Enhancing NWP-Based Reference Evapotranspiration Forecasts: Role of ETo Approaches and Temperature Postprocessing

Abstract

Reference evapotranspiration (ETo) forecasts are essential for estimating irrigation water demand and agricultural water management. However, studies have not examined numerical weather prediction (NWP)–based ETo forecast enhancement with respect to different ETo approaches and climate zones in the Indian subcontinent. In this study, we use two probabilistic postprocessing techniques (PPT), namely, nonhomogeneous Gaussian regression (NGR) and Bayesian model averaging (BMA), and assess their performance in enhancing NWP-based ETo forecasts at short to medium-range time scales (1 to 7 days) over different climate zones in the Indian subcontinent. Weather variables from NWP model outputs are used to estimate the ETo forecasts. Two ETo approaches, namely, the food and agriculture organization (FAO)-Penman Monteith (PM) and temperature-based Hargreaves-Samani (HS) methods, are utilized for ETo estimation. The effectiveness of PPTs in enhancing the ETo forecasts using these approaches is also evaluated. Further, hydrologic forecasting studies have traditionally used postprocessed temperature forecasts toward forecasting of ETo in hydrologic models. However, the rationale of this approach is debatable. In this study, we also evaluate if the postprocessing of temperature forecasts produces comparable ETo forecast performance relative to the postprocessing of the ETo forecasts. Results revealed that raw ETo forecasts from both NWPs perform poorly, especially in the northern (polar zone) regions. Further, wind speed and solar radiation were found to be the dominant variables contributing to low ETo forecast skill over the region. The forecasts using the HS method were found to be less skillful than the forecasts from the PM approach. Postprocessing results indicate that both the PPTs are able to considerably enhance ETo forecast skill across all the climate zones and the NGR approach outperforms the BMA technique. The postprocessing was especially able to enhance the skill of forecasts in northern (polar zone) regions where the raw ETo forecast skill was particularly low. The estimation of ETo forecasts using temperature postprocessed ETo forecasts (EToT) revealed that temperature postprocessing does not considerably improve the accuracy of the EToT forecasts. Outcomes of this study have implications for hydrologic forecasting, irrigation water management, and development of irrigation-based decision-making systems in the Indian subcontinent.