A combined dynamical and bias correction technique for generating probabilistic daily rainfall forecasts over Indonesia

Abstract

Daily rainfall forecast has high importance in Indonesia because of its supporting role in various sectors. However, highresolution forecasts with many ensembles require high computing costs that hamper the development of regional weather/climate forecast, especially in Indonesia. This study aims to develop a daily operational ensemble forecasts with adequate validity and relatively low-computing cost to forecast rain occurrences over Indonesia. The model ensemble forecasts contain 21 ensemble members; each of them is obtained from Global Ensemble Forecast System (GEFS) data. To justify the forecast results, this study used Global Precipitation Measurement (GPM) satellite data as a comparison. The forecasting procedure is as follows: First, Global Forecast System data are used as the initial condition of WRF to obtain regional scale forecast. Second, the output from WRF is used to obtain the correction factor by simple delta method. Third, the correction factor is then used to downscale all ensemble members of GEFS data to regional scale. Finally, using all ensemble members, probability of rain occurrences is then calculated. The rain occurrences are divided into six categories: sunny (no rainfall), cloudy, light rainfall, medium rainfall, high rainfall, and very high rainfall. The processes mentioned above are automated, and the outputs are issued daily. Results show that the model forecasts are consistent with GPM satellite data and have adequate forecast skills, especially over land areas. Moreover, the forecasts have a relatively short running time of approximately 22 hours; without the need of supercomputers. These results feature the possibility of low-computing cost of model ensemble daily weather forecasts over Indonesia.