The age and hydrological history of Blue Lake, South Australia

Abstract Three sediment cores from the Blue Lake, a groundwater fed lake of volcanic origin in South Australia, have been investigated using a range of chemical and isotopic parameters. The 14C activity of both the inorganic and organic carbon fractions of the sediment decreases monotonically with depth. The rate of change with depth is greater for near-surface samples with an apparent hiatus in sedimentation rate at about 7000 yr B.P. Estimates of age for the precipitated authigenic carbonate, after correction for dilution with dead carbon from the groundwater, agree well with calculated ages from the organic carbon fraction of the sediment. We suggest the lake is much older than previously proposed using other dating techniques. Variations in the δ13C and δ18O composition of the authigenic carbonate reflect different residence times of dissolved inorganic carbon and water in the lake caused by changes in the lake level. During periods of hydrologic steady-state, it is suggested that relative changes in the temperature of the lake can be seen in δ18O changes in authigenic carbonate. Blue Lake has been undergoing sedimentation for at least 28,000 years, including two lengthy periods of hydrologic steady state. The lake, for a large proportion of its existence, was much shallower while for the last 7000 years has maintained a level close to the present one. The influence of pumping from the lake for urban water supply during this century is reflected in the isotopic composition of carbonate in the sediment.


Introduction
Blue Lake is the largest of three lakes formed in a volcanic complex in south-east South Australia (Fig. 1). Due to its importance as the major source of water for the city of Mount Gambier, there is considerable interest in establishing its water balance. (Turner et al., 1984;Ramamurthy et al., 1985). In addition, there has been considerable effort establishing the geological history of the volcanic complex and that of the surrounding area (Sheard, 1978;Barton and McElhinny, 1980;Barbetti and Sheard, 1981;Blackburn et al., 1982).
This study uses a combination of stable isotope (6i3C and 6180 values for carbonate and organic carbon fractions), and 14C dating of lake sediment to establish a hydrological history of the lake from its formation to the present-day.

The age of the Crater Lakes
Researchers have approached the dating of the crater lakes in this area by either dating material associated with the volcanic activity when the lake was thought to have formed (i.e dating the volcanits) or dating sediment from within the lake.

Dating the volcanics
The volcanic complex, in which the Blue Lake is situated, lies on the western border of the Newer Volcanic region covering a large proportion of central and western Victoria and the south east of South Australia.
Included in this group of Newer Volcanics until recently were the volcanic complexes at Tower Hill and Mt. Eccles located approximately 150-200 km east of the Mt. Gambier complex. These areas were thought to be formed 7000-8000 years ago (Gill, 1972). A group of 16 eruptive centres is located approximately 30 km to the northwest of the Blue Lake complex. Although definite ages are unknown for these "Older Volcanics", they may range from lo6 to 20,000 yr B.P. (Solomon, 1951;Sprigg, 1952;Firman, 1969).
The Blue Lake complex and that of Mt. Schank, located 10 km to the south of the Blue Lake, lie within sands stratigraphically above those of the northern group and are therefore thought to be younger than the Older Volcanics ( Barbetti and Sheard, 1981). A radiocarbon age of 18,100~350 yr B.P. (Polach et al., 1978) has been given for a sample of soft charcoal below the associated tuff from Mt Schank. However, dating by thermoluminescence of volcanic debris in the area suggests an age of about 5000 years (Smith and Prescott, 1987). Sheard (1978) suggested that the age of formation of the Mt Gambier complex was less than 5000 yr B.P. from 2 dates for charcoal associated with volcanic tuff around the volcanic complex while Blackburn et al. (1982) found a considerable range in 14C age (4OO%SOOO yr B.P.) for a further 11 samples of charcoal.
Hence, from dating volcanics associated with the its formation, Blue Lake was believed to be included in the "Newer" volcanics and was thought to be, at most, 7000-8000 years old. Moreover, there is a considerable range in the suggested age of the lake from dating this type of material in this environment.

Dating qf'sediment
There have been few studies dating lake sediments from the "Newer" volcanics region because of difficulties finding large enough samples containing organic carbon to date (pre-AMS) and/or finding a method to correct for 14C dilution via dead carbon (see appendix). Barton and McElhinny ( 1980) found ages for organic sediment in a shallower lake in the same volcanic complex as Blue Lake increased consistently up to about 6000 yr B.P. with 14C ages of about 14,000 and 38,000 yr B.P. in layers of aragonite and calcareous tuff respectively recorded at greater depths.
At the time of publishing, the older ages were discounted because of the possibility of 14C "dilution" with dead carbon. More recently, evidence from dating lake sediment (Head et al., 1991) suggests an age of formation of between 20,000 and 30,000 yr B.P., considerably older than the age suggested from dating volcanics.
Clearly, there is a trend in these studies for the sediment to apparently pre-date the formation of the lake as determined from dating volcanics.

