NOAA ESRL GSD Earth Modeling Branch Boulder, CO USA 6 th Conf on Weather/Climate/New Energy Economy January 2015 HRRR/RAP assimilation/model improvements for wind/solar energy guidance Stan Benjamin Joseph Olson Eric James Jaymes Kenyon Curtis Alexander John Brown Tanya Smirnova Ming Hu New wind/solar observations enabling model improvements (tower, SURFRAD) Major improvements in wind/solar (cloud) guidance in 2015 from HRRRv2/RAPv3 Rapid Refresh version 3 Scheduled NCEP implementation Jun2015 ESRL RAPv3 update occurred 1 Jan 2015 High-Resolution Rapid Refresh V2 Scheduled NCEP implementation Jun2015 ESRL HRRRv2 update occurred 1 Jan

Rapid Refresh and HRRR NOAA hourly updated models (situational awareness for energy, aviation, severe weather, etc.) AMS-2015 – 6 th EnergyJanuary 2015 HRRR/RAP upgrades for energy 2 13km Rapid Refresh (RAP) (mesoscale) 3km High Resolution Rapid Refresh (HRRR) (storm-scale) RAP HRRR Version 2 -- NCEP implement 25 Feb 2014 Version 3 – GSD Planned NCEP –June 2015 Initial -- NCEP implement 30 Sept 2014 Version 2 – GSD Planned NCEP –June 2015

3 HRRR (and RAP) Future MilestonesHRRR Milestones Wind Forecast Errors in RAP and HRRR - A problem for wind energy / PBL/ surface forecasts WFIP – first extensive evaluation of HRRR/RAP forecasts w/ tower wind obs. Substantial improvement from RAP over RUC, but… Remaining wind errors evident WFIP Northern Study Area (NSA) Results from Wilczak et al, BAMS, 2015 Model forecast errors limit usefulness of renewable energy

4 HRRR (and RAP) Future MilestonesHRRR Milestones Eric James Energy Tues 1145am Cloud Deficiency in RAP and HRRR - A problem for solar energy / PBL/ convection forecasts

4-h forecast of composite reflectivity: valid 00Z 18 Jun 2014 HRRR (and RAP) Future MilestonesHRRR Milestones Convective Evolution 17 June UTC 18 June 2014 Real-Time Forecast Temperature Pseudo-Obs Enabled Observations Coleridge, NE Control run develops too much high-based convection that grows upscale Data assimilation change improves timing and evolution of convection 17 th ARAM5 January 2015HRRR Implementation 5 AMS-2015 – 6 th EnergyJanuary 2015 HRRR/RAP upgrades for energy 5

ModelData Assimilation RAP (13 km) WRF-ARWv3.6+ incl. physics changes Physics changes: Grell-Freitas convective scheme Thompson MP -- Aerosol-aware MYNN PBL -- cloud/non-loc mixing RUC LSM -- MODIS seasonal LAI -- Reduced wilting point Shallow cu parm w/rad feedback RRTMG radiation scheme Direct and diffuse GHI components Merge with GSI trunk Radial velocity assimilation Mesonet assimilation RARS data assimilation Radiance bias correction Pseudo-PBL obs for temperature Improved 2m Temp/Dewpoint Background Estimate Low-reflectivity precip building HRRR (3 km) WRF-ARWv3.6+ incl. physics changes Physics changes: Thompson MP - Aerosol-aware MYNN PBL -- cloud/non-loc mixing RUC LSM -- MODIS seasonal LAI -- Reduced wilting point Shallow cu parm w/rad feedback RRTMG radiation scheme Direct and diffuse GHI components Numerics changes: 6 th order diffusion in flat terrain Merge with GSI trunk 3-km hybrid assimilation Hydrostatic rebalance after analysis Radial velocity assimilation Mesonet assimilation Pseudo-PBL obs for temperature Improved 2m Temp/Dewpoint Background Estimate Low-reflectivity precip building Full cloud/precip hydrometeor assim RAPv3/HRRRv2 Changes ( ) Strong cloud /solar / near-sfc wind forecast impact

HRRR (and RAP) Future MilestonesHRRR Milestones  Grell−Freitas “scale-aware” shallow cumulus (subgrid-scale) Accounts for nonlocal mixing of heat & water vapor from condensed (non-precipitating) subgrid plumes “Scale-aware” mass flux scheme Produces subgrid cloud water & ice  coupled to radiation  MYNN boundary-layer cloud fraction (subgrid-scale) Function of local and PBL-mean RH, surface heat flux, resolution Activates in absence of resolved / parameterized cloud fraction in column Diagnoses subgrid cloud water & ice (specifically for radiation coupling)  RUC-LSM  Reduced wilting points  Maintain cropland at wilting point  Aerosol-aware Thompson cloud microphysics More smaller clouds over land Physics updates to reduce RAP/HRRR biases 7

