1. AERONET in the context of aerosol remote sensing from space and aerosol global modeling Stefan Kinne MPI-Meteorology, Hamburg Germany

2. Overview AERONET statistics – a sampler AERONET and satellite data Validation of Satellite Data (  ) Evaluation or Regional Representation (  ) AERONET and global modeling Evaluation of Model Simulation (  ) Data on composition and vertical profile (  ) state in global modeling and AeroCom

3. AERONET statistics monthly average properties a sampler for three sites: GSFC (near Washington DC) ‘urban’ Mongu (Zambia) ‘biomass’ [Jul-Nov] Cape Verde (west of Sahara) ‘dust’ – measured properties (aot, alfa) – derived properties (absorption, size) – value-added properties (forcing, lidar ratio) locally – aerosol can be defined well

4. urban absorption: 10*aot*(1-  0 ) lidar ratio: for spheres  too small for non-spheres ToA Forcing: clr-sky [W/m2]

5. biomass absorption: 10*aot*(1-  0 ) lidar ratio: for spheres  too small for non-spheres ToA Forcing: clr-sky [W/m2] 70% PDF value 30% PDF value

6. dust absorption: 10*aot*(1-  0 ) lidar ratio: for spheres  too small for non-spheres ToA Forcing: clr-sky [W/m2]

7. the consistency among data allows combination for global assessment this example: seasonal avg. for aerosol absorption [  * (1-  0)] interesting… … AERONET indicates more absorption by aerosol over Europe than over east- US US Europe

8. AERONET and other aerosol data-sets Aeronet data can ‘help’ (  ) Aeronet data can ‘learn’ (  ) – examples are introduced next

9. AERONET  satellite data satellite aot data (aerosol optical thickness or depth) what is available ? what is best? SatelliteAdvantageDisadvantage --------------------------------------------------------------------------------------------------------- AVHRRhistoric recordcalibration, not over land TOMShistoric record 50km pixel, height or absorption needed MODISsmall pixel failure over deserts MISRaltitude info temporally spare POLDERshort record, land: less sensitive to large sizes SEAWIFSnot over land, no IR channels GOES orhigh temporallack of detail with broad MSG resolutionbands, land limitations

10. aot - global yearly averages with all data available  normalized by model to offset sampling biases  AERONET has lower aots than satellite retrievals a clear-sky bias?

11. comparisons or annual pattern Mo: MODIScomposites: Mi: MISR12:Mo,Mi To: TOMS13:Mo,To Av: AVHRR Po: POLDERAe:Aeronet difficult to depict a best global retrieval  composite needed a MODIS (ocean) MISR (land) combination seems promising … … but differences to AERONET still exist

12. local comparisons to AERONET choice : MI / MO still … generally larger than AERONET, particular in urban regions …but seasonal data show biomass burning aots are too low

13. seasonal comparisons at AERONET

14. first impressions MODIS best choice over the oceans … but too low in dust outflow regions (high aot  clouds) MISR most complete land cover … while biased high over oceans MODIS (ocean) / MISR (land) combination the ‘best’ satellite product is generally larger than AERONET … but too low during biomass burning open issues: are AERONET aot smaller due to a clear-sky bias? what can be said about the quality of retrievals of low aot in remote regions (of no AERONET sites?) is it ‘fair’ to compare point data with regional data?

15. satellite data  AERONET use spatial information of satellite data – to relate local measurement detail to coarse gridded data-sets coarse resolution data in global modeling how ? – compare averages for different scales agreement … indicates a ‘useful’ site bias: ‘useful’ site after a bias adjustment highly variable (season/years) : leave off comparison … unless secondary data exist

16. “scaling” Comparison of –300*300km data –100*100km data –10*10km data GSFC (urban) –20% above the regional average Mongu (biomass) –good match for the biomass season (Jul-Nov)  at the bottom are AERONET-MODIS comparisons (2001) note: MODIS statis- tics are very poor! MODIS AERONET

