Toward Improving the Representation of Diurnal Cycle of Precipitation in Climate Models - PowerPoint Presentation

1. Toward Improving the Representation of Diurnal Cycle of Precipitation in Climate ModelsShuaiqi TangPacific Northwest National Laboratory, Richland, WA, USA6/18/2022Collaborators: Shaocheng Xie, Cheng Tao, Hsi-Yen Ma, and many others

2. OutlineCommon issues in simulating diurnal cycle of precipitation (DCP) Representation of DCP in CMIP6 modelsSingle-column model intercomparison to understand processes related to DCP biasLate afternoon precipitation over landNocturnal precipitationA new convective trigger to improve DCPSummary

3. Climate models struggle to simulate DCPRain too early over ocean and land (Betts et al., 2002; Guichard et al., 2004; Yang and Slingo, 2001 ; Dai et al., 2006) Fail to capture the nocturnal peak observed in downstream of major mountains, like the central U.S. (Liang et al., 2004; Wang et al., 2005; Lee et al., 2007)ARMCMIP5JJA Diurnal Cycle of Precipitation in CMIP5 at ARM SGP SiteBlack: ARM observations Grey lines: CMIP5 model results Zhang, Tang et al. (2020) BAMSAmplitudePhaseCovey et al., (2016) J. Cli.LandOceanTRMMCMIP5TRMM

4. Representation of Diurnal Cycle of Precipitation in CMIP6

5. CMIP6 ModelsAMIP Simulations: 15 CMIP6 models, 6 of them also participated in CMIP5 CMIP5 models generally use a resolution lower than their CMIP6 counterpartsCMIP6 Model for AMIP RunsModelsinstituteHori. Res (lon × lat)# of vert. levels*ACCESS-CM2Commonwealth Scientific and Industrial Research Organization (CSIRO) and Bureau of Meteorology (BOM), Australia192 x 14485 (38)*BCC-CSM2-MRBeijing Climate Center, China320 × 16046 (26)CNRM-CM6-1Centre National de Recherches Météorologiques, France256 × 12891E3SM-1-0Department of Energy, USA360 × 18072*EC-Earth3The European Centre of Medium Range Weather Forecast (ECMWF)512 × 256(320x160)91 (62)*FGOALS-g3Institute of Atmospheric Physics, ChineseAcademy of Sciences, China180 x 80(128x60)30 (26)GFDL-CM4NOAA Geophysical Fluid Dynamics Laboratory, USA144 × 9033*IPSL-CM6A-LRInstitute Pierre-Simon Laplace, France144 × 143(96x96)79 (39)KACE-1-0-GNational Institute of Meteorological Sciences, Korea Meteorological Administration, Korea192 × 14485*MIROC6University of Tokyo, Japan256 x 12881 (40)MPI-ESM1-2-HRMax Planck Institute, Germany384 x 19247MRI-ESM2-0Meteorological Research Institute, JMA, Japan320 × 16080NESM3Nanjing University of Information Science and Technology, China192 × 9647SAM0-UNICONSeoul National University, Korea288 × 19230TaiESM1Research Center for Environmental Changes, Taiwan, China288 × 19230*Indicates that the model also participated in CMIP5. Numbers in parenthesis are the resolutions used in CMIP5

6. ObservationsDataTime periodSpatial coverageHorizontal resolutionTime resolutionReferenceTRMM 3B42 V71998-201350°S – 50°N0.25° x 0.25°3hr(Huffman et al. 2007)CMORPH_V1.01998-201860°S – 60°N0.25° x 0.25°1hr(Joyce et al. 2004)IMERG2001-201890°S – 90°N**0.1° x 0.1°30min(Huffman et al. 2019)ARMBESGP: 1993-2018ENA: 2014-2018TWPC2: 1998-2010MAO: 2014-2015Single pointSingle point1hr(Xie et al. 2010)VARANALSGP: 2004-2016MAO: 2014-2015SGP: 30x30 grid boxMAO: 20x20 grid boxSGP: ~3°x3°MAO: ~2°x2°SGP: 1hrMAO: 3hrSGP: (Tang et al. 2019; Xie et al. 2004)MAO: (Tang et al. 2016)Satellite DataGround-base Data

7. Diurnal Harmonic Phase & Amplitude CMIP6 modelsLandOceanOBS: a late afternoon peak after 18 LSTModel: peaks between 10 and 16 LSTOBS: early morning peak 06-07 LSTModel: peaks between 03-05 LSTOBSOBSTang et al. (2021) J. ClimateSummer season (1996 – 2005) for both Hemispheres (Jan for SH and July for NH) from AMIPAmplitude (mm/hr)Phase (local hour)

8. Diurnal Harmonic Phase & Amplitude CMIP6 vs. CMIP5LandOceanCMIP6 shows improved phase over CMIP5 in multi-model mean. The improvement even more clear for those models where their performance can be tracked from CMIP5 to CMIP6 Tang et al. (2021) J. ClimateFor those models that participated in both CMIP5 and CMIP6 Summer season (1996 – 2005) for both Hemispheres (Jan for SH and July for NH) from AMIP

9. 1260/2418E3SM-1-0EC-Earth3FGOALS-g3GFDL-CM4IPSL-CM6A-LRKACE-1-0-GNESM3SAM0-UNICONTaiESM1MIROC6MPI-ESM1-2-HRMRI-ESM2-0CMORPHIMERGTRMMACCESS-CM2BCC-CSM2-MRCNRM-CM6-1Diurnal Harmonic PhaseSummer season (1996 – 2005) for both Hemispheres (Jan for SH and July for NH) from AMIPMost models fail to simulate nocturnal precipitation at certain regions: central U.S., Northern Argentina, Western Africa, ...Tang et al. (2021) J. Climate

