 |
| Titel |
The use of tendencies for post-processing. Analysis of the YOTC forecast data. |
| VerfasserIn |
Bert Van Schaeybroeck, Stéphane Vannitsem |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250050365
|
|
|
|
|
|
| Zusammenfassung |
| We assess the usefulness of model tendencies as predictors for post-processing
meteorological forecasts. A case study is performed using the YOTC dataset available
ECMWF which includes tendencies of the operational deterministic forecast. We have trained
and verified different seasons in 2008 and 2009 for forecasts up to 36 hours. These forecast
data are compared with the ECMWF analysis data and with observations at the different
synoptic stations in Belgium. We focus on forecast improvement for variables such as
surface pressure, two-meter temperatures and wind velocity at ten meter. Using a
selection procedure to sort out predictors from a list including tendencies and a list
without tendencies, we find that a substantial increase of predictability arises from
the use of tendencies. With an operational post-processing purpose in mind, we
evaluate the usefulness of predictor selection for each lead time. Also we attempt
to relate the selected tendency predictors to the model physics and we compare
post-processing techniques including MOS, EVMOS (Vannitsem, 2009) and Tikhonov
regression.
References
[1]Â Â Â Vannitsem S., 2009: A unified linear Model Output Statistics scheme for
both deterministic and ensemble forecasts,Quart. J. Roy. Meteorol. Soc., 135:
1801-1815. |
|
|
|
|
|
|