Conditional verification of all COSMO countries: first results

1. Conditional verification of all COSMO countries: first results COSMO General Meeting, September 2012, Lugano by the members of WG Verification

2. Objectives of conditional verification • Contribute to COSMO model development • Improve the understanding of forecast errors • Identify possible sources of errors in COSMO • Contribute to guidelines on how to use COSMO forecasts

3. Feedback loop within COSMO

4. Common conditional verification results • Cloud cover clearly stratifies the COSMO forecast error of T2M (no matter which diagnostic) • Under observed clear sky conditions, the mean error has a pronounced daily cycle: • All models underestimate daytime T2M and overestimate nighttime T2M • Under observed overcast conditions, this behaviour is not observed T2M for overcast conditions T2M for clear sky conditions RMSE ME Autumn 2011, ME and RMSE for many COSMO models Spring 2012, ME and RMSE for many COSMO models

5. Searching for clues… • Additional stratification to look at cases with stable boundary layer → distinguish between dynamical and radiation dominated processes

6. COSMO-ME Conditional VerificationT2m when observed TCC ≤ 25%

7. COSMO-ME Conditional VerificationT2m when observed TCC ≤ 25% & wind speed ≤ 2 m/s

8. COSMO-7 Conditional VerificationT2m when forecast TCC ≤ 25%

9. COSMO-7 Conditional VerificationT2m when forecast TCC ≤ 25% & wind speed ≤ 2.5 m/s

10. Searching for clues… • Additional stratification to look at cases with stable boundary layer → distinguish between dynamical and radiation dominated processes • → in calm wind conditions, the underestimation of the daily temperature amplitudeis even more pronounced→ overestimated thermal mixing (minimal diffusion coefficient?) • Nighttime overestimation from insufficient radiative cooling? Thermal conductivity of the soil? • Daytime underestimation from underestimated sensible heat flux? Impact of soil moisture?

11. T2M with fcst soil moisture condition ME, RMSE dry conditions ME, RMSE wet conditions ME, RMSE no conditions Similar systematic error properties as for cloud cover conditions, but phase slightly shifted and different data sample → new box opened…

12. Precipitation with CAPE conditions ETS Unstable (CAPE>=50J/kg) Stable (CAPE<50J/kg) FBI Overestimation of preci amount for lower thresholds in high CAPE cases and ETS a bit reduced WG5 COSMO General Meeting, Lugano 2012

13. Precipitation with CAPE conditions FBI Unstable (CAPE>=50J/kg) Stable (CAPE<50J/kg) ETS

14. Precipitation with CAPE FBI Higher performance (in terms of ETS, also POD) in stable conditions, but CAPE condition displaced → find appropriate time period, other condition Unstable (CAPE>=50J/kg) Stable (CAPE<50J/kg) ETS

15. Conclusions • Conditional verification provides us with tools for analysing rather complex COSMO model errors • Use of intensive measurement sites (e.g. sensible and latent fluxes for clear sky temperature error, soil moisture for temperature and dewpoint error) and radiosoundings • Identify suitable stratifications for precipitation (e.g. appropriate time integrations for CAPE or convective time scale, so far no success with weather classes) • Tight interaction with WG3 and others

17. CAPE>50 CAPE<50 Very high POD values for unstable conditions, FAR not so different WG5 COSMO General Meeting, Lugano 2012

18. 2mT, Td with dry or wet soil conditions Fall Winter Spring W_SO Water content of first soil layer(kg/m2) 1cm. Td: Higher error in dry soil and larger underestimation 2mT: Higher error in wet soil and larger understimation Fall Winter Spring WG5 COSMO General Meeting, Lugano 2012

19. TD2M with soil moisture condition ME, RMSE dry conditions ME, RMSE wet conditions ME, RMSE no conditions

20. FF10M with grid point height ME, RMSE all stations ME, RMSE gp height > 800m ME, RMSE gp height < 800m

21. FF10M scatter plot Winter 2012