Seasonal variability of Indonesian rainfall in ECHAM4 simulations and in the reanalyses: The role of ENSO

A study on the skill of the atmospheric general circulation model ECHAM version 4 and two reanalyses in simulating the Indonesian rainfall is presented with comparisons to 30 years of rain gauge data. The reanalyses are those performed by the European Centre for Medium-Range Weather Forecasts and of the National Centers for Environmental Predic- tion jointly with National Center for Atmospheric Research. This study investigates the skill of the reanalyses and ECHAM4 with regard to regional, annual and interannual variability of rainfall and its responses to El Ni˜ no-Southern Oscillation (ENSO) events. The study is conducted at two resolutions, T42 and T106. A new regionalization method called the double correlation method is introduced. With this method, the Maritime Continent is divided into three climate regions, the south mon- soonal, the northwest semi-monsoonal and the Molucca anti-monsoonal region. Except over Molucca, the reanalyses and ECHAM4 simulate these annual rainfall patterns quite well. The three regions are used to study the variability of Indonesian rainfall and to mea- sure the skills of the reanalyses and ECHAM4. The skill of rainfall simulations in Indonesia depends on the region, month and season, and the distribution of land and sea. Higher sim- ulation skills are confined to years with ENSO events. Except for the region of northwest Indonesia, the rainfall from June (Molucca) and July (south Indonesia) to November is in- fluenced by ENSO, and it is more sensitive to El Ni˜ no than La Ni˜ na events. The observations show that the Moluccan region is more sensitive to ENSO, receives a longer ENSO impact and receives the earliest ENSO impact in June. The ENSO impact will diminish in Decem- ber. It is found that the reanalyses and the climate model simulate the seasonal variability better than the monthly one. The seasonal skill is the highest in June/July/August, followed by September/October/November, December/January/February and March/April/May. The correlations usually break down in April (for monthly analysis) or in spring (for seasonal analysis). In general the performance of ECHAM4 is poor, but in ENSO sensitive regions and during ENSO events, it is comparable to the reanalyses. The introduction of a higher resolution land-sea mask improves the model performance. Besides rainfall variability, signatures of the ENSO impact, the spring correlation breakdown and annual cycles are better represented by the higher resolution model