Global and Regional Retrieval of Aerosol from MODIS

Transcription

1 Global and Regional Retrieval of Aerosol from MODIS Why study aerosols? CLIMATE VISIBILITY Presented to UMBC/NESDIS June 4, 24 Robert Levy, Lorraine Remer, Yoram Kaufman, Allen Chu, Russ Dickerson modis-atmos.gsfc.nasa.gov aerocenter.gsfc.nasa.gov QuickTime and a TIFF (LZW) decompressor are needed to see this picture. IPCC,2 AIR QUALITY HUMAN HEALTH EARTH HEALTH Aerosol Measurements IN-SITU (Perturbed) REMOTE (Ambient) Aerosol Types/Properties Satellite: Scattering -> AOD/Size Pump/Filter: Size/Concentration Of interest here Nepholomete/PSAP: Scatt/Absorb->Concentration/Size Sunphotometer: Extinction/Scattering -> AOD/Size Fine Area UV BGR NIR Coarse 2.5µm IR

2 Aerosol Scattering ~ 9% total Extinction Depends on number, size and composition of intervening aerosol Spectral optical properties of aerosol AVHRR DUST Small aerosol DUST α α = πd λ λ Large aerosol SMOKE SMOKE Maximum efficiency when aerosol size wavelength Spectrally dependent Optical depth τ MODIS MODIS The Satellite Signal Target = surface (properties assumed) RGB: April, 9 2 Gas + Aerosol scattering (path radiance) e τ (/ µ s + / µ v ) Direct Transmission (of surface albedo) clear-sky conditions. Adapted from Tanre et al. (979), (98), (983). Terra: Winter 2 Aqua: Summer 22 Scientific Data Atmosphere: Cloud and Aerosol Ocean: Color, Chlorophyll, Temp Land: Vegetation, Change, Fires Aerosol Retrieval Over Land: ( km x km) τ (AOD) at 3 λ Aerosol size/type Over Ocean: ( km x km) τ (AOD) at 7 λ Aerosol size/type e τ (/µ v ) t d (µ s ) Indirect Transmission (adjacency effect) T (µ s )t d (µ v ) + multiple I actions Multiple Reflection T (µ ) = e τ / µ + t d (µ) T = Transmission τ= optical depth µ= cosθ T d = direct trans 2

3 Aerosol is transparent to Mid-IR, thus the surface properties are observed. Surface reflectance in the visible is function of Mid-IR τ.66 [ρ.66.5ρ 2. ] τ.47 [ρ.47 5ρ 2. ] Aerosol type/optical properties are assumed based on season and location Single channel retrievals are performed in red and blue, then dust is added to fit spectral dependence. λ(µm) Aerosol over Land Y. J. Kaufman Modeled and Observed Reflectance from MODIS July 2, 4:5: τ 865 =.48 Reflectance Aerosol Over Ocean - Inversion r_eff =. r_eff =.5 r_eff = r_eff = 5 Salt: r_eff =.98 Salt: r_eff =.48 Salt: r_eff =.98 Dust: r_eff =.48 Dust: r_eff = 2.5 Measured Reflectance Rayleigh Reflectance "S 4: B 6: Ratio= Wavelength Use reflectance in 6 wavelengths to invert τ λ Constrained by 4 fine mode and 5 coarse mode aerosol models. Inversion chooses fine and coarse mode, plus relative concentration Surface optical properties are modeled Dust and Haze in East Asia March 2, 2 AOD Level > Level 3 (daily) Seperating dust and smoke Level 2 x km retrievals (irregular lat/long) 5 minute granules Level 3 (daily) º x º (regular lat/long) 5 minute granules tiled Statistics are produced Mean, Stddev, Pixel Count, Histo Quality Control / Confidence 3

4 Level 3 Daily > Level 3 Monthly Level 3 (monthly) º x º (regular lat/long) Daily values are averaged Statistics are produced Mean, Stddev, Pixel Count, Quality Control / Confidence AOT MODIS AOD Validation Comparison with sunphotometer AERONET AOT at Cart Site for 7/7/2, showing a -hour segment (box) centered on the Terra overpass time.35 AOT_2.3 AOT_87 5 AOT_67 AOT_5.5 AOT_44. AOT_38.5 AOT_34 3:2 4:24 5:36 6:48 8: 9:2 Time of Day (GMT) Comparison with Climate models 5 km lan d km km 5 km ocean AERONET sites AERONET PI s: B. Holben, C. McLain, D. Tanré ocean land both Global MODIS AOD Validation Plume Transport and Spatial Variability MODIS AOT (66 nm) y = x R =.92 points 5 points 25 points 5 points % of retrievals over OCEAN fall within expected uncertainty OCEAN 66 nm N = 252 AERONET AOT (66 nm) ocean MODIS AOT (66 nm) y = x R =.68 3 points 5 points 75 points 32 points LAND 66 nm N = AERONET AOT (66 nm) land 7% of retrievals over LAND fall within expected uncertainty Remer et al., May 4, 2 Sunphotometer AOD.67 2 Interesting Plumes 2-23 King et al., 23 4

5 MODIS (column) vs PM (surface) Monthly mean correlations of.9 in Southeast (Wang & Christopher) Correlation nationally (Engle-Cox et al.) A. CHU MODIS AOT (66 nm) But land retrieval is not perfect y = x R =.68But 3 points 5 points 75 points 32 points land LAND 66 nm N = Positive offset Slope less than one Offset in blue (47 nm) even worse, like.2. Only 6% within error bars over U.S. East Coast AERONET AOT (66 nm) MUST STUDY REGIONALLY! Land retrievals too high for low AOD Customize Surface reflectance ratios: August, 2 Globally Assumed RGB: Aug.47 AOD.6.47/2. ~ 5.3. Current work. Wallops AERONET AOD =.8.66/2. ~ /2. ~.45.66/2. ~.62 Locally Assumed 5

6 Urban/Industrial Model: US East Coast Current Proposed A new aerosol model May improve fits!!! Surface reflectance in the visible is function of Mid-IR, but maybe variable ρ.66 = f(ρ 2. ) ρ.47 = f(ρ 2. ) Aerosol type/optical properties are assumed based on season and location Aerosol is not assumed transparent in Mid-IR. 3 wavelength inversion performed combining red blue, and IR, fitting the spectral dependence. Polarization included Fewer Assumptions! New land algorithm λ(µm) Conclusions Aerosols are perfectly sized to interact with solar radiation. MODIS has high spectral and spatial resolution in solar bands. We exploit aerosol optical spectral dependence to retrieve optical depth and column-averaged size information from MODIS observations. Comparisons with ground-based sunphotometer validate the MODIS products, globally. Regionally, there is correlation with MODIS column aerosol and surface PM measurements MODIS over land can be improved by customizing surface reflectance relationships, updating the assumed aerosol model dataset, and using a new inversion algorithm. 6