Prediksi curah hujan ekstrim secara spasial (studi kasus: curah hujan bulanan di Kabupaten Indramayu)

Abstract

Global climate change could increase the extreme events such as extreme rainfall, extreme temperatures, and windstorm intensity. The study of extreme rainfall prediction is necessary to be conducted to minimize the impact of the global climate change. Thus, statistical method is needed to explain the extreme rainfall events. Extreme rainfall events are usually measured for multiple locations, so it is necessary to use spatial extreme modeling to predict the extreme rainfall which takes into account the spatial correlation in the modeling. In spatial, nearest neighbors have a fairly close relationship, therefore the spatial correlation is used to determine how closely the relationship between locations. In this study, spatial correlation was calculated by using madogram and extremal coefficient. Madogram is a modification of semivariogram and extreme value distributions. Extremal coefficient is a metric characterization of tail dependence. Extreme spatial modeling were analyzed with copula approach and max-stable process. Copula approach assumes the marginal distribution of extreme values follow the uniform distribution. Max-stable process transform the marginal distribution of the extreme values to the Fréchet distribution. In this study, copula approach is used to spatial extreme modelling. The data used were monthly rainfall in the period of 1979-2008 at 15 weather stations in Indramayu Regency. To see the stability of the parameters, the data were divided into three groups: the data for 1979-2003, 1982-2006, and 1984-2008. Modeling for prediction of extreme rainfall using rainfall data of 1979-2007 and the data in 2008 are used for model validation. For the estimation of extreme rainfall rainy season, the data used is the data for 1979-2006, while the rainfall rainy season data of 2007 and 2008 are used for model validation. Madogram and extremal coefficient showed spatial correlation in Indramayu Regency. Copula parameter estimation for 1979-2003, 1982-2006, 1984-2008 year interval showed that the parameter estimate of extreme rainfall in Indramayu Regency tends to be stable. The results of this study showed that the best prediction value for annual period was for the next 12 months and this prediction is quite relevant to be used in the field. In the rainy season, the best prediction was for the next 6 months and this prediction is quite relevant to be used in the field.