Quantifying the Effects of Upstream Farm Dams on inflows into the Gaborone Dam in Botswana: An integrated approach

Abstract:

One of the major challenges of hydrological modelling in semiarid areas is the high spatial and temporal variability of rainfall and subsequent associated hydrological processes, coupled with an inherent non-linearity of response between rainfall and runoff. The problem often gets worse due to a lack of instrumentation of good spatial coverage, which increases input errors and uncertainties when spatial rainfall estimates are made from limited observations for use as input to rainfall-runoff models. This particular problem is well documented for many catchments in the world, including the semiarid southern Africa and has largely promoted the use of lumped models over distributed models in data scarce areas which often fail to adequately represent hydrological processes, and, thus, in addressing key water resources management issues at sub basin levels. One of the major issues these models are unable to address, is the effect of upstream land use changes on flow regimes in the downstream watershed. The Gaborone dam catchment located within a 20 km radius from Gaborone city in Botswana has been experiencing challenges of reduced inflows into the dam, despite some recorded heavy storms in the head streams and within the catchment. Recent studies indicate that there are more than 200 farm dams spread across the 400 km² catchment which may have led to reduced inflows into the dam, representing a main source of water supply to the greater Gaborone area. However, due to insufficient rainfall recording instruments and flow gauging stations in the catchment, no studies had been able to adequately address runoff generation processes and associated inflow dynamics in this important catchment. Through the present study, an experimental hydrological site has been established, consisting of five automated weather stations and two gauging stations to capture spatial rainfall and flow variability within the catchment. This study has taken an integrated approach by considering (i) a multi-site rainfall modeling approach based on generalized linear models, as well as (ii) land use and remote sensing products as input to (iii) a process-based distributed hydrological model known as JAMS/J2000. As shown by improved calibration results, preliminary results are encouraging that dominant processes are better represented.