250 likes | 391 Views
Verification of COSMO-2 with a mobile wind profiler. Validation studies of project CN-MET. Christophe Hug, Pirmin Kaufmann , Guy de Morsier WG5 Meeting 7 th of September 2009. Project CN-MET I – The Context.
E N D
Verification of COSMO-2 with a mobile wind profiler Validation studies of project CN-MET Christophe Hug, Pirmin Kaufmann, Guy de Morsier WG5 Meeting 7th of September 2009
Project CN-MET I – The Context • MeteoSwiss project conducted for the Swiss Federal Nuclear Safety Inspectorate ENSI • Replacement (and improvement) of the current emergency preparedness tool for the Swiss nuclear power plants (NPP) • Current tool is based on empirical profiles and tower measurements (Windbank) to provide input for dispersion model • The tower will be progressively dismantled starting in September 2009 • COSMO-2 data replace the measurements as input for dispersion models of ENSI
CN-MET II – The network St. Chrischona Leibstadt Kleindöttingen Beznau Schaffhausen Grenchen Uetliberg Mühleberg Mühleberg Payerne Gösgen Stockeren Bantiger
CN-MET III – The Requirements • Accuracy requirement set by the specification document for the CN-MET project (HSK-ENSI 2005) • Wind direction (dd) for wind speed > 2 m s-1 95% quantile within ± 20° • Wind speed from 2 m s-1 to 10 m s-1 95% quantile of the absolute error within ± 1 m s-1 • Wind speed larger than 10 m s-1 95% quantile of the relative error within ± 10%
Wind profiler in Kleindöttingen (KDT) • H = 205 m high-mode • H = 72 m low-mode • Measurement: • 675 – 4773 m MSLhigh-mode • 440 – 1452 m MSLlow-mode • Height of site: 321 m MSL Manufacturer specifications: wind direction ± 10°wind speed ± 1 m/s
Verification method • Interpolation of • wind profiler data onto model levels (KDT) or • model data onto wind profiler levels (ZES) • Verification of wind speed and wind direction • 30 min time step (KDT) • 60 min time step (ZES) • Separate verification for profiler high-mode and low-mode • Focus on the forecast lead times +00 to +06 h
Normal distribution ? Both wind speed and wind direction distribution fail the Lilliefors test. Normal probability plots make it obvious for wind direction. For wind direction and speed there are too many outliers.
KDT – High-mode wind direction High-mode Wind direction ff ≥ 2 m/s
ZES – High-mode wind direction High-mode Wind direction ff >= 2 m/s
KDT – Low-mode wind direction Low-mode Wind direction ff ≥ 2 m/s
ZES – Low-mode wind direction Low-mode Wind direction ff >= 2 m/s
KDT – High-mode wind speed High-mode Wind speed ff <10 m/s
ZES – High-mode wind speed High-mode Wind speed ff < 10 m/s
KDT – Low-mode wind speed Low-mode Wind speed ff ≤ 10 m/s
ZES – Low-mode wind speed Low-mode Wind speed ff < 10 m/s
KDT – High-mode realtive wind speed High-mode Wind speed ff ≥ 10 m/s
Histograms and scatterplots – all levels 0’ Wind direction High-mode ff ≥ 2 m/s Wind direction High-mode ff ≥ 2 m/s
Case 14 – 15 Aug. 2008 (outlier) Foehn in green
Verification in KDT with WP (dd) LEI BEZ
Verification in KDT with WP (ff) LEI BEZ
Conclusions • The verification scores of COSMO-2 with wind profiler depends on the location of the profiler. • In average COSMO-2 forecasts are able to reproduce the situation in KDT and ZES • The tails of the error distribution is larger than for a normal distribution, due to outliers that are always part of a numerical forecast. • Comparison with tower measurements shows similar or mostly better results than with wind profiler. Thank you.