The surface albedo of land surfaces are usually unknown and vary with wavelength.
Dark, dense vegetation has a very low reflectance in the blue and red regions.
This has allowed the development of algorithms to retrieve aerosol optical depth
over dark vegetation (called dark targets). Dry or unvegetated surfaces,
such as deserts, and snow or ice covered surfaces reflect too much light to
be used for aerosol retrieval with current methodologies. These areas are masked
out in the MODIS data products showing aerosol loading over land.
In the images below, global gridded results are shown for sample months of
the MODIS Optical Depth over Land and Ocean product (from MOD08_M3).
The values over land masses are the aerosol model corrected optical
depths (explained below). Notice how the masked areas (black w. red stippling)
around
the globe vary in size and location with the changing of the seasons (images
courtesy of Goddard Space Flight Center, NASA).
June
2001
Sept. 2001
Dec. 2001
Mar 2002
May 2002
The MODIS spectral channels used in the aerosol retrieval over land are:
0.47 µm (500 m resolution) 0.66 µm (250 m resolution) 2.1 µm (500 m resolution) 3.8 µm (1 km resolution)The latter two channels are used for the determination of surface reflectance. The 11 µm window channel is used to correct for IR emission from the Earth surface. Other gaseous absorption channels, such as the near-IR (~1 µm) water vapor channel (for the correction of total precipitable water vapor) and the 9.6 µm channel (for total ozone correction) are also used. In Part I we provide a simplified description of the steps followed in the retrieval of aerosols over land. Part II consists of a more detailed explanation of the theory and methodology behind the land algorithm.
The satellite measured reflectances at 2.1 and 3.8 µm (r*2.1 and r*3.8) are used to determine which pixels are sufficiently dark to be used in the algorithm. Why? First, we should understand that reflectance of light by land surfaces across the solar spectrum is correlated to some extent. Soils usually reflect more light as wavelength increases, and the correlation between the reflectances slowly decreases as the wavelength span increases. However, parallel processes affect the reflectance in the visible (eg. 0.47 and 0.66 µm) channels and in the mid-IR channels (2.1 and 3.8 µm). Vegetation decreases reflectivity in the visible channels, due to chlorophyll absorption, but also in the mid-IR channels due to absorption by liquid water associated with the plant. Wet soil has a lower reflectance in the visible channels due to light trapping, but also in the 2.1 and 3.8 µm channels due to the liquid water absorption. Moreover, surface roughness, shadows and inclinations decrease the reflectance across the whole solar spectrum. This means that if we know the surface reflection of light in the mid-IR, we can estimate the surface reflection of light in the visible wavelengths. Second, we should understand that the longer the wavelength, the smaller the effect that aerosol scattering has on the reflectance of light measured in space. Except for dust, the aerosol effect on the radiance measured from space decreases with wavelength as l-1 to l-2. Longer wavelengths (2.1 or 3.8 µm) are less sensitive to aerosol scattering than visible wavelengths (since these wavelengths are much longer than the size of most aerosol particles) but are still sensitive to land surface characteristics. The reflectance of mid-IR observed by the satellite can be assumed to be essentially the reflectance of the land surface, and not of particles in the atmosphere above it.
1. Pixels are grouped into grid boxes. When the spatial resolution of the channel is 1 km, there are 10x10 pixels per grid box. For channels with higher resolution (500m and 250m), more pixels are included in the grid boxes. 2. A series of criteria involving the magnitude of satellite observed mid-IR reflectance (r*2.1 and r*3.8) are applied to each grid box, and the number of pixels (N) that satisfy each criterion is found. The first criterion in the series that yields an N > 5% of the pixels in the grid box is chosen as the winning criterion. The criteria are used only over land surfaces excluding open water, clouds, ice and snow. 3. rs0.47 and rs0.66 are obtained. The criteria:
first priority : If N > 5% when 0.01 £ r*2.1 £ 0.05, then rs0.47 = r*2.1/4, and rs0.66 = r*2.1/2
second priority : If N > 5% when r*3.8 £ 0.025, then rs0.47 = 0.01, and rs0.66 = 0.02
third priority : If N > 5% when 0.01 £ r*2.1 £ 0.10, then rs0.47 = r* 2.1/4, and rs0.66 = r*2.1/2
fourth priority : If N > 5% when 0.01 £ r*2.1 £ 0.15, then rs0.47 = r*2.1/4, and rs0.66 = r*2.1/2 pixels with r*2.1 > 0.15 are not used.
