A surrogate weighted mean ensemble method to reduce the uncertainty at a regional scale for the calculation of potential evapotranspiration

Abstract

We propose a weighted ensemble approach using a surrogate variable. As a case study, the degree of agreement (DOA) statistics for potential evapotranspiration (PET) was determined to compare the ordinary arithmetic mean ensemble (OAME) method and the surrogate weighted mean ensemble (SWME) method for three domains. Solar radiation was used as the surrogate variable to determine the weight values for the ensemble members. Singular vector decomposition with truncation values was used to select five ensemble members for the SWME method. The SWME method tended to have greater DOA statistics for PET than the OAME method with all available models. The distribution of PET values for the SWME method also had greater DOA statistics than that for the OAME method over relatively large spatial extent by month. These results suggest that the SWME method based on the weight value derived from the surrogate variable is suitable for exploiting both diversity and elitism to minimize the uncertainty of PET ensemble data. These findings could contribute to a better design of climate change adaptation options by improving confidence of PET projection data for the assessment of climate change impact on natural and agricultural ecosystems using the SWME method.