A Note on COART and Its Input

By Zhonghai Jin

Brief Description

The coupled Ocean Atmosphere Radiative Transfer (COART) model is established on the Coupled DIScrete Ordinate Radiative Transfer (Coupled DISORT or CDISORT) code, which was developed from DISORT, a publicly distributed software for radiative transfer by NASA. The major difference between CDISORT and DISORT is that CDISORT considers the refractive index change of the media (e.g., at air-water interface). Therefore, CDISORT can be applied to radiative transfer problems in the coupled atmosphere-ocean system, while DISORT has to consider the ocean surface as boundary. The detailed theoretical derivation of the equations and solutions was given in Jin and Stamnes [1994]. However, the Jin and Stamnes solution was for flat ocean. Calm ocean is very rare. Therefore, CDISORT was further modified late to include ocean surface roughness based on the Cox-Munk surface slope distribution [Jin et al., 2006]

Since the refractive index and the surface roughness are included in the radiative transfer solution in CDISORT, COART treats the ocean just as additional "atmospheric layers" but with significantly different optical properties. Hence, the ocean surface albedo can be calculated as an output parameter instead of an input as required in DISORT. However, it reduces to a conventional atmospheric radiative transfer model (i.e., atmosphere-surface system) by setting the ocean depth as zero and providing the surface albedo. This tool is designed to simulate radiance (including water-leaving radiance) and irradiance (flux) at any levels in the atmosphere and ocean. See the examples of model calculations.

The Input/Output

1. General Inputs and Calculation Types

Most input parameters are straightforward from the web input form. One of the parameters which should be treated with caution is the Spectral Resolution when you choose to calculate the integrated radiation within a spectral region. Because the atmospheric absorption is based on LOWTRAN7 band model, which has a resolution of 20cm-1 (approximately 0.5nm at wavelength 500nm; 8nm at 2.0um; and 30nm at 4.0um), to obtain the maximum fidelity is to use the same resolution as LOWTRAN. To set a resolution less 20cm-1 would be overkill, while setting resolution too coarse (say 0.3um in the visible) is also pretty useless. The resolution becomes important at wavelengths where atmospheric gas absorption is strong. However, using resolution as small as 20cm-1 (actually not necessary in most spectra) would take too much time if the spectral range to be integrated is wide. Therefore, the option for integrated radiation is recommended only for narrow band calculations (e.g., in channels of AVHRR, SeaWiFS and MODIS) but not for broadband unless you can afford to wait.

The option, "Broadband shortwave flux", is specifically provided for fast calculation of the broadband solar radiation (0.20-4.0um). A fixed 26 band model is used and the k distribution for atmospheric absorption in each band was generated from HITRAN-2000 database. Because the HITRAN 2000 has larger atmospheric absorption than its previous version which LOWTRAN7 was based on, the downwelling surface solar flux could be several W/m2 smaller than that based on LOWTRAN and integrated over the same spectral range with proper resolutions.

When computation time is a concern for you, it is recommended to select the input option - "use reduced number of atmospheric layers", which will reduce the computation time by a third by sacrificing less than 1% of accuracy in most spectra. But this option is not recommended for wavelenths less than 0.35(um), for thermal radition and for the strong atmospheric absortion bands. You can try and compare with results based on full resolution of atmospheric profiles.

Radiance calculations require significantly more time than the irradiances. The time required for the band radiance calculations will increase proportionally with the integration steps or the spectral resolution. If you select one of the listed satellite channels, the filter function for that channnel is applied to the band radiance and irradiance automatically. You can specify one or more arbitrary directions or many regular directions for radiance output. If you input more than one output angles, separate the numbers by space. See the definition for radiance output direction here.

Because of the immediate absorption of water, no calculation is needed below the water surface for IR radiation. However, the bidirectional emissivity of ocean surface is calculated interactively based on the surface rooughness and the optical properties of water at each wavelength. You can input the surface (land or water) temperature for thermal radiation. If you leave it empty, the surface temperature will be taken from the standard atmospheric profile.

2. Inputs for Atmosphere

If you don't like the default amounts of trace gases (i.e., total precipitable water, Ozone, CO2, and CH4) defined in the standard atmospheric models, you can change them, but their vertical distributions still follow those in the selected atmosphere.