This study
The Blue Lake has several attributes which make it particularly interesting to study and applicable to carbon dating. The lake is currently oligotrophic (Turner, 1979). The 613C signature of the organic matter in the sediment is considerably depleted (< -26%0, see Fig. 11) indicating that organic material persisting in the sediment had a terrestrial rather than aquatic origin (Gupta and Polach, 1985). Electron microscopy showed that this material was most likely slivers of wood and charcoal loo-500 urn in length and up to 100 urn in diameter. Hence, no dilution effect has been assumed when calculating ages from the i4C activity of organic material. Hence, 14C dating of the organic fraction is possible without consideration of the "dilution effect" (sometimes called the hardwater effect). The carbonate in the lake sediment is largely authigenic and hence, dating of this fraction is possible if an estimate for the dilution in the 14C of the lake water (by incoming groundwater) can be made.
The major aim of this study was to infer historic changes in the water balance of the Blue Lake from the isotopic composition of sediment sampled. Changes in the water balance of the lake were inferred from excursions in the 6i3C and 6180 composition of the sediment while carbon dating of organic and inorganic carbon fractions provided a time frame for these changes. Results of 613C analyses on the carbonate sediment have been used to correct for the reduction in 14C activity of carbonate precipitating in a groundwater fed lake (the "dilution" effect). In this paper, we suggest two methods to estimate this dilution effect with each method being applicable to different sections of the sediment core. Using this approach, two semi-independent estimates of age, based on dating the organic and inorganic carbon fractions, may be made for the sediment at any particular depth.

The environment of Blue Lake
The Blue Lake is located within the Gambier Embayment of the Otway Basin. It has the potential to be fed by two major aquifers: (1) the confined Dilwyn Formation which consists of interbedded sands and lignitic clays and (2) the unconfined Gambier Limestone which is predominantly bryozoal calcarenite (Fig. 2). These are separated by a confining layer of glauconitic silt several metres thick. The hydrogeological description of the region is given by Love ( 1992) and a detailed stratigraphy by Sheard (1978). The lake bed is relatively flat with a mean depth of about 70 m, volume of 37 x lo6 m3 and a surface area of about 6 x 10' m2. The lake is steep-sided with a very small catchment and hence, runoff to the lake is low. Evaporation from the lake has been estimated at 1200 mm yr~' (Turner, 1979) and exceeds the average rainfall by about 500 mm yr -'. The rate of pumping of water from the lake has increased markedly over the last 40 years (Fig. 3) and currently varies between 3-4 x lo6 m3yr-'.
While the water level of the Blue Lake is approximately the same as that of the surrounding unconfined aquifer, just north of the lake the potentiometric surface of the confined aquifer is about 9 m higher than the unconfined (Turner, F. W. J. Leaney et ul.lPalaeogeograph~~,Palaroclimatolog~v,Palaroecology 118 (I995)  1979). Hence, there is hydraulic potential for upward leakage from the confined aquifer water into the lake. The currently accepted estimate for the unconfined:confined groundwater input to the lake is around 4:l (Ramamurthy et al., 1985).

Methods
Three sediment cores, each 5 m in length were taken at equi-spaced locations from the bottom of the lake using a Mackereth corer. The depth of water was 72 m at site 1 (shallowest part of the lake floor) and 78 m at the other two sites. In addition to these cores, a sample of the top 0.30 m of sediment was taken using a cold finger corer.
The Mackereth cores, stored in 50 mm polyvinyl chloride tubes, were frozen then cut lengthwise to give two half cylinders which were stored at a temperature of 3°C to prevent evaporation prior to sampling and analysis. The cold finger core was sampled at 10 mm intervals, dried at 70°C and stored in glass vials upon arrival at the laboratory.

I. Dry bulk density, particle size unalysis and electron microscopy
Measurements of dry bulk density were made on all cores, except for the cold finger core, to permit correction for different degrees of compaction that may have occurred during collection or during sediment accumulation.
For cores 1 and 2, 10 mm slices of sediment were removed at approximately lOO-mm intervals over the length of the core, dried at 70°C for 12 hours and weighed. For core 3, the complete core was divided into 50 mm sections for density measurements.
For particle size analysis, a total of 130 samples were taken from the three cores and allowed to air dry. The relative amounts of sediment in the ranges ~20 urn, 20-200 urn and >200 urn were measured using a particle size settling method described by Lewis (1983) for 100 of the samples with a further 30 samples analysed in more detail (50 urn intervals from O-400 urn). The particle fractions for 10 of the samples were determined by electron microscopy.

r4C analysis on carbonates and organic fractions
Sufficient sediment was sampled to obtain 5 g of inorganic carbon for analysis of 14C. Nine samples were taken from core 1, 11 samples from core 2 and 7 samples from core 3, pretreated with 1 A4 NaOCl to remove organic carbon and analysed for r4C using the method described by Gupta and Polach (1985).
A further 9 sediment intervals from core 2 and two intervals from core 3, each 20 mm, in length were sampled for 14C analysis of the organic carbon fraction. The sediment was pretreated to obtain the acid/alkali insoluble carbon fraction (Gupta and Polach, 1985) and combusted at 900°C in an evacuated quartz tube in the presence of CuO wire to form CO, (Boutton et al., 1983). Graphite targets for accelerator mass spectrometric determination of 14C were prepared by catalytic reduction of CO, according to the method described in Vogel et al. (1987) and subsequently, AMS measurements were made at the Lawrence Livermore National Laboratory Centre for Accelerator Mass Spectrometry (Davis et al., 1990).

