Development of risk-based groundwater operating rules: a case study of Siloam Village, South Africa

Abstract

This study developed operating rules for groundwater supply from a probabilistic (risk-based) approach. Groundwater supply systems are often operated without relating groundwater yield/availability to demand which makes groundwater resource planning and management challenging and unpredictable. Risk-based approaches for developing groundwater operating rules comprehensively incorporate assurance of supply and also account for uncertainty due to model inputs, model structure and climate variability. A groundwater resource unit (GRU) was delineated and its hydrogeological conceptual model developed. Automatic curve matching was used to identify appropriate aquifer models and test solutions for estimating hydraulic characteristics (storativity, transmissivity and hydraulic conductivity) based on Aquifer Test Solver (AQTESOLV) Pro version 4.5. Limited groundwater levels and rainfall data were infilled and/or extended using Output Error-Nonlinear Hammerstein Weiner (OE-NLHW) and non-parametric regression (NPR), respectively. Performances of these models were based on relative error (RE), correlation coefficient (COR), root mean square error (RMSE), coefficient of determination (R2) and Nash Sutcliffe coefficient of efficiency (NSE). A program for generation of monthly groundwater levels for the GRU was coded in FORTRAN based on the revised version of the Pitman model (referred to as GW-PITMAN model). The model was calibrated using groundwater levels from a neighbouring borehole due to lack of observed representative data for the GRU. Validation was done by establishing the realistic nature of simulated runoff, recharge and groundwater levels. A Variable Length Block (VLB) bootstrapping model was used for simultaneous generation of stochastic inputs (rainfall, evaporation and groundwater levels) of the groundwater operating rules model. Operating rules were developed from statistical analysis of 100 base yields for the GRU simulated from 5-year long stochastically generated sequences (with length of 34 years) of rainfall, evaporation and groundwater levels. The hydrogeological conceptual model indicated presence of faults and diabase dykes which influence preferential flow paths and storage of water in the aquifer. Identified aquifer test solutions were found to be suitable for estimation of hydraulic characteristics, since they had generally good model fits and low mean residual errors. Heterogeneous aquifer types were identified though leaky aquifer dominated. Storativity, transmissivity and hydraulic conductivity values ranged from 0.0003-0.060, 0.78-12.3 m2/day and 0.074-0.460 m/day, respectively, indicating limited storage with potential for local groundwater supply for private consumption. Graphical fits for observed and estimated rainfall and groundwater levels were mostly comparable, though scatter plots indicated cases of underestimation and overestimation of observed values. R2, COR, NSE, RMSE and RE values were 0.76 and 0.7, 0.87 and 0.84, 0.75 and 0.68, 3.67 and 3.03 mm and 30 and 29% for both calibration and validation runs, respectively, for NPR model. R2, COR, NSE, RMSE and RE were 0.99 and 0.86, 0.97 and 0.93, 0.99 and 0.84, 0.03 and 0.01 m and 0.08 and 0.11% for both calibration and validation runs, respectively, for OE-NLHW model. The models were therefore found to have efficient calibration and validation, and were thus, suitable for data extension. Estimated groundwater levels, streamflow and groundwater recharge for both calibration and validation runs of the GW-PITMAN model, generally fluctuated with changes in rainfall, indicating that they are realistic. Majority (9 out of 10) of the historical statistics were mostly well preserved by VLB, except for skewness. Historic highest groundwater levels were also not well preserved. Superimposing the cumulative demands on the base yield curves and analysis of percentages of water demands that can be supplied indicated that the groundwater system could not meet the water demands at all times. To promote sustainable multipurpose use of water that can enhance rural livelihoods, allocating water using priority classification was found to be essential. Operating rule curves for groundwater supply were derived using a risk-based approach. The operating rule curves indicated that if priority classification is used all water demands are met up to maximum groundwater level of 25 m. The developed operating rule curves are therefore expected to improve water supply to both domestic and productive water uses, if they are adequately implemented and hence improve livelihoods. The procedures followed in developing risk-based groundwater operating rules for Siloam Village were summarised to assist in their application in any delineated groundwater resource unit. Though minimal infrastructure is available to support implementation of the operating rules, additional monitoring boreholes are required to aid in estimation of average groundwater levels for further calibration and validation of the GW-PITMAN model. Detailed geological and geophysical investigation are required to improve on characterisation of the GRU and its hydrogeological conceptual model. Undertaking a study of this nature in other areas including those which are data-scarce could promote wide implementation of risk-based groundwater operating rules.