Use of radiation / cloud observations to reduce cloud-radiation model errors from 4-h to 4-week forecasts Stan Benjamin 2018 GMAC Joseph Olson, Tanya Smirnova, Shan Sun, Allison McComiskey, Boulder, CO GMD, GSD Kathy Lantz, Chuck Long, Curtis Alexander, Georg Grell NOAA Earth System Research Laboratory, Boulder, CO USA
RAP/HRRR: Hourly-Updating Weather Forecast Suite - June 2018 NOAA/NCEP upgrade 13-km Rapid Refresh (RAPv4) – to 39h (Jun 2018 ) Initial & Lateral Initial & Lateral Boundary Conditions Boundary Conditions 3-km 3-km High-Resolution High-Resolution Rapid Refresh Rapid Refresh Alaska (HRRR-AK) (HRRRv3) – to 36h 36 hr (Jun 2018) (Jun 2018) • More accurate • Runs longer (out to 36h) • Alaska domain added 23 May 2018 - GMAC
Unified ed m mod odel el d developmen ent in NOAA/ESRL L (ESRL divisions: GSD with PSD/GMD/CSD), NCEP, NCAR, etc.) Spatial resolution 3km 13km 10-15km 15-30km Forecast range 1-36h 1-39h 1-10 day Week 2 – 9 month Domain Regional Regional Global Global Model development area HRRR (High-Resolution Rapid RAP (Rapid Refresh) NGGPS – FV3 (current FIM-HYCOM coupled Refresh model) – physics testing with seasonal (for NOAA SubX FIM and FV3) experiment, switch to FV3) Radar/cloud/surface/land, 4D Ensemble DA Data assimilation 3km ensemble DA (Whitaker-PSD, EMC) Dynamic core numerics Use/refinement of WRF-ARW, FV3 – cubed sphere, hybrid vert coordinate FIM - icosahedral PBL/MYNN, cloud microphysics Same as HRRR but GFS physics + Grell- GFS physics + Grell-Freitas Physical (Thompson), RUC land-surface, with Grell-Freitas Freitas cu. Testing of cumulus. Test HRRR suite parameterization scale-aware cumulus HRRR/RAP suite-2017 starting late 2017. Application of inline 2-aerosol – NCEP, testing of 18- “ + gas-phase 18-aerosol and gas- 18-aerosol only so far chemistry aerosol chemistry phase chem Systematic error Clouds, precipitation Clouds Clouds, blocking, Clouds, blocking, precipitation precipitation investigations 23 May 2018 - GMAC
HRRR Users and Applications Example: National Weather Service including Storm and Weather Prediction Centers (SPC and WPC) Aviation Weather Center (AWC) and FAA Command Center National Severe Storms Laboratory (NSSL) and Air Resources Laboratory (ARL) National Centers for Atmospheric Research (NCAR) and Lincoln Laboratory (LL) General Forecasting Renewable Energy (wind/solar) Renewable Energy (wind/solar) Same-Day Decision Support Day-Ahead Decision Support Severe Severe Weather -Watches, Severe Weather Weather Convective Outlooks Day 2 Outlooks QPF / Heavy rainfall/snowfall Heavy rainfall/snowfall hydro watches, National Water Model Day 2 Outlooks Aviation Aviation Tactical Strategic Planning Planning 0–2 hr 2–8 hrs 8–15 hrs 15–24 hrs 24–48 hrs Forecast Length 23 May 2018 - GMAC