Stable isotope analysis
Samples for 6r3C and 6180 analysis were taken at 50 mm intervals for cores 1 and 2, -300 mm intervals for core 3, and at 10 mm intervals for the cold finger core. After pretreatment with 1M NaOCl, -100 mg of each sample reacted with phosphoric acid in a water bath maintained at 30°C (McCrea, 1950) and the CO, evolved, analysed for 13C/12C and 18O/'6O ratios using a VG Micromass 602 D mass spectrometer.
The acid/alkali insoluble carbon samples used for 14C analyses were subsampled then analysed for ii13C composition using a Europa "Tracermass" mass spectrometer after conversion to CO? via a "rob0-prep" automatic preparation line. All data is expressed in d-notation as %I relative to Pee Dee Belemnite (PDB).

Results
Results for all of the analyses and measurements are given here with a more detailed explanation of the combined use of these results to provide a hydrologic history of the lake in subsequent sections.

Core description
Cores 1-3, taken at depths of 72, 78 and 78 m, respectively, display similar features with laminae present from the surface of the sediment to a well defined boundary at -2.00 m for core 1, -2.88 m for core 2 and -2.63 m for core 3. The laminae are narrower and more obvious for the bottom third of this interval and appear compressed over a shorter depth interval for core 1. Below the zone where laminae commence, the material is uniformly dark coloured in all cores.

Dry bulk density measurements
There is a relatively uniform increase in dry bulk density from -400 kg mm3 at the top of the cores to -1400 kg rnp3 at varying depths for the 3 cores, and a uniform density of -800 kg m m3 for the homogenous dark material at the base of the cores (Fig. 4). For each core, a linear relationship between density and depth was approximated using a least squares fit for the sediment down to the uniform material and constant thereafter (as indicated for core 3 in Fig. 4). While it was not possible to measure density for the cold finger core, it is likely to be less than or equal to the minimum values seen for the Mackereth cores. A value of 400 kg m-3 was chosen for this section and all subsequent data was plotted against cumulative mass M (kg rn-') rather than depth.

Purticle size analysis and electron microscopy
Each of the three cores display similar particle size distributions.
The particle size distribution for core 3 is shown in Fig. 5. From the surface to A4 2 1800 kg rnp2 the particle size distribution is approximately constant and dominated by < 200 pm particles.
Between M z 1800-2400 kg m ~', the particle sizes vary considerably. The dark material below the laminae is homogeneous, dominated by particles ~20 pm in diameter. Abrupt changes in some parts of the core suggests disruption during sedimentation, the implications of which will be discussed later.
For the 30 samples analysed in greater detail, there was a bimodal distribution of particles with the majority of particles either < 150 pm or > 250 pm with most material < 1.50 pm. There were few particles in the lOOG200 pm size range. When viewed under the electron microscope, rhomboid calcite crystals were clearly present in particle fractions < 1.50 pm but were rarely seen in fractions >250 pm. Particles larger than 250 pm were typically well-rounded having no well-defined shape. Although individual calcite crystals measured lo-30 pm in diameter, it was common to see several joined together in the loo-150 pm range.
We considered particle fractions ~200 pm to represent carbonate of authigenic origin and particles >200 pm to be detrital. A correction was made to the stable isotope composition and to the 14C activity of the carbonate analysed (see appendix). We acknowledge, in hindsight, that it would have been easier to sieve the samples prior to analysis and then use only the ~200 pm fraction. However, the amount of material >200 pm is < 5% and so the corrections are small, except for the intervals ~2400-1650 kg rn-' and for the top of the core where corrections are up to 25%~ After acidification of the larger particle size fractions, elongated slivers, most likely being charcoal and wood, were seen when viewed through the electron microscope. This represented the majority of organic material analysed for 13C and 14C composition.

Curbon-carbonate rnuteriul
The results for 14C activities from carbonate precipitate after allowance for the inclusion of detrital material, (Aauth), are given in Fig. 11. These values cannot be used directly in age determinations because the 14C activity of the sample may be reduced by dilution with Total Dissolved Inorganic Carbon (TDIC) from the groundwater (see appendix).
Two methods were used to estimate the i4C activity of the TDIC @(lake)) in lakewater at the time precipitation occurred (see appendix for details). The first method assumes that the lake was in a hydrologically steady-state for a sufficient time to establish a linear relationship between age and cumulative mass, at least over some part of the core (the upper part, in this case). The second method follows an approach by Fontes et al. (1983) using the 613C composition of the carbonate precipitate to estimate the 14C activity of the TDIC in the lakewater at the time of precipitation.

