2012 and 2013-15 changes to the Rapid Refresh and HRRR

weather forecast models 31 October 2012

Stan Benjamin Steve Weygandt Curtis Alexander NOAA Earth System Research Laboratory

Boulder, CO FPAW - 2012 31 October 2012 HRRR/RAP Model Status 1

13km Rapid Refresh (RAP) (mesoscale)

13km RUC (mesoscale)

3km HRRR (storm-scale)

High-Resolution Rapid Refresh Experimental 3km nest inside RAP, hourly 15-h fcst

Replaced RUC at NCEP 05/01/12 WRF, GSI, RUC features

Hourly Updated NOAA Weather Models

FPAW - 2012 31 October 2012 HRRR/RAP Model Status 2

RAP: Data assimilation engine for HRRR

3

RAP

Data Assimilation cycle

Observations

Hourly cycling model

HRRR

HRRR Forecast Behavior

2012 Targets

(1) Lower peak bias in

convection over eastern US

(2) Fewer false alarms

(3) Improved timing of

convective initiation

(4) More success maintaining

mesoscale convective systems

(5) More realistic reflectivity

RAP/HRRR Model Development and Evaluation

GSD Program Review 13 March 2012 High-Resolution Rapid Refresh 4 FPAW - 2012 31 October 2012 HRRR/RAP Model Status 4

RAP and HRRR Changes 20112012 Model Data Assimilation

RAP-ESRL (13 km)

WRFv3.3.1+ Physics changes (convection, microphysics, land-surface, PBL) Numerics changes (w-damp upper bound conditions, 5th-order vertical advection) MODIS land use, fractional 3010 min shortwave radiation New reflectivity diagnostic

Soil adjustment Temp-dep radar- hydrometeor building PW assim mods Cloud assim mods Tower/nacelle/sodar observations GLD360 lightning GSI merge with trunk Radial wind assim

HRRR (3 km)

WRFv3.3.1+ Physics changes (microphysics, land-surface, PBL) Numerics changes (w-damp upper bound conditions, 5th-order vertical advection) MODIS land use, fractional 3005 min shortwave radiation New reflectivity diagnostic

3 km/15 min reflect assim 3 km radial wind assim 3 km cloud cycling 3 km land-surface cycling

GSD Program Review 13 March 2012 High-Resolution Rapid Refresh 5 FPAW - 2012 31 October 2012 HRRR/RAP Model Status 5

RAPv2 (2012 HRRR) more accurate than RAPv1 (2011 HRRR, NCEP) for

6

Wind Temperature

6h forecast RMS errors vs. radiosonde observations (weather balloons) - July – September 2012

NOAA High-Resolution Rapid Refresh HRRR Model Forecast of Mid-Atlantic Derecho – 29 June 2012

Composite Reflectivity (dBZ)

Radar observed HRRR forecast

Derecho loop

Radar observed HRRR 12-h forecast

Composite Reflectivity (dBZ)

NOAA Next-Generation RAP / HRRR System Forecast of Mid-Atlantic Derecho – 29 June 2012

Valid 11PM EDT Real-time HRRR forecast

HRRR Experiment with 2011 RAP

2013 – add 3-km 15-min radar assimilation for HRRR

Now: radar data assimilation only in 13km RAP By Jan 2013: continued radar assimilation in RAP (13 km) + 1h radar assimilation in HRRR (at 3 km) before HRRR forecast begins Future: cycling with all obs (including radar) on HRRR (3-km) grid 3DVar and reflectivity-based temperature tendency, hybrid / ensemble assimilation and forecasting

RAP 13 km fcst

obs

radar data

fcst

HRRR 3 km fcst

t0−2 h t0−1 h t0 interpolation

radar data

radar data

radar data

radar data

3DVar + cloud analysis … …

HRRR reflectivity DA • same formulation of

reflectivity-based temperature tendency as in RAP

• no digital filter

HRRR Storm Skill: 20 dBZ radar reflectivity threshold

3-day retrospective period June 2011, forecasts every 2 hours composite reflectivity

verification over eastern half of US (widespread convective storms)

