Assessment of the Seasonal Rainfall Forecast Skills of clWRF and RegCM Climate Models

Abstract

This paper evaluates the ability to forecast monthly and seasonal rainfall in seven climatic regions of Vietnam using the dynamical downscaling method, employing two climate models, clWRF and RegCM, with input data from the global climate model (NCEP CFSv2). The results indicate that the models perform best in the northern regions. However, significant forecast errors occur in the Central Highlands and Southern regions during dry months. The RegCM model provides more accurate rainfall forecasts in the North Central, South Central, and Central Highlands regions, while the clWRF model performs better in the Southern region. Forecast quality varies with lead times. At 5-month lead time, the models show considerably larger errors compared to 1- and 3-month lead times, particularly in September, October, November, and December in the Northwest, Northeast, and Red River Delta regions. Similarly, higher errors happen in January, February, November, and December in the other regions, while in March, April, and May, the models using 5-month lead time exhibit the lowest errors in these regions. The correlation between forecasted and observed rainfall remains low, emphasizing the complexity of seasonal rainfall forecast. Therefore, exploring post-model correction methods is needed to improve forecast quality.