1 Assimilate Column and Surface Data for Air Quality Forecasting and Chemical Reanalysis 13 th JCSDA...

1

Assimilate Column and Surface Data for Air Quality Forecasting and Chemical Reanalysis

13th JCSDA May13-15, 2015, College Park

Pius Lee1, Greg Carmichael2, Yongtao Hu3, Edward Hyer4, Yang Liu5, Dick McNider6, Brad Pierce7, Robert Atlas8, Ted Russell3,

Sid Boukabara9, Youhua Tang1,10, Hyuncheol Kim1,10, Li Pan1,10, Sean Casey9 and Daniel Tong1,10

1Air Resources Laboratory / NOAA, NCWCP, College Park, MD 2College of Engineering, University of Iowa, Iowa City, IA

3School of Civil and Environmental Engr., Georgia Institute of Technology, Atlanta, GA4Naval Research Laboratory, Monterey, CA

5Department of Environmental Health, Emory University, Atlanta, GA6Department of Atmospheric Science, University Alabama, Huntsville AL

7National Environmental Satellite and Information Service (NESDIS), Madison, WI8Atlantic Oceanographic & Meteorological Lab., Miami, FL

9NOAA/NESDIS/Joint Center for Satellite Data Assimilation (JCSDA), College Park, MD 10Cooperative Institute for Climate and Satellite, University of Maryland College Park, MD

Courtesy: Dan Costa “New Directions in Air Quality Research at the US EPA”

2

Dr. L. Uccellini, May14 2015 “Impact-based DSS”


Gridded forecast guidance products• On NWS servers: airquality.weather.gov and ftp-

servers• On EPA servers• Updated 2x daily

Verification basis, near-real time: • Ground-level AIRNow observations • Satellite smoke observations

Twice daily surface O3 concentration forecast NAQFC


4

Constrained

SatelliteProducts

Global Assimilation

+ AQ Assessments

+ State Implementation Plan Modeling

+ Rapid deployment of on-demand rapid-response forecasting; e.g., new fuel type,…, etc.

+ Health Impacts assessments

+ Demonstration of the impact of observations on AQ distributions

+ Ingestion of new AQAST products into operations

http://acmg.seas.harvard.edu/aqast/projects.html

AQAST Project: Air Quality Reanalysis (Translating Research to Services)


WRF_ARW-MCIP-CMAQ forward model

CMAQ 4.7.1

WRF-ARW (LCC)

(42 -P model Layers)

LBC from GFS

Projection ofendo-domain intermitten

sources: Obs’d wild fireand prescribed burns

EPA Emissions Inventory + simple obs-based adjustment

Hourly 3-D Gridded Chemical

Concentration

MCIP

MODIS-AOD-based adjusted IC,

BC: RAQMS

C-grid

C-grid

Column integrated AODC-grid

42 -P met. Layers

C-grid

C-grid


6

Widely used data assimilation schemes in Air Quality (AQ):

3DVAR e.g. in NCEPEnsemble Adjustment Kalmar Filter (KF) e.g. in trained Bias Correction schemesOptimal Interpolation: is a truncated extended KF

Continuity equation for species φi in generalized co-ordinate(Byun and Schere 2006)


Optimal Interpolation (OI)OI simplifies the extended Kalman filter formulation

(Dee et al. Q. J. R. Meteor. Soc. 1998) by limiting the analysis problem to a subset of obs.

Obs far away (beyond background error correlation length scale) have no effect in the analysis.

Injection of Obs through OI takes place at 1800 UTC daily.

)()( 1 HXYOHBHBHXX TTba


Estimate Model Error Statistics w/ Hollingsworth-Lonnberg Method

• At each data point, calculate differences between forecasts (B) and observations (O)

• Pair up data points, and calculate the correlation coefficients between the two time series

• Plot the correlation as a function of the distance between the two stations,

Alternatively 3DVar: M. Pagowski Q.J.R.Meteor. 2010 13th JCSDA May13-15, 2015, College Park

MODIS AOD & AIRNow PM2.5 assimilated

For initial condition adjustment

9

00Z 06Z 12Z 19Z

AIRNOW PM2.5, PM10, Ozone

MODIS AOD (Terra and Aqua)

