Comparison of probabilistic post-processing approaches for improving NWP-based daily and weekly reference evapotranspiration forecasts

Abstract. Reference evapotranspiration (ETo) forecasts play an important role in agricultural, environmental, and water management. This study evaluated probabilistic post-processing approaches, including the nonhomogeneous Gaussian regression (NGR), affine kernel dressing (AKD), and Bayesian model averaging (BMA) techniques, for improving daily and weekly ETo forecasting based on single or multiple numerical weather predictions (NWP) from The International Grand Global Ensemble (TIGGE), including the European Centre for Medium-Range Weather Forecasts (ECMWF), the National Centers for Environmental Prediction Global Forecast System (NCEP), and the United Kingdom Meteorological Office forecasts (UKMO). We found that the NGR, the AKD and the BMA methods greatly improved the skill and reliability of the ETo forecasts compared to a linear regression bias correction method, due to the considerable adjustments on the spread of ensemble forecasts. The methods were especially effective when applied over the weekly NCEP forecasts, followed by UKMO forecasts. The post-processed weekly forecasts had much lower rRMSE (between 8 %–11 %) than the persistence-based weekly forecasts (22 %), and the post-processed daily forecasts (13 %–20 %). Compared with the single model ETo forecasts based on ECMWF, multi-model ensemble ETo forecasts showed higher skill at short lead times (1 or 2 days) and over the southern and western regions of the United States. The improvement was higher at the daily timescale than at the weekly timescale. The NGR and AKD methods performed the best, but the NGR method is more flexible and computationally efficient than the other methods. In summary, the study demonstrated that the three probabilistic approaches generally outperform conventional procedures based on the simple bias correction of single model forecasts, with the NGR post-processing of the ECMWF and ECMWF-UKMO forecasts providing the most efficient ETo forecasting.