Indonesia is developing a numerical weather prediction (NWP) method that is suitable for the geographical conditions in Indonesia, namely INA-NWP. However, the accuracy of INA-NWP is still low and requires statistical postprocessing to improve its accuracy. One of the statistical postprocessing techniques to improve INA-NWP accuracy is to use model output statistics based on regression models, such as partial least squares (PLS), Principal Component Regression (PCR), and ridge regression, etc. These regressions are considered capable of overcoming multicollinearity problems in weather predictions. Accuracy using MOS is still low, so it is necessary to use an ensemble method from several MOS to improve the prediction of INA-NWP. The ensemble method is the Ensemble Model Output Statistics (EMOS). The weakness of EMOS itself is that it is unable to capture spatial dependencies that often occur in meteorological cases. Hence, its combination with the Geostatistical Output Perturbation (GOP) method is needed to capture spatial dependency patterns between research locations. The Spatial Ensemble Model Output Statistics is a combination of EMOS and GOP. In this study, Spatial EMOS were applied to 10 meteorological stations in Surabaya, Indonesia and surrounding areas. The result is that the accuracy of Spatial EMOS is higher than that of PLS, PCR, and Ridge.
Sutikno et al. (Wed,) studied this question.