Modified-INSAT Multi-Spectral Rainfall Algorithm (M-IMSRA) – A New Satellite Rainfall Estimation Algorithm based on Climatic region

Abstract:

A new simple geostationary satellite based hybrid rainfall estimation algorithm called Modified-INSAT Multi-Spectral Rainfall Algorithm (M-IMSRA) has been developed and evaluated in the present study. Mainly, the following two questions have been addressed and accordingly the algorithm has been developed and evaluated:

1. Can simple geostationary satellite based SRE’s perform equivalent to merged techniques which uses all the available satellite datasets? If so/not so, then how best it can perform? 
2. Whether SRE’s perform differently over different climate regions? Whether incorporating topography in SRE will improve the performance equally over all climate regions? 

M-IMSRA incorporates topographic information in the IMSRA (INSAT Multi-Spectral Rainfall Algorithm) algorithm using 20 different variables extracted from Digital Elevation Model by using Least Absolute Shrinkage and Selection Operator (LASSO) technique. The results show that the simple algorithms like M-IMSRA can perform similar to the other highly computationally expensive merged algorithms like TRMM 3B42 and TRMM 3B42-RT over some climatic regions of India. It has been observed that, by incorporating static topographic information the estimates over orographic regions of India like Western Ghats and North-East India has significantly improved with only reduction in the additive bias over other regions. Relative performance of the tested satellite rainfall estimates like TRMM 3B42, TRMM 3B42-RT and M-IMSRA are completely different over different climatic regions, with better performance over moderate rainfall climate regions and relatively poor performance over low and high rainfall climate regions. The obtained results highlight that only one algorithm with same input variables cannot produce better rainfall estimates over all the climatic regions where the driving variables for each region will be different. Therefore, the imminent development of SRE’s must give attention to this fact and consider this to be of vital importance to develop different algorithms for different climate regions separately.