Evaluation and Ranking of different gridded rainfall-interpolation combination for large Himalayan watershed using entropy based weighted sum model.
AGU Fall Meeting Abstracts
Dericks Praise Shukla.
2021-12-01
Rainfall estimates are one of the major input datasets for any type of the hydrological modelling and their accuracy has great impact on the modelling output. This study investigates the best rainfall-interpolation (R-I) combination for re-gridding and rescaling of gridded rainfall dataset. The ranking analysis is applied on total nine combinations of three satellite gridded rainfall data (TRMM-3B42, PERSIAN-CDR & NCEP-CFSR) and three commonly used interpolation techniques (IDW, Kriging, & Spline) over the Upper Beas River Basin (UBRB) in Himalayan region. The Shannon entropy based weighted sum model (WSM) was applied to assess the performance of nine rainfall-interpolation combination based on seven rainfall indices namely Mean rainfall (PCPMM), Rainfall standard deviation (PCPSTD), Rainfall skewness (PCPSKW), probability of consecutive wet days (PR_W1), probability of alternate wet days (PR_W2), No. of rainy days (PCPD) & maximum rainfall (RAINHHMX). Rainfall-Interpolation Combination Skill Score (RICSS), a new entropy based index is developed in this study for the ranking of different rainfall-interpolation combinations. IMD gridded data (is taken as reference for evaluation of performance index using NashSutcliffe efficiency (NSE) for all the R-I combinations. The results showed that IDW method is best suited for TRMM-3B42 & NCEP-CFSR datasets, whereas Kriging is most suitable for PERSIANN-CDR dataset. Spline interpolation has performed worst for all three rainfall datasets; however TRMM-Spline combination is ranked well than the all combinations of PERSIANN-CDR & NCEP-CFSR datasets. NCEP-CFSR dataset was the worst dataset to be used with any interpolation method. Finally, it is concluded that TRMM dataset with IDW interpolation has the best R-I combination showing RICSS value of 0.9903 and should be used for the re-gridding or missing value estimation and hence suitable for spatially distributed hydrological modelling e.g. SWAT or VIC over the similar Himalayan regions. Keywords: UBRB; Rainfall indices; Rainfall-Interpolation combinations; Rainfall-interpolation combination skill score (RICSS); Shannon Entropy; IDW interpolation.