14Z 17Z

18Z


Total AOD assimilation tests

Day

AO

DR

MS

E

0 1 2 3 4 5 6 7 8 9 10 11 12 130

0.05

0.1

0.15

0.2

0.25

Base caseOI forecastOI analysis

Base case: CMAQ4.7.1 without any data assimilationOI forecast: CMAQ results after assimilating previous day AOD observationsOI Analysis: CMAQ results after assimilating same day AOD, for next day forecast

Day

AO

DR

MS

E

0 1 2 3 4 5 6 7 8 9 10 11 12 130

0.05

0.1

0.15

0.2

0.25

Base caseOI forecastOI analysis

Fine mode AOD assimilation Total AOD assimilation

10

18Z

17Z 20Z OI: MODIS total AOD OI: AIRNow pm2.5


11

Transmittance

AIRNow

Cloud-obs Photolysis rates

Isoprene& PAR

done

done

done

Yet to do


12

Analysis field Verification: AIRNOW O3 & PM 2.5 over NE


13

WRF 3.2.1 for meteorological fields• NCEP North American Regional Reanalysis

(NARR) 32-km resolution inputs• NCEP ADP surface and soundings observational

data • MODIS landuse data for most recent land cover

status• 3-D and surface nudging, Noah land-surface

modelSMOKE 2.6 for CMAQ ready gridded

emissions • NEI inventory projected to 2011 using EGAS growth and

existing control strategies • BEIS3 biogenic emissions based on BELD3 database• GOES biomass burning emissions:

ftp://satepsanone.nesdis.noaa.gov/EPA/GBBEP/ CMAQ 4.6 revised to simulate gaseous & PM

species• SAPRC99 mechanism, AERO4, ISORROPIA

thermodynamic, Mass conservation,• Updated SOA module (Baek et. al. JGR 2011) for

multi-generational oxidation of semi-volatile organic carbons

Analysis fields as IC and forecast with Total AOD Assimilation basis for LBCs

12 km

36 km

4 km

IC/BC of Base and AOD_DA cases

DISCOVER-AQGA_Tech4 km


14

Analysis field Verification: AIRNOW O3 & PM 2.5 over BW SIP


PM2.5 24h avg Obs mean Mean bias RMSE Corr. Coef.

L42 12.3 -7.7 10.4 0.4

L42RAQMS 12.3 -5.4 8.5 0.5

L42-Fire1 12.3 -7.6 10.6 0.3

L42-Fire1-OI4 12.3 -1.11 6.65 0.61

CONUS

PM2.5 24h avg Obs mean Mean bias RMSE Corr. Coef.

Base 19.8 -0.67 24.18 0.49

With L42-Fire1-OI4 field to derive LBC

19.8 1.15 5.79 0.53

15

DISCOVER-AQGA_Tech4 km

Application of analysis field for State Implementation Planning


Forecast Model

Simulated Observations

Validation of End-to-End OSSE Set-up

Nature Runs DAS

OSSE Metrics

OSSE TESTBED COMPONENTS -Current

ECMWF GLA .25 T213 GLA .5° T511 (MM5) T799 (WRF)

Validation and Augmentation

Valid.CyclonesMean fieldsCloudsPrecip

Aug.CloudsAerosolsTurbulence

CurrentScatterometerRawinsondeSurfaceAircraftTOVSAIRSAMV

FutureDWLCMISMolnyaUASCL BalloonGPS SoundingBuoy Rocket

- Realism of simulated observations and input- Impact model not too close to NR

GlobalGFSFIM

RegionalWRFHWRFNAMCMAQ

Cyclones, Fronts, JetsAnomaly CorrelationAir Traffic RoutingPrecip.Utility Load Mgmt.

