PCMDI - IPCC

In the 20th century, we have faced condition that green house gases (GHGs) concentration, particularly CO2, in the atmosphere has increased so tremendously. Regarding to the effects, IPCC (2001) has shown many proofs about the raising of global temperature and the changing of other climate parameters such as rainfall, humidity, etc. Generally, it gives specific influence to the distribution of rainfall patterns regionally or even locally. The study based on general idea to learn the relationship of the increasing of CO2 concentration in the atmosphere to the changing of rainfall patterns in Indonesia on regional and local point of views.
In the first part of our research, we were focusing our study in developing regionalization techniques. We have developed and practiced some of statistical downscaling methods and select the best method that statistically produces the best result on reconstructing Indonesian rainfall. The selected method also will be use to estimate the future rainfall based on GCMs scenario data as well as the historical rainfall. The study compared three statistical downscaling methods; they are Principal Component Regression (PCR), Artificial Neural Network (ANN) and Multivariate Adaptive Regression Splines (MARS). These methods were formulated using historical precipitable water (prw) data from different GCM models (CSIRO-Mk3.0, CGCM3.1, ECHAM5 and GFDL-CM2.0) which used to downscale the observation rainfall data in Aceh, Padang, Solok, Lampung, Jakarta, Kupang, Pontianak, and Ambon. We suggest that MARS method has given better performance rather than PCR and ANN methods. The study also found that the reducing of GCMs gridded data using principal component analysis for the six data of observed areas produces >80% of total variants for the first five or six components. The study of historical rainfall shown that in a century DJF rainfall pattern in Padang changed gradually as well as in Aceh on JJA and SON patterns, and in Jakarta on DJF pattern.