1. Satellite observations of AOD and fires for Air Quality applications Edward Hyer Naval Research Laboratory AQAST 3 13-15 June, Madison, Wisconsin 15 June 2012Hyer AQAST 31

2. In This Talk Update: Production of assimilation-ready MODIS AOD product at NASA LANCE Plans for MODIS and VIIRS assimilation-ready products Other activities 15 June 2012Hyer AQAST 32

3. Why Does Assimilation-Grade AOD Matter? Aerosol analysis and forecasting requires AOD for assimilation Assimilation has specific requirements – Minimize outliers – Correct persistent bias – Quantify residual uncertainty Level 2 AOD products are not good enough – Correlated bias – Limited error characterization 15 June 2012Hyer AQAST 33

4. 15 June 2012Hyer AQAST 3 Surface Boundary Condition Issues in over- land MODIS AOD Strong systematic bias with surface albedo Positive or negative biases for different land areas 4

5. 15 June 2012Hyer AQAST 3 Surface Boundary Condition Issues in over- land MODIS AOD Albedo correction calculated empirically Corrects negative bias in S. America, positive bias over arid surfaces Correction reduces occurrence of negative AODs AFTER CORRECTION 5

6. 15 June 2012Hyer AQAST 3 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. AMERICA Regional slope correction improves global correlation of MODIS AOD vs AERONET from r 2 =0.62-0.65 to r 2 =0.71- 0.73 SLOPE vs AERONET CORRELATION Vs. AERONET 6

7. 15 June 2012Hyer AQAST 3 Prognostic Error Estimation Obs. Uncertainty = instrument error + representativeness error Representativeness error = SD of AOD within grid cell Instrument error: – Over ocean: f(AOD, fine mode fraction) – Over land: f(AOD, region) Error = MAX(“noise floor”,linear relation) Below left: Estimated uncertainty for over-land product without corrections – QA=Very Good, Cloud =0 Below right: Error ratio of corrected vs. uncorrected AOD – Australia: filtering changes distribution of AOD, increases noise floor error 7

8. NRL MODIS L3 AOD for Data Assimilation Filtered – QA – Surface – Clouds – Textural Corrected – Surface – Microphysics Gridded – 0.5 degree – 6 hours Error estimation – Empirical based on AERONET 15 June 2012Hyer AQAST 38 Our goal is to make this product available in near real time for any user with an air quality or visibility forecasting application

9. Production at LANCE: Progress 1.Updated code for unified production of over- land and over-ocean AOD 2.Improved resolution from 1 degree to 0.5 degree 3.Changed surface wind correction (over- ocean) from Navy NOGAPS to GMAO GEOS-5 4.Created HDF-EOS output file format 5.Added metadata for traceability 15 June 2012Hyer AQAST 39

10. Production at LANCE: Product 1 Product will be called MCDAODHD Product will be produced with 2-3 hour latency Terra-only product will be produced first (Terra has lower latency) LANCE will make production code available to the public Documentation and sample code for using this product available (C, FORTRAN, IDL, Python) 15 June 2012Hyer AQAST 310

11. Plans for DA-ready AOD products Working on product from NPP VIIRS MODIS Collection 6? – May permit improved spatial resolution Improvements to error estimation – Prognostic error model has skill, but does not account for some known factors (atmospheric path, surface properties) Investigate possible residual snow contamination – Snow filtering is already very strict, but suspicious high AODs continue to appear 15 June 2012Hyer AQAST 311

12. Other activity: NPP VIIRS, fire products Fire products: see Hyer & Westphal poster this afternoon NPP VIIRS: ask me about NPP VIIRS! To right: NRL NexSat (http://www.nrlmry.nav y.mil/NEXSAT.html) GOES animation of New Mexico Gila Firehttp://www.nrlmry.nav y.mil/NEXSAT.html 15 June 2012Hyer AQAST 312