Journal of Hydrology

The proposed approach aims to estimate the flood extent and soil wetness using AMSR-E passive microwave data. The approach is applied over the Mackenzie River Basin, which is situated in northwestern Canada. The methodology is based on the polarization ratio index (PR), which is computed using AMSR-E 37 GHz, vertically and horizontally polarized brightness temperature values. The water surface fraction (WSF), which represents the fraction of flooded soil, was derived on a pixel-per-pixel basis. The fractional vegetation cover was added to the WSF calculation in order to take into account the temporal variation of the vegetation shading effect. The WSF derived from AMSR-E data, WSF(AMSR-E), was compared to those derived from the Moderate-resolution Imaging Spectroradiometer Terra instrument (MODIS-Terra) images (250 m), WSF(MODIS). A rating curve relationship was developed between the observed discharge and WSF(MODIS). It was noted that the WSF obtained from AMSR-E images systematically exceed those from MODIS, as they are formed from a combination of different contributions, including open water surface, flooded area and wetlands, which are abundant in the northern climates. Therefore, a wetness index was defined based on the difference between passive microwave and visible image responses. This index was able to qualitatively describe the temporal evolution of the wetness over the Mackenzie River Basin. The availability of discharge observations and passive microwave data leads to the definition of a consistent wetness index and soil moisture monitoring over the Mackenzie River Basin. A satisfactory agreement was noted between the wetness index, the precipitation, and the temperature values. The wetness index agrees well with the measured soil moisture.

In large scale field experiments over the past 10 years, long wavelength microwave radiometers have been used to map surface soil moisture with considerable success. These experiments, which include PIPE (87 & 89), Monsoon 90, Washita 92 and 94, and HAPEX-Sahel, covered a wide range of climatic regimes. The results from all of them have shown that the microwave emission at the 21-cm wavelength is a strong function of surface (0 to 5-cm) soil moisture. The salient results from these experiments are presented and compared. The use of the surface soil moisture data to determine geophysical parameters such as evaporative fraction, soil evaporation and soil hydraulic properties are discussed

Theoretical considerations are discussed, and design criteria and fabrication of a physical hydrologic model including a storm-simulating device are described. Data from several preliminary experimental tests indicate that the use of physical models of actual watersheds merits further investigation.

Construction and spatial and temporal properties for a 0.25° resolution gridded data set of monthly Penman–Monteith reference evapotranspiration estimates over the territory of the PR China (including Tibet) and adjacent areas (15°N–55°N, 65°E–135°E) for the period 1951–1990 are described. To account for the interaction between climate and the rugged topography of the study area the REGEOTOP procedure was used to incorporate the effects of relief forms into the interpolation.Evapotranspiration rates over much of China show a range of values (annual rates from 550–2300 mm) and variability comparable to precipitation. Monthly evapotranspiration rates are distributed more evenly over the year than precipitation, are out of phase with the summer precipitation peak and in some cases may reach winter rates comparable to those in summer. Hydrological studies based on idealized regular seasonal variation of evapotranspiration may contain considerable errors due to inherent seasonal fluctuations as compared to precipitation.High resolution gridded PET data that account for the influence of topography on climate are required to resolve the spatial heterogeneity of topography and land use in order to allow precise estimates of actual evapotranspiration and run-off. The spatial distribution of runoff appears to have remained fairly constant over most of China during 1951–1990 which stands in contrast to the anticipated increase in hydrological activity under global warming conditions.

A time series model of the ARMA class for seasonal, multisite applications is presented. Methods of estimating model parameters are developed and two methods (maximum likelihood and the method of moments) are compared. For model verification, a data generation experiment provides diagnostic checks at three levels: (1) properties of the model residuals; (2) preservation of short-term statistics (serial and cross-correlation); and (3) preservation of long-term characteristics (drought and reservoir storage properties). The model is then used to model bivariate, monthly river flow on the Yampa and White Rivers in northwest Colorado.

