A Hydrological Drought Analysis of River Mpanga.

Abstract

This study’s main objective was to carry out hydrological drought prediction on R. Mpanga, in the Upper-Mpanga sub-catchment and it was achieved by using a SWAT model to simulate streamflows from 1990 to 2020, and using these flows as an input into 2 stochastic models (a SARIMA and an Exponential Smoothing model) to forecast monthly flows and identifying instances of drought using 3 thresholds i.e., environmental flow, monthly & annual mean of the 30yr streamflow time series. A calibrated and validated SWAT model was used for streamflow simulation. Measured streamflow data from MoWE was obtained and used for calibration (1998-2004) and validation (2005-2010). The model’s performance in terms of the NSE was 0.635 and 0.610 during calibration and validation respectively. In terms of R2 and PBIAS, the model yielded 0.766 & 0.776 and -21.9% & -23.7% during calibration and validation respectively. This indicated that the SWAT model was adequate for the application of streamflow estimation (Moriasi et al., 2007). The stochastic models were calibrated using the flows from SWAT for 3 forecasting scenarios (3- yrs, 4-yrs & 6-yrs). For both models, a 3-yr forecasting scenario was chosen because it yielded the most suitable correlation coefficient, coefficient of efficiency, RMSE, Wilcoxon’s & Levene’s Pvalue. Validation of both models was accomplished by carrying out goodness-of-fit tests on the noise residuals. K-S and Portmanteau Q statistics obtained for both models confirmed normality and time independence of the residuals. Both normality and time-independence signified that the residuals were a white noise process and hence the models were valid and adequate for forecasting. The models were then used to forecast monthly streamflows to a 3-yr lead time. Within the 3-yr forecast, the SARIMA model identified 12 drought instances (3 near normal, 7 moderately dry & 2 severely dry) while the exponential smoothing model identified 21 instances of drought (15 near normal, 3 moderately dry & 3 severely dry). These results show the need for careful planning of relevant and appropriate adaptation measures to mitigate the identified droughts at a catchment scale. Based on these results, suitable drought mitigation strategies were proposed at the macro and micro scale