Evaluation of performance of POAMA forecast system in predicting monthly precipitation over Indonesia

Abstract

Hindcast produced by a model used in a numerical model-based seasonal prediction system is an essential part in the operational seasonal prediction system. This paper is aimed at evaluating the performance of POAMA model from the climatological aspect. The data used in this research are obtained from three variants of POAMA m24 model. The results show that the annual cycle of climatological rainfall averaged over Indonesia is well simulated by POAMA m24, although there is a dry bias in the rainy season and a wet bias in the dry season. From those three variants, m24b model has a relatively low variation of bias against lead-time compared with the m24a and m24c models due to the implementation of flux correction scheme in m24b model. However, the performance of POAMA m24 model with a resolution of T24 is inferior to CFSv2 with a resolution of T126 in simulating the spatial pattern of rainfall over Indonesia. Beside model resolution, convection scheme used in a model also has significant influence. This can be seen from the resemblance between the spatial pattern of total rainfall and that of convective rainfall. Thus, the horizontal resolution of a model and a suitable convection scheme for Indonesian region are the two factors that must form important consideration in the development of Indonesian seasonal prediction system.