Hydrological modifications frequently result in wetland loss and degradation while wetland management, restoration and creation schemes rely upon further hydrological manipulations. These schemes can benefit from models which can accurately represent often complex wetland hydrological situations. Although the potential of the physically based, distributed model MIKE SHE to model wetlands has been demonstrated, a number of inadequacies in its channel flow component have been identified. These include difficulties in representing control structures and simulating inundation from channels. A coupling has been developed between MIKE SHE and the MIKE 11 hydraulic modelling system. This paper reports a coupled MIKE SHE/MIKE 11 model developed for a lowland wet grassland, the Elmley Marshes, in southeast England. Long term monitoring, supplemented by selected secondary sources, provided the necessary input, calibration and validation data. A procedure was developed to evaluate evaporation from ditch surfaces which could not be represented dynamically within MIKE 11. Two consecutive 18-month periods were used for model calibration and validation which were based upon comparisons of observed and simulated groundwater depths and ditch water levels. Model results were generally consistent with the observed data and reproduced the seasonal dynamics of groundwater and ditch water. The close association between flooding and both groundwater and ditch water levels was demonstrated. Topographic depressions are important for the initiation of flooding and are responsible for much of the shallow surface water in areas isolated from ditches. Deeper flooding occurs in areas which are inundated from these ditches. Results suggested that improvements could be made to the MIKE SHE bypass flow routine to enable it to more accurately represent macropore flow associated with soil cracking and swelling. Dynamic calculation of evaporation from ditch water surfaces would enhance the ability of the model to explore alternative water level management and climate change scenarios. The potential use of the model to investigate these scenarios is outlined.

Poor agreement between 3H/3He ages and CFC-11 and CFC-12 ages suggests that CFCs may not be conservative tracers in the Everglades National Park. 3H/3He ages were used to calculate the expected concentration of CFC-11 and CFC-12 in groundwater from wells 2 to 73 m deep. The expected concentrations of CFCs were compared to the measured concentrations and plots of the % CFC-12 and CFC-11 remaining offered no evidence that significant CFC removal was occurring in the groundwater at depths ≥2 m, suggesting that CFC removal occurs at shallower depths. Except where CFC contamination was suspected, CFC-11, CFC-12 and CFC-113 concentrations in fresh surface water were nearly always below solubility equilibrium with the atmosphere. Measurements of CFC-11, CFC-12 and CFC-113 in pore water indicate a 50–90% decrease in concentration 5 cm below the groundwater–surface water (GW–SW) interface. In the same 5 cm interval CH4 concentrations increased by 300–1000%. This suggested that CFCs were removed at the GW–SW interface, possibly by methane-producing bacteria. CFC derived recharge ages should therefore be viewed with caution when recharging water percolates through anoxic methanogenic sediments.

The distribution of groundwater and dynamic fluctuations in groundwater levels have direct impacts on the eco-environment of arid areas. Investigations of groundwater recharge in arid areas are typically limited by a lack of adequate meteorological and hydrogeological records. This study focuses on groundwater recharge in a seasonally flooded arid area within the Tarim Basin, China, with the aim of analyzing the influence of groundwater and topography on the response characteristics of overland flow. We conducted a simulation using a coupled MIKE SHE/MIKE 11 model over 112 days of the flood season and calculated the average water balance. Based on the properties of the multi-scale recognition of wavelets and the self-comparability of fractals in analyzing the detailed characteristics of groundwater diffusion and fluctuations in groundwater levels, a hybrid fractal–wavelet method was used to explain the recharge response associated with overland flow, distance from midstream, topography, and flooding depth. The results of the model simulations are generally consistent with observed data, indicating that the hybrid fractal–wavelet method is able to recognize the detailed characteristics of the groundwater response. Furthermore, the groundwater levels show a significant relationship with the orientation of the riverway prior to flooding. When flooding occurred, the groundwater levels showed a rapid response to changes in the depth of long-term overland flow. A total of 71.31% of the study area showed a strong correlation between groundwater level and the distance from midstream. The results demonstrate that the relationship exists for fluctuations in groundwater levels of more than 1.4 m. Variations in the height of the water table were significantly influenced by topographic elevation; in contrast, slope and aspect had little effect. In conclusion, the above results indicate that the proposed methodology is applicable for the management of water resources in arid regions. The modeling and hybrid fractal–wavelet method study allowed quantification of the processes affecting groundwater levels and provided an insight into their implications in exploring groundwater level management.

