Journal of Hydroinformatics (Jul 2023)
Statistical blending of global-gridded climatological products: an approach to inverse hydrological model
Abstract
The growing use of global-scale environmental products in hydro-climatic modeling has increased the variety of their applications and the complications of their uncertainties and evaluations. Researchers have recently turned to statistical blending of these products to achieve optimal modeling. The proposed statistical blending in this study includes five large-scale and satellite precipitation (CHIRPS, ERA5-Land of ECMWF, GPM (IMERG), TRMM, and Terra) and evapotranspiration (GLEAM, SSEBop, MODIS, Terra, and ERA) products committed in three modeling scenarios. The blending procedures are organized using a conceptual water balance model to achieve the best precipitation and evapotranspiration results for the conceptual production of streamflow using hydrological inverse modeling. Based on the results, the proposed blending procedures of precipitation and evapotranspiration improved the performance of the model using different statistical metrics. In addition, the results show the conformity of the pattern and behavior of the blended precipitation calculated using the moving least square method in the study area. This happened by changing the estimation based on in situ values, particularly in cold months considering the orographic/snow effects. The combining method provides a good fusion procedure to improve the realistic estimation of precipitation and evapotranspiration in ungagged watersheds as well. HIGHLIGHTS Five precipitation and evapotranspiration products have been used to blend as new input sets.; Statistical linear blending method has been applied to determine the model's inputs via the concept of inverse hydrology.; 100 combinations of precipitation–evapotranspiration sets have been statistically combined to calibrate the models.; The proposed method provides a good fusion method to improve the estimation of climatological signals.;
Keywords