First implemented with RAPv2 / HRRR 2013 – better convection, wind forecasts Improved mixing length formulations to flexibly change behavior across stability spectrum Improved surface layer scheme (customize to PBL ) Further enhancements for RAP v3 / HRRR v2 -- cloud mixing, coupling to different shallow cumulus schemes RAP / HRRR use of MYNN PBL scheme UTC Wind speed bias (m/s) RAPv1 RAPv2 RAPv3 NightAfternoon Morning Evening NightAfternoon Morning Evening RAPv1 RMS wind vector error (m/s) RAPv2 RAPv UTC MYNN-modifiedMYJ TKE (color bars) 8

Example: RAP cold-start tests without/with aerosol-aware cloud microphysics. More small-scale cloud with more CCN over land. NCEP RAPv3/HRRRv Changes Use of forecast aerosol fields to have prognostic cloud- condensation nuclei (CCN). 9

Thinner soil layer in energy / moisture budgets Potential for increased near-surface diurnal cycle Reduced warm bias at night, cold bias in day Updates to RUC Land Surface Model Increased roughness Z 0 for forests, cropland, urban New formulation to compute effective roughness length Z 0eff in the grid box (exponential) ======================= Version 3 (partial list) Seasonal variations of Z 0 for MODIS cropland category Seasonal variations of LAI based on the current vegetation fraction and variability of this parameter for different vegetation types OLD ~ 8 m 0 (cm) NEW ~ 8 m 0 (cm) RAP improvements Increased number of levels in soil domain – 9 levels Version th ARAM - AviationJanuary 2015Physics RAP / HRRR 10 AMS-2015 – 6 th EnergyJanuary 2015 HRRR/RAP upgrades for energy 10

ComponentItems GSI Data Assimilation Canopy water cycling Temp pseudo-innovations thru model boundary layer More consistent use of surface temp/dewpoint data GFO Convective Parameterization Shallow cumulus radiation attenuation Improved retention of stratification atop mixed layer Thompson Microphysics Aerosol awareness for resolved cloud production Attenuation of shortwave radiation MYNN Boundary Layer Mixing length parameter changed Thermal roughness in surface layer changed Coupling boundary layer clouds to radiation RUC Land Surface Model Reduced wilting point for more transpiration Keep soil moisture in croplands above wilting point HRRR (and RAP) Future MilestonesHRRR Milestones HRRR Model Warm/Dry Bias Mitigation Total Resource Change: Approximately ~6% increase in runtime for aerosols 17 th ARAM5 January 2015HRRR Implementation 11 AMS-2015 – 6 th EnergyJanuary 2015 HRRR/RAP upgrades for energy 11

8 Aug.2014 Reduced wilting point 13 Aug Change for cropland soil moisture 15 Aug Q2m diagnostic for data assim Testing RAPv3 LSM Changes NCEP Production Suite Review2-3 December 2014Rapid Refresh / HRRR 12 NCEP RAPv2 GSD test RAPv3 Temperature bias (K) Dewpoint bias (k) Warm Cold Moist Dry

NCEP OPER RAPv2 GSD RAPv3 pre-NCEP Warm Cold Temperature bias (K) Dewpoint bias (k) Temperature RMS (K) Dewpoint RMS (K) Moist Dry Overall: Rapid Refresh 12h Surface verification 13 East US July 2014

NCEP OPER RAPv2 GSD RAPv3 pre-NCEP Warm Cold Temperature bias (K) Dewpoint bias (k) Temperature RMS (K) Dewpoint RMS (K) Moist Dry Overall: Rapid Refresh 12h fcst -Surface diurnal verification 14 East US July 2014 RAPv3-RAPv2 Evening Night Morning Afternoon Evening Night Morning Afternoon

NCEP OPER RAPv2 GSD RAPv3 pre-NCEP too high too low 10m Wind bias (m/s) 10m wind RMS vector error (m/s) Overall: Rapid Refresh 12h Surface verification 15 East US July th ARAM - AviationJanuary 2015Physics RAP / HRRR 15 AMS-2015 – 6 th EnergyJanuary 2015 HRRR/RAP upgrades for energy 15

NCEP OPER RAPv2 GSD RAPv3 pre-NCEP too high too low DIURNAL Variation 10m Wind bias (m/s) DIURNAL Variation 10m wind RMS vector error (m/s) Overall: Rapid Refresh 12h fcst – 10m wind diurnal verification 16 East US July th EnergyJanuary 2015Physics RAP / HRRR 16 AMS-2015 – 6 th EnergyJanuary 2015 HRRR/RAP upgrades for energy 16 NightAfternoon Morning Evening NightAfternoon Morning Evening