We describe field observations and numerical simulations of the environmental tracers 3H, 85Kr, CFC-113 (C2Cl3F3), SF6 and Ne in groundwater. The field site is a well-characterised shallow aquifer in central Germany, consisting of basalts of Miocene age, overlain by up to 15 m of loess deposits. A two-dimensional numerical model was used to simulate tracer transport at the study site. Simulated and observed tracer concentrations show acceptable agreement for most wells and tracers. Due to the variable thickness of the loess cover, residence times of 3H in the unsaturated zone are highly variable with values ranging from 1 to more than 30 years. This effect explains the observed variability of 3H in the saturated zone. Excess air in groundwater requires correction of measured concentrations of the dissolved gas tracers. A maximum excess of the SF6 content in water compared to the theoretical solubility equilibrium concentration of 28% was observed. A novel iterative method is used to correct for excess air. CFC-113 transport seems to be retarded. On the basis of the effective porosity for SF6, 85Kr and 3H transport, a retardation factor of R=1.5 for CFC-113 with respect to SF6, 3H and 85Kr can be derived. Together with non-adsorbing tracers, such as 85Kr or SF6, CFC-113 can therefore serve as a reactive retardation tracer exploring mean sorption characteristics of the aquifer material. Extrapolation of CFC-113 sorption characteristics to organic contaminants with comparable physicochemical properties (e.g. chlorinated hydrocarbons) offers an opportunity for improved assessment of the behaviour of this important group of contaminants in groundwater.

The most complete historical series of instrumental data available, spanning more than a century, on rainfall, temperature and outflow of a karst spring obtained from gauging stations in the south of the Iberian peninsula were analysed by means of spectral and correlation analyses and continuous wavelet analyses. Annual periodicity of the rainfall and temperature distributions was constant over more than 100 years, although weaker (6-month) periodicities have also been observed, as well as rainfall and temperature periodicities of 5 and 2.5 years, which have also been recorded in other areas of Europe. These multiannual scale components can be explained by climatic variations or effects described in the literature in connection with the North Atlantic Oscillation (NAO) and are likely to be the same as the climate variability at decadal to annual scale detected in several proxy data from geological records. No long-term trends in the distribution of precipitation and temperature were detected.

On the 12th and the 13th November 1999, the Aude region received amounts of rainfall that locally exceeded 500 mm. This extreme rainfall event resulted in one of the century's most significant floods in the Aude river catchment area and produced remarkable flash floods in some of its tributaries. Peak flood discharges from many upstream watersheds (area smaller than 100 km2) exceeded 10 m3/s/km2. An extensive post-flood investigation was carried out in 2000 to collect information concerning the floods in seven of these upstream tributaries and to analyse the hydrological behaviour of the watersheds during this extreme rainfall event. As well as analysing the available rainfall data and estimating peak discharges on the basis of high water marks, information related to the timing of the floods, as provided by eyewitnesses, was gathered. A simple hydrological model, based on the SCS method and the kinematic wave equation, was used to guide the hydrological interpretation. This study revealed some interesting aspects of the hydrological rainfall-runoff relation during flash flooding that could be important for estimating flood frequencies and flood forecasting: (1) around 200 mm of rainwater were retained in the catchments without contributing to the flood flow, (2) the response of the watersheds to the intense storm bursts was late and relatively sudden, (3) no significant effects resulting from different types of land use could be identified. These conclusions, based on inaccurate data, were validated for the entire Aude river catchment. They also concur with some other reported flash flood data.

A method for estimating flash flood peak discharge, hydrograph, and volume in poorly gauged basins, where the hydrological characteristics of the flood are partially known, due to stage gauge failure, is presented. An empirical index is used to generate missing hourly rainfall data and hydrologic and hydraulic models performs the basin delineation, flood simulation, and flood inundation. The peak discharge, hydrograph, and volume, derived from the analysis of measured hydrographs in a number of non-flood causing rainfall events with operating stage gauge, were used for calibration and verification of the simulated stage-discharge hydrographs. An empirical equation was developed in order to provide the peak discharge as a function of the total precipitation, its standard deviation, and storm duration. The peak discharge for a flash flood case based on the empirical equation was in close agreement with the results from a number of consolidated methods. These methods involved hydrological and hydraulic modeling and peak flow estimates based on Manning’s equation and post flash flood measurements of the maximum water level observed at the control cross-section, for the 13–14 January 1994 flash flood in the Giofiros basin on the island of Crete, Greece. This method can be applied to other poorly gauged basins for floods with a stage higher than that defined by the rating curve.

