Temperature and humidity forecast via univariate partial least square and principal component analysis

Abstract

Indonesian Meteorology, Climatology, and Geophysics Agency (BMKG) uses Numerical Weather Prediction (NWP) for short-term weather forecast but it gives biased result. Therefore, this study implements Univariate Partial Least Square (PLS) as Model Output Statistics (MOS) for temperature and humidity forecast. This study uses the maximum temperature (Tmax), minimum temperature (Tmin), and relative humidity (RH) which are called response variables and NWP as predictor variable. The results show that the performance of the model based on Root Mean Square Error of Prediction (RMSEP) are considered to be good and intermediate. The RMSEP for Tmax in all stations is intermediate (0.9-1.2), Tmin in three stations is good (0.5-0.8), and humidity in three stations is also good (2.6-5.0). The prediction result from the PLS is more accurate than the NWP model and able to correct an 89.94% of the biased NWP for Tmin forecasting.