Using method I
Using data from all cores, plots of A4 against the log of the 14C activity of the authigenic carbonate (In A,,,,(M)) are linear from the surface to M-1650 kg me2 suggesting a constant sedimentation rate and hydrological steady-state during this time interval.
Hence, for this part of the core, method 1 is appropriate for determining the 14C activity of the TDIC. Using a least-squares fit for the 14C data for the interval M=O-1650 kg mP2, we estimate values for the sedimentation rate, k, and A( lake) as 0.229 kg mm2 y-l and 71% Modern Carbon (pMC) respectively. (r2 =0.98, Fig. 6). The error calculated from counting statistics alone is less than 100 yr (la) which is small compared to the overall error using this method of analysis. The overall error associated is estimated at between 500 and 1000 years as seen in the comparison between ages using actual sample activities (solid Using method 2 Fontes et al. (1983) proposed a method for calculating A(lake) at the time of precipitation from the 6r3C composition of precipitate for lakes where there was incomplete re-equilibration between gas and aqueous phase CO,. The lakes were described as being only weakly mixed, had a relatively short carbon residence time, a pH in the vicinity of 778 and immobilization of the organic material in the sediments of the basin without recycling of carbon. A description of the method is described in the Appendix.
The Blue Lake is consistent with most of the criteria proposed for this method and has a pH range from h 8.0 to 8.4. An estimate for A( lake) using this method of 77-79X modern carbon (pMC ) at the time of sampling agrees well with values measured by Turner ( 1979) of 8 1 k 9 pMC (see Appendix).
This suggests that this method provides a reasonable determination of A( lake) and hence age of the carbonate at any depth in the Blue Lake sediment.
The major error using this method arises from estimating the relative contributions of confined and unconfined groundwater to the lake, as these aquifers have significantly different 14C activities. In principle, this range may have varied from totally confined to totally unconfined groundwater input and hence we have calculated an age range based on these extremes (Fig. 11, Ages 2 and 3). Such an error estimate is much more realistic than that calculated from counting statistics alone for dating carbonate sediment. Estimates of the ages for sediment, assuming no hardwater effect, have been included for comparison with corrected ages (Fig. 11, Age 4).

Carbon-14 organic material
Results from dating bulk organic fractions of the sediment are also presented in Fig. 11. This material has a predominantly terrestrial rather than aquatic origin as seen from electron microscopy and the 13C composition.
Standard counting errors range from 50 to 200 years for these samples.
Ages determined from 14C analyses for organic and carbonate samples are presented as a function of cumulative mass in Fig. 7 and Fig. 11. Ages indicated by the solid line assumes a constant sedimentation rate and value for A (lake) (method 1). The joined bars indicate ages calculated using the method of Fontes et al. (1983) with the younger age estimated assuming groundwater input totally from the confined aquifer and the older age assuming input totally from the unconfined aquifer (method 2). The estimates using methods 1 and 2 agree well for the top 1650 kg m-' suggesting the mechanism for calcium carbonate precipitation and the estimate for A(lake) proposed by Fontes et al. (1983) is appropriate for the Blue Lake. Furthermore, there is good agreement between ages determined from 14C analyses of organic material and the carbonate sediment, except at around 6000-7000 yr B.P. where the 14C ages determined on organic material are scattered and appear younger than the carbonate material.

Stable isotope data
Results for 613C and #*O analyses on samples from cores 1, 2, 3 and the cold finger core are given (Figs. 8 and 9)  were obtained for detrital limestone samples ( 10 samples). The results for core 3 analyses ( 14 samples shown as asterisks in Fig. 8) are similar to those for core 2 (70 samples) at the same cumulative mass and, in the following, the results from core 2 are discussed. Comparing the results for the cold finger core with the other cores indicates that approximately the top 20 kg me2 (50 mm) of the sediment was lost during sampling using the Mackereth corer. Both cores 1 and 2 show similar 613C and 6r80 profiles although samples for core 1 were only analysed to a cumulative mass of about 2400 kg mm2 and show compressed excursions in parts of the core.
For convenience, results of analyses are presented for four distinct sections of the cores believed to represent different evolutionary stages of the lake (Figs. 8 and 9). The first section concentrates on the uniform dark sediment at the bottom of the core; the second section, the laminae to M-1750 kg mm2 (-28,000&-7000 yr B.P.); the third section, the laminae from M-1750 kg m p2 to M= -80 kg rnp2 (-7000 yr B.P.--1900 AD). The final section examines the surface sediment collected via the cold finger (M-0%30 kg m m2) and shows the impacts of managementinduced changes to the water balance of the lake since pumping began early this century.
The 6r3C compositions are -0.7% at depths greater than M = 2050 kg me2 for core 1 and M= 2500 kg rnp2 for cores 2 and 3 (i.e. section 1; >29,000 yr B.P.). The 6r80 composition of sediment samples for the same interval ranges from -0.0 to 0.7% for all cores. This is approximately that measured for detrital limestone. For section 2 (-29,000. -7000) there is little change in 6r3C composition until the top where all cores show -3% decrease. Two negative excursions are seen in the 180 composition of the sediment for this section.
For the interval M~SO~1750 kg mm2 (section 3, 50 -7000 yr B.P.), the 613C composition is approximately constant except for an -1.5% positive excursion at M= 1250 kg mm2 (5400 yr B.P.). The 6180 composition for samples in this section generally increase toward the surface reaching a plateau value of around 0.5% at about 3000 yr B.P. The cold finger core (section 4) shows a negative shift of -3%0 in both 6i3C and #*O towards the top of the profile. 6 auth is determined by the 6r3C composition of the lake at the time of precipitation (& from Fontes et al., 1983) and the fractionation between the carbonate and bicarbonate phases E,. If Bauth is close to the 613C composition of the groundwater, 6,, , the carbon budget of the lake is dominated by groundwater inflow. However, if it is close to the isotopic composition of the bicarbonate in equilibrium with the atmosphere, &, then the carbon budget is dominated by exchange with the atmosphere.
The ratio of carbon input from the groundwater to the lake relative to carbon exchange with the atmosphere is dependent on many factors but in this study we have considered the lake surface area to volume ratio to be most important.
As the Blue Lake has very steep sides and is almost cylindrical (Fig. 2), this ratio approximates to the depth and we suggest that more negative values for aauth may represent periods when the lake was deeper and/or displayed a shorter mean water residence time. Alternatively, periods when the lake was shallow should result in an enriched signature for dauth.