Recent concern for problems of soil degradation and the offsite impacts of accelerated erosion has highlighted the need for improved methods of estimating rates and patterns of soil erosion by water. The use of environmental radionuclides, particularly caesium-137 (137Cs), as a means of estimating rates of soil erosion and deposition is attracting increasing attention and the approach has now been recognised as possessing several important advantages. However, one important uncertainty associated with the use of 137Cs measurements to estimate soil erosion rates is the need to employ a calibration relationship to convert the measured 137Cs inventory to an estimate of the erosion rate. Existing calibration procedures are commonly subdivided into empirical relationships, based on independent measurements of soil loss, and theoretical models, that make use of existing understanding of the fate and behaviour of fallout radionuclides in eroding soils to derive a relationship between erosion rate and the reduction in the 137Cs inventory relative to the local reference value. There have been few attempts to validate these theoretical calibration models and there is an important need for such validation if the 137Cs approach is to be more widely applied. This paper reports the results of a study aimed at validating the use of a simple exponential profile distribution model to convert measurements of 137Cs inventories on uncultivated soils to estimates of soil erosion rates. It is based on a small (1.38 ha) catchment in Calabria, southern Italy, for which measurements of sediment output are available for the catchment outlet. Because there is no evidence of significant deposition within the catchment, a sediment delivery ratio close to 1.0 can be assumed. It is therefore possible to make a direct comparison between the estimate of the mean annual erosion rate within the catchment derived from 137Cs measurements and the measured sediment output. In undertaking this comparison, account was taken of the different periods covered by the measured sediment output and the erosion rate estimated using 137Cs measurements. The results of the comparison show close agreement between the estimated and the measured erosion rates and therefore provide an effective validation of the use of the 137Cs approach and, more particularly, a profile distribution calibration model, to estimate soil erosion rates in this small catchment. Further studies are required to extend such independent validation to other environments, including cultivated soils, and to different calibration procedures.

The turnover of radioactive caesium was studied experimentally and theoretically in a forested catchment that was covered by snow during the wet deposition of radioactive nuclides from the Chernobyl accident. The study spans from 1 week before to 8 years after the deposition event. A fraction of the catchment is covered by a mire (16010). From the edge of the mire a stream channel runs to the outlet of the catchment. Two phases of decreasing activity concentration in the stream water were found in addition to a positive influence of runoff on the activity concentration in the stream. The half-lives for 137CS in the stream water corresponding to the early and the later phase were estimated by non-linear regression to be 6.5 days and 4 years, respectively. During the first phase, which corresponded to the initial snow melt in 1986, 6.8% of the deposition was lost from the catchment, whereas the slow secondary loss during the following 8 years was 1.8%. The main contribution to the yearly discharge of 137Cs occurred during spring and autumn when the areal contribution to saturated surface runoff was highest. The remaining deposition in soil a few years after the fallout was significantly lower in the mire than in the surrounding forest. By using the calculated activity concentration of 137Cs in the stream water together with the remaining deposition in the different biotopes and information on stream flow for the catchment it was concluded that the loss originated from the mire. During the initial phase 44% of the deposition was lost from the mire, and during the following years the yearly loss was 30% from the fraction that constantly undergoes saturated surface runoff and 2% from the drier fractions of the mire. Until the end of the study it was not possible to demonstrate any loss from the recharge areas (podzol and cambisol soils), which means that physical decay will govern the decrease in activity in these areas.

A 13C and 14C isotope subroutine called ISOTOP has been written for WATEQF, a FORTRAN IV version of the water-analysis treatment program (WATEQ) translated and modified by Plummer et al. from the original PL1 version by Truesdell and Jones. This subroutine performs a series of simulations on water samples where 13C and 14C data are available by assuming various initial pH— −log (PCO2) values under open-system conditions and computing the δ13C of the CO2 gas phase to be compared with the actual water sample's value and an age for each simulation applied. Some of the models use total Ca plus Mg as a measure of whole-rock carbonate dissolved under closed-system conditions while others make corrections for gypsum dissolution and ion-exchange reactions.In addition, the subroutine may be used to compute age differences between pairs of water samples along the same flow path using congruent and incongruent carbonate dissolution models. In the simulations applied, carbonate and bicarbonate ion pairing and complexing are considered.

