ABSTRACT
The Indonesian convective system plays an important role in the dynamics of global climate through the meridional Hadley circulation and the zonal Walker circulation. Such climate effects are important for Indonesia, a country that economically very much depends on the agricultural sector, which depends strongly on water availability and rainfall. Rainfall characteristics are also crucial in regards to other natural hazard such as forest fire. Important characteristics are total seasonal rainfall, its inter-annual variability, and the timing of arrival of the dry and wet season.
The Indonesian seasonal rainfall is the result of monsoon influence whose inter-annual variability is closely linked to the El Nino Southern Oscillation (ENSO). Accordingly, the Indonesian Bureau of Meteorology and Geophysics forecasting scheme, which is based solely on rainfall characteristic, will be currently expanded to take ENSO into consideration.

Study of seasonal rainfall variability is restricted due to limited data availability. Detailed knowledge of the spatial and temporal structure of rainfall-ENSO relationships necessary for an ENSO-based forecasting model is not available. For these reasons, this study constructed a high quality monthly rainfall data set for Indonesia that is suitable for climate research and compiled a reliable and homogenous monthly rainfall series from 85 stations spanning the period from January 1879 to March 1999. This data set is divided into two sub data sets, i.e. a 63-station and a 85-station data set which have been beneficial not only for this study but also a number of other independent studies on Indonesian rainfall.

Using that data set, the spatial and temporal variability of Indonesian rainfall has been studied. The results show that the largest inter-annual variability is observed for the dry and dry-to-wet season transition rainfalls. These two seasons exhibit clear spatial coherence so that rainfall forecasting using a single large-scale predictor like SOI may be possible. It has also been found that the dominant periodicity of rainfall is similar to that of ENSO. Decadal variations in seasonal rainfall are also observed demanding caution in utilizing any statistical forecasting model either relying on rainfall itself or another large-scale predictor.
Detailed analysis of Indonesian rainfall-ENSO relationship showed regional and temporal details not previously documented. The association is found to be strongest for the dry and dry-to-wet transition season rainfalls but less clear for the wet season rainfall. Regarding the possibility of using the preceding value of SOI to predict rainfall, it was found that it varies with season and region. It was also found that the rainfall-SOI association varies with season.

Finally, the thesis provides evidence of the impacts of El Nino-rainfall variability on agricultural production and outlines some potential for crop forecasting. As expected for Indonesia, El Nino events are often associated with reduced crop production. However, the intensity of the impact varies with ENSO event, crop type and location. There is scope for some provinces and for some seasons to predict likely crop production from observations of previous rainfall and SOI. For crop production, the dry-to-wet transition rainfall can be as important as the other seasons rainfall.