Improving Efficiency of Flash Flood Forecasting System in Thailand Based on the Dynamic Antecedent Moisture Condition-III (D-AMCIII) and 18 UTC of WRF-ROMS's Rainfall Forecast

Abstract

Flash floods are rapidly evolving local floods that occur on small catchment scale and steep terrain areas. The severity and initiation of floods are influenced by the catchment characteristics and the current state of the catchment. The development of a flash flood warning system is very important to reduce the damage and losses. Hydro-Informatics Institute (HII) has developed A Simple Flash Flood Forecasting System (FFFWS) using multi-criteria analysis and data overlaying technique was developed using 3 simple concepts consist of 1) the selecting high potential flash flood areas using the Flash Flood Potential Index (FFPI) which depends on physical properties of catchment, 2) the updating soil moisture state from AMC classes were calculated by 5-days accumulated rainfall from satellite, and 3) the forecasting of flash flood risk-areas using 24 hrs. predicted rainfall from WRF-ROMS. Due to the previous version of the flash flood forecasting system used only one trigger value for 5-day accumulated rainfall threshold, known as Fixed-AMC III (F-AMC III), and 24-hour forecast rainfall from the 12 UTC for the entire country. To enhance the system's forecasting efficiency, this study investigated adjusting the 5-days accumulated rainfall threshold appropriately for each region by applying two trigger values that vary according to current daily rainfall compared to normal daily rainfall (30-year average), known as Dynamic-AMC III (D-AMC III), and using forecast rainfall data from the 18 UTC in the rainy season. For the evaluation of forecasting performance before and after these improvements was conducted by comparing the Probability of Detection (POD), False Alarm Ratio (FAR), and Critical Success Index (CSI) across six heavy rainfall events in the Northeastern and Southern regions of Thailand. The results showed that the statistical forecasting metrics generally improved, with POD values increasing significantly.