifield, L. K., Ophel, T. R. &Bird, J. R. (1991) Evaluation of the use of chlorine-36 in recharge studies. In: International Symposium on the Use of Isotope Techniques in Water Resources Development (Vienna, Austria, March, 1991), IAEA-SM319/2, IAEA, Vienna.