Courtesy: Atmospheric process representation Kathy Lantz - GMD necessary for weather prediction (incl. HRRR/RAP) There is an fundamental connection between Atmospheric solar radiation Processes forecasts (both diffuse and direct) and low- level wind forecasts . Surface Net Radiation Aerosol-Cloud Interactions Energy available for SH, LH, and ground heat flux Drives turbulent mixing, PBL formation, low- level winds, clouds 23 May 2018 - GMAC
Grid Grid Vertical Vertical Pressure Boundary Model Run at: Domain Initialized Points Spacing Levels Coordinate Top Conditions GSD, North 953 x Sigma-Isob Hourly RAP 13 km 50 10 mb GFS NCO America 834 Hybrid (cycled) Hourly (pre- GSD, 1799 x Sigma-Isob HRRR CONUS 3 km 50 20 mb RAP forecast hour NCO 1059 Hybrid cycle) Radiation Cloud Cumulus Turb Land-sfc Model Version Assimilation Radar DA LW/SW Microphysics Param /PBL scheme RUC MYNN WRF-ARW GSI Hybrid 13-km DFI, RRTMG/ Thompson GF + RAPv4/ v3.8.1, RAP v3.8.1, v3.8.1+ Ensemble to 0.85 20-min LH RRTMG Aerosol v3.8.1 Shallow 2mT/snow, EDMF/cl mosaic HRRRv3 RUC MYNN WRF-ARW GSI Hybrid 3-km RRTMG/ Thompson v3.8.1, HRRR None v3.8.1, v3.8.1+ Ensemble to 0.85 15-min LH RRTMG Aerosol v3.8.1 2mT/snow, Summary of EDMF/cl mosaic 6 th Order Model Horiz/Vert Scalar Upper-Level Diffusion SW Radiation Land Use MP Tend Time- Changes Advection Advection Damping Option Diffusion Update Limit Step MODIS Positive- w-Rayleigh Yes RAP 5 th /5 th Full (2) 20 min Seasonal, 0.01 K/s 60 s Definite 0.2 0.12 VIIRS GVF MODIS Positive- w-Rayleigh Yes, 0.25 15 min with HRRR 5 th /5 th Full (2) Seasonal, 0.07 K/s 20 s Definite 0.2 no slope SW-dt VIIRS GVF 23 May 2018 - GMAC
GSD model verification vs. observations SURFRAD Aircraft RAOBs SOLRAD CERES cloud radiation observations Surface METARs Precipitation Radar reflectivity Including StageIV NSSL MRMS Ceiling/cloud obs
Summer July 2016 / Jan 2017 – Downward SW error – models minus from CERES (W/m 2 , day 1) Winter General problem – excessive downward short-wave radiation, too little resolved and subgrid clouds • RAP – 13km • HRRR – 3km • Global FIM/HYCOM – borrows from cumulus physics for RAP (Grell-Freitas conv) • Part of NOAA SubX subseasonal experiment 23 May 2018 - GMAC
Goal: More accurate weather guidance (via improved process understanding, modeling, assimilation) Bias – HRRR 0.5” precip / 6h Bias – 12h HRRR 2m temp forecast valid 00z Progress toward lower biases 23 May 2018 - GMAC
Candidates for w arm bias / too-little cloud Vertical Vertical Pressure Boundary Model Run at: Domain Grid Points Grid Spacing Initialized Levels Coordinate Top Conditions GSD, North Sigma-Isob RAP 953 x 834 13 km 50 10 mb GFS Hourly (cycled) NCO America Hybrid Hourly (pre- GSD, 1799 x Sigma-Isob HRRR CONUS 3 km 50 20 mb RAP forecast hour NCO 1059 Hybrid cycle) Radiation Cumulus Turbulence Land-sfc Model Version Assimilation Radar DA Cloud Microphysics LW/SW Param PBL scheme MYNN RUC v3.8.1, WRF-ARW GSI Hybrid Ensemble to 13-km DFI, RRTMG/ Thompson Aerosol RAP GF + Shallow v3.8.1, 2mT/snow, v3.8.1+ 0.85 20-min LH RRTMG v3.8.1 EDMF/cl mosaic MYNN RUC v3.8.1, WRF-ARW GSI Hybrid 3-km RRTMG/ Thompson Aerosol HRRR None v3.8.1, 2mT/snow, v3.8.1+ Ensemble to 