Prediction of Daily Rainfall in Dodokan Watershed Based on Statistical Downscaling Model: An Effort to Manage Watershed Ecosystems Mustika Hadijati1*, Irwansyah 1 , Nurul Fitriyani 1

Abstract

Climatic conditions in watersheds, especially rainfall, can affect ecosystems in the watershed. Changes in ecosystem conditions can affect the condition of the watershed. It also affects the condition of water resources in the watershed. Therefore, developing a rainfall model that provides an accurate rainfall prediction is necessary to manage the watershed ecosystems. The rainfall model was developed using the statistical downscaling method. This method utilizing global-scale GCM output information, so data pre-processing is needed. The CART algorithm is one method that can be used for data pre-processing to lessen the GCM data dimension. Moreover, the functional form of the rainfall relationship with GCM data was determined using nonparametric kernel regression. The final result obtained a rainfall model that provides fairly good prediction accuracy with a relatively small RMSE value