Analyzing Rainfall Patterns in Khon Kaen Province using Satellite-based CHIRPS Rainfall Data and Unsupervised Clustering

Abstract

Khon Kaen Province is a significant area for sugarcane cultivation in Thailand, where most plantations rely on rainfed agriculture that often faces drought and prolonged dry spells, negatively impacting crop growth and yield. Understanding annual rainfall patterns is crucial for agricultural planning and water management. This study aimed to classify rainfall patterns in Khon Kaen Province using monthly CHIRPS rainfall data from 1981 to 2023. Data validation against the Khon Kaen Meteorological Station (WMS Code 48381) showed high accuracy, with a correlation coefficient (r) of 0.89 and PBIAS of -5.94%. Silhouette Analysis determined the optimal clustering into four groups. The K-means and Hierarchical Clustering methods were compared, with the Davies-Bouldin Index (DBI) indicating that Hierarchical Clustering is the most suitable method. Four distinct patterns were identified: Group 1 with sustained high rainfall from mid to late year (13.95%), Group 2 with intense rainfall concentrated in September (30.23%), Group 3 with consistently low rainfall throughout the rainy season (23.26%), and Group 4 with low early-year rainfall that increases by mid-year (32.56%). Group 4 was identified as the most vulnerable to low rainfall, potentially affecting crop yield, with a high risk of drought and dry spells. These findings can be applied to optimize cropping calendars, plan supplementary irrigation, and assess yield potential based on rainfall patterns.