1. APR 20021 CRM simulations of the development of convection – some sensitivities Jon Petch Richard Forbes Met Office Andy Brown ECMWF October 29 th 2003 GCSS/ARM Broomfield

2. APR 20022 Content Motivation  Results from an NWP limited area model Impacts of resolution, domain size and 2D/3D differences  Results from the GCSS shallow cumulus »ARM diurnal cycle case  Results from EUROCS »Weakly forced convection »Stronger forced convection  Preliminary results from LBA case Summary

3. APR 20023 Motivation NWP model now run at 12 and 1 km resolution  Event forecasts (Wimbledon etc…)  Flood prediction Sensitivities in NWP simulations  Resolution has been shown to impact timing of convection  Parametrizations also shown to impact timing of convection Timing/location of convection is key in flood prediction  Too late (too far down stream – wrong catchment area?)  Too soon (on upslope for orographic rain – definitely wrong catchment area)

4. APR 20024 Motivation Bridge the gap between LES/CRM and NWP models  Resolution impacts  2D/3D differences Benchmark simulations from LES and CRM  Parametrization/tuning for 1 and 12km simulations »Microphysics (e.g. impacts of rain advection or graupel) »Sub-grid condensation

5. APR 20025 NWP Model Experiments Impacts of rain advection on 1km NWP simulation

6. APR 20026 1 km 3D NWP model

7. APR 20027 2 and 12 km NWP model 2 km 12 km Vertical cross sections of w and ice over Wales Thanks to HumphreyLean (Met Office)

8. APR 20028 Bridging the gap between experiments LES resolution experiments  ARM/GSCC WG1: [10’s to 100’s m] high resolution simulations of the development of shallow cumulus CRM resolution experiments  EUROCS: [100’s to 1000 m] 2D and 3D CRM experiments of the diurnal cycle of deep precipitating convection  LBA: [100’s m] 2D and 3D CRM simulations of the development of convection High resolution NWP  1 to 12 km 3D simulations of weather events over U.K  LBA: 1 km case?

9. APR 20029 LES results The development of shallow cumulus

10. APR 200210 Results from shallow case – high resolution tests TKE Cloud liqud water path

11. APR 200211 Results from shallow case – high resolution tests Cloud fraction Cloud top

12. APR 200212 Results from shallow case – low resolution tests TKE Cloud liqud water path

13. APR 200213 Results from shallow case – low resolution tests Cloud fraction Cloud top

14. APR 200214 Shallow case summary Basic information  Cloud forms around 12 local time  Cloud fractions peak around 20% between 1 and 4pm Resolution study  100 m and less shows convergence  200 m not too bad  Worse than 200 m is poor or forms no cloud  Other models gave similar results at default 100 m (not shown) Domain and 2D/3D differences  2D not tested but had given strange results in BOMEX case  Domain was 6.4km x 6.4 km

15. APR 200215 Some questions (I don’t plan to answer all of them!) Can we begin to bridge the gap between this shallow case and deeper cases? Model related questions:  What is the impact of 250 m to 100m or higher horizontal resolution?  What is the impact of smaller domain?  What is impact of going from 2-D to 3-D

16. APR 200216 More questions Simulation related questions  What is the impact of large scale forcing?  What is impact of interactive radiation?  How important are the initial conditions?  How important are the surface fluxes? Points of interest  At what point does the shallow cumulus deepen?  How long does it stay shallow?  How long does it take to rain after it goes deeper?  How long does it take to reach upper troposphere?  Do interesting things occur at freezing level?

17. APR 200217 Deeper convection EUROCS – diurnal cycle of deeper convection

18. APR 200218 EUROCS simulations LES -> CRM Model  Met Office CRM/LEM (as used in shallow case)  120 vertical levels (15 km domain)  100m or higher vertical resolution  Forced with standard semi-idealised case Simulation  2D »250 m ensemble from 15 runs (250 km domain) »100 m (160 km domain)  3D (100x100 points) »100 m (10 km x 10 km) »250 m (25 km x 25 km)

19. APR 200219 Initial conditions

20. APR 200220 EUROCS surface fluxes cf. ARM Case 3

21. APR 200221 Impact of NSF on basic 2D runs: rain rate New Surface Fluxes

22. APR 200222 Rain rate and w max (NSF)

23. APR 200223 Cloud fraction and total hydrometeor (NSF)

24. APR 200224 Water vapour transport (NSF) 2D 3D

25. APR 200225 In-cloud liquid water (NSF)

26. APR 200226 Cloud fraction (NSF)

27. APR 200227 Double domain size

28. APR 200228 Increased moisture and rectangular domain

29. APR 200229 Increased moisture, bigger domain but lower resolution (1km)

30. APR 200230 Vertical water vapour transport 3D 250m (tot)3D 1km (tot) 3D 250m (res)3D 1km (res) (Res)

31. APR 200231 78910 2D (250km) 3D av (60km) 3D plan (60km) 7 8 9 10

32. APR 200232 8 3D 2D

33. APR 200233 3D 2D 9

34. APR 200234 8910 89 89 89 7km 5km 1.5km 2D

35. APR 200235 100m v 250m cloud sizes

36. APR 200236 Poor representation of the formation of clouds in 1km simulations

37. APR 200237 Summary of EUROCS case CRM runs are very similar to shallow case  timing of initial cloud and peak; cloud fractions; LWP Shallow phase lasts for 2-3 hours but water contents are low 3D produces less deep cloud and rain than 2D at 250m and 100m resolution  no very large domain 3-D runs (50km x 50 km)  some sensitivity to domain size between 25 and 50 km 3D and 2D produce similar amounts of cloud and rain at 1km resolution  even with smaller domains (i.e. ones used at high res.)  much of the transport of water vapour is by sub grid processes right up to the time of significant precipitation

38. APR 200238 LBA LES->CRM->NWP Diurnal cycle again Early days yet! Some problems setting Up NWP model for this case

39. APR 200239 LBA Differences from EUROCS

40. APR 200240 LBA v EUROCS [SH and LH]

41. APR 200241 LBA runs

42. APR 200242 LBA runs

43. APR 200243 2D 3D differences

44. APR 200244 LBA – comparison with NWP

45. APR 200245 Rain 5 hours CRM NWP 100 km

46. APR 200246 Final summary A range of simulations and model set ups are being used to improve convective development in high resolution NWP models:  Benchmark simulations »some experiments have resolved all important scales »which microphysical processes dominate? »impacts of sub-grid sensitivity Along the way:  Understand sensitivity to domain size, resolution and 2D/3D differences in LEM/CRM »2D still needed to allow many runs and very detailed microphysics and radiation