engleski

Evaluation of Kalman Filter based post-processing methods: focusing on the coastal region of Croatia

Analog-based methods have already shown good results in improving the NWP for point-based forecasts of temperature and wind speed and are thus implemented at the Croatian Meteorological service. Similarly, the algorithm based on the Kalman Filter (KF) shows potential for further forecast improvement if real-time measurements are available, especially when combined with the analog method. After the optimal parameters of the KF algorithm are found, several variations of the KF-based forecast are verified, with a special focus on the coastal area. The four forecasts presented in this work are KF (KF applied to the raw NWP time series), KFAN (KF applied to the simple analog method time series), KFAS (KF applied to NWP forecasts in analog space), and KF- KFAS (KF applied to the KFAS time series). The analog-based forecast is generated with optimal predictor values and additional correction for extreme values. The raw NWP model used is the ALADIN model with a 2 km horizontal resolution. The KF variations reduce the overall RMSE by simultaneously improving the correlation and reducing systematic errors. The KFAN and KF-KFAS forecasts show the best results overall. Nevertheless, the other KF variations have specific advantages. The benefits of including the KF algorithm in the post-processing chain are evident even for rare and more extreme events. Finally, although the results for the coastal area somewhat differ from those obtained for other terrain types, it is evident that including the KF algorithm in the forecasting process shows a great potential for the implementation in the operational suite.

post-processing, Kalman filter, analog method

nije evidentirano