You are here

Novel methods for better flood prediction

Prediction of rainfall and flood is very important for densely populated country like India. Scientists at the Indian Institute of Science (IISc), Bangalore have proposed a new method for predicting flood magnitude, which is more effective than conventional methods. The team consisted of Prof. V.V. Srinivas and Dr. Bidroha Basu of Department of Civil Engineering.

Flood prediction has other uses in addition to providing flood warning to residents. “Flood estimates are often necessary for design and risk assessment of water control structures (barrages, dams, levees) and conveyance structures (culverts, storm sewers, spillways) in river basins”, explains Prof. Srinivas. Flood estimates also help in economic evaluation of flood protection projects, land use planning and management, and flood insurance assessment.

The Central Water Commission (CWC) and state government agencies have installed several stream gauges to measure the flow of water in rivers at various places. Based on data gathered from these gauges over the years, scientists have been making predictions about floods. However, not all places have these gauges. Records could be limited at some gauges. This becomes a hurdle for flood prediction.

It is a practice to predict flood magnitude at those places using theoretical methods. The idea underlying the theoretical methods is to search for gauged catchments which are similar to catchment of the location in question. Scientists make predictions based on data pooled from the gauged catchments, assuming flood behaviour in the ungauged catchment would be similar to that in gauged catchments. Accuracy of the prediction depends on how well we are able to identify similar gauged catchments, and how we pool flood related data from the similar catchments to perform frequency analysis. The task of identification of similar catchments is called regionalization, and the task of performing frequency analysis using pooled information is called Regional Flood Frequency Analysis (RFFA). Various methods are available for regionalization and RFFA. The IISc team has proposed a method called Entropy Based Clustering Approach (EBCA) for regionalization, and a novel approach to perform RFFA.

The CWC classified geographical area of India into spatially contiguous regions (groups of similar catchments) for use in prediction of floods. The IISc team performed a case study on Godavari, Krishna, Cauvery and Mahanadi river basins to test potential of their approach in flood prediction. They compared flood estimates obtained using the EBCA based approach with those obtained by earlier methods. They considered the peak flow records of 126 gauges in these river basins, and used data from shuttle radar topography mission (SRTM) to identify catchments of the gauges. Scientists used other catchment attributes depicting their physiography (drainage area, slope, main water channel length, watershed shape indicators), meteorology (amount of rainfall) and hydrogeology (soil characteristics, baseflow), and catchment location indicators (latitude and longitude of watershed centroid), to group catchments into seven homogeneous regions. The regions were homogenous with respect to the attributes and frequency distribution of floods, and they are not necessarily contiguous spatially.

The team observed that errors in flood estimates for ungauged sites were least when the EBCA method is used for regionalization and the proposed approach is used for RFFA. The errors were higher when the analysis was performed using the CWC regions and those formed using the standard region-of-influence method. “There is a lot to explore further on this topic. One option is to consider instantaneous peak flow values at stream gauges, instead of currently available daily flows. However, instantaneous peak flow data is not yet available for the study area. We also plan to use ECBA method to be able to predict extreme rainfall and low flows”, signs off Prof. Srinivas.

With potential methods to aid in accurate flood prediction, the reliability of existing water control and conveyance structures could be assessed properly in order to mitigate the likely damage to crops and livesdue to future flood events.

About the publication

The paper “Regional Flood Frequency Analysis Using Entropy-Based Clustering Approach” appeared in the Journal of Hydrologic Engineering in March 2016.



Prof. V. V. Srinivas, Professor, Dept. of Civil Engineering, Indian Institute of Science, Bangalore, Karnataka 560012, India (corresponding author). E-mail:

Bidroha Basu, IISc Research Associate, Dept. of Civil Engineering, Indian Institute of Science, Bangalore, Karnataka 560012, India. E-mail: