AND HIGH IMPACT WEATHER WORKSHOP 09 February 2012 Use of Rapid Updating Meso‐ and Storm‐ scale Data Assimilation to Improve Forecasts of Thunderstorms and Other High Impact Weather: The Rapid Refresh and HRRR Forecast Systems NOAA/ESRL/GSD Curtis Alexander, David Dowell, Steve Weygandt, Stan Benjamin, Ming Hu, Tanya Smirnova, Patrick Hofmann, Haidao Lin, Eric James and John Brown
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
WARN ON FORECASTAND HIGH IMPACT WEATHER WORKSHOP
09 February 2012
Use of Rapid Updating Meso and Storm scale Data ‐ ‐Assimilation to Improve Forecasts of
Thunderstorms and Other High Impact Weather: The Rapid Refresh and HRRR Forecast Systems
NOAA/ESRL/GSDCurtis Alexander, David Dowell, Steve Weygandt, Stan Benjamin, Ming Hu, Tanya Smirnova, Patrick Hofmann, Haidao Lin, Eric James and John Brown
ESRL/GSD/AMBModeling System Overview
3
Hourly Updated NOAA NWP Models
13km Rapid Refresh (RAP) (mesoscale)
13km RUC (mesoscale)
3km HRRR (storm-scale)
RUC – current oper Model, new 18h fcst every hour
High-Resolution Rapid Refresh Experimental 3km nest inside RAP, new 15-h fcst every hour
Rapid Refresh (RAP) replaces RUC at NCEP in 2012 WRF, GSI with RUC features
NOAA/ESRL/GSD/AMB Models
Model Version Assimilation Radar DFI Radiation Microphysics Cum
Param PBL LSM
RUC N/A RUC-3DVAR Yes RRTM/Dudhia Thompson Grell-
DevenyiBurk-
Thompson RUC
RAPWRF-ARW v3.2+
GSI-3DVAR Yes RRTM/Goddard Thompson G3 +
Shallow MYJ RUC
HRRRWRF-ARWv3.2+
None: RR I.C. No RRTM/
Goddard Thompson None MYJ RUC
Model Run at: Domain Grid Points
Grid Spacing
Vertical Levels
Vertical Coordinate
Boundary Conditions Initialized
RUC GSD, NCO CONUS 451 x
337 13 km 50 Sigma/ Isentropic NAM Hourly
(cycled)
RAP GSD,NCO
North America
758 x 567 13 km 50 Sigma GFS Hourly
(cycled)
HRRR GSD CONUS 1799 x 1059 3 km 50 Sigma RR Hourly
(no-cycle)
Model Configurations
HRRRPrimary CoSPA
(FAA)
NCEPESRL/GSD/AMB
RAPDev
RAPPrimary
HRRRDev
RUCOper
RAPNCO
RAPDev2
RAPRetro
HRRRRetro
Retrospective Real-Time
AMB RAP/HRRR verification system
NWS
ESRL/GSD/AMB Model Verification
Ceiling Vis 24-hr Precip (03-80 km)
Reflectivity(03-80 km) Upper-Air Surface
0.5 – 3.0 kft 0.5 -5.0 mi 0.01 – 3.00 in 15-45 dBZ Temp RH Wind Height Temp Dewpt Wind
CSI, Bias, PODy/n, FAR, TSS, HSS, Fcst Ratio, Obs Ratio RMS and Bias
Convection(04 km, 80 km) Other Weather Events?
0 – 100 % 0 – 100 %
CSI vs Bias, Reliability, ROC
Deterministic
Probabilistic
Spring 2011 Hourly HRRR Initialization from RAP
HourlyRAP
LateralBoundaryConditions
Interp to 3 km grid
HourlyHRRR 15-h fcst
Initial Condition
Fields
11 z 12 z 13 z
Time (UTC)
AnalysisFields
3DVARObs
3DVARObs
Back-groundFields
18-h fcst 18-h fcst
1-hrfcst
DDFI DDFI
1-hrfcst
18-h fcst
1-hrfcst
Interp to 3 km grid
15-h fcstUse 1-h
old LBC to reduce
latency
Use most recent IC (post-DFI)
to get latest radar info Reduced
Latency:~2h for 2011
- Hourly cycling of land surface model fields - 6 hour spin-up cycle for hydrometeors, surface fields
Rapid Refresh Partial Cycling
RAP Hourly cycling throughout the day
RAP spin-upcycle
GFSmodel
GFSmodel
RAP spin-upcycle
00z 03z 06z 09z 12z 15z 18z 21z 00z
Observationassimilation
Observationassimilation
ESRL/GSD/AMBModel Analysis/Forecast Bias
Inherent in Model Physics?Non-Optimal Use of Obs?
Limitations of Data Assimilation Method?
Need Accurate Mesoscale Environment for High Impact Prediction
RAP vs. RUC surface – cold season
COLD
W
ARM
LOW
H
IGH
2m Temper-ature (K)
2m Dew Point (m/s)
RAPRUC
3-week comparison9 – 30 Jan 2012Eastern US only
RUCRAP
Diurnal bias variation6-h fcst
RUC daytime quite cool, RAP good
Both too moist, especially during
day
Steve Weygandt
RAP vs. RUC surface – warm season
COLD
W
ARM
LOW
H
IGH
2m Temperature (K)
2m Dew Point (m/s)
RAPRUC
2-month comparison20 April – 20 July 2011
Eastern US only
RUCRAP
Diurnal bias variation6-h fcst
RUC daytime slightly cool, RAP warm, esp.
overnight
Both too moist, especially at night
RUC worse than RAP
Steve Weygandt
13-km CONUSComparison2 X 12 hr fcstvs. CPC 24-h analysis1 May – 15 July 2011Matched
RUC/RR drier/moister below 650 mb, opposite aboveRR a closer fit to the observations by 6 hrs in lower tropWater vapor is the dominant source of the RH bias
Reflectivity DA in RAP/RUC increases HRRR forecast skillHRRR bias depends strongly on parent model
CSI 40 km
RUC->HRRR RadarRAP->HRRR Radar
RAP->HRRR No RadarRUC->HRRR No Radar
RAP->HRRR RadarRAP->HRRR No RadarRUC->HRRR RadarRUC->HRRR No Radar
BIAS 03 km
Optimal
Reflectivity DA on 3-km (HRRR) Grid
interpolation from RAP,hydrometeor specification
t060 min t045 min t030 min t015 min t0
reflectivity-based temperature tendency
NSSL3D
mosaic
HRRR composite reflectivityDavid Dowell
t045 min
t030 min t015 min t0
HRRR (3-km) grid produces convective storms explicitlyReflectivity-based temp. tendencies are applied during sub-
hourly cycling (forward model integration only, no digital filtering)
ESRL/GSD/AMBHRRR Forecast Post-Processing
Time-Lagged (TL) Ensemble
Spatio-temporal scale of probabilities
10-11 hr fcst
09-10 hr fcst
08-09 hr fcst
11-12 hr fcst
10-11 hr fcst
09-10 hr fcst
Forecasts valid 21-22z 27 April 2011 Forecasts valid 22-23z 27 April 2011
All six forecastssummed to formprobabilities valid22z 27 April 2011
HRRR 11z Init
HRRR 12z Init
HRRR 13z Init
The HCPF and HTPFHRRR Convective/Tornado Probabilistic Forecast (HCPF/HTPF) Estimate likelihood of convection and tornado production
• Intensity – ≥ 25 m2 s-2 Maximum Updraft Helicity 2-5 km AGL• Time – Two hour search window centered on valid times• Location – 45 km (15 gridpoints) search radius of each point• Members – Three consecutive HRRR initializationsHTPF = # grid points matching criteria over all members
total # grid points searched over all members
Caveats:Refinement needed to discriminate surface/elevated rotationEvaluate additional environmental fields that support tornado formationEvaluate additional diagnostic fieldsTornado scale not resolved
Case Selection from 2011
Date Region#
Tornado Reports
Max SPCRisk
Category*
Max SPCTornado
Probability*
14-Apr-2011 OK-AR 38 MDT 15%
15-Apr-2011 MS-AL 146 MDT 15%
16-Apr-2011 NC-SC-VA 139 HIGH 30%
26-Apr-2011 TX-AR-LA-MS-AL-TN 126 HIGH 30%
27-Apr-2011 MS-AL-TN-GA 292 HIGH 45%
28-Apr-2011 NC-SC-VA-MD 15 SLGT 10%
22-May-2011 OK-MO 75 MDT 15%
23-May-2011 22 MDT 10%
24-May-2011 OK-TX-AR 57 HIGH 45%
25-May-2011 MO-IL-IN 127 HIGH 30%
*Note: Maximum SPC forecast categories through 20z Day 1
ESRL/GSD/AMBHRRR Deterministic Skill
In Outbreaks
Tornado PeriodsStronger Forcing 14-16 April 2011
CSI
25 dBZ, 40 km 40 dBZ, 40 km25 dBZ, 03 km 40 dBZ, 03 km
03 hr fcsts 06 hr fcsts
09 hr fcsts 12 hr fcsts
25 dBZ, 40 km 40 dBZ, 40 km25 dBZ, 03 km 40 dBZ, 03 km26-28 April 2011
Valid 22z 27 April 20111630z SPC Tornado Probability
27 April 2011 Storm Reports
Tornado = Red DotsTornado Probability (%)
27 April 2011
Tornado Probability (%)Reflectivity (dBZ)
HTPF 13z + 09hr fcstValid 22z 27 April 2011
27 April 2011 Storm Reports
Tornado = Red Dots
Observed Reflectivity22z 27 April
22 May 2011HTPF 13z + 11hr fcstValid 00z 23 May 2011
1300z SPC Tornado Probability
22 May 2011 Storm Reports
Tornado = Red DotsTornado Probability (%)
22 May 2011
Tornado Probability (%)Reflectivity (dBZ)
22 May 2011 Storm Reports
Tornado = Red Dots
HTPF 13z + 11hr fcstValid 00z 23 May 2011
Observed Reflectivity00z 23 May
23 May 2011HTPF 13z + 11hr fcstValid 00z 24 May 2011
1300z SPC Tornado Probability
23 May 2011 Storm Reports
Tornado = Red DotsTornado Probability (%)
23 May 2011HTPF 13z + 11hr fcstValid 00z 24 May 2011
1300z SPC Tornado Probability
23 May 2011 Storm Reports
Tornado = Red DotsTornado Probability (%)
With MUCAPE LPL < 50 mb AGL
23 May 2011
Tornado Probability (%)Reflectivity (dBZ)
23 May 2011 Storm Reports
Tornado = Red Dots
HTPF 13z + 11hr fcstValid 00z 24 May 2011
Observed Reflectivity00z 24 May
24 May 2011HTPF 13z + 11hr fcstValid 00z 25 May 2011
1630z SPC Tornado Probability
24 May 2011 Storm Reports
Tornado = Red DotsTornado Probability (%)
24 May 2011
Tornado Probability (%)Reflectivity (dBZ)
HTPF 13z + 11hr fcstValid 00z 25 May 2011
24 May 2011 Storm Reports
Tornado = Red Dots
Observed Reflectivity00z 25 May
Summary
50
HRRR Challenges from a WOF/HIP perspective
Reduce latency from 2 hrs to 1 hr, 30 min, etc…?Tradeoff between assimilating latest mesoscale observations (more latency) and some radar observations (less latency)
Model moisture (and other) bias dominates convective forecastbehavior: initial condition error, model error, both?
Mesoscale environment remains a strong driver ofstorm-scale predictability
Weaker (stronger) forcing and smaller (larger) favorableenvironments do translate to lower (higher) skill forecasts
Small (large) run-to-run variability doesn’t always translate to a higher (lower) skill forecast