A Machine Learning Approach to Re-Classification of Climate Zones Based on Multiple Rain Features Over India

Climate change studies require a large number of parameters to be handled simultaneously and machine-learning techniques are now-a-days widely used to solve this kind of situation. One of the interesting problem in climate studies to identify the homogeneous regions for trend analysis as it is rather impractical to study the time series for each measurement points. The concept of spatial clustering is applied to the TRMM satellite data to improve the classification of homogeneous rain zone. Instead of relying on the rain rate only, a number of spatial rain features are included in re-classification of the regions. The present approach clearly improves the decadal trend detection as opposed to the existing climatology as evident from the preliminary results presented in this paper.