Classification of River Basins Using Artificial Neural Network

Hydrological homogeneity is often assumed whenever there is lack of data in a basin and hydrologic analysis is carried out utilizing the data from neighboring basins. However, the neighboring basins may not be hydrologically homogeneous with the basin under consideration. The identification of hydrological homogeneity of basins or clustering of basins based on homogeneity demands a high degree of subjective judgment. Although an expert integrates multivariate, nonlinear, and unquantifiable factors quite well based on his subjective judgment, a different expert may not reproduce the same results. Hence, there is a need to have a rational procedure for clustering or grouping of basins based on hydrometeorological homogeneity. In this study, the application of artificial neural network for clustering the basins on the basis of hydrological homogeneity is investigated. First, an attempt is carried out to check whether the classifications in the data hyperspace have any physical meaning or not. Subsequently, it is attempted to check whether the clustering with factors that affect runoff has any effect in runoff values of each cluster. The statistics presented indicates that there is congregation about the cluster center. Finally, use of clustering of basins based on homogeneity in data hyperspace is investigated.