Spatio-Temporal Mining of Core Regions: Study of Rainfall Patterns in Monsoonal India

Natural events like climate, disease, etc., and man-made events like theft have a great impact in the regions where they occur. Hence, there is a need to assess the behavior of these events -- regions where they occur, the patterns they exhibit etc., to help manage them suitably. In addition, events that are dynamic in nature make it even more difficult to extract or understand such behavior. Our work here, proposes a method to achieve this goal of detecting the regions, called Core regions or Cores, influenced by an event over a time period by using a combination of watershed delineation, neighborhood analysis and frequent item mining. The method involves both a spatial analysis step to detect focal points and a spatio-temporal analysis over the entire data time period T to identify core regions. Further, the cores are classified as Cores with Contiguous points (CC) and Cores with defined Radius (CR) based on the type of neighborhood, and Cores with Highly Dominating points (CHD), Cores with Less Dominating points (CLD) and Cores with No Dominating points (CND) based on frequency of occurrences. The frequent/predominantly occurring focal points capture the localized behavior of an event whereas the neighborhood constraints capture the nature (dynamic/non-dynamic) of the event. In this work, core regions of monsoonal rainfall are detected over a total period of 56 years (1951-2006). Due to the dynamic nature of rainfall, it is observed that CR shows better results than CC. Also, out of the seven CR detected in Central & Peninsular India, three of them exhibit CHD (highly localized behavior).