Section I (uniform dark sediment)
The stable isotope composition of this section is uniform with depth which suggests that this material was homogeneous during deposition. However, as no cores sampled to the bottom of the sediment, we cannot conclude whether or not the sediment is homogeneous over the entire sediment record. The most likely scenario resulting in the formation of such uniform sediment is the

3(
880(sed) o/o0 rel PDB S13C(sed) %O rel PDB Fig. 9. 'T and "0 composition of carbonate sediment function of cumulative mass for the cold finger core. mixing of volcanic debris and material from the as a aquifer matrix in a mudpool after cessation of geothermal activity. Electron microscopy does not show any evidence for the precipitation of calcite in this section. Almost 90% of the sediment is ~20 urn and well rounded which is consistent with the hypothesis above. Under such conditions, the sediment in this section would be the last to settle and would consist of uniform small particles. Hence, although this material consists of small particles, we believe the carbonate in it is detrital limestone, mechanically eroded during the formation of the lake. The isotopic composition of the sediment in this interval being close to that of detrital limestone gives support to this hypothesis.
Because most of the inorganic carbon in this section is likely to be detrital in origin, it is impossible to estimate the age from the carbonate fraction. Similarly, the 613C and 6180 compositions for sediment in this part of the core are unlikely to provide useful information on the evolution of the lake. The 14C age of organic carbon for the sample at the top of this section probably provides the most reliable estimate for the time at which the mud pool ceased its activity. As such, it also provides a minimum age of formation of the lake (ie >28,000 yr B.P.). 6.2. Section 2 luminue above the uniform dark sediment to A4 = 1750 kg m ' / > 28,000-7000 yr B. P. ) Laminae are clearly present for this section of the core, although they are neither frequent nor regular enough to correspond to seasonal sedimentation or precipitation. More likely, they represent periods ranging from years to possibly centuries during which time precipitation took place at a greater than average rate for this section. These periods were separated by similar intervals during which precipitation was much slower. The large changes seen in the particle size distribution in this part of the core suggests significant changes in the amount of detrital material washed into the lake during this time. This is consistent with a shallower lake and the potential for more erosion during this time from the large steep-sided catchment than seen in subsequent periods. The analyses for cores 2 and 3 are similar while core 1 shows evidence of a compressed or possibly truncated depositional record. The lower rate of sedimentation (Fig. 7) may be explained by the fact that the surface level of core 1 is approximately 6 metres higher than that of the other cores. In times of very low lake level, the core 1 site may have been exposed and been subject to erosion or at least a much slower rate of deposition.
If this was the case, then the amount of sediment in this section of core 1 is likely to be less than the other 2 sites. It is difficult to determine whether the sediment record for core 1 is complete although. from visual observations of the core, there are no obvious discontinuities.
The agreement seen in the sedimentation rate for the other two cores and their lower elevation suggests that the records for these are complete.
In the following discussion, results from the stable isotopic analysis of core 2 are discussed as sediment from this core appears to represent a more complete history of the lake (Fig. 8).
The 613C of the authigenic carbonate in the precipitate has a mean value of -0.7%0 for most of this section. This is more positive than that for most of the precipitate suggesting a much shallower lake during this period. Near the top of this section of the cores (M=2000-1750 kg rnm2; 15,000-7000 yr B.P.) the 613C of the authigenic carbonate material decreases markedly to about -2%0 suggesting the lake became deeper during this time.
The 6i80 of authigenic carbonate is dependent of the temperature and the water balance of the lake. The historical water balance of the lake is not known well enough to make quantitative estimates of the palaeo lake temperatures.
However, it is possible to use the 6180 data of the carbonate sediment to suggest qualitative palaeo-climatic changes during periods where the water balance of the lake is constant.
Under steady-state conditions positive excursions for the #*O composition of the carbonate sediment are indicative of lower lake water temperature while negative excursions indicates a period during the evolution of the lake when the lake temperature was higher. For the section of the core M=2500-2000 kg mm2 (28,000-15,000 yr B.P.), the 613C composition of the sediment suggests that the lake was consistently shallow or well-mixed and hence, any changes in the 6180 composition of the carbonate sediment should be related to temperature changes. The depleted values seen in the al80 composition of the carbonate sediment at about M= 2400 kg rnp2 (28,000 yr B.P.) suggest a relatively warm lake at this time. The reasonably constant 6i3C composition (Fig. 8) for this section suggests the lake water balance was in a relatively steady-state for most of this period. The ratio of the exchange between the lake and atmosphere to the input inorganic carbon from the aquifer system, as indicated from the Lth composition, suggests a much deeper lake during this period than that prior to 7000 yr B.P.
There is evidence of a disruption to the water balance of the lake at about 5000 yr B.P. (M= 1250 kg m-") as indicated by an increase in 613C in the sediment of about 2%0. The apparent reduced water inflows temporarily changed the water balance of the lake and probably lasted for several centuries. There are no obvious changes in the 6180 of the sediment around 5000 yr B.P. However, 6180 values go through a minimum between 7000 and 5000 yr B.P., suggesting that the lake reached its maximum level at a time when temperature was higher than present. Wasson and Donnelly ( 1991) suggest that 5000-7000 yr B.P. was the warmest period in SE. Australia during the late Holocene.
Another explanation for the depleted 6180 and to a lesser extent 613C composition of the sediment at this time would be an increase in the precipitation/evapotranspiration ratio in the area. This is consistent with data for other lakes in S.E. Australia that recorded maximum depths during this period (Wasson and Donelly, 1991). It is probable that the combined effects of high temperatures and increased precipitation resulted in the depleted isotopic signature of the sediment during this time.

