Water Balance Modelling of the Tandjari Reservoir in Burkina Faso
Summary
The construction of reservoirs is a widely used strategy for dealing with limited water availability by capturing runoff. In Burkina Faso reservoirs supply water for; irrigation in agriculture and for both human consumption and livestock. Despite reservoirs being used for a century in Burkina Faso, data on the water balance of reservoirs and their upstream catchments is scarce and knowledge of the importance and applicability of these data is insufficient. In addition, increasing water demands and climate change are calling upon adequate statistics with regard to water availability in reservoirs.
Determining the reservoir water balance and enhancing knowledge on relevant input parameters for a hydrologic model, trend analysis of water utilization, and the impact of possible climatic and water demand changes have been the main goals of this research.
This study focused on the Tandjari reservoir, located in southeast Burkina Faso, for which the various components of the water balance were determined. The reservoir water balance describes the change in water storage, which is dependent on rainfall runoff, groundwater inflow, rainfall on the water surface, evaporation, water consumption, discharge and infiltration into the reservoir bottom. The distributed hydrologic model “Soil and Water Assessment Tool” (SWAT) was used to simulate the water balance components over time. The water balance was simulated for the past five years (2012-2016) and for a future period of ten years (2037-2047). The latter was based on climate and water demand scenarios. Model calibration and validation were based on observed reservoir storage (2012-2016), and uncertainties in model output were quantified using the program “Sequential Uncertainty Fitting Algorithm” (SUFI-2).
The calibration showed that the best simulation fitted the observed storage generally well. 62% of the simulated reservoir storage was within the 95 Percent Prediction Uncertainty (PPU) band. The quality of the model, indicated by the thickness of the 95PPU band and referred to as the R-factor, was 0.49 (out of a perfect 0 and quite reasonable around 1). The acceptable model performance is reflected by the values of the Kling-Gupta efficiency (KGE) and the coefficient of determination (R²), which were respectively 0.83 and 0.70. The PBIAS value of -0.0 % indicated that the model did not over- or underestimate the observed reservoir storage. Validation of the model, for which a different dataset was used, showed that the 95PPU band bracketed 75% of simulated data and the thickness of the band, indicated by the R-factor, was 0.44. The KGE and R² for the validation period (May 2017 to December 2017) were respectively 0.78 and 0.93. The improved behavioral parameters for the validation indicate that the model reliability improved by using a dataset that was obtained from a weather station, which had a better location with respect to the catchment. Moreover the observed reservoir data used in the validation procedure were obtained by automated water level recordings instead of manual readings, which minimized bias due to human intervention.
The water balance, modelled over a period of five years (2012-2016), showed that more than half of the annual reservoir outflow evaporated, about a quarter infiltrated, and less than one-fifth of the total outflow was used for consumption. Infiltration does not need to be considered as direct loss, since nearby villages with wells are likely to benefit from the infiltration, which ensures year-round groundwater recharge. About 1.4 million cubic meters (MCM) of water is annually infiltrating, however, no conclusions can be drawn on how much of this quantity is actual accessible by groundwater abstraction. The general trend in reservoir water utilization was that by far most water was withdrawn by the drinking water company for the water supply to the people in Fada N’Gourma, followed by irrigation water for agriculture and finally water withdrawals by local residents and livestock. The reservoir was not over-allocated or under stress. The final-to initial-volume ratios, indicate the reservoir is not emptying; hence water is withdrawn from the reservoir at a sustainable rate. At the minimum observed level, the reservoir’s capacity was reduced with about 70%, which was enough to sustain important activities including fishing. However, climate change and increasing water demands ask for careful monitoring of the water balance. As water demand gets doubled by 2030, it is expected that during a period of drought the reservoir cannot meet the water demand with the risk that the reservoir empties during the dry season. Climate change could have a worsening effect if the frequency and intensity of droughts will increase. The interruption of water withdrawals over a long period will adversely affect local residents who are depended on this resource.
The results of this study highlight the importance of data collection and analysis, in order to gain understanding of the reservoir behavior and its adjacent catchment in a hydrological context. The quantification of the water balance components is an important task for water management authorities because; it will support the development of operational strategies for water supply and water allocation, it also offers support to the design and implementation of water policy.