Ensemble Approach for Drought Monitoring and Forecasting in Southeast Asia with Mekong Drought Crop Watch Tool

Drought has been frequent and severe in some parts of Southeast Asia due to increased climate variability forced by climate change. Though drought is occurred when the rainfall is well below normal for a prolong period (meteorological drought), it has a cascading effect leading to shortage of water (hydrological drought) and soil moisture (agriculture drought), creating adverse impacts in climate sensitive sectors and socio-economic hardships to countries, communities and the people.

Drought forecasting has been challenging due to uncertainty involved in rainfall predictions as the forecast skills are drastically declining beyond two weeks of forecasting. Therefore, many initiatives have been underway to improve sub-seasonal to seasonal (S2S) as well as seasonal forecasting skills, which would help to improve drought forecasting in future.

Mekong Drought Crop Watch (MDCW), a tool developed under the SERVIR-Mekong project is used to test forecasting skills of droughts in Lower Mekong region with different rainfall forecast products (ensemble) of North American Multi-Model Ensemble (NMME). MDCW has been developed on Regional Hydro-Extreme Assessment System (RHEAS) platform, which has VIC hydrological model and DSSAT crop model, to capture cascading effect of drought and also to ensure seamless forecasts to support climate sensitive sectors in the region.

NMME products, which are available in ClimateSERV (https://climateserv.servirglobal.net/) contain 34 different ensemble members under two models CCSM4 and CFSV2. Ensemble rainfall and temperature products are used in MDCW to check sensitivity of the MDCW and to select most suitable ensemble product or combination of ensemble products to improve future drought forecasting in the region. The skills of forecasting are tested with several drought indicators as well as water balance variables and validations of MDCW are done with earth observation data, which are remotely sensed and in-situ observed data from countries in the region. Forecast skills of meteorological variables are estimated using an error metric which includes, bias, RMSE, correlation, POD. The skill of drought indicators is estimated using POD, intensity and geographical extent.