Stable isotope data for dissolved inorganic carbon (DIC), carbonate shell material and cements, and microbial CO2 were combined with organic and inorganic chemical data from aquifer and confining-bed pore waters to construct geochemical reaction models along a flowpath in the Black Creek aquifer of South Carolina. Carbon-isotope fractionation between DIC and precipitating cements was treated as a Rayleigh distillation process. Organic matter oxidation was coupled to microbial fermentation and sulfate reduction. All reaction models reproduced the observed chemical and isotopic compositions of final waters. However, model 1, in which all sources of carbon and electron-acceptors were assumed to be internal to the aquifer, was invalidated owing to the large ratio of fermentation CO2 to respiration CO2 predicted by the model (5–49) compared with measured ratios (two or less). In model 2, this ratio was reduced by assuming that confining beds adjacent to the aquifer act as sources of dissolved organic carbon and sulfate. This assumption was based on measured high concentrations of dissolved organic acids and sulfate in confining-bed pore waters (60–100 μM and 100–380 μM, respectively) relative to aquifer pore waters (from less than 30 μM and 2–80 μM, respectively). Sodium was chosen as the companion ion to organic-acid and sulfate transport from confining beds because it is the predominant cation in confining-bed pore waters. As a result, excessive amounts of Na-for-Ca ion exchange and calcite precipitation (three to four times more cement than observed in the aquifer) were required by model 2 to achieve mass and isotope balance of final water. For this reason, model 2 was invalidated. Agreement between model-predicted and measured amounts of carbonate cement and ratios of fermentation CO2 to respiration CO2 were obtained in a reaction model that assumed confining beds act as sources of DIC, as well as organic acids and sulfate. This assumption was supported by measured high concentrations of DIC in confining beds (2.6–2.7 mM). Results from this study show that geochemical models of confined aquifer systems must incorporate the effects of adjacent confining beds to reproduce observed groundwater chemistry accurately.

Measurements of the carbon-14 concentration of water in 33 boreholes in a saline unconfined aquifer in southern Australia have been interpreted to show that local recharge under natural conditions has a mean annual value which falls between 0.1 and 0.2 mm yr−1.The chloride concentration of groundwater is log-normally distributed suggesting that its source is rainfall. The estimate of mean annual recharge to the aquifer made using chloride is 0.25 mm yr−1.

Delineation of the spatial distribution of ground water that is several thousands of years old can provide an important piece of the puzzle in the evaluation of long-term, ground water resource sustainability under pumping conditions. Ground water for municipal and local water supplies within the Palouse Basin of eastern Washington and northern Idaho are derived almost entirely from two basalt aquifer systems. Decades of continual water level declines in the deeper aquifer system in response to interstate pumping have suggested that this high transmissivity, low storativity aquifer system is being mined. Average water level declines in the deeper aquifer system have been on the order of 0.46 m (1.5 ft) per year for the past 40-plus years. Carbon-isotopic measurements on dissolved inorganic carbon were used to provide information on the relative ages of the ground water pumped from various locations within the basin. Mass balance modeling of hydrochemical reactions together with steep, downward vertical gradients throughout the basin suggest that the ground water being withdrawn currently required several thousand years to traverse spatially variable sequences of loess, fluvial sediments, and basalt flows with associated sedimentary interbeds, that overlie the primary producing zones in the basin. Restrictions to the vertical movement of recharge water such as low permeability, overburden deposits, well-developed fragipans, and thick, low permeability, sedimentary interbeds increase in the eastern portion of the Palouse Basin compared to the western portion. This has resulted in accumulation of the oldest ground water in the topographically highest areas of the basin adjacent to the basin divide.

Helium isotope and concentration characteristics were determined for a suite of groundwater samples from the Amadeus Basin in Central Australia. Two study areas include a wellfield south of Alice Springs, and the Dune Plains and Mututjulu aquifers near Uluru. Measurements of 36Cl/Cl and 14C on the same sample suite enable us to assess the relative applicability of the three groundwater chronometers over a range of anticipated groundwater residence times (ages), and to investigate possible causes of discordant ‘ages’ derived from the different groundwater dating techniques.

Repetitive sampling for radioactive isotope analysis can be used to more accurately estimate the age of groundwater. Groundwater was sampled in wells along the Coastal Aquifer of Israel for tritium, 14C, and 13C analysis in the 1970s as well as the 1990s.Examination of the tritium values from two sampling periods on the tritium bomb curve enabled us to overcome the ambiguity of tritium data, in which a single value could yield two ages, and allowed a better age estimate. Therefore, repetitive sampling over an extended time interval was found to be effective in assisting calculations of groundwater ages through the use of the tritium curve over the 20-year period. The age, obtained by repetitive sampling can be used to determine the dilution factor (Q) of the initial 14C value in the aquifer. This corrected initial value can, thus, be utilized for better age determination of older samples in the deeper sub-aquifers.In most cases, based on the presence of detectable levels of tritium throughout most of the aquifer, the groundwater in the coastal aquifer was determined to contain a young component with a transit time of less than forty years. However, the 14C levels of most of the same groundwater samples were found to be lower than atmospheric levels (34–102 vs. >100 pMC), probably due to water–rock interactions rather than decay due to old age. Some older waters exist in the lower sub-aquifers of the deeper wells (120–140 m) where both lower 14C values (42–69 pMC) and very low levels of tritium were found (0.1–0.7 TU).

