Improving flash flood early warning systems in North-Eastern part of Bangladesh

Abstract

In Bangladesh, the north-eastern part is highly exposed to pre-monsoon flash floods occurring in almost every year resulting in extensive damage of crops and livelihood opportunities. Flash flood forecasting and effective dissemination of early warnings can be a vital tool to reduce flash flood vulnerabilities in the affected areas. Flood Forecasting and Warning Center (FFWC) under the Bangladesh Water Development Board (BWDB) has recently started providing the forecast of flash flood water levels for 25 water level stations. This study evaluated the exiting flash flood forecasting of 14 water level stations to evaluate the forecast performance based on 6 forecast verification criteria, namely, Accuracy or Fraction Correct (FC), Threat Score (TS) or Critical Success Index (CSI), Hit Rate (HR) or Probability of Detection (PoD), Probability of Occurrence (PoO), False Alarm Rate (FAR), and BIAS Score. To evaluate the forecast performance, daily observed and forecasted water levels data for 3-day lead were taken from FFWC for the pre-monsoon season (1st April-15th May) of 2017 and 2018. At the peak water level, when water level rises and falls sharply, the forecast performs poorly resulting in higher forecast errors. Performance evaluation results have been found for Accuracy or FC of 0.29-0.88, TS or CSI of 0.14-0.83, HR or PoD of 0.21-0.92, PoO of 0.23-1.0, FAR of 0.0-0.77, and BIAS of 0.33-2.10, respectively. Forecast performed comparatively better for the station of flood plain areas e.g. Sylhet, Sherpur, Moulvibazar, Ballah, Habiganj, and Jariajanjail stations compared to flash flood areas stations e.g. Kanaighat, Sunamganj, Amalshid, Sheola, Manu Rly Bridge, Lourergorh, Durgapur, and Sarighat stations. The forecast has shown the best performance at Sherpur station while the poorest performance of forecasting has been found at Sarighat station throughout the study. To disseminate Early Warnings (EWs) at the community level, a detailed understanding from national to the local level hasbeen done applyingParticipatory Rural Appraisal (PRA) tools such as Key Informant Interview (KII) and questionnaire survey. The field results revealed that Mikingusing handheld microphones is the best option to disseminate EWs at community level. A 3 to 5-daylead time in forecasting and warning is acceptable for the flood vulnerable communities to protect crops from damage. Upazila Agriculture Officer can play a vital role for the dissemination of EWs at local level. Finally, a dissemination framework has been developed in congruence with the Standing Orders of Disaster (SOD) for the study areas to effectively disseminate flash flood EWs and forecasting at community level in an enhanced manner.