170 likes | 324 Views
4th COSMO meeting 25-27 September 2002 Warsaw. QPF verification of DWD-LM and LAMI model using high resolution non-GTS data in Piedmont region and Northern Italy. Massimo Milelli Elena Oberto Renata Pelosini Paolo Bertolotto. Working Group 5: Verification and Case Studies
E N D
4th COSMO meeting 25-27 September 2002 Warsaw QPF verification of DWD-LM and LAMI model using high resolution non-GTS data in Piedmont region and Northern Italy Massimo Milelli Elena Oberto Renata Pelosini Paolo Bertolotto Working Group 5: Verification and Case Studies (Co-ordinator C.Cacciamani ARPA-SMR)
Overview: the W.P. 5.3.1 since Athens 2001 • Verification of the QPF based on the Lokal DWD model using high resolution non-GTS data coming from: • Piedmont Region (from 01/2000 to 06/2002) • Italian Regions (Trentino, Veneto, Emilia Romagna, Marche, Liguria, Piemonte) from 03/2001 to 06/2002 For Piedmont region we have averaged over warning areas; for Italian region over boxes sized 0.50° *0.50° (about 50 km * 50 km) that represent the measurement of a mean basin.
Piedmont “warning areas” • 11 “basins”, 3000 km2 each • 1-2 ECMWF grid points • 60 Lokal Modell grid points • Each “basin” is a group of neighbouring hydrological catchments
Verification over Piedmont basins average in 24h (01/2000-06/2002) • LM12 is slightly better than LM00 • overestimate for low thresholds for the first 24h • greater underestimate of LM00 for high thresholds • The first 24h are better than the second (decline with time)
Verification over Piedmont basins average in 24h (01/2000-06/2002) • Good results (all the points are in the upper left part) • LM12+24: good skills for all thresholds (%HR>>%FAR) • skill decreases with time
Situation of data set for Italy: data used data not yet useful data soon available
Verification over Italy, box 0.5° average in 24h (03/2001-06/2002) Underestimate for thresholds > 10 mm Are these stations representative of high precipitation? About 45 grid point/box compare with 4/5 station point: is there any smooth of the forecasted signal ?
Verification over Italy, box 0.5° average in 24h (03/2001-06/2002)
2 Sensitivity: we have carried out a parametric study of the averaging mesh size, in order to find the optimum area for QPF evaluation. Data from: • Piedmont Region (from 01/2000 to 06/2002) • Italian Regions (Trentino, Veneto, Emilia Romagna, Marche, Liguria, Piemonte) from 03/2001 to 06/2002
Sensitivity over Piedmont, boxes average in 24h 01/2000-06/2002 • AA perform slightly better
Sensitivity over Piedmont, boxes average in 24h 01/2000-06/2002 • verification sensitive to the definition of the areas
Sensitivity over italian regions, boxes average in 24h 03/2001-06/2002 • Sensitive to area size • not all the boxes considered • all the boxes taken into account
Sensitivity over italian regions, boxes average in 24h 03/2001-06/2002
3Verification of LAMI model using high resolution non-GTS data from Piedmont region, warning areas averaged, over period from 06/2002 to 08/2002 • Lami12 is better than Lami00 • the first 24h overestimate more than the second 24h
A comparison of the two models: • Lami +24 overestimates more than LM +24 • Lami +48 has a different behaviour with respect to the thresholds and it is worse than LM +48
Conclusions • Generally, the 12 runs (LAMI and LM-DWD) have better skills more inertia of the atmosphere at 00 UTC in the triggering of the precipitation • The model performance decreases with time (known !) • Lami first 24h overestimates the precipitation: probably due to the missing data assimilation cycle • Sensitivity tests show a great dependency of the QPF skill on area definition; this definition has to be based on the morphology and on the climatology of severe events • the QPF skill for hydrogeological risk assessment over Piedmont is good (importance of working with end-user targeted verification)
Precipitation thresholds for warning over Piedmont basins averaged in 24h 01/2000-12/2001