0.85 15-min LH RRTMG v3.8.1 EDMF/cl mosaic 6 th Order Horiz/Vert Scalar Upper-Level Diffusion SW Radiation MP Tend Model Land Use Time-Step Advection Advection Damping Option Diffusion Update Limit MODIS Positive- w-Rayleigh Yes RAP 5 th /5 th Full (2) 20 min Seasonal, 0.01 K/s 60 s Definite 0.2 0.12 VIIRS GVF MODIS Positive- w-Rayleigh Yes, 0.25 HRRR 5 th /5 th Full (2) 15 min with SW-dt Seasonal, 0.07 K/s 20 s Definite 0.2 no slope VIIRS GVF 23 May 2018 - GMAC
Physics Development Emphasis: Sub-Grid Clouds Explicit (Resolved) Clouds/Precipitation Sub-Grid (Unresolved) Clouds RAP and HRRR use the Thompson microphysics scheme with 5 hydrometeor types Adapted from Fig. 2 of Tompkins (2009) Assume a PDF of s Parameterize: (saturation ratio) assume subgrid PDFs for thermodynamic variables Chaboureau and Bechtold (2002) Retrieve Cloud Fraction, Cloud Condensate 23 May 2018 - GMAC RAPv3/HRRRv2
HRRR/RAP: Physical Processes and Representations Model Process Change/Addition Component • Mixing length MYNN PBL/ Turbulent o Scale-aware 3d-Blended TKE Diffusion o Z-less • 1D → 3D as f ( Δx ) → 0 • Multi-plume Non-local MYNN Mass-flux • TKE transport Turbulent • Momentum transport Transport • Scale-aware • Scalar roughness Surface RUC LSM/ • M-O alternatives Fluxes MYNN Sfc Layer • 3D surface stress Thompson Aerosol • Subgrid scale clouds Clouds / Chaboureau- • Coupled to radiation Bechtold • prognostic Vertical Numerics/ • Hybrid WRF-ARW Coordinate, Dynamics w’q’ w’u’ w’T’ Vertical Coordinate Advection Wind Farm • Wind direction effects Turbine Drag Parameterization • Power calculation. 23 May 2018 - GMAC
Improvements to mass-flux scheme: not enough coverage 5-h GHI forecasts valid 1800 UTC 19 August 2015 Visible Imagery Previous version (v3.9) 23 May 2018 - GMAC W m -2
Improvements to mass-flux scheme: better 5-h GHI forecasts valid 1800 UTC 19 August 2015 Visible Imagery Updated Version 23 May 2018 - GMAC W m -2
12h HRRR v2 /v3 downward SW vs. SURFRAD – Mean 15-21z HRRR-oper HRRR-exp Bias Mean abs err New HRRR Better / worse SURFRAD geographical diversity 23 May 2018 - GMAC critical for model evaluation
12h HRRR Downward SW bias vs. SURF/SolRAD – Mean 15-21z SURFRAD geographical diversity critical for model evaluation Similar excessive downward SW (HRRR) Similar excessive downward SW (HRRR) for different SURFRAD stations for SURFRAD vs. SOLRAD obs. 23 May 2018 - GMAC
Downward SW radiation - FIM-HYCOM vs. CERES FIM-HYCOM uses Grell-Freitas deep/shallow 23 May 2018 - GMAC convection (same as RAP). • Similar downward SW bias for all 4 weeks. • Warm season: model too high SW over land, cold ocean stratocu zones, high-latitude ocean.
HRRR Improved Convective Forecasts Experimental HRRRv3 13 hr fcst Composite Reflectivity Observations Operational HRRRv2 13 hr fcst Valid 00 UTC 17 May 2017 00 UTC 17 May 2017 Valid 00 UTC 17 May 2017 More Accurate Convection Along Weakly Forced Dryline 23 May 2018 - GMAC Reflectivity Verification
Recommend
More recommend