COSMO General Meeting – Moscow 06-10 Sept 2010 Some results from operational verification in Italy...
Transcript of COSMO General Meeting – Moscow 06-10 Sept 2010 Some results from operational verification in Italy...
COSMO General Meeting – Moscow 06-10 Sept 2010
Some results from operational verification in Italy
Angela Celozzi - Federico GrazziniMassimo Milelli - Elena Oberto
Adriano Raspanti - Maria Stefania Tesini
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs ECMWF CROSS MODEL COSMOI7 vs ECMWF CROSS MODEL COSMOME vs COSMOIT COSMOME –Upper Air COSMOI7 – Upper Air (OBS and
Analysis) CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 CROSS MODEL COSMOME vs COSMOIT COSMOME –Upper Air COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME vs ECMWF Temperature
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME vs ECMWF Dew Point Temperature
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME vs ECMWF Mean Sea Level Pressure
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME vs ECMWF Total Cloud Cover
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME vs ECMWF Wind Speed
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 CROSS MODEL COSMOME vs COSMOIT COSMOME –Upper Air COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOI7 vs ECMWF Dew Point Temperature
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOI7 vs ECMWF Mean Sea Level Pressure
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOI7 vs ECMWF Wind Speed
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOI7 vs ECMWF Total Cloud Cover
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 CROSS MODEL COSMOME vs COSMOIT COSMOME –Upper Air COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
TemperatureCOSMOME vs COSMOIT
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Mean Sea Level PressureCOSMOME vs COSMOIT
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Dew PointTemperatureCOSMOME vs COSMOIT
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Total Cloud CoverCOSMOME vs COSMOIT
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Wind SpeedCOSMOME vs COSMOIT
SON
JJA
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Conclusion
• COSMO - ME generally better than IFS, except MSLP
• COSMO – I7 better or almost the same than IFS
• Comparison COSMO-ME and COSMO-IT shows improvements for High-Res.
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs COSMOIT CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 COSMOME – Upper Air COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME –Upper AirTemperature
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOME –Upper AirWind Speed
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs COSMOIT CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 COSMOME –UpperAir COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOI7 –Upper Air Temperature
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMOI7 –Upper AirWind Speed
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMO I7 COSMO BackUp ECMWF
MSL FC+48 – Each model is verified against its own analysis
Shaded contouring every 0.5 hPa, starting from 0.5. Red and blue lines representPositive/Negative bias, every 0.5 hPa
Spatial distribution of mean absolute error (MAE), computed over MAM 2010
UPPER-AIR (against analysis)
COSMO General Meeting – Moscow 06-10 Sept 2010
Forecast Step
MAE growth with forecast step, computed over COSMOI7 domainSpring 2010 (MAM 2010) – All models and analyses are interpolated onA regular grid at 0.25 * 0.25 deg of h-resolution. Everyone against its own analysis. MSL
UPPER-AIR (against analysis)
COSMO General Meeting – Moscow 06-10 Sept 2010
Spatial distribution of mean absolute error (MAE), computed over MAM 2010
T850 FC+48 – Each model is verified against its own analysis
Shaded contouring every 0.5 C°, starting from 0.5. Red and blue lines representPositive/Negative bias, every 0.5 C°
COSMO I7 COSMO BackUp ECMWF
UPPER-AIR (against analysis)
COSMO General Meeting – Moscow 06-10 Sept 2010
MAE growth with forecast step, computed over COSMOI7 domainSpring 2010 (MAM 2010) – All models and analyses are interpolated onA regular grid at 0.25 * 0.25 deg of h-resolution. Everyone against its own analysis. Z 700 hPa
UPPER-AIR (against analysis)
COSMO General Meeting – Moscow 06-10 Sept 2010
Conclusion
