Kalman filter sensitivity tests for the NWP and analog-based forecasts post-processing

The goal of this study is to perform a detailed sensitivity test to find the optimal value of the variance ratio r for four different post-processing forecasts that use the Kalman filter (KF) for the point-based wind gust predictions. The four forecasts analyzed in this paper 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 wind gust impacts the severity of wind-related events. It is usually not a prognostic but a diagnostic variable in NWP models, which makes it an excellent candidate for post-processing. The results suggest that for the KF and KFAS forecasts the r value of 0.01 should be used, whereas for the KFAN and KF-KFAS forecasts the r value should be set to 0.001. The proposed values are considered optimal since they lead to excellent results for the overall data, and the results remain satisfactory even for strong wind. It, however, needs to be mentioned that the usage of different r values always comes with certain trade-offs. The r values smaller than proposed can sometimes slightly improve the overall result, but often lead to considerably worse results for the strong wind. On the other hand, the r values greater than proposed often show good quality of strong wind forecasts but lead to worse results overall. Even so, if the focus is set on extreme events, higher r values than proposed could be considered, but with caution.