1. © Crown copyright Met Office DICE project Adrian Lock and Martin Best

2. © Crown copyright Met Office Apologies For not looking at submitted data in sufficient detail until last week Especially to Maria whose model output from stage 1 missed stage 3 as a result For sending and resending stage 3a forcing All the other problems that I’ve now forgotten!

3. © Crown copyright Met Office Outline of DICE These stages test: LSM and SCM stand-alone performance against observations (stage 1) What is the impact of coupling? (stage 2) How sensitive are different LSM and SCM to variations in forcing? (stage 3)

4. © Crown copyright Met Office CASES-99 case study Steeneveld et al (2006) Field experiment in Kansas, USA Copy Steeneveld et al (2006) and run a 3 day simulation from 2pm local time on 23 rd October 1999 (to 26 th ) Recall GABLS II ran for from 2pm on 22 nd for 2.5 days Clear skies throughout Gives 3 nights of varying character intermittent turbulence continuous turbulence very stable, almost no turbulent fluxes

5. © Crown copyright Met Office CASES-99 location

6. © Crown copyright Met Office CASES-99 location Poulos et al 2002

7. © Crown copyright Met Office CASES-99 location Main 60m tower

8. © Crown copyright Met Office CASES-99 location Google streetview Looking north Looking SW “The area consisted of relatively flat homogeneous terrain (average slopes are 0.2°) with a relatively dry soil and lacks obstacles in the near surroundings.” Steeneveld et al

9. © Crown copyright Met Office CASES-99?

10. © Crown copyright Met Office CASES-99 photos Joan Cuxart

11. © Crown copyright Met Office CASES-99 photos Joan Cuxart

12. © Crown copyright Met Office CASES-99 photos Poulos et al 2002 – 60m tower

13. © Crown copyright Met Office CASES-99 photos Website – from 60m tower Looking south Looking SSW

14. © Crown copyright Met Office Stage 1a set-up

15. © Crown copyright Met Office DICE set-up Atmosphere model forcing

16. © Crown copyright Met Office Single column model Forcing for surface-obs-forced and coupled Based on Steeneveld et al (2006) who used: Time-varying geostrophic wind (uniform with height) Large-scale horizontal advective warming (~4K/day), for day 2 only Subsidence Relaxation to observed sondes above the BL top

17. © Crown copyright Met Office How much subsidence? From sondes (descent of base of moist layer): D=5x10 -6 s -1 (w=-Dz) until 26 th October, as in Steeneveld et al

18. © Crown copyright Met Office Large-scale advection problems (1) Lack of daytime warming on first afternoon Suggestive of cold advection not picked up by Steeneveld et al’s mesoscale analysis using MM5

19. © Crown copyright Met Office Large-scale advection problems (2) Wind stress and 10m wind drop simultaneously on evening of 25 th Oct, as does the “geostrophic” wind Something other than surface drag is slowing the wind For DICE I’ve taken it to be horizontal advection U at 800m (Steen et al) U average below 3km

20. © Crown copyright Met Office Large-scale forcing for DICE So what to do about it? Easy enough to run SCM with relaxation to sondes What timescale? Doesn’t really allow model to develop the PBL structure it wants – gets imposed by sonde

21. © Crown copyright Met Office Large-scale forcing for DICE Use a simple budget analysis to derive large- scale advection, eg for temperature: Sonde evolution From SCM radiative transfer scheme Estimated from sonde evolution From SCM with prescribed surface heat flux Use relaxation to sondes in SCM for this calculation

22. © Crown copyright Met Office Large-scale forcing for DICE To reduce the signature of the Met Office PBL scheme(!): Integrate PBL increment over PBL depth + 250m and redistribute uniformly over that depth Finally use a simple 1-2-1 filter in time and space on the resulting advection increment

23. © Crown copyright Met Office Large-scale forcing (5) Do the same for moisture:

24. © Crown copyright Met Office Large-scale forcing (6) …and momentum: Strong nocturnal jet

25. © Crown copyright Met Office Large-scale forcing summary Used in all atmosphere simulations (stages 1b, 2, 3b) Use the following large-scale forcing Time-varying geostrophic wind (uniform with height) Large-scale horizontal advective tendencies for T, q, u, v Subsidence for T, q No relaxation Note that radiation should be switched on in all simulations (as well as all other physics!)

26. © Crown copyright Met Office Surface forcing for stage 1b (atmosphere-only simulation) Surface flux observations from Oscar Hartogensis, Wageningen U (via Gert-Jan) Average observations to 30 minutes and smooth (1-2-1 filter) Observed surface sensible and latent heat fluxes Observed u * Ideally explicitly specify surface stresses too Alternatively set c D 1/2 = u * / V 1 for observed u * and model V 1 Observed surface T and albedo For upward LW, SW (or use observed – did anyone?)

