GOESR ABI Aerosol Algorithms GOESR Algorithm Working Group
GOES-R ABI Aerosol Algorithms GOES-R Algorithm Working Group Aerosol, Atmospheric Chemistry and Air Quality (AAA) Application Team 1
Presentations · Suspended Matter/Aerosol Optical Depth Algorithm – Istvan Laszlo, STAR · Aerosol Detection Algorithm – Shobha Kondragunta, STAR · Proving Ground and User Interaction – Shobha Kondragunta, STAR 2
Suspended Matter (SM) Aerosol Optical Depth (AOD) Presented by Istvan Laszlo With contributions from Mi Zhou, Pubu Ciren, and Hongqing Liu 3
SM/AOD Retrieval: Physical Basis · The aerosol portion of the atmospheric radiation (aerosol reflectance) observed by satellites is determined by the amount and type (size, shape and chemical composition) of aerosol. · Over dark surfaces, aerosol reflectance increases with increasing amount of aerosol (as measured by AOD) → Used for estimating AOD · The spectral dependence of aerosol reflectance is a function of aerosol type. → Used for estimating aerosol type (model) model 1: urban; model 2: smoke; top: 0. 64 μm, bottom: AOD=0. 4; solar zenith angle = 40 o; view zenith angle = 40 o; relative azimuth = 180 o 4
SM/AOD Algorithm: Features of the GOES-R/ABI SM/AOD algorithm: · · Based on the MODIS/VIIRS heritages Separate algorithms for land water Uses multiple channels to estimate AOD and aerosol type Advantages » » A lot of ground work has already been done with MODIS Has been tested in an operational environment Potential synergy with MODIS/VIIRS aerosol product Estimates aerosol type · Disadvantages » » Sensitive to radiometric error (multi-channel retrieval) No retrievals over bright surface (sun-glint, bare soil, desert) Dependence on aerosol model assumptions Over land, uses Lambertian surface model and spectral regression with large variance for surface albedo, which can lead to large AOD error for not dark enough surface 5
SM/AOD Retrieval: Illustration of Methodology · AOD and aerosol model corresponding to calculated reflectances best matching observed ones are selected as solutions. · AOD and model is from the “minimum” residual between observed and calculated spectral reflectances. · Residual 2 < Residual 1, so retrieved AOD ≈ 1. 0 and aerosol model is model 2. aerosol model 1 aerosol model 2 TOA reflectance in red band · Aerosol retrieval is accomplished by comparing observed spectral reflectances with calculated ones. Illustration of aerosol retrieval concept Residual 1 Observation Residual 2 Retrieved AOD 550 TOA reflectance in blue band 6
SM/AOD Algorithm Input Sensor Input 1 2 3 4 5 Wavelength Range (μm) 0. 45 -0. 49 0. 59 -0. 69 0. 846 -0. 885 1. 371 -1. 386 1. 58 -1. 64 Central Wavelength (μm) 0. 47 0. 64 0. 865 1. 378 1. 61 Central Wavenumber (cm-1) 21277 15625 11561 7257 6211 Sub-satellite IGFOV (km) 1 0. 5 1 2 1 6 2. 225 - 2. 275 2. 25 4444 2 7 8 9 10 11 12 13 14 15 3. 80 -4. 00 5. 77 -6. 6 6. 75 -7. 15 7. 24 -7. 44 8. 3 -8. 7 9. 42 -9. 8 10. 1 -10. 6 10. 8 -11. 6 11. 8 -12. 8 3. 90 6. 19 6. 95 7. 34 8. 5 9. 61 10. 35 11. 2 12. 3 2564 1616 1439 1362 1176 1041 966 893 813 2 2 2 2 2 16 13. 0 -13. 6 13. 3 752 2 ABI Band land only both land ocean Sample Use AOD Land and Ocean AOD Land Ocean Estimate land surface reflectance ocean only 7
SM/AOD Mathematical Description Calculation of TOA Reflectance The satellite-observed reflectance (ρtoa) is approximated as the sum of atmospheric (ρatm) and surface components (ρsurf) TOA reflectance surface contribution atmospheric contribution • Calculated reflectances account for transmission and absorption of radiation in the atmosphere and reflection at the surface. • Atmospheric reflectances and transmittances are pre-calculated using the 6 S RTM (Vermote et al. , 1997) and stored in LUT for speed. • Surface reflectance of ocean is calculated; that over land is retrieved. → Separate algorithms for aerosol retrieval over ocean and land. 8
SM/AOD Mathematical Description Atmospheric Contribution Calculation of atmospheric reflectance term gas transmittance atmosphere LUT top of atmosphere O 3, O 2, CO 2, N 2 O, CH 4 molecules, aerosol, H 2 O bottom of atmosphere ρR+A : reflectance due to molecules (R) and aerosol (A) together – calculated with 6 S RTM and stored in LUT ρR : reflectance due to molecules – calculated in the code following 6 S; P 0 and P are standard and actual pressures, respectively T : gas transmittance (parameterized) 9
