A seismic intensity map, which describes ground motion distribution due to an earthquake, is crucial for disaster evaluation after the event. The ShakeMap system, developed and disseminated by the USGS, is widely used to generate intensity maps in many countries. The system utilizes a semi-variogram model to interpolate the measured intensities at seismic stations spatially. However, the default semi-variogram model embedded in ShakeMap is based on data from high seismic regions, which may not be suitable for the Korean Peninsula, categorized as a low-to-moderate seismic region. To address this discrepancy, this study aims to develop the region-specific semi-variogram model using local records and a region-specific ground motion model (GMM). To achieve this, we followed these steps: 1) collected records from significant earthquake events in South Korea, 2) calculated residuals between the observed intensities and predictions by the GMM, and 3) created semi-variogram models using weighted least squares regression to better fit short separation distances for PGA, PGV, SA0.2, and SA1.0. We compared the developed semi-variogram models with conventional models embedded in ShakeMap. Validation tests showed that the region-specific semi-variogram model reduced the mean squared error of intensity predictions by approximately 3.5% compared to the conventional model.