Development of a novel Weighted Average Least Squares-based ensemble multi-satellite precipitation dataset and its comprehensive evaluation over Pakistan

Abstract

Ensemble multi-satellite precipitation datasets (ESPDs) are alternative to satellite-based precipitation products (SPs), which tend to reduce the errors, combine advantages of individual SPs, and have higher accuracy for hydrological applications. The current study proposes and evaluates a dynamic WALS-ESPD developed using the Weighted Average Least Square (WALS) algorithm, which has 0.25° spatial and daily temporal resolutions across glacial, humid, arid and hyper-arid regions of Pakistan during 2000–2015. WALS-ESPD is developed using three SPs, Tropical Rainfall Measurement Mission (TRMM) Multi-Satellite Precipitation Analysis (TMPA) 3B42V7, Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks–Climate Data Record (PERSIANN-CDR), Climate Prediction Center MORPHing technique (CMORPH), and one re-analysis product, Era-Interim. Mean Bias (MB), Mean Absolute Error (MAE), unbiased Root Mean Square Error (ubRMSE), Correlation Coefficient (R), Kling-Gupta efficiency (KGE score), and Theil's U are used to evaluate the performance of WALS-ESPD both spatially and temporally. Moreover, the skill scores of statistical metrics are used to assess the WALS-ESPD performance against two previously developed ESPDs, DBMA-ESPD and DCBA-ESPD. TMPA dominated all SPs with average weights of 0.317, 0.341, 0.314, and 0.326 across the glacial, humid, arid and hyper-arid regions. TMPA dominated pre-monsoon (30.26%) and monsoon (35.82%) seasons, while PERSIANN-CDR dominated post-monsoon (27.58%) and winter (29.82%) seasons. WALS-ESPD performed relatively poor across the glacial and humid regions, and during monsoon and pre-monsoon seasons. Skill scores of WALS-ESPD against DBMA-ESPD show better performance of WALS-ESPD in all four regions, especially across the glacial region with the maximum MB, MAE, and ubRMSE scores of 27.36%, 28.34%, and 27.67%, respectively. Meanwhile, WALS-ESPD performed better than DCBA-ESPD in the whole glacial region and most part of other regions, while DCBA-ESPD dominated WALS-ESPD at few stations across humid, arid, and hyper-arid (south-east) regions.