AD)
The amount of carbonate in this section of the core was insufficient to allow 14C dating (AMS was unavailable at time of analysis). Despite this, several changes in the stable isotope composition of the sediment can be explained by the history of pumping from the lake over the last century.
The rate of pumping from the lake during this century (Fig. 3) changed in three stages: (1) a gradual increase in the rate from 1900 until about 1950 (from 0.1 to 0.8 x lo6 m3.yp1), (2) a more rapid increase from 1950 until about 1970 (from 0.8 to 4 x lo6 m3 yy') and (3) an approximately constant rate of withdrawal since then. Similarly, three regimes may be seen in the (513C composition for this section of lake sediment (Fig. 9); (1) a slow but gradual decrease in the 613C composition of the sediment from M-80 kg me2 to M-20 kg m -* followed by (2) a much more rapid decrease from M-20 kg m-* to M-6 kg me2 and (3) a constant composition from there to the surface. The decrease in 613C of lake sediment nearer the surface reflects a reduction in the mean residence time of water in the lake resulting in 613C and 6180 approaching that of groundwater. However, we suggest that in this instance, the increase in groundwater flow during this period is induced by pumping from the lake rather than a change in the lake level.
If the three stages of pumping do indeed correspond to the three 613C regimes, then approximately 80 kg rnp2 of sediment has precipitated in 80 years with a mean sedimentation rate of 1 kg -2 m y-l for this period. This is four times the sedimentation rate suggested prior to pumping and is not unexpected given increased groundwater inflow and hence, TDIC input induced by pumping.

Hydrological historic of the Blue Luke
From dates defined in this study, Blue Lake formed >28,000 yr ago. For the period 28,000~15,000 yr B.P., the lake was, on average, shallower than today and at times was probably less than 10 m at the deepest point. At the start of this period, we suggest that the lake water temperature was relatively high and gradually decreased reaching a minimum between 22,000 and 15,000 yr B.P. Sedimentation rate during this time was at most 0.03 kg rn-* y ' From 15,000 years ago, the level of the Blue Lake increased to a level slightly above that seen at present by about 7000 yr B.P. From -7000 yr B.P. until pumping from the lake commenced at the start of this century, the lake was in hydrological steady-state with a sedimentation rate of -0.224 kg rn-' y-'. During this time, it is likely that the lake water temperature was highest between 7000 and 5000 yr B.P. and decreased until around 4000 yr B.P. It has remained approximately constant since that time. There is evidence that the water balance was disrupted possibly for a period of 5OOGlOOO years at about 5000 yr B.P. This time corresponds to several estimates of 14C age of charcoal associated with volcanic tuff in the immediate area (Blackburn et al., 1982;Fergusson and Rafter, 1957) and with estimates for volcanic activity from thermoluminescence dating by Smith and Prescott (1987). Hence, it is possible that this disruption may have been a result of renewed volcanic activity in the area.
Since pumping began in about 1900 AD, the hydrological balance of the lake has been perturbed and groundwater inflow has increased. Current sedimentation rates are approximately 1 kg me2 yr-' considerably higher than that prior to pumping.
It is likely that. as pumping rates have levelled off, the lake has reached a new hydrologic steady-state.
The hydrological history of the Blue Lake presented here may be compared with the inferred aquifer response to the lowering of sea level lowering during the last glacial (Y 18,000 yr B.P., Love, 1992). Love suggested that piezometric heads of the 2 aquifers at the Blue Lake site (Fig. 10) would have been considerably lower than at present for the period prior to 7000 yr B.P. with a minimum at about 18,000 yr B.P. The correlation between calculated groundwater levels and lake level, as indicated from the 13C signature of the carbonate sediment, suggests that the lake level has been, and continues to be, dominantly groundwater controlled. Moreover, the correlation between sedimentation rate and depth of water suggests that the precipitation of calcite (which contributes the majority of sediment) is and was controlled primarily by the water depth and thus is due chiefly to mixing of inflowing groundwater with lake water. Prior to approximately 15,000 yr B.P., we suggest that the lake probably did not exceed 10 m in depth and at this stage, precipitation at the site of core 1 would have been significantly less than at the other 2 sites as observed.
Changes in lake water temperature, suggested only for the time interval when the carbonate (Love 1991 ) I I 20 000 10000 0 Age yrs BP precipitate displayed constant 613C compositionindicating a constant water balance but changing 6180 concentration-are relative only. However, the periods of inferred high lake temperature agree well with data summarised by Wasson and Donnelly (1991) (Fig.  10).

