Calibrating weather forecasting in Indonesia: The geostatistical output perturbation method

Abstract

The Numerical Weather Prediction (NWP) was developed by the Meteorological, Climatological, and Geophysical Agency in Indonesia for the purpose of weather forecasting, however, it comes with a high level of bias. This purpose of this study therefore was to improve this model with the use of Geostatistical Output Perturbation (GOP), implemented in the conformal-cubic atmospheric model (CCAM) on NWP data from the eight meteorological stations in Indonesia, i.e. Kemayoran, Priok, Cengkareng, Pondok Betung, Curug, Dermaga, Tangerang and Citeko stations. The findings indicated exponential as the best distribution model for analyzing temperature in Indonesia using GOP. Also, locations which are considerably far away from other locations could have significant impact on the accuracy of the weather forecasts. In this case, Citeko station has quite different characteristics location considering the fact that it is located on higher elevation compared with other stations. Therefore, the exclusion of Citeko station produced better forecasting in terms of accuracy and precision, increasing to about twice the result when the station was included in the analysis.