Improving Short-Term Weather Forecasting using Support Vector Machine Method in North Barito

Flooding is a recurring issue in North Barito Regency due to the overflow of the Barito River. Weather forecasts in the region rely mainly on Numerical Weather Prediction (NWP) models, which often fail to capture local details due to their grid-based homogenization. To address this limitation, statistical techniques such as Model Output Statistics (MOS) can enhance NWP outputs by representing local conditions more accurately . MOS establishes statistical relationships between response variables (predictands) and predictor variables derived from NWP outputs, enabling operational applications without the need for advanced instruments. This study utilizes rainfall data from 2021-2022 from the Beringin Meteorological Station in North Barito as the response variable, while data from the Integrating Forecasting System (IFS) model serve as the predictor variables. The Support Vector Machine (SVM) method is employed to identify the relationship between predictor and response variables. By integrating the MOS technique with the SVM method, this research aims to improve the accuracy of weather forecasting, particularly for short-term predictions in North Barito. This approach demonstrates the potential to enhance localized weather predictions by addressing the limitations of conventional NWP models. The results indicate a consistent reduction in RMSE across all experiments conducted. Furthermore, the SVM model showed notable improvements in bias values and exhibited a stronger correlation compared to the original outputs from the IFS model. The percentage improvement (%IM) in rainfall forecasts, following correction using the SVM model, increased by 5.03%. The percentage improvement (%IM) in rainfall forecasts, following correction using the SVM model, increased by 5.03% for use as a predictor variable in the applied SVM method. In contrast, a combination of surface pressure, temperature across various layers, and rainfall proved to be the the most effective input variables for enhancing the accuracy of weather forecasting in North Barito using the SVM model.