Conclusions
The combined use of r4C, 13C and 180 analyses, primarily on the carbonate fraction of sediment from the Blue Lake, has enabled a reconstruction of its hydrological history. The major conclusions from this study are: -The age of formation of the Blue Lake has been estimated, using a combination of r4C and stable isotope data, to be at least 28,000 yr B.P. This is much older than that suggested by other workers (Sheard, 1978;Blackburn et al., 1982).
-14C analyses on carbonate material may be used to date sediment providing the ages are (1) Age estimated using Method I.
(2) Age estimated using method 2 with groundwater input assumed to be wholly from the confined aquifer.
(3) Age estimated using method 2 with groundwater input assumed to be wholly from the unconfined aquifer (4) Age estimated assuming no dilution (hardwater) effect. Fig. 11. Age estimates for carbonate and organic material from lake sediment. corrected for detrital carbon and the i4C dilution from groundwater input. -The two methods used for estimating the "dilution" effect are valid subject to the limitations presented in this paper. The appropriate conditions for using each method are based primarily on the i3C composition of the sediment. -Lake levels for the Blue Lake were much lower than at present prior to 7000 yr B.P. and lake levels for the Blue Lake have been, and continue to be, groundwater controlled. -The residence time of water in the lake has shortened since pumping began due to induced groundwater flow. Sedimentation rate for this lake is dependent on the depth of water in the lake with greater depths resulting in greater authigenic carbonate production. Several studies, have used proxy records to establish past variations in lake levels and infer palaeoclimatic changes such as increased precipitation or evaporation rates. This study shows that it is necessary to ensure that the lakes used in such studies are not groundwater controlled and that such lakes are not merely responding to changes in groundwater level that are, in turn, responding to other factors which may not be closely related to regional climatic changes. Laboratories and the authors wish to acknowledge the efforts of J. Davis and his colleagues. Further thanks go to S. McClure from the CSIRO Div. Soils for electron microscopy of the samples.

Appendix-estimating the amount of detrital carbon when dating carbonate sediment
When using carbonate material to date sediment, some of inorganic carbon in the sediment may not be authigenic carbonate precipitate but limestone which has washed from the walls of the lake. The Gambier Limestone was formed during the late Tertiary and hence has zero 14C activity. Incorporation of this material into sediment and subsequent analysis for 14C for the inorganic fraction would result in an apparent age older than that of the authigenic carbonate.
While it is difficult to remove this material quantitatively, a more accurate 14C activity for the authigenic carbonate may be determined by removing the larger fractions of the sediment prior to analysis. If this is not done prior to analysis, the measured carbon-14 activity of the sample may be corrected after analysis by assuming the coarse fraction (> 200 urn) to be inorganic carbon from the Gambier Limestone incorporated in the sediment. In this study detrital material was considered to be >200 urn. The where f& is the sediment fraction greater than 200 lull. For this study, the correction to 14C activity for most of the samples is less than 5%. However, for one sample in section 2, the correction is 33%.

Estimation of the 14C activity of the dissolved bicarbonate in the lake, A( lake), at the time of precipitation
The 14C activities of authigenic carbonate in sediment samples (Aauth) cannot be used directly calculate the age of sediment because the activity of the dissolved bicarbonate from which the calcium carbonate has precipitated (the dilution effect) is unknown for all but the present time. This is sometimes referred to as the hardwater effect. If the mean residence time of water in a lake is long, 14C exchange between the HCO,-in the lake and atmospheric COZ will result in the 14C activity of the HC03-approaching that of dissolved bicarbonate in equilibrium with atmo- spheric CO, (minimum dilution effect). If the mean residence time of water is short and input of groundwater HCO,-dominates, the i4C activity of the HC03-in the lake will be much less and approach that of the groundwater HCO,-(maximum dilution effect). Two possible approaches to estimate this effect can be used.

