Hide/Show Apps

Evaluating and merging model- and satellite-based precipitation products over varying climate and topography

Amjad, Muhamma
Before using the satellite- and model-based precipitation retrievals in hydrological studies, their uncertainty assessment is crucial. Improving their performance accuracy is another important issue worth consideration. This study first evaluates and intercompares a set of nine precipitation products (2 satellite estimation-based, 2 model reanalysis-based, and 5 model forecast-based products) over varying climate and topography by using the in-situ observed precipitation data as truth. The products were then merged, in the form of two groups, using two different merging techniques: 1. Taking ensemble mean (i.e., simple merging); 2. Taking ensemble mean after rescaling them by a linear regression method. The merged products obtained were statistically and hydrologically evaluated and inter-compared with the individual products using the same in-situ precipitation data and the observed surface runoff data, respectively. The results show that the errors in the products increase, while their correlation with the observed data decreases with the increasing terrain complexity. Comparatively, wetness and terrain slope have a more prominent role in the error variability of the products. Both the merging methods improve the errors and correlations of the products not only over the entire study area, but all its sub-regions classified based on the wetness, elevation, and terrain slope. Simple merging improves the precipitation detection ability of the individual products, while the merging after rescaling the products improves their random errors and correlation with the observed data. Overall, the merged products show better efficiency in regenerating the observed runoff data as compared to the individual products.