Premium
Combination of Kalman filter and an empirical method for the correction of near‐surface temperature forecasts: Application over Greece
Author(s) -
Anadranistakis Manolis,
Lagouvardos Kostas,
Kotroni Vassiliki,
Skouras K.
Publication year - 2002
Publication title -
geophysical research letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.007
H-Index - 273
eISSN - 1944-8007
pISSN - 0094-8276
DOI - 10.1029/2002gl014773
Subject(s) - kalman filter , mean absolute error , environmental science , econometrics , meteorology , computer science , statistics , mathematics , mean squared error , physics
In response to the need of improved daily operational weather forecasts, a method has been developed and applied for the correction of operationally forecasted near surface temperatures in Greece. Four distinct corrections are applied. The first two are based on Kalman theory making two different assumptions on the dependence of the forecast to the observed temperature. The third correction is based on an empirical method while the fourth provides a corrected temperature combining the corrected temperatures from the previous three methods, with dynamic weights assigned to each of them. The analysis showed that the combined method gives always the best results. Large mean absolute errors of temperature forecasts as high as 6.2°C are suppressed down to 1.5°C after correction, showing that a large part of the forecast error has a systematic character which can be corrected by the method described in this paper.