Evaluation of coupled versus uncoupled cloud physics and radiation in WRF Gregory Thompson Mukul Tewari, Kyoko Ikeda, Sarah Tessendorf, Courtney Weeks Research Applications Laboratory National Center for Atmospheric Research Jason Otkin Cooperative Institute for Meteorological Satellite Studies University of Wisconsin-Madison Fanyou Kong Center for Analysis and Prediction of Storms University of Oklahoma 1 st World Weather Open Science Conference, Montreal 17-21 Aug 2014
Motivation: OU-CAPS ARW ensembles SPC/NSSL Hazardous Weather Testbed old MCS decaying cloud mostly composed of snow was essentially transparent next day s convection triggered much too early since minimal cloud cover seen by radiation caused by cloud ice versus snow categorization Hurricane WRF (HWRF) tests supported by Developmental Testbed Center (DTC) Thompson et al (2008) microphysics considered for HWRF HWRF operational version used GFDL radiation scheme Hurricane Earl (2010) examples
HWRF 4 Earl test simulations Ferrier microphysics GFDL radiation Thompson microphysics GFDL radiation Longwave, outgoing top of atmos Thompson microphysics uncoupled RRTMG radiation Thompson microphysics coupled RRTMG radiation
Goal: Improve cloud-radiation treatment Convective initiation by improved T-sfc forecasts Longevity/duration of convection due to cloud longwave radiative forcing Improve tropical cyclone track/intensity prediction? AND Demonstrate cloud indirect effects
Methodology code issues: Previous assumed cloud water size(s)
Methodology code issues: Previous assumed ice size(s) ice and snow added
Solution: Compute cloud water droplet, ice, and snow effective radii 3 rd moment divided by 2 nd moment (Volume/Area) water droplets are simple ice and snow are more complex due to crystal geometry mass-dimensional and number density assumptions are already known in microphysics scheme pass explicitly calculated radii from microphysics to radiation radiation interface code calculates optical depth, single-scattering albedo, asymmetry parameter in clouds using explicit radii variables
Example radiative effective size Radiative effective radius: cloud droplets (k=16 from bottom) Radiative effective radius: cloud ice (k=44 from bottom) Radiative effective radius: snow (k=37 from bottom) 5 to 15 microns 20 to 70 microns 50 to 250 microns
Methodology: Sensitivity Experiments OU-CAPS WRF ensemble members pre-2013 used RRTM or Goddard radiation schemes Cintineo et al (2014) noted specific cloud forecast biases 2013 & 2014 used RRTMG scheme nomenclature: control used standard/uncoupled code m25 used prototype coupled code; inadvertent snow too small m30 used corrected snow radius (in WRFv3.5.1) Verification GOES vs. Synthetic Satellite longwave infrared U.S. Climate Reference Network downward shortwave radiation Near surface temperature (METARs and USCRN) comparison m25 m30
Results: GOES vs. Synthetic-Sat GOES (ch4) longwave IR SynthSat Control SynthSat m30 SynthSat m25
Results: cloud-top temp 20 bins
Results: shortwave radiation
Results: near surface temperatures Model and Obs Temp Bias Ctrl m25 m30 Mean Error (standard error) 1.01 (0.01) 1.45 (0.01) 1.01 (0.01) RMSE 2.57 2.74 2.59
Results: m25 daytime temperature drift
Aerosol 1 st indirect effect 1 st aerosol indirect effect More aerosols result in more droplets with overall smaller size. Smaller droplet effective radius increases cloud albedo (+5.4% outgoing shortwave, TOA). Increased cloud thickness in higher aerosols resulting in higher (0.47%) downward longwave IR and lower (-0.11%) outgoing LW-IR, TOA. See Poster#1107 Thompson, G. and T. Eidhammer, 2014: A Study of Aerosol Impacts on Clouds and Precipitation Development in a Large Winter Cyclone, J. Atmos. Sci., early online release.
Conclusions The inadvertent small snow effective radii experiment caused clouds to become too opaque Surface temperature cool bias worsened in m25 compared to control or m30 Shortwave radiation reaching the ground decreased too much The coupled microphysics & radiation experiment exhibited expected behavior Smaller ice crystals made clouds more opaque to shortwave Smaller water droplets (due to higher CCN) increased cloud albedo (1 st indirect effect) Typical model errors are far larger than changes in uncoupled vs. coupled experiments Control and m30 were very similar to each other Finding signal/meaning in small differences was challenging when overall errors were large Experiment m30 did no harm and produced proper physical behavior
Acknowledgements We gratefully acknowledge the support/contributions of: NSF & Short-Term Explicit Prediction (STEP) program FAA Aviation Weather Research Program office Roy Rasmussen (NCAR-RAL) Stan Benjamin et al (NOAA/ESRL/GSD) Ligia Bernadet, Shaowu Bao (DTC) Dave Gill, Jimy Dudhia, Wei Wang, Michael Duda and the entire WRF development team Thank you!!