Critical Success Index (CSI) Bias

upscaled to 40-km grid native 3-km grid

2013 - with 3-km radar DA without 3-km radar DA

2013 - with 3-km radar DA without 3-km radar DA

Bias = 1.0

Composite Reflectivity

2200 UTC 11 June 2011

0-h fcst with 3-km radar DA

observations

0-h fcst without 3-km

radar DA

explicit precipitation from RAP

1000 km

Composite Reflectivity

2200 UTC 11 June 2011

0-h fcst with 3-km radar DA

0-h fcst without 3-km

radar DA

storms developed during HRRR cycling; locations were forced partly by reflectivity-

based heating

observations

1000 km

Composite Reflectivity

2300 UTC 11 June 2011

1-h fcst with 3-km radar DA

1-h fcst without 3-km

radar DA

mature convective systems benefit particularly from

subhourly radar DA

observations

1000 km

Composite Reflectivity

0100 UTC 12 June 2011

3-h fcst with 3-km radar DA

observations

3-h fcst without 3-km

radar DA

convective systems maturing;

errors apparent

1000 km

Composite Reflectivity

0100 UTC 12 June 2011

3-h fcst with 3-km radar DA

observations

3-h fcst without 3-km

radar DA

better representation of convective

character and location

1000 km

Composite Reflectivity

0400 UTC 12 June 2011

6-h fcst with 3-km radar DA

observations

6-h fcst without 3-km

radar DA

differences persist (cumulative effects

following from radar and/or other DA)

1000 km

• Current – 1 computer running HRRR – NOAA/ESRL – Boulder – Current reliability: 97% for last 12h months (allowing up to 3h

gaps) • 2013-14 – 2 computers running HRRR – interim solution

– Boulder – computer 1 (jet) – Fairmont, WV – computer 2 (zeus) – Expected reliability to increase further to 98.5-99% via

coordination of downtimes for Boulder vs. Fairmont computers • 2015 – NCEP running HRRR

– NOAA/NCEP computing budget – will allow no increase before 2015

• Conclusion: Interim HRRR computing for 2013-15 on 2 sites to provide “research regular” HRRR from NOAA for NWS, FAA, DOE/energy users

HRRR Transition to NCEP

FPAW - 2012 31 October 2012 HRRR/RAP Model Status 17

HRRR 12 hr fcst availability Includes all missed/incomplete runs

Jet (HRRR feeding CoSPA) Zeus (HRRR backup) FPAW - 2012 31 October 2012 HRRR/RAP Model Status 18

Increasing reliability from HRRR running on zeus (Fairmont, WV computer)– for 2013

13km Rapid Refresh and 3km HRRR hourly updated weather models

ESRL – experimental version • RAPv1 – used in 2011

– Initialized 2011 HRRR – effective but too many storms

• RAPv2 – used in 2012 – Initialized 2012 HRRR – Better use of surface obs / radar,

storm bias eliminated • HRRR – 2012

– Major improvement over 2011 HRRR, storm coverage/accuracy

• HRRR – 2013 – 3km/15min radar assimilation – Initialized from RAP-2013 – Available 45 min earlier, much more

accurate 0-15h storm forecasts, more reliable 2-computer

NWS-NCEP - operational

• Implemented 1 May 2012 • RAPv2 - Scheduled to be

implemented in June 2013

• HRRR – estimated 2015

19

NOAA Next-Generation Global Model Development

FIM

iHYCOM

atmosphere

ocean iHYCOM – Icosahedral Hybrid Coordinate Ocean Model

- Matched grid design to FIM for coupled ocean- atmosphere prediction system

- Experimental testing at ESRL, Navy development

FIM – Flow-following finite volume Icosahedral Model

– “soccer-ball” grid design for uniform grid spacing

– New 14-day forecast twice daily – Real-time experimental at ESRL

Hurricane Isaac 7-day forecast track by FIM9 (15km) from Friday 24 Aug 18z run

Observed track for Isaac

Initial Position of Isaac – 18z 24 Aug

Eyjafjallajökull 2010: Comparison of FIM-Chem volcanic ash forecasts with London VAAC at 1800Z, April 17 and 0000Z, April 18

Why NOAA Weather Modeling Matters

Innovation to meet crucial societal needs: – Protecting lives and property

– Improving quality of life Aviation / Severe weather / Energy tornadoes hurricanes wind events floods winter storms air quality

What makes it possible - Improved model formulation

- Improved data assimilation - Increased computer power

HRRR

FIM