Several standard aerosol models are listed for selection. The tropospheric and stratospheric aerosol models are based on those in MODTRAN. The total aerosol loading is constrained by your inputted visiblity, column Aerosol Optical Thickness (AOT) at 500nm, or AOT at 550nm. You also have an option to directly input the aerosol optical properties (.i.e., AOT, single scattering albedo (SSA), and asymmetry factor (g)) at up to 10 wavelengths. The wavelengths must be in micrometer(um) and in increasing order. If you input any aerosol optical property at only one wavelength, this optical property will be considered as constant for all wavelengths. Otherwise, the optical property at the calculation wavelength will be obtained either by interpolation if the calculation wavelength falls between, or by using the inputted value at the nearest input wavelength if the calculation wavelength is beyond the range you specified. It is not necessary to fill all the elements in the table because the undefined optical properties will be fit in with the selected boundary layer aerosol model. If you checked the box to upload the aerosol scattering phase function, the asymmetry factor specified by either the aerosol model or your input in the table will be ignored. The vertical aerosol profile also follows that in the specified aerosol model.

There are three types of clouds are listed for selection: water cloud, spheric ice cloud, and non-spheric ice cloud. Inputs for cloud include the bottom and top heights (km) above surface, effective radius (RE) and either liquid (ice) water path (LWP) or cloud optical depth (TAU). But RE (in um) represents the effective size (Fu, 1996) for non-spheric ice. When "No cloud" (default) is selected, all inputs for cloud are ignored. The optical properties of water cloud are based on Hu and Stamnes (1993); the spheric ice cloud is also based on mie calculations; the non-spheric ice cloud is based on Fu's (1996) parameterization for Cirrus clouds.

3. Inputs for Ocean

Although the model can take arbitary varieties of oceanic constituents (as long as you can provide the optical properties for each), the input on the web is limited. By default, the CDOM absorption in ocean is correlated with Chl as the parameterization in Morel (1991) and Gordon-Morel (1983). But you can override this if you input the DOM absorption at 440nm (a440DOM (m-1)). Then DOM absorption at wavelength, lamda, will be
aDOM(lamda) = aDOM(440) * exp[-0.014(lamda-440)]
You can also override the default parameterization for ocean absorption (Morel, 1991) by inputing the absorption for the soluble (DOM) and particulate materials directly.

The marine particle scattering coefficient, bP(lamda), is calculated as

bP(lamda) = b0(550/lamda)n*[Chl]k
Where Chl is the chlorophyll concentration. bo, n, and k are parameters you need to input. These three numbers are [0.3, 1.0, 0.62] in the Gordon-Morel(1983) parameterization; [0.416, 1.0, 0.766] in Morel-Maritorena(2001) for upper ocean case 1 waters; and about [0.45, 0.6, 0.62] (default) for case 2 waters in the Gould(1999) measurements. However, you can input any scattering coefficient for ocean particulates at a particular wavelength (lamda) by varying the bo, n, and k, for example, setting n=0 for wavelength independent partice scattering, and k=0 for chlorophyll independent scattering.

There are several phase functions for ocean particle scattering listed for selection. Three of them are based on the Petzold's (1972) measurements for the three types of sea waters. One is based on Morel-Maritorena(2001) for case 1 waters, which has backscattering ratio (bb/b) depending on Chl and wavelength automatically but smaller than other phase functions listed. If you select the Fournier-Forand (1999) phase function, you have to specify bb/b (in range of 0-0.5).

The measurements and parameterizations for ocean particle scattering and absorption are highly diverging. On the other hand, the water-leaving radiance, surface albedo in visible spectrum, and radiance and irradiance in the ocean are highly dependent on those ocean optical properties. Therefore, if your calculation is one of those sensitive to ocean optics, you have to pay attention to those parameters that define ocean optical properties. The various input parameters above have given you the flexibility to input any ocean optical properties.

Surface roughness and the sun glint it induces are included as default, but you can shut down these by selecting the "flat ocean" option or no wind input. See how the wind affect the ocean surface albedo and sunglint here?

If you input the ocean depth as 0, all the input/output setting for ocean will be ignored. That, in fact, reduces the atmosphere-ocean system to the atmosphere-land system. Then the bottom albedo will be assumed as the Lambert surface albedo.

If you still have question after reading this note, please send email to me. But I can't promise you that I will answer every question due to my time limit. If you are interested in any coorporation, have any good comments, or find any bugs, welcome to send me an email too.

Reference for COART:

Jin, Z., T.P. Charlock, K. Rutledge, K. Stamnes, and Y. Wang, Analytical solution of radiative transfer in the coupled atmosphere-ocean system with a rough surface. Appl. Opt., 45, 7443-7455 (2006).

(Last updated on January 11, 2008.)

Back to COART model online