Evaluation of radiance data assimilation impact on Rapid Refresh

forecast skill for retrospective and real-time experiments

Haidao LinSteve WeygandtStan Benjamin

Ming HuCurtis AlexanderPatrick Hofmann

Assimilation and Modeling Branch

Global Systems DivisionNOAA Earth System Research

LabBoulder, CO

Cooperative Institute for Research in the Atmosphere

Colorado State University

http://rapidrefresh.noaa.gov

Presentation Outline

1. Background on Rapid Refresh (RAP) system

2. Data introduction and selected channels

3. Retrospective experiments AIRS data impact Other radiance data impact

(AMSU-A, HIRS, and MHS)

4. Real-time radiance data impact in RAP

(with hybrid EnKF assimilation)

5. Real-time RAP radiance data availability issues

6. Summary and future work

– Advanced community codes (ARW model, GSI analysis)– Key features for short-range “situational awareness”

application (cloud analysis, radar DFI assimilation) RAP guidance for aviation, severe

weather, energy applications

Background on Rapid RefreshNOAA/NCEP’s hourly updated model

RAP version 1 -- NCEP since Spring 2012

RAP version 2 -- Planned NCEP Late 2013– DA enhancements (Hybrid –

EnKF using global ensemble)

– Model enhancements(MYNN PBL, 9-layer LSM)

Improved mesoscale guidance, GSD version parent for HRRR

Rapid Refresh Hourly Update Cycle

1-hrfcst

1-hrfcst

1-hrfcst

11 12 13Time (UTC)

AnalysisFields

3DVAR

Obs

3DVAR

Obs

Back-groundFields

Rawinsonde (12h) 150NOAA profilers 35VAD winds ~130PBL profilers / RASS ~25

Aircraft (V,T) 3500 – 10,000METAR surface 2000 -2500Mesonet (T,Td) ~8000Mesonet (V) ~4000Buoy / ship 200-400GOES cloud winds 4000-8000METAR cloud/vis/wx ~1800

GOES cloud-top P,T 10 km res.Satellite radiances (AMSUA, HIRS, MHS)Radar reflectivity 1 km res.

Data types – counts/hr

Partial cycle atmospheric fields – introduce GFS information 2x per dayFully cycle all land-sfc fields

- Hourly cycling of land surface model fields - 6-hour spin-up cycle for hydrometeors, surface fields

RAP Hourly cycling throughout the day

RAP spin-upcycle

GFSmodel

RAP spin-upcycle

GFSmodel

00z 03z 06z 09z 12z 15z 18z 21z 00z

Observationassimilation

Observationassimilation

Rapid Refresh Partial Cycling

RAP Radiance impact Testing

RETROSPECTIVE REAL-TIME

Model WRF-ARW v3.2 WRF-ARW v3.4

Analysis GSI 3DVAR GSI hybrid EnKF

(80 member global)

Cycle 3-h, NO part. cycle 1-h with part. cycle

Satellite Ideal (no latency) Actual latencyCoverage

Expts Control (conv. obs) Control (conv. obs)

CNTL+ AIRS only CNTL + all 4 sat. CNTL+AMSUA only

CNTL+HIRS onlyCNTL+MHS onlyCNTL+all 4 sat.

Radiance Data• AIRS (Focused work)

– High vertical resolution– Temperature and moisture information – Not yet operationally used in RAP,

used experimentally in GSD real-time run

• AMSU-A (Operationally used in RAP)– Temperature information

• HIRS (Operationally used in RAP)– Temperature information– Moisture information (channels 10-12)

• MHS (Operationally used in RAP)– Moisture information

Radiance Channels Selected for RAP• AIRS (remove high peaking channels)

– Aqua: 68 channels selected from 120 GDAS channel set

• AMSU-A (remove high peaking and surface channels) – metop-a: channels 4-6, 8-10– noaa_n15: channels 4-10– noaa_n18: channels 4-8, 10– noaa_n19: channels 4-7, 9-10

• HIRS (remove high peaking and ozone channels)– noaa_n19: channels 4-8, 10-15– metop-a: channels: 4-8, 10-15

• MHS– noaa_n18: channels 1-5;– noaa_n19: channels 1-5; (not available for retro run on 2010)– metop-a: channels 1-5;

Retrospective Experiments

AIRS impact

Settings for AIRS Retrospective Runs • Extensive retro run for bias coefficients spin up

• May 8 – May 16, 2010, 3-h AIRS radiance data with bias coefficients cycled (the very first bias coefficients came from GDAS global system)

• For each analysis cycle, 30 successive GSI runs are completed with bias updated from the previous GSI runs.

• Control run (CNTL) – Conventional data only• 3-h cycling run, 9 day retro run (May 8 2010 – May 16 2010)

• AIRS experiment one (AIRS Ex. 1) -- NO BIAS SPIN UP• CNTL + AIRS radiance data (60 km thinning in GSI)

• Use bias coefficients from GDAS

• Use the 68 selected channel set for RAP

• AIRS experiment two (AIRS Ex. 2) – WITH BIAS SPIN UP• CNTL + AIRS radiance data• Use updated bias coefficients from the extensive retro run• Use 68 selected channel set

AIRS Bias Correction Assessment

After BC

Before BC

channel 252 (CO2 channel ~672h Pa

Channel 1382 (water vapor channel ~866 hPa

With BCSpin-up

Without BCSpin-up

Number of Observations Used

Water vaporCarbon dioxideSurface

o Without BC spin-up * With BC spin-up

Better BC more obs used improved BC

AIRS Forecast Impact (against raob, 100-1000 hPa)

