Remote sensing of ocean color: Assessment of the water-leaving radiance bidirectional effects on the atmospheric diffuse transmittance for SeaWiFS and MODIS intercomparisons

The portion of the radiance exiting the ocean and transmitted to the top of the atmosphere (TOA) in a particular direction depends on the angular distribution of the exiting radiance, not just the radiance exiting in the direction of interest. The diffuse transmittance t relates the water component of the TOA radiance to that exiting the water in the same direction. As such t is a property of the ocean–atmosphere system and not just the atmosphere. Its computation requires not only the properties of the atmosphere but the angular distribution of the exiting radiance as well. The latter is not known until a determination of the water properties can be made (which is the point of measuring the radiance in the first place). Because of this, it has been customary to assume an angular distribution (uniform upward radiance beneath the water surface) in the computation of t, which is referred to as t⁎. However, it is known that replacing t with t⁎ can result in an error of several percent in the retrieved water-leaving radiance. Since the error depends on sun-viewing direction, this error could be particularly important when water-leaving radiance from two or more sensors in different orbits are compared. Even given an estimate of the angular distribution of the water-leaving radiance, computation of t using full radiative transfer theory in an image processing environment is not practical. Thus, we developed a first-order correction to t for bidirectional effects in the water-leaving radiance that captures much of the variability of t with viewing direction. The correction computed across a SeaWiFS scan line shows that a t⁎-induced error of as much as 5–6% could occur near the edges of the scan; however, limiting the scan to polar viewing angles (θ) < 60° reduces the error to ~ 1%. Direct application to SeaWIFS and MODIS (AQUA) suggests that the bidirectionally-induced error in the diffuse transmittance will result in an error less than about 1% in the comparison of their normalized water-leaving radiances, as long as θ is less than about 60°. We conclude that, given this constraint, the normalized water-leaving retrievals from these two sensors at a given location can be merged without regard for the bidirectionally-induced error in the diffuse transmittance, as the resulting uncertainty is well below that from other sources. It is important to note that this result is likely to apply to any other polar-orbiting sensor with equatorial crossing times (similar to SeaWiFS and MODIS) between 1030 and 1330 h (local time).