we carried out the qpf verification of the three model versions (cosmo-i7, cosmo-7, cosmo-eu) with...
TRANSCRIPT
we carried out the QPF verification of the three model versions (COSMO-I7, COSMO-7, COSMO-EU) with the following specifications:
• From January 2006 till July 2007
• Dataset: high resolution network of rain gauges coming from COSMO dataset and Civil Protection Department about 1300 stations
• Method: 6h or 24h averaged cumulated precipitation value over meteo-hydrological basins
Last results on precipitation verification over Italy(Elena Oberto, Marco Turco - ARPA Piemonte)
• We performed the seasonal trend over the last year (mam’06- mam’07)
• We performed the daily trend
we present COSMO-I7 verification results over a long period (from 2003 till now) over Italy and a comparison between COSMO-I7 and COSMO-I2
COSMO-I7 COSMO-7 COSMO-EU
BIAS D+2 10mm/24h
200601-200707
BIAS D+2 10mm/24h
200601-200707
BIAS D+2 10mm/24h
200601-200707
The BIAS has a quite similar pattern for the three versions for North Italy: in general there is an overestimation over the mountain areas and an underestimation over the lowlands. For central and South Italy COSMO-7 and COSMO-EU are similar: there are more cases of underestimation, while COSMO-I7 presents more cases of overestimation.
(the values in the Abruzzo region maybe are due to observed data problems)
COSMO-I7 COSMO-7 COSMO-EU
POD D+2 10mm/24h
200601-200707
POD D+2 10mm/24h
200601-200707
POD D+2 10mm/24h
200601-200707
The best values are in the North but the three versions have different skill; COSMO-7 has very good values over alpine chain; COSMO-EU has good values in the Northwest. In general, the Eastern side presents the lowest values.(the low POD values in the Abruzzo region maybe are due to observed data problems)
COSMO-I7 COSMO-7 COSMO-EU
FAR D+2 10mm/24h
200601-200707
FAR D+2 10mm/24h
200601-200707
FAR D+2 10mm/24h
200601-200707
The three versions have quite similar pattern skill: the worst values are in the South, central, and mountains areas. Slightly more false alarm for COSMO-I7. (the high FAR values in the Abruzzo region maybe are due to observed data problems)
COSMO-EUCOSMO-7
Seasonal comparison between COSMO-EU/COSMO-7 (D+2)
mam06, jja06, djf7: COSMO-7 better than COSMO-EU;
mam07, better bias for COSMO-EU, better pod COSMO-7
It seems there is a better skill in latest three seasons, less FAR, more POD
COSMO-7COSMO-I7
Seasonal comparison between COSMO-7/COSMO-I7 (D+2)
mam06, son06, djf07, mam07: COSMO-7 better than COSMO-I7
It seems there is a better skill in latest three seasons, less FAR, more POD
COSMO-EUCOSMO-I7
Seasonal comparison between COSMO-EU/COSMO-I7 (D+2)
djf07 and mam07: COSMO-EU better BIAS than COSMO-I7 but worse POD;
jja06, better COSMO-I7 than COSMO-EU
It seems there is a better skill in latest three seasons, less FAR, more POD
COSMO-7
Seasonal comparison: MAM 2007 (for-obs)/obs %
COSMO-I7 COSMO-EU
Balance of negative and positive error distributed all over the territory !
COSMO-7COSMO-I7COSMO-EU
Daily trend from Jan06 to Aug07 General remarks:
•bias>1 (particularly for COSMO-I7)
•It is evident a sort of diurnal cycle
COSMO-7COSMO-I7COSMO-EU
35
Open question: for low thresholds the bias peak occurs during midday, but for high thresholds it is shifted to midnight. Why??
(For high thresholds the events mainly occurred during spring-summer so the precipitation have a great convective component…)
FOCUS ON COSMO-I7: SEASONAL TREND
for the period from 200212 to 200706 (DJF’03- MAM’07)
FOCUS ON COSMO-I7: SEASONAL TREND
for the period from 200212 to 200706 (DJF’03- MAM’07)
In general there is not a remarkable trend. It has been chosen the 20 mm/24h threshold because of a slightly positive trend.
There is a seasonal cycle with generally better skills during autumn and worse skills during summer.
D+2 is worse than D+1
COMPARISON COSMO-I7/COSMO-I2: 200705-200707
BIAS; D+1 (+00/+24H); PERIOD: 200705-200707
BIAS; D+2 (+24/+48H); PERIOD: 200705-200707
+ COSMO-I7
COSMO-I2
+ COSMO-I7
COSMO-I2
Above 10 mm/24 the bias
difference is statistically significant: COSMO-I2
overestimates more then COSMO-I7
COMPARISON COSMO-I7/COSMO-I2: 200705-200707
POD; D+1 (+00/+24H); PERIOD: 200705-200707
POD; D+2 (+24/+48H); PERIOD: 200705-200707
+ COSMO-I7
COSMO-I2
+ COSMO-I7
COSMO-I2
statistically significant
differences for high thresholds with better POD
for COSMO-I2
statistically significant
differences for medium-high
thresholds with better POD for
COSMO-I2
COMPARISON COSMO-I7/COSMO-I2: 200705-200707
FAR; D+1 (+00/+24H); PERIOD: 200705-200707
FAR; D+2 (+24/+48H); PERIOD: 200705-200707
+ COSMO-I7
COSMO-I2
+ COSMO-I7
COSMO-I2
statistically significant
differences for medium
thresholds with greater false
alarm numbers for COSMO-I2
….to sum up
•The error spatial pattern shows a general underestimation on the
plain and overestimation on the mountain.
•The seasonal trend over the last year (mam’06- mam’07) seems to
have a better skill in the latest three seasons..positive trend!
•We performed the diurnal cycle: moving from low to high
precipitation amount the overestimation peak is shifted from midday to
midnight open discussion
•COSMO-I7 verification results over a long period (from 2003 till
now) over Italy do not show a worsening but the improvement is only
relative to high thresholds.
•The comparison between COSMO-I7 and COSMO-I2 during last
three months generally shows a better scores for COSMO-I7