Extreme rainfall is a phenomenon that occurs because of the high intensity of rainfall which causes many negative impacts. For example, is flooding that can ruin crops for farmers. To minimize this hazard, rainfall prediction is carried out. One methods that can be used is spatial extreme value theory using max stable approach and copula approach. From it, we will get the return level (predictive value) rainfall by paying attention to the location elements in it. This study discusses the max-stable approach using the Smith model and the copula approach using the Copula Gaussian model. Parameter estimation used is Maximum Likelihood Estimation (MLE) method. Because it takes into spatial elements, the trend surface model is also used. At the end of the study, Akaike Information Criterion (AIC) is used to compare and select the best model that can be used in predicting rainfall. Spatial extreme value theory will be applied to rainfall data in East Java which consists of nine rainfall observation stations. The results of spatial parameter estimation show that the highest of rainfall intensity value will occur around rainfall observation station namely Sawahan which is located in Nganjuk district in a period of two years. From AIC, it can be said that the model using copula Gaussian is the best model compare with max-stable approach as it provides the smallest AIC values.