HYDRO-METEOROLOGICAL FREQUENCY ANALYSIS FOR INVESTIGATING RECENT KERALA FLOODS

Abstract

This research study aims to investigate the climatic drivers, particularly precipitation, behind extreme hydrological events in three major rivers of Kerala: Bharathapuzha, Muvattupuzha, and Kallada. These rivers pass through densely populated areas of Kerala, making it crucial to understand the relationship between precipitation and flood events. The research utilized a dataset consisting of daily flow from gauge stations and precipitation data (0.1o x 0.1o) spanning 20 years (2001-2020) within the watersheds. The analysis employed a bivariate binary classification frequency model, using daily precipitation data divided into event and antecedent precipitation as input variables. This model serves as a straightforward flood prediction model based solely on daily precipitation inputs. Correlation analysis was conducted to examine the relationship between river flow and various precipitation metrics, including event and antecedent precipitation. The goal was to determine the lag-time or warning time available to mitigate flood damages. By deriving a flood threshold through a linear equation, the model successfully distinguished storm events from ordinary events. The model's performance was evaluated for the first 15 years and validated for the subsequent 5 years by comparing the predicted flood events with historical flood events. This evaluation highlighted the model's efficacy in predicting floods while also revealing its limitations. The results demonstrated that in addition to event precipitation, antecedent precipitation (representing antecedent soil moisture levels) played a significant role in causing floods. This finding emphasized the importance of incorporating threshold processes that can differentiate various antecedent conditions in a watershed. By considering thresholds for both event precipitation and antecedent precipitation, a binary threshold system can enhance flow prediction and forecasting by effectively separating flood events from ordinary events. Furthermore, the developed model has the potential to be utilized for flood prediction based on inputs from weather forecasts. This application would allow for timely and accurate flood forecasting, enabling proactive measures to mitigate flood-related damages.