aircraft spiral on july 20, 2011 at 14 utc validation of goes-r abi surface pm2.5 concentrations...

17
Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA), Chuanyu Xu (IMSG), and Pubu Ciren (IMSG) AERONET data provided by NASA (PI: Brent Holben) Aircraft data provided by UMD (PI: Jeff Stehr) AIRNOW data provided by EPA (PI: John White)

Upload: bridget-owen

Post on 26-Dec-2015

221 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Aircraft spiral on July 20, 2011 at 14 UTC

Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data

Shobha Kondragunta (NOAA), Chuanyu Xu (IMSG), and Pubu Ciren (IMSG)AERONET data provided by NASA (PI: Brent Holben)

Aircraft data provided by UMD (PI: Jeff Stehr)AIRNOW data provided by EPA (PI: John White)

Page 2: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

GOES-R Advanced Baseline Imager (ABI) Aerosol Products

• Aerosol Optical Depth (AOD)

• Suspended Matter (µg/cm2)• Aerosol detection

(qualitative flag for smoke and dust)

• ABI Products generated using proxy data– MODIS (Aqua and Terra)

radiances– Theoretical radiances

calculated from CMAQ 3D aerosol fields

2 km

10 km

All validation work reported in this presentation is based on the 10 km ABI products

Page 3: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

• Suspended matter is calculated from ABI AOD using:

Where mc is mass concentration, τ is ABI AOD, k is mass extinction efficiency (cm2/µg) and h is height.

– k is a function of aerosol type and and h is assumed to be 3 km (from 6S model used to create LUTs)– Errors in ABI aerosol type (k) and h being a non-derivable parameter can lead to errors in mc

GOES-R ABI Suspended Matter

hkmc

0 1 2 3 4 5 60

1020304050607080

Mass Extinction Efficiency (cm2/µg) for Different Aerosols

GenericUrbanSmokeDust

Aerosol Optical Thickness

Inve

rse

of k

Page 4: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Validating the ABI Aerosol Products

• AOD validation (routinely done)– AERONET measurements

• Suspended matter validation (spot check verification only)– EPA AIRNOW observations

• Knowledge of aerosol type and height not available.

– Aircraft observations• Capture vertical profile of aerosols• Information on scattering vs. absorption can be used as a proxy for aerosol

type• Limited to data between surface and 3 km. Aerosols aloft are not captured

• In this study, we attempt to use aircraft, AIRNOW, and AERONET data to assess if aircraft data can be useful to evaluate GOES-R SM product.

Page 5: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Study Domain• Aircraft data

– July 2011 (13 flight days with multiple spirals each day).

• GOES data– 4 km (nadir) and 30 min temporal

resolution.• GOES-R data

– GOES-R ABI algorithm run on MODIS (Aqua and Terra) 10 km (nadir) resolution radiances as proxy.

• AIRNOW data– Hourly surface PM2.5 measurements.

• AERONET data– 15 minute AOD measurements.– DRAGON network (10s of stations

within 100 km x 100 km over eastern US.

AOD

MatchupsGOES/GOES-R vs. AERONET: ± 15 min; nearest GOES/GOES-R pixelGOES/GOES-R vs. Aircraft: ± 30 min; nearest GOES/GOES-R pixelGOES/GOES-R vs. AIRNOW: ± 30 min; nearest GOES/GOES-R pixel

Page 6: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

AIRCRAFT vs. GOES AODs

Bias 0.089

RMS 0.137

Integrated aerosol extinction profiles

Page 7: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Aircraft vs. AERONET AODs

Bias 0.187

RMS 0.203

Page 8: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

GOES-R ABI vs. Aircraft AODs

Agreement between aircraft and AERONET AODs is similar to aircraft and GOES AOD results (previous slide) indicating that during this time period (July 2011), the aerosol was not necessarily confined to PBL. Note aircraft measures between surface and 3 km while both GOES and AERONET measurements are for total column.

Bias 0.095

RMS 0.229

Page 9: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

GOES-R ABI vs. AERONET AODs

GOES-R ABI AODs retrieved using MODIS (Aqua and Terra) as proxy agree with AERONET observations better compared to aircraft AODs. This could be due to the fact that aerosols were not necessarily confined to PBL. The scatter plots also show some strange features (highlighted) that need further investigation. ABI AOD, however, meets GOES-R specification set at 0.04 for AOD ranging between 0.2 and 0.8.

Bias 0.037

RMS 0.109

Could be issues with surface reflectance retrievals at low optical depths.

Page 10: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

GOES-R ABI vs. AIRNOW PM2.5

Bias 13.53

RMS 14.5

Page 11: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

GOES-R ABI vs. Aircraft Mass Concentrations

Aircraft particle number density measured at the lowest altitude converted to mass concentration using density of aerosol typical for eastern US and measured particle size

Bias 8.94

RMS 24.30

Bias -4.97

RMS 11.49

Page 12: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Left panel: AOD converted to mass concentration using aerosol type that the ABI algorithm identifiedRight panel: AOD converted to mass concentration using urban aerosol type

The two data points (highlighted) changed marginally but no significant change to the overall results. Ideal way to observe the impact is to run the ABI algorithm using urban

aerosol model LUT and then convert the AOD to mass concentration using urban aerosol type. That work will be done in the future.

Aerosol Type

Page 13: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Aerosol Height

Page 14: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Aircraft data: Lowest Layer

Aircraft data: Lower than 500m

Aircraft data: Lower than 1000m

Aircraft data: Column Average

Density 1.7 g/cm3

Aircraft data: Lower than 1000m

Page 15: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Aircraft data: Column Average

Aircraft data: Lowest Layer

Aircraft data: Lowest 500 m

Aircraft data: Lowest 1000 m

Density 1.0 g/cm3

Page 16: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Profile for one spiral case on July 18, 2011 Profile for one spiral case on July 20, 2011

Profile for all spiral cases (averaged)

Variability in aerosol concentration as a function of height.

Page 17: Aircraft spiral on July 20, 2011 at 14 UTC Validation of GOES-R ABI Surface PM2.5 Concentrations using AIRNOW and Aircraft Data Shobha Kondragunta (NOAA),

Conclusions• AOD comparisons

– GOES-R ABI AOD correlates well with aircraft AOD (integrated extinction profile) but biased high (~0.1). Potential reasons:

Aircraft data extend only up to 3 km ABI algorithm picked the wrong aerosol model (type) Uncertainty in surface reflectance retrieval

• PM2.5 comparisons– GOES-R ABI estimates of surface PM2.5 correlate well with AIRNOW (ground) and aircraft

observations but biased high. The bias between ABI and AIRNOW is ~13.5 µg/m3 and bias between ABI and aircraft is 9 µg/m3 when the aircraft data are averaged for the whole column.

– These comparisons are sensitive to assumed density of aerosol as well as the aerosol height.

• Limitations of aircraft data– Uncertainty in converting particle number density to mass concentration (UMD working

on calibrating this conversion)– Aircraft ceiling of ~3 km– Limited geographic coverageAnalysis indicates that although suspended matter product from ABI has reasonable

values, it is biased high. Knowing aerosol height is more important than aerosol type. ABI AOD meets specification when compared to AERONET but not when compared to aircraft AOD but the sample size is very small for aircraft data.