• COSMO – ME and COSMO-I7 have a general good result in upper air Verification
• COSMO-ME seems better, but improvement from MAM for COSMO-I7 (bug in AOF file until march)
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs COSMOIT CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 COSMOME –Upper Air COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional Verification Temp – TCC obs <=35%
Worse behaviour for all the seasonsCompare to no condition model
SON
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional Verification Temp – TCC obs >=75%
SON
MAM
DJF
Better behaviour for all the seasonsCompare to no condition model
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional Verification MSLP – MSLP >=mean
SON
MAM
DJF
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional Verification Tdew – Wind Speed (Obs) <=2 m/s
SON
MAM
DJF
Almost indifferent to the condition in obs space
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional Verification Tdew – Wind Speed (fcs) <=2 m/s
SON
MAM
DJF
Worse behaviour for all the seasonsIn fcs space
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional Verification
Temp – Prec +06 <= 0,5 Temp – Prec +06 <=10
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional VerificationTemp – MSLP >=mean
SON
MAM
DJF
Worse behaviour in DJF for RMSESimilar fo other seasons
COSMO General Meeting – Moscow 06-10 Sept 2010
Conditional VerificationTemp – MSLP <=mean
SON
MAM
DJF
General better behaviour for all theSeasons compare to NC
COSMO General Meeting – Moscow 06-10 Sept 2010
Conclusion
• Comparison between NC and Cond verification seems effective in most of the cases
• A standard set of Conditions should be decided by WG5 and produce on regular basis
COSMO General Meeting – Moscow 06-10 Sept 2010
Verification
CROSS MODEL COSMOME vs COSMOIT CROSS MODEL COSMOME vs ECMWF CROSS MODEL ECMWF vs COSMOI7 COSMOME –Upper Air COSMOI7 – Upper Air CONDITIONAL VERIFICATIONS Long Term and Seasonal Precipitation
COSMO General Meeting – Moscow 06-10 Sept 2010
Seasonal trend - low thresholds• All the versions present a seasonal cycle with an overestimation during summertime (except COSMO-7 and I2)• COSMO-7 and I2 underestimate• Overestimation error decreases in D+2 (spin-up effect vanished)
QPF verification of the 4 model versions at 7 km res. (COSMO-I7, COSMO-7, COSMO-EU, COSMO-ME) with the 2 model versions at 2.8 km res. (COSMO-I2, COSMO-IT)Dataset: high resolution network of rain gauges coming from COSMO dataset and Civil Protection Department 1300 stationsMethod: 24h/6h averaged cumulated precipitation value over 90 meteo-hydrological basins
COSMO General Meeting – Moscow 06-10 Sept 2010
• Very light improvement trend
• Seasonal error cycle: lower ets during winter and summertime
• no significant differences between D+1 and D+2
• Last winter (very snowy particularly in Northern Italy): low ets value (D+1 and D+2) model error or lack of representativeness of the rain gauges over the plain during snowfall ?
Seasonal trend - low thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
• Very light improvement trend
• Seasonal cycle with more false alarms in summertime (particularly for I7)
• no significant differences between D+1 and D+2
Seasonal trend - low thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
• Quite stationary during last seasons
• I2 has very low values during summer
• no significant differences between D+1 and D+2
Seasonal trend - low thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
Driving model comparison: ECMWF/COSMO-ME/COSMO-IT, low thresholds
• ECMWF tendency to forecast low rainfall amounts big overestimation, big false alarms, very low ets, quite good pod
• Better prediction for COSMO-models (no strong differences between ME and IT)
COSMO General Meeting – Moscow 06-10 Sept 2010
• ECMWF tendency to forecast low rainfall amounts big overestimation, big false alarms, very low ets, quite good pod
• Better prediction for COSMO-models BUT bad performance during summertime
Driving model comparison: ECMWF/COSMO-ME/COSMO-IT, low thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
• Slight bias reduction during latest seasons
• Last winter: all the versions overestimate (probably due to lack of representativeness of the rain gauges over the plain during snowfall)
• Strong COSMO-7 underestimation BUT slight improvement during latest seasons
