Quantitative Estimation of Rainfall from Remote Sensing Data Using Machine Learning Regression Models

To estimate rainfall from remote sensing data, three machine learning-based regression models, K-Nearest Neighbors Regression (K-NNR), Support Vector (SVR), and Random Forest (RFR), were implemented using MSG (Meteosat Second Generation) satellite data. Daytime nighttime data a rain gauge are used for model training validation. optimize the results, outputs of models combined weighted average. The combination (hereafter called Com-RSK) markedly improved predictions. Indeed, MAE, MBE, RMSE correlation coefficient went 23.6 mm, 10.0 40.6 mm 89% SVR to 20.7 5.5 37.4 94% when combined, respectively. Com-RSK is also compared few methods classification in estimation, such as ECST Enhanced Convective Stratiform Technique (ECST), MMultic technique, Convective/Stratiform Rain Area Delineation (CS-RADT). show superior performance ECST, CS-RADT methods.The two products estimates, namely CMORPH CHIRPS. results indicate that performs better than CHIRPS according CC (coefficient correlation). A comparison with types precipitation estimation products, global product, regional near real-time performed. Overall, methodology developed here shows almost same product exhibits methods.