1. Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Experiments at MeteoSwiss : TERRA / aerosols Flake Jean-Marie Bettems / MeteoSwiss COSMO GM Roma, September 5 th, 2011

2. 2 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Test suite configuration COSMO-7 4 years assimilation cycle, 2007 (2x) – 2009, initialized with IFS (including soil) 2 daily 72h forecasts in January 2008, June 2008 and June2009, initialized at 00 and 12UTC 2 configurations, reference (513/515) and test (514/516) Reference matches the operational configuration Test configuration differs by the use of a new TERRA configuration (hydrological LBC, exponential root, heat cond.), new external parameters for TERRA (in particular NDVI based climatology of LAI and PLCOV, emissivity and min. stomatal resistence maps), new climatology for aerosol species, FLake

3. 3 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Soil types

4. 4 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Soil water content, reference

5. 5 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Soil water content, test

6. 6 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Soil water content Bottom layer reflects difference between lower boundary conditions (free drainage in reference, water reservoir in test). Equilibrium state is not reached even after 4 years, in particular in the reference (CH. Schaer communication: requires 5 to 10 years spin-up, but after spin-up reference configuration does not dry up). Similar time series of top and mid levels in reference, but not in the test configuration (effect of exponential roots? Root depth typically within 0.3 and 0.7 [m]). In summer, test configuration has a drier top level but a moister mid level.

7. 7 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Verification of surface pressure SYNOP, COSMO-7 domain, test vs. reference Surface pressure is an integrated quantity (air mass). Surface pressure is assimilated → no differences in analysis. Higher surface pressure with new configuration Significant reduction of mean error in June 2008 and June 2009, increase of mean error in winter 2008 June 2008June 2009 Mean error [Pa] Mean abs. error [Pa] Lead time [h]

8. 8 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Higher near surface temperature with new configuration, specially during day time (not shown here). Test is drier in June forecasts, wetter in January forecast and in assimilation cycle. No daily cycle of error. Mixed impact on verification scores, depending on the period of interest. June 2008June 2009 T, Mean error [K] TD, Mean error [K] Lead time [h] Verification of 2m temperature and humidity SYNOP, COSMO-7 domain, test vs. reference

9. 9 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Rel. hum. bias, +72h [%] Verification of troposphere TEMP, COSMO-7 domain, test vs. reference June 2009 Pressure [hPa] Geopot. bias, +72h [m] Temp. bias, +72h [%] Geopotential and humidity are consistently changed in the whole boundary layer, unlike the temperature which is mainly affected at screen level.

10. 10 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Verification - Summary The new configuration impacts the whole boundary layer. Higher surface pressure O(20Pa), higher screen level winds O(0.1m/s), warmer and drier screen level in summer O(0.2K). Mixed impact on verification scores (surface pressure, precipitation, screen level temperature, wind and humidity, cloud cover, boundary layer), depending on the period considered. Further tuning of the model could maybe help getting an impact which is more consistently positive. Some shortcomings of this study: Too short spin-up for the initial conditions chosen for this experiment (IFS) Location dependent analysis is missing Deeper analysis is missing (fluxes...)

11. 11 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Modelling of lake temperature Only the lake surface temperature is considered here. The reference configuration uses the climatology of the Limmat near Baden as proxy for the lakes considered in this study. The test configuration uses the FLake model cold start on 01.01.2007 lake depth is the only external parameter, all other FLake parameters are fixed.

12. 12 COSMO GM, Data Pool / Roma, Sep. 5 th, 2011 Observations of lake temperature – Leman pl gl

13. 13 Test suite TERRA and FLake FLake – Model versus Observations LEMAN WESTi = 231, j = 155Lake depth = 50m

14. 14 Test suite TERRA and FLake FLake – Model versus Observations LEMAN EASTi = 234, j = 156Lake depth = 50m

15. 15 Test suite TERRA and FLake FLake – Model BODENSEEi = 267, j = 174Lake depth = 9.8m

16. 16 Test suite TERRA and FLake FLake – Impact on T_SO and T_2M T_SO ref T_SO FLake T_2M ref T_2M FLake ∆ 6-7° ∆ 2-3°

17. 17 Test suite TERRA and FLake FLake – Impact on T_SO and FF_10M T_SO ref T_SO FLake FF_10M ref FF_10M FLake ∆ 6-7°

18. 18 Test suite TERRA and FLake FLake – Learnings Behaviour of surface temperature in FLake very much depends on the choice of the lake depth. Differences in surface temperature between reference and test can be significant, larger than 5°C, in particular for shallow lakes. When comparing both approaches against observations (Leman), no clear winner. Both methods provide reasonable results, within ±5°C, with FLake too warm and Limmat too cold. One benefit of FLake is a plausible representation of the inter-annual variability of the surface temperature; significant day to day variability is also observed, but no verification is available. Another benefit of FLake is the universality of this solution (but it is important to have a lake depth which is reasonably correct). Impact on 2m temperature is much reduced and spatially localized; impact on 10m wind is spatially localized (at least for the case considered).

19. 19 Test suite TERRA and FLake FLake – Remarks Differences in observed values at the two Leman locations, which can be larger than 5°C for many weeks and with the shallow part of the lake colder than the deeper part, illustrate the importance of water currents. This is a limitation of a 1D model such as FLake.

20. 20 Test suite TERRA and FLake Thank you for your attention!