17 Rapid Refresh 12h Upper-air verification NCEP OPER RAPv2 GSD RAPv3 pre-NCEP Temperature RMS (K) Relative Humidity RMS (%) Wind RMS vector (m/s) CONUS Vs. raobs July 2014 RAPv3 better RAPv3 better RAPv3 better

Results: SW-down Verification Observations (1 hr averages, all 14 SURFRAD stations) RAPv3-preliminary - RAP-Dev2 (13 km) 12 hr fcst RAPv3 - RAP-Dev3 (13 km) 12 hr fcst RAPv2 - RAP-Oper (13 km) 12 hr fcst Oper: does not include subgrid clouds or LSM mods (WRFv3.5.1) Dev2: has improved subgrid clouds and sh. cu scheme (WRFv3.5.1) Dev3: Dev2 enhancements + LSM wilting point mod (WRFv3.6) 18 Much better SW radiation agreement with SURFRAD obs with RAPv3-final than RAPv2

HRRR (and RAP) Future MilestonesHRRR Milestones 3000 ft 1000 ft 500 ft 29 Jan 30 Jan31 Jan Control Prototype Diagnostic (w/ subgrid stratus) Ceiling – new cloud-fraction diagnostic for RAP/HRRR Results: Probability of Detection (12-h forecasts, CONUS)

HRRR (and RAP) Future MilestonesHRRR Milestones Example: 30 January 2013 Control (12-h forecast valid 2000 UTC) Reflectivity Mosaic (2000 UTC) GOES Visible (3-h loop ending 2000 UTC) Source: UCAR kft (AGL)

HRRR (and RAP) Future MilestonesHRRR Milestones Example: 30 January 2013 selected ceiling reports versus 12-h ceiling forecasts (valid 2000 UTC) Control Prototype Cloud Fraction-Based Diagnostic kft (AGL) KDIK: OVC090 KRAP: BKN028 KUNU: OVC007 CYWG: OVC th ARAM - AviationJanuary 2015Physics RAP / HRRR 21

Winds – Substantial RAPv3 improvement for both upper-air and near-surface, for all seasons. Cloud – Improved SW-down radiation from physics changes and upcoming cloud-fraction diagnostic. Summary: RAPv3/HRRRv2 vs. RAPv2/HRRRv1 for wind / solar energy guidance RAPv3/HRRRv2 changes Substantial for some parameterizations (MYNN PBL, RUC LSM, Thompson cloud) Data assimilation improvements also drive RAPv3/HRRRv2 (ens, radar, sfc) Major impact on wind and solar forecast accuracy PBL, cloud, LSM, radiation work in concert New observations (tower, SURFRAD/solar, including proprietary data) critical for evaluation and assimilation. NCEP OPER RAPv2 GSD RAPv3 pre-NCEP

ESRL – experimental version RAPv3 – GSD testing in 2014 Final RAP-primary impl - 1/1/2015 – Will initialize 2014 ESRL-HRRR(v2) – Improved PBL, LSM, cu-parm, DA – WRFv3.6.1 w/ Thompson/NCAR aerosol-aware microphysics HRRRv2 – GSD testing in 2014 – Final HRRR-primary impl - 1/1/2015 – Initialized by 2014 RAP (v3) – Improved radar assimilation, hybrid assimilation, PBL/cloud physics RAPv4 – GSD testing in 2015 – Hourly RAP ensemble data assimilation HRRRv3 – GSD testing in 2015 – Target: Improved 3km physics + improved data assimilation. NWS-NCEP - operational – Planned dates Implement June 2015 Implement 2016 HRRR (and RAP) Future MilestonesHRRR Milestones RAP/HRRR Implementation Map 23

Evolution of hourly updated NOAA modeling Feb 2014 –Rapid Refresh v2 – oper at NCEP Sep 2014 – HRRR v1 – oper at NCEP 2015 – RAPv3 and HRRRv2 – est. June impl. Model – improved physics – MYNN PBL / LSM/snow, aerosol-aware cloud μphysics numerics improvements, updated WRF Assimilation – radar assim for light precip, lightning, 2m T/Td forward model, rad wind  Improved surface/lower trop forecasts, convective environment, clouds North American Rapid Refresh Ensemble NMMB, ARW cores Hourly updating with GSI-hybrid EnKF (4DEnVar?) Initially 6 members, 3 each core, physics diversity (stochastic only or with RAP/NAM/NCAR physics suites) Hourly forecasts to 24-h NMMB +ARW members to 84-h 4x/day Rapid Refresh  NARRE 2016 – RAPv4/HRRRv – Ensemble Rapid Refresh/NAM – NARRE (w/ hybrid 4d-ens/var DA) 2019? – Ensemble HRRR – HRRRE – ( ultimately with hourly ~3km ensemble DA) HRRR Future NCEP 24