Sub-seasonal extreme rainfall prediction in the Kelani River basin of Sri Lanka by using self-organizing map classification

Abstract

The availability of several multi-model and ensemble sub-seasonal forecasts online has generated a growing interest in extreme rainfall prediction and early warning. Developing countries located in the tropics like Sri Lanka are good examples of complex meteorological zones where early warning system progress is crucial for flood damage mitigation. This study investigates the potentials and advantage of the recently available Sub-seasonal to Seasonal (s2s) database provided by a consortium of weather forecasting institutes using self-organizing map classification. The results (1) highlight the relation between teleconnection indexes such as the Madden–Julian Oscillation and the spatiotemporal rainfall pattern, (2) illustrate that heavy rainfall event frequencies depend on the type of the cluster, (3) find that the performance of s2s forecasts varies among cluster and (4) provide corrective bias coefficient to forecast water volume in the basin for each cluster. This study highlights the interest of s2s forecast for extreme rainfall prediction and advocates for the release of real-time s2s data that can provide useful information for early warning in developing country such as Sri Lanka.