Classification and Flooding Potential Assessment of Subbasins of a Tropical River Using Cluster Algorithms

Abstract

Three clustering algorithms, K-means clustering analysis (KCA), fuzzy cluster analysis (FCA), and density-based spatial clustering of applications with noise (DBSCAN), are applied to classify the 13 subbasins of the Mahe River, southwest India, based on 13 morphometric parameters of each. Suitable validation indices, such as Davies–Bouldin and Calinski–Harabasz indices, have been used to select the optimal number of clusters using KCA and FCA techniques. All three analyses have yielded three clusters, with subbasins 3–8 forming the first one. These constitute 23% of the total basin area of the Mahe. SW 12 forms a grouping of its own. The rest, SW 1–2, 9–11, and 13, form the third cluster. The first cluster corresponds to the subbasins identified as most susceptible to flooding. Cluster 3 encompasses the subbasins falling in the “Moderate” and “Least” categories with respect to the risk of flooding. The subbasin 12 (< 1 km2) exhibits a deviant morphometric pattern likely due to its specific topographical and network characteristics. The study reveals that cluster algorithms are effective in ranking and prioritizing subbasins of a river based on their potential for natural hazards like flooding. Moreover, the DBSCAN averts the use of cluster validation indices to determine the optimum clusters without compromising the results. All these methods would be beneficial in chalking out suitable management measures for different subbasins of a river based on their potential for any given hazard.