Method 1
Estimating the dilution efSect during periods when the lake approximates hydrological steady-state While uncertainties in the dilution effect limits the use of inorganic i4C data in dating carbonate precipitate, a relatively accurate estimate for this value may be made if the hydrological balance of the lake has remained steady for a considerable length of time.
If the hydrological balance of any lake is in a steady state for a length of time then it is likely that during that time, precipitation rate, sedimentation rate and the activity of the TDIC in the lake will be constant.
This being true then the cumulative mass, M kg mm2 will be proportional to the age t(y) of the sediment and hence: Also the age may be determined radiometrically as
> where A( lake) = steady-state activity of TDIC in the lake (pMC) Eq. 4 holds true for any particular depth in the core profile. However, this equation is a good approximation for any discrete sampling interval providing the sampling interval and range of sediment age in the sampling interval is small compared to the overall age of the core section. The cumulative mass at which the sample is taken may then be approximated by the mean of the cumulative mass for the core segment. In this study, the core segments sampled spanned at most 100 kg m ~' (450 yr) compared to the 1650 kg m ' (7000 yr) for section 3 of the cores.

Combining
Eqns. 3 and 4: M= -8033 In Aauth(M) + 8033k in A( lake) Hence, if a steady state situation prevails, cumulative mass, M, should be proportional to the log of the activity of the sediment, ie In&,,,(M)). Hence, M should be linear with In&,,,(M)), and the constant sedimentation rate constant can be calculated. The mean 14C activity can be calculated from the intercept on the cumulative mass axis (= exp(intercept/8033k)).
For this study, 613C for the authigenic carbonate is approximately constant for all core sections from M = 0 to -1650 kgm-' suggesting a period of hydrological steady-state during this time.

Method 2
Using the 613C composition of authigenic carbonate precipitate The following equation was derived by Fontes Estimation of the "dilution effect" may also be et al. (1983) to determine A(lake).
made by consideration of the i3C balance of the lake. The 613C signature of the sediment reflects the 13C of the lakewater which, in lakes with low biological activity, in turn reflects the ratio of exchange of CO, between the lake and the atmosphere compared to input of TDIC from the groundwater.
A( lake) = (6) 6CO2(,, is the 613C composition of atmospheric CO2 at the time of precipitation and is considered to be -6.4% for the Mackereth cores (pre-1900) and -8% for the top of the cold finger core (Leuenberger et al., 1992).

ACO2w
is the activity of atmospheric CO, at the time of precipitation and is considered to be 100 pMC for the Mackereth cores and 130 pMC for the top of the cold finger core (Leuenberger et al., 1992) &,, is the 613C composition of the dissolved bicarbonate in the groundwater. A,, is the 14C activity of the dissolved bicarbonate in the groundwater expressed as percent Modern Carbon (pMC). &,=4232/T("K)-15. 10 (Mook, 1980) sg= -9483/T("K)+23.89 (Mook, 1980) Possible temporal variations in A,, and Sbs increase the error associated with estimating A(lake) using this method. Blue Lake is fed by both confined and unconfined aquifers, the relative contributions of which are likely to have changed during the history of the lake.
The unconfined and confined aquifers have Bbs --13.0 and --11.3%0, respectively with values for the confined aquifer more uniform than those measured in unconfined (Love, 1992). Values for the 14C concentration are again very uniform for the confined aquifer (A,_ -3 pMC) in the vicinity of the lake. However, for the unconfined aquifer A,, varies markedly (20-40 pMC) with a mean value -30 pMC. The error associated with estimation of A(lake), if Abs is unknown, becomes less important as the relative fraction of atmospheric CO, exchange to groundwater derived carbon increases (i.e. as bauth becomes more positive). To indicate the extent of this induced error we have calculated all ages ( Fig.7 and Fig. 11) on the basis of both total input from the unconfined aquifer (older age) and total input from the confined aquifer (younger age).
Following the estimation of A( lake) using either method 1 or 2, the age of the carbonate precipitate expressed as years before present (yr B.P.) is then calculated by the following equation Age(yr B.P.) = -8033 In [A.,,,/A(lake)] No allowance has been made for the possible addition of 14C-free CO, to the atmosphere due to volcanic activity although for some time intervals this may well have been a possibility.
Testing the 14C dilution correction estimated using method 2 Seven samples of water from the Blue Lake collected during 1976/1977 had a measured 14C activity for DIC of 81+9 pMC (Turner, 1979). This may be compared to the estimate for A(lake) using 13C data from the carbonate precipitate at the top of the cold finger core Eq. 6. 6 auth of surface precipitate (Fig. 9), has been estimated at -5%0. The ratio of the unconfined:confined groundwater input at this time is estimated at between 0.7:0.3 and 0.8:0.2 resulting in a mean 14C activity for groundwater entering the lake of between 22-25 pMC (Ramamurthy, 1983;Turner et al., 1983). The value used for 6 C02(a) and for ACozcaj are -8%0 and 130 pMC, respectively (Gupta and Polach, 1985). If these values are substituted into Eq. 6, A(lake) of Blue Lake at this time is estimated to be between 83-93 pMC, in good agreement with that measured by Turner ( 1979).