In the absence of pure piston-flow, the 14C decay ages of groundwater can be related to groundwater residence times only in the context of a flow or mixing model. Discrete-state compartment models offer a means of constructing flow models that can simulate flow regimes ranging from pure piston-flow to perfect mixing and can thus be used to relate 14C decay ages to residence times. A three-dimensional steady-state flow model of a portion of the Tucson Basin aquifer, south-central Arizona, U.S.A., was constructed using the theory of discrete-state compartment models and calibrated with the spatial distribution of adjusted 14C decay ages. The model provides estimates of groundwater residence times (about 100–15,000 yr.), vertical flow in the aquifer and long-term average annual recharge (0.048 km3). Although the amount of time spanned by the model is long and presents problems with respect to the constancy of hydrologic processes, it may suggest a novel use for such models: that of delineating the paleohydrology of a particular region.

Estimation of groundwater recharge in arid and semi-arid areas is difficult due to the low amount and variability of recharge. A combination of radiotracers investigation based on simple mixing models allows direct investigation of relatively long-term renewal rates of an aquifer. The recharge process of the shallow Continental Terminal aquifer in the Iullemeden basin (Niger) was investigated using a geochemical and isotopic approach. This study investigates the area in the one degree square of Nianey (13–14°N, 2–3°E). In this area, recharge is highly heterogeneous and mainly occurs through a drainage system of temporary streams and pools during the rainy season. Heterogeneity of the recharge is reflected through the wide variation in electrical conductivity and oxygen-18 content of the groundwater. The carbon-14 activity range for most of the groundwater falls between 69 and 126 pmc showing pre and post-aerial thermonuclear test recharge. Two renewal rate models have been investigated: the first one models a well-mixed reservoir and the second one is derived from a piston flow model, in which mixing is in equal proportions. Major ions in tritium data analyses allow exclusion of non-representative samples and confirm the carbon-14 renewal rate estimations. Both models give similar results for the relatively low renewal rate investigated in the area. Using carbon-14, the mean annual rates of groundwater renewal range from 3 to 0.03% of the aquifer volume with a median of 0.1%. Assuming the median is representative of the overall renewal rate of the area, the recharge rate is in the order of 5 mm a−1. The shallow aquifer recharge extends from the last small humid period (around 4000 a) up to now. High recharge rates are found in depressions whereas low recharge occurs below the plateaux.

The calculation of model 14C ages for groundwater often relies on a correction being made for the incongruent dissolution of aquifer carbonate. Correction equations are particularly sensitive, in certain ranges, to variation in the δ13C value input for dissolving carbonate. This study uses a combined and δ13C approach to define the source and mean δ13C value of dissolving carbonate in the Lincolnshire Limestone aquifer of eastern England. This δ13C value is most appropriately used in 14C correction equations and yields the most potentially accurate groundwater model ages.

Stable isotopes (δ2H, δ18O and δ13C) and radiocarbon (14C) have been used in conjunction with chemical data to evaluate recharge mechanisms, groundwater residence time and palaeohydrology within the confined Dilwyn sand aquifer in the Gambier Embayment of the Otway Basin. This aquifer does not receive recharge down-gradient of the hydraulic hinge-line and data have been interpreted along two discrete flow lines. The mean residence time of groundwater (determined by 14C) in the confined aquifer from the hydraulic hinge position to the sea (a distance of about 50 km) along an inferred flow path is approximately 12 800 years. The corresponding hydraulic travel time calculated from Darcy's law is approximately 49 000 years. The apparent discrepancy may be a result of eustatic sea-level lowering during the last glacial. Because the groundwater system is hydraulically connected to the sea, lower sea-level would result in increased gradients and a decrease in groundwater residence time. Variations in stable isotopic composition along flow lines suggest a number of recharge mechanisms. Stable isotope data indicate progressive depletion of 2H and 18O in the groundwaters over the past 30 000 years. Groundwaters older than about 10 000 years B.P. were recharged either during a cooler climate climatic regime (lower precipitation/evapotranspiration and temperature) and/or the vapour source(s) had travelled over greater continental mass than those recharged over the past 10 000 years. Lower chloride concentrations in some of the older groundwaters indicate lower evapotranspiration rates in the recharge areas before 10 000 years ago.