CurrentGSI4DVAREnKFH*Wind

Targeting Observation schemes

16


AQ study using 2 GFS simulations

18

• CONTROL• T382(~38km); 3D-Hybrid (Var & EnKF); No AIRS_G13• Simulated observations for July-August 2005 (T511 ECMWF Nature Run

(NR)) assuming 2012 observation system• All instruments (conventional, GPS radiation occultation, radiance)

operational in July-August 2012, with addition of Special Sensor Microwave Imager Sounder (SSMIS)-F16,F17,F18

• Random-errors added to all radiance observations using modified version of R. Errico’s (GMAO) error-addition code

• Two week spin-up, 47-day experiment period (20050716-20050831)• AIRS

• As CONTROL but w/AIRS_G13 (AIRS instrument in the location of GOES-13, 75°W)

• Simulated from T511 NR by Z. Li, U. Wisconsin, using Stand Alone Radiation Transfer Algorithm (SARTA) (compared to CRTM for JCSDA-simulated radiances)

• Random-errors added using expected error distribution for AIRS_AQUA

Atlas et al. BAMS 2011 doi:2010BAMS2946.1


Preliminary VSDB Results

19

• Right: RMSE for 500 hPa geopotential height (forecast hour on horizontal axis)

• Lower figure: difference between mean RMSE, AIRS-CONTROL

• Red boxes: 95% confidence interval; counts outside these bounds are considered statistically significant

• Comparisons done with respect to T511 NR

• Here, the experiment with AIRS_G13 shows significant reduction in RMSE for days 1, 2

• Full results can be viewed on JCSDA website: http://www.jcsda.noaa.gov/vsdb/users/scasey/prs382hwa/vsdb.php

• I created a “rough scorecard” based on the Version 16 results (next slide)

CONTROLAIRS

CONTROL


Courtesy: Sean Casey

http://www.jcsda.noaa.gov/vsdb/users/scasey/prs382hwa/vsdb.php

http://www.jcsda.noaa.gov/vsdb/users/scasey/prs382hwa/vsdb.php

20

O3 on 20050805

CONTROL- standard GFSAIRS-assisted GFS

Localized improvement in r & other metrics

O3 on 20050805

O3 on 20050803 O3 on 20050803

Spatial distribution of correlation coefficient for surface O3 on Aug 3,5 2005


21

Configured AQ forward-model vertical structure matching that of GFS The reanalysis forward model is tested and used to generate July 2011 analysis fields for MDE for SIP modeling – GATech partnered SIP modelers Data Set assimilated: RAQMS (MLS, OMI O3, MODIS AOD); HMS Fire; GOES

cloud fraction for photolytic rate correction; MODIS AOD; AIRNow O3, PM2.5

Next sets: lightning NOx; OMI SO2; PAR adjustment by retrievals Verification to include data from O3 lidar network Begin transition of assimilation algorithm to GSI Production mode generation of analysis field for 2010 (HTAP Collaboration)OSSE accomplishment: Performed GFS-WRF-CMAQ sensitivity runs to investigate AIRS-assisted

GFS AIR-assisted GFS-WRF-CMAQ does improve surface O3 and PM2.5

forecast The degree of improvement is small compared to Chemical Data

assimilation on CMAQ The Chemical Data Assimilation (CDA) on PM2.5 benefits stronger than that

for O3

Summary


2222

EXTRA SLIDES

Contact: [email protected]://www.arl.noaa.gov/


mailto:[email protected]

This study aims to build a prototype User friendliness and reliability are keys to success: e.g..

Goal: A user friendly downloadable archive


MODIS (Moderate Resolution Imaging Spectroradiometer) AOD

http://terra.nasa.gov/About/Orbit: 705 km, 10:30 a.m. descending

node (Terra) or 1:30 p.m. ascending node (Aqua)

Swath Dimensions:

2330 km (cross track) by 10 km (along track at nadir)

Spatial Resolution:

250 m (bands 1-2)500 m (bands 3-7)1000 m (bands 8-36)

Courtesy :NESDIS

24

National correlation map between AIRNow measurement and MODIS AOD

Typically good correlation betweensurface PM2.5 and AOD retrievedby MODIS


25

Analysis field Verification: AIRNOW O3 & PM 2.5 over CONUS


1 Assimilate Column and Surface Data for Air Quality Forecasting and Chemical Reanalysis 13 th JCSDA...

Documents

Transcript of 1 Assimilate Column and Surface Data for Air Quality Forecasting and Chemical Reanalysis 13 th JCSDA...