for air quality applications Edward Hyer Naval Research Laboratory AQAST Meeting Research Triangle Park, NC 16 November 2011

Transcription

1 Assimilation grade MODIS AOD for air quality applications Edward Hyer Naval Research Laboratory AQAST Meeting Research Triangle Park, NC 16 November 2011

2 Aerosol Data Assimilation for Air Quality Applications Model provides: coverage over domain Forecast Satellite observations: can correct magnitude of features Provide limited help for missed sources MODIS RGB NAAPS Natural MODIS AOD NAAPS + NAVDAS

3 Numerical Modeling for Aerosol Pollution Applications Model uses meteorological forcing to add/advect/remove particles Gives shape to pollution features Uses information on location+timing+magnitude of sources Model must distribute aerosol vertically Injection Boundary layer depth Vertical mixing A forward model is still necessary, with any foreseeable AOD data source. NAAPS Natural NAAPS + NAVDAS

4 Satellite Aerosol Data for Air Quality Applications AOD data from satellite are necessarily sparse: Cloud Sunglint Surface Retrieval of AOD is under determined More radiance at sensor from sources other than particle scattering = more uncertainty Satellite data do not provide domain coverage, and AOD cannot be retrieved from all satellite data A forward model is necessary for domain coverage MODIS RGB MODIS AOD Preparation of observation for assimilation entails: Elimination of outliers; Reduction of systematic bias; Reduction of random noise; Numerical estimation of observation uncertainty

5 NRL AOD provides significant benefits in error reduction and characterization Uncertainty using QA filters only is high in much of N. America NRL L3 AOD uncertainties are 5-40% lower over N. America Fraction of data within 20%+/-0.05 is 80% for both Terra and Aqua MODIS HYER et al. Atmospheric Measurement Techniques, 2011

6 Data Assimilation for Air Quality Applications Finer resolution of obs. = greater complexity of features that DA can help resolve Aggregation of obs = necessary to reduce random noise MODIS AOD AOD provides no vertical information AOD does not capture speciation of aerosol particles NAAPS Natural NAAPS + NAVDAS

7 Can AOD Assimilation Help with NAAPS surface PM vs TEOM monitor 24 hour PM10 from Singapore and Malaysia DOE 2006 severe haze event from burning in Sumatra and Borneo Model run: Pro: timing of many events Con: missed events Con: magnitude far too low Model + MODIS AOD Events more realistically simulated PM10? Yes.

8 Can AOD Assimilation Help with NAAPS surface PM vs TEOM monitor 24 hour PM10 from Singapore and Malaysia DOE 2006 severe haze event from burning in Sumatra and Borneo Model run: Pro: timing of many events Con: missed events Con: magnitude far too low Model + MODIS AOD Events more realistically simulated PM10? Yes.

9 Can AOD Assimilation Help with PM10? Yes. NAAPS surface PM vs TEOM monitor 24 hour PM10 from Singapore and Malaysia DOE 2006 severe haze event from burning in Sumatra and Borneo Model run: Pro: timing of many events Con: missed events Con: magnitude far too low Model + MODIS AOD Events more realistically simulated BINTULU OBS NAAPS NAAPS+MODIS

10 Can AOD Assimilation Help with PM10? Yes. NAAPS surface PM vs TEOM monitor 24 hour PM10 from Singapore and Malaysia DOE 2006 severe haze event from burning in Sumatra and Borneo Model run: Pro: timing of many events Con: missed events Con: magnitude far too low Model + MODIS AOD Events more realistically simulated KUALA LUMPUR OBS NAAPS NAAPS+MODIS

11 NRL L3 AOD data can be obtained from GODAE bin/datalist.pl?data_type=all&parameter=all&provider=nrl&meta=go#nrl_modis_l3 Latency 24 hours Can retrieve land/ocean, Aqua/Terra separately Currently staged 2006 present Access via FTP/HTTP

12 Next Steps NRT distribution of product through LANCE Code transfer approved by NRL Within government only LANCE will port code to their system LANCE will make ported code publicly available ETA early 2012 Improvement of products for other users Half degree resolution: available for research Finer resolution: need modeling partner to test Additional sensors: VIIRS, AVHRR

13 Backup Slides

14 Surface boundary condition bias MODIS c5 is a simultaneous retrieval of surface reflectance and AOD Flaws in surface reflectance estimate create geographic artifacts Bias affects low AOD Can be corrected with climatology Will still have problems with phenology BEFORE AFTER CORRECTION

15 Microphysical bias MODIS c5 uses a coarse climatology of aerosol optical properties Two problems: Bias over regions Uncaptured variability Above: Bias has general regional trends Below: High correlation + variable bias = uncaptured variability in aerosol properties Result: NRL L3 product has a regional correction applied N. AFRICA S. AFRICA