L1B Clear Flag Methodology

L. Strow
Updated: Jun 18, 2001
Document URL: http://asl.umbc.edu/pub/airs/exercises/clearmethod.html

Definitions

Our computed radiance, R_c is given by

Rc = eB(Ts) tatm + ó õ B dt

where T_s is the NCEP sea surface temperature, and t_atm is the atmospheric transmittance computed with the forward model using the NCEP model fields for temperature and water vapor.

RsC = eB(Ts)

is the radiance if the atmospheric transmittance was zero, and

Robs = eB(Tst) tatmt + ó õ Bt dtt,

is the observed radiance, where the superscripts denote ``true'' quantities. (R<sub>c</sub> - R<sub>s</sub><sup>c</sup>) is our estimate of the ``atmospheric'' part of R_obs, which we subtract from R_obs in order to estimate just the sea surface emission part of R_obs, or eB(T^t_s). T^t_s(est) denotes our estimate of the true sea surface temperature. Therefore, our final estimate of the sea surface temperature from an observed radiance is

BT é ê ë Robs - (Rc - Rsc) e ù ú û = Tst(est)

where BT[] turns radiance into brightness temperature. (R_c -R_s^c) is very small for 2616 cm^-1(few tenths of a K), but up to 4K or more at 900 cm^-1.

Clear Flag

I compute T^t_s(est) for the 900 and 2616 cm^-1channels. The clear flag is derived from FOVS that return true for the following three tests:

1. Is Mitch Goldberg's L2 flag (fov_clear_flag) set?
2. | T^t_{s,2616 cm^-1}(est) - T_s | < 1K
3. | T^t_{s,2616 cm^-1}(est) - T^t_{s,900 cm^-1}(est) | < 0.5K

Since we know the true sea surface temperature Test-#2 above is quite good. However, in practice T_s will have 0.5 - 1K errors, and Test-#3 might be quite important, although it is less sensitive. Test-#2 will help get rid of outliers that could make it through Test-#3.