1. 1 Convection parameterization for Weather and Climate models Hua-Lu Pan and Jongil Han NCEP/EMC With help from EMC physics team Myong-In Lee and Sieg Schubert (NASA/GMAO) Soo-Hyun Yoo and Jae Schemm (NCEP/CPC)

2. 2 Climate modeler’s interest in convection parameterization Maintenance of the Hadley Cell, the Walker Cell, the prediction of ENSO, and climate response to increasing CO2 Removal of model biases Double ITCZ Weather in the climate models Mid-latitude disturbances are realistic Tropical variabilities are too weak for synoptic scales, MJO and ENSO Tropical cyclogenesis investigated through nested mesocale models Recent work using cloud resolving models to simulate the climate Indications are that the MJOs get stronger and the models can generate tropical storms

3. 3 Weather modeler’s interest in convection parameterization Tropical storm tracks Tropical storm genesis Summer precipitation over land Predictions of meso-scale convective systems In recent years, MJO and ENSO as well At NCEP, the global weather model is used for weather and climate Parameterized convection can do the job if we continue to work on the problem

4. 4 CFS (coupled) Simulations 64 Level (0.2 hPa) vs 28 Level (2.0 hPa) Atm. ENSO SST cycles Nino 3.4 SST Anomalies Observed Coupled Red: monthly bias 28 Level Atm 64 Level Atm

5. 5 CFS simulations using a T126 version of the model

6. 6 Testing with CMIP Runs (variable CO2) OBS is CPC Analysis (Fan and van den Dool, 2008) CTRL is CMIP run with 1988 CO2 settings (no variations in CO2, current operations) CO2 run is the ensemble mean of 3 NCEP CFS runs in CMIP mode –realistic CO2 and aerosols in both troposphere and stratosphere Processing: 25-month running mean applied to the time series of anomalies (deviations from their own climatologies)

7. 7 Examples of ENSO events T62 CMIP Simulated El Nino 2015-2016Simulated La Nina 2017-18 Real El Nino 1982-1983Real La Nina 1988-1989

8. 8 Observed “AMIP” (forced by monthly SST) Forced by Climatological SST Coupled 64 Level 19791980198119821983 ENSO Event 19791980198119821983 MJO Events 2025 2026202720282029 200 hPa Divergence

9. 9

10. 10 Tropical disturbances Since 1995, the GFS system started to produce easterly waves in the forecast and some of the easterly waves deepen into tropical storms Since 2001, the tropical cyclone genesis false alarm rate greatly diminished In longer forecasts, the GFS has a very active tropics

11. 11 GFS_2003082400f000

12. 12 GFS_2003082400f012

13. 13 GFS_2003082400f024

14. 14 GFS_2003082400f036

15. 15 GFS_2003082400f048

16. 16 GFS_2003082400f060

17. 17 GFS_2003082400f072

18. 18 GFS_2003082400f084

19. 19 GFS_2003082400f096

20. 20 GFS_2003082400f108

21. 21

22. 22

23. 23 What Needs to be Done? Parameterization of convection is still needed for the next 5-10 years Continue to develop and improve the physical basis for coded algorithms determining current performance –Improvements need to perform as well (or better) for both weather and climate models –Improvement areas Trigger Closure Cloud momentum mixing Cloud model

24. 24 Trigger function Most mass-flux schemes use closure as trigger … Whenever the column is unstable ‘enough’, convection starts –Modifications to delay onset mostly use environmental conditions such as RH Mesoscale modelers look at parcel buoyancy when lifting a parcel through inversion

25. 25 Trigger in the GFS GFS uses the parcel concept to check for level of free convection –Simplified trigger requires lifted parcel to have level of free convection within 150 hPa –Often delays the onset of convection

26. 26 Phase (local time) of Maximum Precipitation (24-hour cycle) Five-member ensembles driven by Climatological SST forcing (1983-2002 avg)

27. 27 SE GP Diurnal Cycle of Rainfall – Ensemble Mean and Spread Obs(HPD) GFDL NCEP NASA (spread) (ensemble mean)