17. needed scaling activities for different spatial domains a data-base of simultaneous satellite retrievals over AERONET sites is needed satellite requirements: small (~1km) pixel retrievals at regional coverage sufficient data (for seasonal /annual dependence) coverage of all AERONET sites (incl. desert sites) MODIS and MISR data are a start … although their smallest pixels size at 10.0 and 17.6 km is too large to represent ‘truly’ local characteristics

18. AERONET  global modeling pick 20 sites (well spread globally) compare aerosol optical depth sub-micron sized aerosol optical depth aerosol mass (note AERONET: wet mass, Models: dry mass) sub-micron sized mass refractive index, imaginary part … identify large disagreements identify poor concepts eliminate poor concepts (or poor models)

19. white rings : AERONET aot smaller than model average black rings : AERONET aot larger than model average color-coded compositional fractions as predicted by global models at selected AERONET sites

20. aerosol optical depth comparisons

21. beyond annual aot averages differences among (15) models and to AERONET are better understood on a monthly basis –test seasonality major ‘enhancements’ MARCHdust (Africa and Asia) JUNEurban aerosol (NH) SEPTEMBERbiomass burning (SH) DECEMBER biomass burning (trop. Africa) –identify monthly outliers in modeling

22. aerosol optical depth in March

23. aerosol optical depth in June

24. aerosol optical depth in September

25. aerosol optical depth in December

26. aerosol optical depth comparisons for selected month

27. anthropogenic = smaller sizes = SU + OC + BC

28. optical depth comparisons for selected month for submicron size aerosol

29. aerosol mass comparisons (Models: dry, AERONET: wet) AERONET over- estimates expected in humid conditions

30. aerosol mass comparisons for selected months (Models: dry, Aeronet: wet)

31. mass comparisons for submicron size aerosol (Models: dry, AERONET: wet) Anthropogenic = smaller sizes ? = SU + OC + BC ?

32. aerosol refractive index imaginary part (absorption) (Models: dry, AERONET: wet)

33. modeling and AeroCom AeroCom http://nansen.ipsl.jussieu.fr/Aerocom –validate against data! surface concentrations (IMPROVE, EMEP, GAW) surface remote sensing (AERONET, EARLINET) remote sensing from space (MODIS, MISR) –14 groups participate so far A: ‘best as you can’ – simulation B: yr 2000 simulation with prescribed emissions C: yr 2000 simulation with pre-industrial emissions – to address anthropogenic ‘forcing’

34. global modeling  AERONET compositional is the dominant component captured? detail  is the seasonality captured? is the anthropogenic fraction correct? Aeronet component combined totals vertical integrated column properties Modeling detail on compositional mixture detail on vertical distribution …still modeling though

35. 3 examples of AERONET sites dominated by a particular type dust biomass (carbon) urban (sulfate) Aerosol modules agree better on total aot than in terms of aot composition AERONET can help identify unusual model behavior

36. Summary large differences in component aerosol modules in global modeling annual averages hide larger seasonal differences global averages hide larger regional difference component totals hide larger diff. among compo. integrated properties (e.g. forcing) hide larger differences on a sub-level basis (e.g. mass) AERONET provide constraints to models for aot (and locally even for component aots) for other aerosol properties (e.g. size, mass, absorption) only at lower certainty

37. Message anthropogenic impact of aerosol on climate needs to be better quantified (reduce uncertainties) uncertainties in aerosol forcing (the end product in modeling) do not represent ‘actual’ uncertainties model differences at intermediate processing steps and on different scales are much larger AERONET can provide needed constraints … …especially in conjunction with space data and value-added modeling (e.g. inversions)

38. Outlook tighter and stronger links (via selective sampling) between AERONET and remote sensing from space are needed to extend detail on size and composition to remote sensing from space to identify sites with regional representation needed for evaluations in global modeling complementary information on vertical distribution is highly desirable... and we keep our fingers crossed for the A-train

40. extras

41. AERONET statistics statistics for 1998-2001 absorption = 10* aot*(1-  0 )