LAUSR

This study presents an improved GIS-based regression model that incorporates an adaptive effective radius algorithm into the structure of previous regression model (Kongju National University Regression Model, KNU/RM), named of IGISRM. The performances of IGISRM were evaluated at various spatial resolutions to select the most suitable grid-spacing considering accuracy and computational efficiency. In addition, the performances between IGISRM and KNU/RM were inter-compared based on various performance measures related to amount and occurrence of precipitation and extreme indices. The results showed that both regression models produced the compatible performance in amount-related measures. However, IGISRM outperformed KNU/RM in occurrence-related measures mainly by the adaptive effective radius. Especially, IGISRM improved the ability in reproducing spatial distribution of precipitation events when the spatial variability in precipitation occurrence highly exists. In contrast, KNU/RM predicted overly wet days due to drizzle effect. Such results indicate that IGISRM has the better skill to capture the heterogeneity of spatial distribution of precipitation. In addition, IGISRM showed higher skill in reproducing the extreme indices, in particular those related to wet and dry spell of precipitation, number of wet days and annual precipitation intensity mainly due to over-predicted wet days of KNU/RM induced by the longer radius of influence circle.