Normalize Errors

EN = (CNTL – EXP)

CNTL

Temperature

Relative Humidity

Wind9-day retro

average

Control run: conventional data only

AIRS Ex. 2(w/ BC spin-up)

AIRS Ex. 1(w/o BC spin-up)

+1%

-1%

0%

Retrospective Experiments

AMSU-A impactMHS impactHIRS impact

All impact (AIRS + AMSU-A + MHS + HIRS)

Settings for other Radiance Retrospective Runs

• Extensive retro run for bias coefficients spin up• Control run (CNTL) – conventional data only

• 3-h cycling run, 9 day retro run

(May 8 2010 – May 16 2010)• AMSU-A experiment – WITH BIAS SPIN UP

• CNTL + AMSU-A radiance data• MHS experiment – WITH BIAS SPIN UP

• CNTL + MHS radiance data • HIRS experiment – WITH BIAS SPIN UP

• CNTL + HIRS radiance data

AMSU-A Forecast Impact

Control run: conventional data only

9-day retro average

Forecast verification against raob. for 100-1000-hPa layer over a national domain

AMSU-A data from NOAA_15, 18, 19 and METOP-A

Temperature Relative Humidity

Wind

MHS Forecast Impact

Control run: conventional data

9-day retro average

Forecast verification against raob. for 100-1000-hPa layer over a national domain

Temperature

Relative Humidity

Wind

MHS data from NOAA_18 and METOP-A

HIRS Forecast Impact

control run: conventional data

Forecast verification against raob. for 100-1000-hPa layer over a national domain

9-day retro average

HIRS4 data from NOAA_19 and METOP-A

Temperature

Relative Humidity

Wind

Comparison of Radiance Impact

Control run: conventional data

Forecast verification against raob. for 100-1000-hPa layer over a national domain

9-day retro average

TemperatureRelative Humidity

Wind

AIRS

AMSU-A

HIRS

MHS

All

24-h (2 X 12h) Precipitation VerificationCSI by precip threshold(avg. over eight 24h periods)

AIRS

CNTL (conventional data )

MHS

HIRS

AMSU-ASlight improvement

for heavy precipitation thresholds from

radiance data

MHS data have largest positive impact for heavy precipitation prediction

Real-time Experiments

All impact (AIRS + AMSU-A + MHS + HIRS)

Real-Time RAP Runs

• Real-time RAP hybrid systems on Zeus: • 1-h cycling with partial cycle• real-time data

• Time period: May 29 – June 4 2013• RAP dev_sat_1: Control run (CNTL)

• conventional data only• RAP dev_sat_2: radiance experiment run

• conventional data + radiance data

(AIRS, MHS, HIRS, AMSU-A)

Real-Time Impact

real-time control run: conventional data

Temperature

Relative Humidity

Wind Forecast verification against raob. for 100-1000-hPa layer over a national domain

6-day real-time average

Data Availability Issues

Estimating fraction of data used

Initial examination of Regional ATOVSRetransmission Services (RARS) data

Estimating fraction of data used

short data cutoff times combined with long data availability latency times leads to minimal satellite data availability

W = Data Window Time

L = Data Latency Time

C = Data Cutoff Time

W = 180 minL = 60 minC = 30 min

(W/2 - L + C)/W

= 33% obsused

after cutoffdata

latencycutoff time

Diagram and equationfollowing Steve Lord

Samplecase

data window initial time03z02z 04z 05z 06z

dataavailable

0130z 0230z 0330z 0430z Obs time

Fraction of data used given by:

Real-Time Data Availability -- RARS

AMSU-A channel 3 from NOAA_18

Real-Time RAP

IDEAL -- No latency/cutoff

RARS feed (not used in real-time yet)

18Z May 29, 2013

Assuming +/- 1.5 h time window

AMSU-A RARS Data Coverage

12Z 13Z 14Z 15Z 16Z 17Z

18Z 19Z 20Z 21Z 22Z 23Z

RAP used +/- 1.5 hour data window

RARS +/- 1.5 hour data window

RAP used +/- 1.5 hour data window

RARS +/- 1.5 hour data window

May 29, 2013, NOAA_19

1741

Summary Retrospective runs

AIRS Focused-- Positive impact for short-range skill after BC spin-up;

improved BC result in more data assimilated

Overall (AIRS, AMSU-A, MHS, HIRS)-- Temperature: Positive impact for AIRS (largest) and MHS;

negative impact from HIRS data-- Moisture: Positive impact from all; max for MHS (3.5% for

3-h fcst); descending order: MHS, AIRS, AMSU-A, HIRS, -- Wind: Positive impact from AMSU-A, AIRS; the largest

impact from AMSU-A (more than 2.5% at 12-h forecast)-- Precipitation: Small positive impact at high threshold

for MHS

Summary Real-time runs

• Small positive impact overall (near-neutral for temperature)

• Largest positive impacts for 3-h forecast:-- moisture ~ 0.8%-- wind ~0.8%-- temperature ~0.6%

• Real-time impact smaller than retrospective impacts(actual data vs. ideal data)

• Real-time Data latency/cutoff issues for rapidly updating model system

Future Work• Investigate negative/neutral impacts for AMSU-A/HIRS data

• Model top problem -- Increase RAP model top and model levels (for experiment

purpose) -- Blending GFS fields

• Real-time data latency problem-- Use RARS feed-- Partial cycle (more waiting time)

• Retro tests with new RAP hybrid system

• Tropical cyclone prediction using RAP

• New data/instruments-- METOP-B, CrIS/ATMS

• Continue retro/real-time RAP radiance assimilation testing, implement enhancements into operational RAP.