10. Long-term single-column model (SCM) intercomparison

11. ExperimentsMulti-year SCM simulationsSGP (2004-2015 summer months)MAO (2014-2015) (GoAmazon 2014/5)10 participating SCMsEAMv1 (E3SM SCM, Bogenschutz, Tang et al., 2020), EAMv1-SILHS, EAMv1-trigger, SCAM5, SCAM6, SAM0-UNICON, SKIM, CMC, SMCPCP, TaiESM1Initialized by the ARM Variational Analysis (VARANAL) products with prescribed large-scale forcing (Zhang and Lin 1997; Xie et al., 2004; Tang et al., 2016; Tang et al., 2019) and surface turbulent fluxesHindcast simulations without relaxation of temperature and moistureSGPMAO

12. SGP (2004-2015 summertime)MAO (2014-2015)Overall performancemost models miss the nocturnal peakTang et al. (2022) QJRMS

13. Different types of convective systems occur in different time of the dayLocal occurring system (LOS)Costal occurring system (COS)Basin occurring system (BOS)Afternoonwet and dry seasonsLate night/morningwet seasonmostly nighttimemostly wet seasonTang et al. (2016) ACP

14. Afternoon PrecipitationPeak Pr > 1mm/dayPeak hours*: 1pm (11am) – 8pmPeak Pr > 1.5 Pr outside Peak HrsAll the models trigger convection too earlySimulated precipitation peak time is more spread at MAO than at SGPDiurnal precipitation magnitude varies with modelsTang et al. (2022) QJRMS

15. Afternoon Precipitation – Onset TimePrecip onset time: the first hour (t) starting from 6am with pr(t)>1 and pr(t-1)<1 mm/day. meanmedian75%25%Better performance at SGP than at MAOGCMs usually release CAPE too soon, and/or miss the transition of shallow-to-deep convection and the gradual moistening of the free troposphere.Parameterizations that unify shallow and deep convection help delay convection onset.Tang et al. (2022) QJRMS

16. Nocturnal PrecipitationPeak Pr > 1mm/dayPeak hours: 00Z – 07ZThese three models are able to capture convective instability above PBLThe nocturnal precipitation from other models are primarily from stratiform precipitationSKIMCMCTaiESM1Tang et al. (2022) QJRMS

17. A new convective trigger to improve diurnal cycle of precipitation

18. A New Convective TriggerXie and Zhang (2000)Unrealistically strong coupling of convection to surface heatingA new convective triggering mechanism (dCAPE&ULL, Xie, Tang et al., 2019, JAMES):Use a dynamic constraint, the dynamic Convective Available Potential Energy (dCAPE), for preconditioning of the convection-favoring environment and prevent CAPE from being released spontaneously (Xie and Zhang 2000)New trigger: CAPE>0 & dCAPE>0Implement the Unrestricted air parcel Launch Level (ULL) scheme to capture elevated convection (Wang et al. 2015)Implemented the new trigger in E3SM which uses the deep convection scheme developed by Zhang and McFarlane (1995), i.e., the ZM scheme.Lee et al. 2008Convection elevatedNocturnal PrecipitationSurface decoupled Elevated convection is not consideredXie et al. 2014Surface decoupled

19. Sensitivity to different model parameterizations/setupsTang et al. (2022) QJRMSThe revised convective trigger (Xie et al., 2019) improves both afternoon and nocturnal precipitation in the diurnal cycle in E3SMUnified convection scheme may help delay afternoon peak over land, but not necessary for capturing the nocturnal peakChange of other schemes (SCAM5 to SCAM6) have minor change on DCPUNICONNew TriggerSILHSSILHSNew TriggerUNICONSILHS: A high-order turbulence closure scheme (CLUBB) for turbulence, shallow and deep convection. All precipitation is from cloud microphysics (Griffin and Larson, 2016; Thayer-Calder et al, 2015)UNICON: A sub-grid vertical transport scheme for a unified treatment of turbulence, shallow and deep convection (Park 2014a,b)E3SM defaultE3SM default

20. More on E3SMv1.trigger (dCAPE&ULL)CTLDCAPE&ULLTRMMXie, Tang et al. (2019) JAMESdCAPE reduces the “too frequent, too weak” problemULL is the key to capture nocturnal elevated convectionA substantial improvement in the phase of the diurnal cycle although its amplitude is still weaker than the observations The improvement is seen globally

21. SummaryClimate models are still suffering the long-standing errors on DCPToo early precipitation peak over ocean with the observed peak in the early morning Too early precipitation peak over land with the observed peak in the late afternoon Missing nocturnal precipitation peak over some regionsClear improvements are seen in CMIP6 models compared to CMIP5 versionsMainly due to improvements in cloud/convection parameterizationsRecent developments in convection parameterizations are shining lights on reducing the long-standing model errorRelax unrealistically strong coupling of convection to surface fluxes (the dCAPE&ULL trigger) or unified treatment of turbulence, shallow and deep convection help delay the afternoon peak (UNICON, SILHS)Model capability to capture mid-level convection is the key for nocturnal precipitation

22. precipitation - CRH (column-integrated relative humidity) relation SCMs simulate more (less) precipitation than OBS when CRH is low (high).SCMs are more likely to produce high-CRH but less likely to produce moderate-CRH.

23. Total heating/dryingRegardless of convective fraction, models have similar heating/drying levels as in observations.In SCMs: Q1: T tendency from all physical schemes Q2: (neg) Q tendency from all physical schemesMany models produce weaker heating and drying. The remaining instability is released later in the day, causing a morning-time precipitation peak.Tang et al. (2022) QJRMS