27. © Crown copyright Met Office Stage 1b results atmosphere only simulations forced by observed surface fluxes

28. © Crown copyright Met Office Models ModelContact scientist InstituteStages submitted LevelsSensitivity tests AromeEric BazilleMeteo FranceAll60/70resolution ArpegeEric BazilleMeteo FranceAll60/70resolution ECEARTHReinder RondaWageningenSCM only91LAI GDPS3.0Ayrton ZadraCMCAll79 GFDLSergey MalyshevPrincetonAll24 GISS_E2Ann Fridlind, Andy Ackerman GISSAll40 IFS/HTESSELIrina Sandu, Gianpaolo Balsamo ECMWFAll137LAI MESO_NHMaria JimenezUIBAll85Bare soil UM/JULESAdrian Lock, Martin Best Met OfficeAll70Vegetation WRF-NOAHWeiguo WangNUISTAll60Lots! WRFWayne AngevineNOAA?119PBL scheme CAM5, CLM4David LawrenceNCAR1a, 1b? PBCMPierre GentineColumbiaNot yet

29. © Crown copyright Met Office NUIST sensitivity tests Stage 1a Stage 1b

30. © Crown copyright Met Office Model grids Solid lines = control model Dotted/dashed lines = experiment L70! L60 Control lowest level

31. © Crown copyright Met Office Stage 1b: basic check Is LS forcing working correctly? Time evolution at 1500m (above PBL) No radiation

32. © Crown copyright Met Office Stage1b surface fluxes LW up – emissivity differences? Surface SW down differ be nice to check TOA SW down

33. © Crown copyright Met Office Stage 1b: SCM forced by obs fluxes Driving with observed u * too successfully achieved Two models show small departures One model has apparent minimum u * of 0.1ms -1

34. © Crown copyright Met Office Stage 1b Clouds and rain! Cirrus (9 km) Fog ?

35. © Crown copyright Met Office Stage 1b: SCM forced by obs fluxes PBL evolution Some oddities Quite consistent features more gradual growth on 25 th than 26 th shallow SBL on 1 st and 3 rd nights big differences over SBL depth on middle night

36. © Crown copyright Met Office Profile comparison with sondes Sondes available at 2am, 6am, 2pm, 10pm = every 8 hours plus a 2am sonde

37. © Crown copyright Met Office Stage 1b: PBL evolution at 300m SW’ly wind implies from the pond? Warm bias develops in most models

38. © Crown copyright Met Office Stage 1b Near surface evolution 20m55m

39. © Crown copyright Met Office Stage 1b Profiles – first night, pot. temperature

40. © Crown copyright Met Office Stage 1b Profiles – first night, moisture

41. © Crown copyright Met Office Stage 1b Profiles – first night, winds Missing data!

42. © Crown copyright Met Office Stage 1b Profiles – first night, heat fluxes

43. © Crown copyright Met Office Stage 1b Profiles – second night, winds: LLJ?

44. © Crown copyright Met Office Stage 1b Profiles – second night, temperature

45. © Crown copyright Met Office Stage 1b Profiles – second night, heat fluxes and stress

46. © Crown copyright Met Office Stage 1b Profiles – third night, temperature

47. © Crown copyright Met Office Stage 1b Profiles – all 3 nights

48. © Crown copyright Met Office Stage 1b Profiles – final state

49. © Crown copyright Met Office Stage 1b summary Simulations successfully completed Overall doing a reasonable job Lots to look at SCM forcing issues: I‘m not happy with the wind forcing Can geostrophic wind be set better? simply use 1-3km average instead of below 3km? Remove fine-scale structure in wind ICs and forcing? Subsidence slightly too weak? Check cirrus on 26 th (is it spurious?) Check SW TOA between models Check temperature budgets for all models What should be happening with near surface night-time moisture (LHF~0)?

50. © Crown copyright Met Office Stage 2 results Coupled atmosphere-land models

51. © Crown copyright Met Office Stage 2: coupled land-atmosphere Cloud and rain! Still foggy Still get cirrus (9 km)

52. © Crown copyright Met Office Stage 2: coupled land-atmosphere Near surface wind and wind stress Stress overdone for a given windspeed? U * OK when winds are too weak U * too strong when winds are OK

53. © Crown copyright Met Office Stage 2: coupled land-atmosphere Surface fluxes SHF too large on stable nights LHF generally far too large by day (and night?)

54. © Crown copyright Met Office Stage 2: coupled land-atmosphere Net surface fluxes SW net +LW net -H-L = G

55. © Crown copyright Met Office Stage 1 vs 2 Bulk PBL sensitivity Coupling removes some oddities Some significant differences in PBL height…

56. © Crown copyright Met Office Stage 1 vs 2 Bulk PBL sensitivity Some significant differences in PBL height… Linked to surface Bowen ratio?

57. © Crown copyright Met Office Stage 1 vs 2 Bulk PBL sensitivity Extract mean PBL depth and EF for early afternoon…

58. © Crown copyright Met Office PBL depth sensitivity Stage 2 – stage 1b, control simulations More evaporation gives shallower PBL and vice versa ◊= 24 th x= 25 th

59. © Crown copyright Met Office Stage 2 sensitivity tests Some sensitivity of PBL depth to EF PBL depth depends on other things too! ◊= 24 th x= 25 th Small symbol=test

60. © Crown copyright Met Office Stage 1 vs 2 Bulk PBL sensitivity More spread between models in stage 2 Interesting lack of spread within (some) models

61. © Crown copyright Met Office Stage 2 Little LHF sensitivity to increasing PBL humidity

62. © Crown copyright Met Office Stage 2 summary Excessive SHF on stable nights Daytime LHF mostly greatly overdone “the grass is dead” (Larry Mahrt) coupled simulations don’t appear sensitive to atmospheric humidity Significantly more spread between models in coupled simulations (stage2 vs 1b) Night-time u * much better (weaker) in coupled that stand- alone LSM driven by 55m data (where MO not applicable) Need lowest model level << 55m

63. © Crown copyright Met Office Thank you