SM/AOD Mathematical Description Surface Contribution Calculation of surface reflectance term gas transmittance atmosphere LUT land ocean reflectances Total (direct+diffuse) downward and upward transmittance TR+A and spherical albedo SR+A of molecular and aerosol atmosphere are calculated with 6 S RTM and stored in LUT 10
SM/AOD Mathematical Description Ocean Surface Reflectance Water reflection includes three components: • Water-leaving radiance (Lambertian) Whitecap effective reflectance Wind speed (m/s) • Whitecap (Lambertian) ABI Channel (wavelength in µm) • Sunglint (bi-directional) 1 (0. 47) 0. 2200 0. 0148 2 (0. 64) 0. 2200 0. 0013 3 (0. 865) 0. 1983 0. 0 5 (1. 61) 0. 1195 0. 0 6 (2. 25) 0. 0475 0. 0 ρwc corresponds to constant chlorophyll concentration (0. 4 mg m-3) 11
SM/AOD Mathematical Description Ocean Surface Reflectance Term 1 Term 2 Sunglint calculated Term 3 Term 5 Term 4 • Formulation follows 6 S RTM • Cox and Munk (1954) ocean model • Constant salinity (34. 3 ppt) • Fixed westerly wind direction Sunglint LUT All , , and from atmosphere LUT 12
SM/AOD Mathematical Description Land Surface Reflectance a Surface reflectances in the visible and NIR ABI channels · Lambertian reflection is assumed. · Surface reflectances at 0. 47 (ρ0. 47) and 0. 64 μm (ρ0. 64) are estimated from those at 2. 25 μm (ρ2. 25). · Use NDVI to separate vegetation- and soil-based surface types (VIIRS approach) Mid-IR NDVI · For vegetation-based surface · For soil-based surface 13
SM/AOD Mathematical Description Selection of Dark Pixel · Land – select pixels with low SWIR reflectance: » 0. 01 ≤ ρ 2. 25 μm ≤ 0. 25 · Ocean – avoid areas effected by glint: » glint angle θg > 40 o – θg is the angle between the viewing direction θv and the direction of specular reflection θs: θs θg Φ Z θs θv θg= cos-1( cosθs cosθv + sinθs sinθv cosΦ ) 14
SM/AOD Mathematical Description Aerosol Models LAND: Four aerosol models: dust, smoke, urban, generic (MODIS C 5, Levy et al. , 2007) Single scattering albedo and asymmetry parameter as a function of wavelength for the four land aerosol models WATER: Four fine mode and five coarse mode aerosol models (MODIS C 5) Single scattering albedo and asymmetry parameter as a function of wavelength for the fine (left) and coarse mode (right) models over ocean. 15
SM/AOD Mathematical Description SM/AOD Retrieval over Land • Retrieve ρ2. 25 , AOD and aerosol model simultaneously by matching the observed TOA reflectance of the reference channel 0. 47µm and calculate the corresponding residuals at 0. 64µm for each of the four aerosol models Lookup Table Satellite & Ancillary Data Each aerosol model calculate TOA reflectance at 0. 47µm Y match 0. 47 um observation ? retrieved AOD calculate residual at 0. 64µm N Increase AOD at 550 nm where residual is calculated as: ● Select the aerosol model and AOD with the minimum residual as the “best” solution 16
SM/AOD Mathematical Description SM/AOD Retrieval over Ocean • TOA reflectance is assumed to be a linear combination fine and coarse mode aerosols Retrieve AOD and fine mode weight for each combination of candidate fine and coarse aerosol models. ● Lookup Table Satellite & Ancillary Data For each fine & Coarse model combination calculate TOA reflectance in ABI channel Increase AOD at 550 nm match 0. 87 μm obs. ? Y retrieved AOD N calculate residuals in channels 2, 5&6 residual Minimum residual? retrieved AOD & Weight & residual Change fine mode weight where residual is calculated as: ● Select the AOD and combination of fine and coarse modes with minimum residual as the “best” solution. 17
SM/AOD Mathematical Description Size Parameter and SM ● The Ångström exponent (α) is used as proxy for particle size: ● Large/small values of Ångström exponent indicate small/large particles, respectively. ● The Ångström exponent is calculated from AODs and two pairs of wavelengths (MODIS heritage): ● SM: The retrieved AOD is scaled into column integrated suspended matter in units of µg/cm 2 using a mass extinction coefficient (cm 2/µg) computed for the aerosol models identified by the ABI algorithm. 18
SM/AOD Algorithm Verification Comparison with MODIS/Terra ABI AOD MODIS/Terra aerosol reflectances are used; 03/15/2012 MODIS-ABI AOD 19
SM/AOD Algorithm Verification Comparison with AERONET Land • Retrievals are from MODIS Terra and Aqua from 2000 -2009 • All available AERONET stations • AOD at 550 nm Water • Same overall performance of MODIS and ABI over land • Slightly smaller overall ABI bias over water 20