Seasonal trend - high thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
• Low values during summertime
• In general, quite stationary error since son2008 up to now
• All the versions present a jump around son2008: ets increases from 0.2-0.4 up to 0.3-0.5 (cosmo-I7: son2008 introduction of 4.3 version with new T2m diagnostic)
• Skill decreases with forecast time
Seasonal trend - high thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
Seasonal trend - high thresholds
• Slight far reduction during last two years BUT high values during summer2009 and winter 2010 (probably due to lack of representativeness of the rain gauges over the plain during snowfall)
• Small far increase last spring
COSMO General Meeting – Moscow 06-10 Sept 2010
• Low probability of detection during intense convective events
• It is noticeable the improvement since son2008 BUT a subsequent worsening during 2009 and the first half of 2010
• Skill decreases with forecast time
Seasonal trend - high thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
• ECMWF difficulty to forecast high rainfall amounts bias around 1 BUT big false alarms, very low ets and pod
• Better prediction for COSMO-models
Driving model comparison: ECMWF/COSMO-ME/COSMO-IT, high thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
• ECMWF difficulty to forecast high rainfall amounts bias around 1 BUT big false alarms, very low ets and pod
• Better prediction for COSMO-models
Driving model comparison: ECMWF/COSMO-ME/COSMO-IT, high thresholds
COSMO General Meeting – Moscow 06-10 Sept 2010
Diurnal cycle - low thresholds
• Little initial spin-up (especially for I7 and I2)
• No strong performance differences among the versions
• Slight diurnal cycle
• Slight worsening with forecast time
COSMO General Meeting – Moscow 06-10 Sept 2010
Diurnal cycle - high thresholds
• Little initial spin-up vanished with threshold increasing
• No strong performance differences among the versions except COSMO-7 underestimation
• Pronounced worsening with forecast time
COSMO General Meeting – Moscow 06-10 Sept 2010
COSMO-7 COSMO-I7 COSMO-ME
COSMO-EU COSMO-I2 COSMO-IT
Bias, 10mm/24h
200812-201005
• Systematic overestimation over Alpine areas, especially in the western part and in Veneto/Trentino-Alto Adige (incorrect representation of flow interaction with alpine chain during westerlies and north-easterlies ?)
• COSMO-7 underestimates especially in southern Italy (border of the domain ?)
• COSMO-I7 overestimates the Adriatic areas (especially during north-easterly flow forecasters experience)
• COSMO-I2 underestimates, COSMO-IT overestimates
COSMO General Meeting – Moscow 06-10 Sept 2010
Far, 10mm/24h
200812-201005
COSMO-7 COSMO-I7 COSMO-ME
COSMO-EU COSMO-I2 COSMO-IT
• More false alarms over Sardinia and Alpine areas, especially in the western part and in Veneto/Trentino-Alto Adige (incorrect representation of flow interaction with alpine chain during westerlies and north-easterlies ?)
• More false alarms for COSMO-I7 (and the other ones) over the Adriatic areas (especially during north-easterly flow forecasters experience)
COSMO General Meeting – Moscow 06-10 Sept 2010
Relative error %, SON 2009
COSMO-7 COSMO-I7 COSMO-ME
COSMO-EU COSMO-I2 COSMO-IT
• Too precipitation amount over Alpine areas, especially in the western part and in Veneto/Trentino-Alto Adige (incorrect representation of flow interaction with alpine chain during westerlies and north-easterlies ?)
• Few QPF for COSMO-7, COSMO-I2 and COSMO-IT
• Quite good QPF for COSMO-I7, COSMO-ME and COSMO-EU
COSMO General Meeting – Moscow 06-10 Sept 2010
Relative error %, DJF 2010
COSMO-7 COSMO-I7 COSMO-ME
COSMO-EU COSMO-I2 COSMO-IT
• Few precipitation amount (all the versions except COSMO-I2) in the Padana Plain: very snowy winter also in plain areas where there are no heated rain gauges lack of representativeness
• Few QPF for COSMO-7 in southern Italy
COSMO General Meeting – Moscow 06-10 Sept 2010
Relative error %, MAM 2010
COSMO-7 COSMO-I7 COSMO-ME
COSMO-EU COSMO-I2 COSMO-IT
• Too precipitation amount over Alpine areas, especially in the western part and in Veneto/Trentino-Alto Adige (incorrect representation of flow interaction with alpine chain during westerlies and north-easterlies ?)
• Few QPF for COSMO-7, COSMO-I2 and COSMO-IT
• Quite good QPF for COSMO-I7, COSMO-ME and COSMO-EU