Performance Evaluation of a Regional Climate Model in Simulating Rainfall over Indonesia

While the issue of rainfall variability and prediction highly affect the livelihood of the population in Indonesia, in which farming and commercial fishing both have been the main economic activities, climate model performance in simulating rainfall over the Maritime Continent is yet to show a decency and is still considered to be a great modeling challenge. The region is a densely populated chain of islands with complex atmospheric and oceanic features. Nonetheless, change in future rainfall tendencies is to be expected, and with the increasing demands for the information of future rainfall pattern changes from various governmental and private sectors, such as mining, transportation, health, regional planning, forestry and disaster risk-reduction, an analysis for the future trend of precipitation on a detailed scale is necessary. In this study, we utilize a regional climate model and a CMIP5 future climate projection GCM to verify if and under which conditions the regional model will perform better than its forcing GCM at representing historically observed precipitation over Indonesia. The results and findings could be the basis of the utilization of the downscaled CMIP5 data for future projection regional rainfall analysis.

We investigated the sensitivity of simulated rainfall towards different convective physics parameterizations in the WRF regional model using an MIROC5 projection dataset as the forcing. A series of 1-year period simulations under different ENSO phases for the past years were done using different cumulus physics settings at the spatial resolution of 20 km. The results were then compared with satellite and ground-based observational datasets. Preliminary results suggest that the regional model has shown to produces a more realistic detail in terms of spatial distribution with respect to the orographic influences. Different convective parameterizations also play a significant role in simulating rainfall as some options tend to produce more wet biases over another. It was also found that the results varied over different parts of the domain with different rainfall patterns, which are depending on the prominent modes of circulation affecting the area. Regional model results have a better temporal correlation in regions with “monsoonal” type rainfall pattern, primarily driven by the Asian-Australian monsoon, compared to the “equatorial” or “local” types. The results are also shown to be more realistic in illustrating the consecutive rainfall events by significantly reducing the overestimation of wet and/or dry spell that was found to be prolonged in the GCM simulated rainfall. However, as with the GCM still hardly able to represent the effect of ENSO variability on its simulated rainfall, the regional simulation result is possessing the same nature and still yet to capture inter-annual extreme events.