28. 28 Closure Traditional closure for climate models –Rate of adjustment of the column CAPE (or cloud work function) to the final state For meso-scale models –Final state has convective instability eliminated (CAPE elimination) –Moist adiabat (after accounting for liquid and/or frozen water) For GFS closure –Approaches CAPE elimination when the atmospheric state is ‘disturbed’ –Modifies ‘climate CAPE’ with the ambient vertical motion

29. 29 Cumulus Momentum Mixing (CMM) Has a remarkable effect on tropical storm genesis –Without CMM Most of the tropical disturbances develop vorticity centers due mostly to grid-scale heating. Grid-scale “resolved” convection Too many disturbances –With CMM Parameterized convective heating is smaller Most important, vortex development is restricted only to the ‘real’ storms

30. 30 Further Remarks on CMM Leads to weaker storms –Vorticity –Sea-level pressure –Since the model tropics selects the real storms, hurricane track prediction in global model is improved. In-cloud parcel momentum is not conserved –Pressure gradient term must be parameterized –Further study with cloud-resolving models may help Impacts mid-latitude MCC evolution in coarse- grid models

31. 31 Simple cloud model in SAS The cloud model in the A-S scheme is a simple one. We should be able to add better physics in it. Currently, there is no cloud water generated other than at the detrainment level, so convective cloud needed in radiation is either made up or missing.

32. 32 Mesoscale modeling How does parameterized convection (and for that matter turbulence) work when the resolution goes from 40 km to 4 km? –The “convergence” problem for convective parameterization (Arakawa) –With the GFS trigger Air column in disturbed regions becomes very moist CAPE is reduced i.e. moist adiabat is approached Parameterized convection plays a diminishing role Grid-scale convection “takes over” Convective momentum mixing continues to exert influence on the intensity of the tropical storms

33. 33 Observed Precip RSM 50 km GFS 40 km 24 h Forecasts 12 UTC 31 May 2004 OPS (Eta)

34. 34 WRF (OPS) Orig. BMJ WRF-SAS Total Prec 8 km WRF-SAS Total Prec 4 km

35. 35 WRF-SAS Total Prec 16 km WRF-SAS Total Prec 4 km WRF-SAS Conv Prec 16 km WRF-SAS Conv Prec 4 km

36. 36 Impact Of Microphysics 5 km (RSM) Ferrier (GFS Next) GFS OPS

37. 37 New shallow convection Use a bulk mass-flux parameterization Based on the simplified Arakawa-Shubert (SAS) deep convection scheme, which is being operationally used in the NCEP GFS model Separation of deep and shallow convection is determined by cloud depth (currently 150 mb) Main difference between deep and shallow convection is specification of entrainment and detrainment rates Only precipitating updraft in shallow convection scheme is considered; downdraft is ignored

38. 38 Siebesma & Cuijpers (1995, JAS) Siebesma et al. (2003, JAS) LES studies

39. 39 CTL New shallow convection scheme

40. 40

41. 41 PBL & Low clouds combined (CFS run) ISCCP Control Revised PBL & new shallow convection

42. 42 Revised PBL + New shallow (Winter, 2007) NH(20N-80N)SH(20S-80S) 500 hPa Height Anomaly Correlation

43. 43 12-36 hrs 36-60 hrs 60-84 hrs Precipitation skill score over US continent

44. 44 Revised PBL + New shallow (Summer, 2005) NH(20N-80N)SH(20S-80S) 500 hPa Height Anomaly Correlation

45. 45 12-36 hrs 36-60 hrs 60-84 hrs Precipitation skill score over US continent

46. 46 NINO3.4 OIV2

47. 47 NINO3.4 set22

48. 48 NINO3.4 set28b

49. 49 Summary Contribution of parameterized convection important until sub-4 km resolution is reached To continue to improve NCEP’s real time applications, convective parameterization should continue to be developed Climate models can benefit from better parameterized convection in next 5-10 years Improvement areas –Convective trigger (+PBL) –Convective momentum transport –Refining physical basis for closure –Better cloud model within the convection scheme Approach must be physically-based –CRMs can be useful for specific problems (e.g. CMM) –To run at the many resolutions required, the scheme has to be physically based