| Aerosol sub-category |
mean particle radius of number distribution
|
mean particle radius of volume distribution
|
standard deviation of ln(r)
|
column volume of particles per column cross section
|
single scattering albedo
|
| Water soluble * | 0.005 | 0.176 | 1.090 | 3.050 | 0.96 |
| Dust-like | 0.500 | 17.60 | 1.090 | 7.364 | 0.69 |
| Soot | 0.0118 | 0.050 | 0.693 | 0.105 | 0.16 |
| First Criterion (Q = scattering angle) | Second Criterion (ratio between single scattering path radiances, Lp) | Third Criterion (location, season) | then use this MODEL |
| 40° £ Q £ 150° | Lp-red/Lp-blue > 0.90 | globally | Dust |
| 40° £ Q £ 150° | Lp-red/Lp-blue < 0.72 | North America and Europe (year round) (100W-50E, 30N-70N) | Industrial/Urban Aerosol |
| South East Asia (year round) (105E-150E, 15N-45N) | Industrial/Urban Aerosol | ||
| Central America and Africa (May - Nov) (110W-50E, 0-30N) | Industrial/Urban Aerosol | ||
| South America and Africa (Dec-Apr) (110W-50E, 65S-0) | Industrial/Urban Aerosol | ||
| Central America and Africa (Dec-Apr) (110W-50E, 0-30N) | Biomass Burning | ||
| South America and Africa (May-Nov) (100W-50E, 65S-0) | Biomass Burning | ||
| Rest of the world | Biomass Burning | ||
| 150° £ Q £ 168° | Lp-red/Lp-blue > (0.90) - (0.01)(Q - 150°) | globally | Dust |
| 150° £ Q £ 168° | Lp-red/Lp-blue < 0.72 | North America and Europe (year round) (100W-50E, 30N-70N) | Industrial/Urban Aerosol |
| South East Asia (year round) (105E-150E, 15N-45N) | Industrial/Urban Aerosol | ||
| Central America and Africa (May - Nov) (110W-50E, 0-30N) | Industrial/Urban Aerosol | ||
| South America and Africa (Dec-Apr) (110W-50E, 65S-0) | Industrial/Urban Aerosol | ||
| Central America and Africa (Dec-Apr) (110W-50E, 0-30N) | Biomass Burning | ||
| South America and Africa (May-Nov) (100W-50E, 65S-0) | Biomass Burning | ||
| Rest of the world | Biomass Burning |
Once each grid cell has an aerosol model assigned to it, the surface reflectances (rs0.47 and rs0.66) plus the satellite observed reflectances (r*0.47 and r*0.66) are used again with the new model parameters to create a look-up table. From the look-up table, new values of optical depth (t0.47new, t0.66new) are obtained. Just like the Continental Model, the other models are based on real world measurements and assign values for the size and volume distributions, single scattering albedos, refractive indices, asymmetry factors and phase functions.
| Aerosol category |
mean particle radius of number distribution |
mean particle radius of volume distribution |
standard deviation of ln(r) |
column volume of particles per column cross section |
single scattering albedo |
| Biomass Burning |
rg(µm) |
rv(µm) |
s |
Vo(106cm3/cm2) |
wo (0.67 µm) |
| Accumulation | 0.61 | 0.130 | 0.500 | -2.4 + 45t | 0.90 |
| Stratospheric | 0.380 | 0.510 | 0.310 | 0.984 | 0.98 |
| Coarse | 1.0 - 1.3t | 6.0 - 11.3t + 61t2 | 0.69 + 0.81t | 2.4 - 6.3t + 37t2 | 0.84 |
| Industrial/urban |
rg(µm) |
rv(µm) |
s |
Vo(106cm3/cm2) |
wo (0.67 µm) |
| Accumulation 1 | 0.036 | 0.106 | 0.60 | -2.0+70t-196t2+150t3 | 0.96 |
| Accumulation 2 | 0.114 | 0.210 | 0.45 | 0.34-7.6t+80t2-63t3 | 0.97 |
| Stratospheric | 0.430 | 0.550 | 0.29 | 0.73 | 0.98 |
| Sea Salt | 0.990 | 1.300 | 0.30 | -0.16+4.12t | 0.92 |
| Coarse | 0.670 | 9.500 | 0.94 | 1.92 | 0.88 |
Dust |
rg(µm) |
rv(µm) |
s |
Vo(106cm3/cm2) |
wo (0.67 µm) |
| 1st mode | 0.0010 | 0.0055 | 0.755 | 6.0x10-6 | 0.015 |
| 2nd mode | 0.0218 | 1.230 | 1.160 | 1.0 | 0.95 |
| 3rd mode | 6.2400 | 21.50 | 0.638 | 0.6 | 0.62 |
So, we have: t0.47new, t0.66new, t0.47cont and t0.66cont.
Now what?