Prediction uncertainty of evapotranspiration deficit index

Abstract

Understanding uncertainty in drought characteristics is vital in water resource management. To demonstrate an effective presentation of the uncertainty in agricultural drought derived from the evapotranspiration deficit indices (ETDIs), this study simulated ETDIs using the general ETDI formula with 1000 parameter combinations at different locations in the Ruvu River basin, Tanzania. Then, principal component analysis was used to reduce ETDI simulations to proxies of simulations. Thirdly, the evaluation of Kendall tau correlations between the proxies of simulations and ETDI simulations produced behavioural simulations, which were used to calculate the 95 per cent prediction uncertainty (95PPU) of ETDIs and median ETDI. Finally, the 95PPU of ETDIs and median ETDI were characterized into drought durations. The findings show that the 95PPU bands of ETDIs and drought durations were very narrow but enclosed their median values. Therefore, the 95PPU and median from this approach comprehensively present drought characteristics derived from several ETDIs.