Abstract:
Climate change can exacerbate flooding. The objective of this study was to evaluate the need to update stationary assumption in flood frequency analysis under changing climate in the Upper Awash basin (UAB), Ethiopia. The basin is affected by recurrent floods as caused by climate variability and land use change. The hydrological model HEC-HMS was calibrated and validated against observed discharge during 1996–2005 and 2006–2012 period at Tefki, Melka Kuntire and Hombole gauging stations. Predicted discharge for 2050s (2041–2070) were obtained by forcing the calibrated HEC-HMS with bias corrected outputs from four GCMs (HadGEM2, MPI, EARTH and IPSL) for the RCP4.5 scenario. Generalized Pareto (GP) distribution was fitted to Peak over Threshold (POT) series for flood frequency analysis. In this paper, the parameter estimation methods of the maximum likelihood (ML) and generalized maximum likelihood (GML) were compared based on Akaike and Bayesian information criterion. The intermediate future (2050s) flood quantiles were compared to the current (1971-2000) flood quantiles. Consequently, a flood frequency curve was developed for baseline and future flood data after non-stationarity test was carried out using non-parametric Mann Kendell test. The model revealed that satisfactorily reproduced the volume, pattern and peak of the observed hydrograph with RVE of -4.06%, NSE of 0.69 and R2 of 0.73 during validation period over the study area at Hombole outlet. The result also indicates that the ML method is superior to a GML method, with lowest value of Akaike and Bayesian information criterion. The findings show that the magnitude of low-return floods will increase which may have the probability of more occurrence over the basin while the magnitude of higher return floods will decrease in the future because of climate change. This is caused by the increase in heavy rainfall magnitudes and extended wet spells in 2050s future period. The RCP4.5 scenario has an influence to increase the flood magnitude and frequency which may also increase the flood severity in UAB. Results in this study suggest that it is important to update stationary assumption for more accurate flood estimates using multi-models' simulations.
