Regional Flood Frequency Analysis Using Fuzzy c-Means Clustering Algorithm for West-Flowing Rivers in Kerala, India: Fid-Bau Portal

Regional Flood Frequency Analysis Using Fuzzy c-Means Clustering Algorithm for West-Flowing Rivers in Kerala, India

Jothiprakash, V. / Drissia, T. K. / Anitha, A. B.

Now-a-days frequent occurrence of flood and its consequent devastating damages are becoming a common phenomenon all over the world. It is seen that for a given rainfall and its frequency (return period), there exists a good correlation of flood magnitude among various discharge stations in a given region. This led to regional flood frequency analysis (RFFA), several methods have been developed and applied to regionalize the stations having similar hydro-meteorological and basin characteristics. The RFFA of 43 watersheds of west-flowing rivers in Kerala, India has been carried out using fuzzy c-means (FCM) clustering method. This study explored the use of a radar plot for clustering the basin characteristics. These plots helped in the selection of attributes to find flood quantiles. The present study derived twelve feature vectors (basin characteristics) from radar plots. The optimum number of clusters has been identified through sensitivity analysis by varying the number of clusters from 4 to 13, and by studying the cluster validity indices. The L moment-heterogeneity (H) test was used to test the homogeneity of optimal clusters formed. The FCM algorithm has resulted in five homogeneous regions after analyzing 60 scenarios with various combinations of feature vectors and clusters. The flood quantiles estimated for each of the homogenous region indicate that FCM-based RFFA results are good up to 100 years return period. It is hoped that the flood quantile equations developed in this study will help the field engineers for predicting flood quantiles in ungauged sites of the west-flowing rivers, Kerala, India.