Aerosol Detection (Smoke & Dust) Presented by Shobha Kondragunta With contributions from Pubu Ciren 21
Aerosol Detection Sensor Inputs Future GOES Imager (ABI) Band Nominal Wavelength Range (μm) Nominal Central Wavelength (μm) Nominal Central Wavenumber (cm-1) Nominal sub-satellite IGFOV (km) Sample Use 1 0. 45 -0. 49 0. 47 21277 1 Dust/Smoke 2 0. 59 -0. 69 0. 64 15625 0. 5 Dust/Smoke 3 0. 846 -0. 885 0. 865 11561 1 Dust/Smoke 4 1. 371 -1. 386 1. 378 7257 2 Dust 5 1. 58 -1. 64 1. 61 6211 1 Dust/Smoke 6 2. 225 - 2. 275 2. 25 4444 2 Smoke 7 3. 80 -4. 00 3. 90 2564 2 Dust/Smoke 8 5. 77 -6. 6 6. 19 1616 2 9 6. 75 -7. 15 6. 95 1439 2 10 7. 24 -7. 44 7. 34 1362 2 11 8. 3 -8. 7 8. 5 1176 2 12 13 14 15 9. 42 -9. 8 10. 1 -10. 6 10. 8 -11. 6 11. 8 -12. 8 9. 61 10. 35 11. 2 12. 3 1041 966 893 813 2 2 16 13. 0 -13. 6 13. 3 752 2 Input for both Dust and smoke Input for smoke Dust/Smoke Input for dust 22
Physical Basis of the Algorithm · Aerosols, surface, and clouds have different spectral and spatial characteristics » Aerosol and surface signals can be separated through analysis of spectral differences in BTs and reflectances » Cloud mask information is passed on by the cloud algorithm but internal tests for additional cloud screening and snow/ice have been implemented · Thresholds based on simulations and observations from existing satellite instruments. 23
Physical Basis of the Algorithm Clear Sky Thin Dust Thick Dust 24
Physical Basis of the Algorithm · Spectral (wavelength dependent) thresholds can separate thick smoke, light smoke, and clear sky conditions clear smoke Clear Regime Heavy smoke Thick Smoke Regime 25
Aerosol Detection – Example Global Smoke/Dust Flags (May 26, 2008) Mongolia desert dust ABI smoke/dust detection algorithm is tested by using MODIS as proxy data smoke dust Biomass Saharan burning desert dust 26
Routine Validation Tools · · Product validation: using CALIPSO Vertical Feature Mask (VFM) as truth data (retrospective analysis not near real time. Data downloaded from NASA/La. RC) Tools (IDL) » Generates match-up dataset between ADP and VFM along CALIPSO track, spatially (5 by 5 km) and temporally (coincident) » Visualizing vertical distribution of VFM and horizontal distribution of both ADP and VFM » Generating statistics matrix 27
Percentage of Pixels (%) ”Deep-Dive” Validation Tools 28
Proving Ground and User Interaction Shobha Kondragunta With contributions from P. Ciren, C. Xu, H. Zhang 29
Air Quality Proving Ground (AQPG) http: //alg. umbc. edu/aqpg/ · NOAA has created the AQPG – a subset of the GOES-R Proving Ground – focusing on the aerosol products that will be available from the ABI. · Goal: build a user community that is ready to use GOES-R air quality products as soon as they become available. • This distinction is important because the air quality community has very different needs than the majority of NOAA users (NWS meteorologists). • AQPG is using simulated GOES-R ABI data for training and interaction with the user community. 30
Proxy ABI Aerosol Optical Depth · AOD indicates areas of high particulate concentrations in atmosphere · AOD is unitless; high AOD values (yellow, orange, red) indicate high particulate concentrations · Clouds block AOD retrievals 31
Proxy ABI Aerosol Type • New product - not available with current GOES imager • Qualitative and untested • Useful for distinguishing between smoke and dust but can be noisy, especially at low AOD values 32
Proxy ABI Synthetic Natural Color (RGB) • No green band on ABI • Algorithm development underway to improve RGB product 33
Haboob (intense dust storm) over Pheonix, Arizona in the evening of July 5, 2011. Photo by Nick Ozac/ The Arizona Republic MODIS RGB Image (bottom left) and Aerosol Optical Depth (bottom right) the next morning during Terra overpass show widespread dust. Neither Aqua nor Terra captured the event as it happened on July 5 th because it happened at the night fall
Compared to a single snapshot of Terra overpass (bottom) the morning after haboob, 30 -min refresh rate movie of GOES shows changing dust plume features. However, note the noise in GOES data. For GOESR, 5 -min refresh rates with good quality “MODIS-like retrievals” will be the norm to track episodic events such as dust storms and smoke plumes. . Widespread dust over Phoenix on July 6 th : the remnant of the haboob
http: //www. star. nesdis. noaa. gov/smcd/spb/aq/ NOAA’s IDEA Site (dynamic flat webpages)
- Slides: 36