Investigating the Performance of Bias Correction Algorithms on Satellite-Based Precipitation Estimates

摘要

Quantification and correction of error in Satellite-based Precipitation Estimates (SPEs) is indispensable for accurate climate predictions and reliable hydrological applications. The present study aims to evaluate the performance of two widely used bias-correction algorithms, i.e., Quantile mapping based on an empirical distribution (QME) and Linear scaling (LS), for application on SPEs. The performance of real-time and gauge-corrected versions of Tropical Rainfall Measurement Mission (TRMM) Multi-Satellite Precipitation Analysis (TRMM 3B42 and TRMM 3B42RT) are analyzed over India for 13 years (2001-2013) duration. The total bias in the TRMM datasets are initially estimated and further subjected to error components analysis, wherein, the total biasis disintegrated into three components: hit bias (H), missed precipitation (M) and false precipitation (F). Further, the QME and monthly LS algorithms are developed and applied to the TRMM datasets. The bias-corrected TRMM datasets are later subjected toerror component analysis and the actual reduction in the magnitude of error components after bias correction is investigated. The results of the study highlight the presence of significant bias in TRMM datasets over India. Among the bias correction methods, the LS method outperformed the QME method in representing the average bias over India. The QME method reduced the missed precipitation error significantly; however, it increased the false alarm error in the SPEs, especially over regions of high rainfall. The Hit bias was positive in case of QME method and negative in case of the LS method. The present study highlights the significance of reduction of bias by considering individual error components, rather focusing the same on total error.