Benthic classification and IOP retrievals in shallow water environments using MERIS imagery

Confidence Measure of the Shallow-Water Bathymetry Map Obtained through the Fusion of Lidar and Multiband Image Data

“…As shown in the literature [23,25] and as desired, the discrete or line-type bathymetry product from ICESat-2 is now expanded to form a bathymetry map. Overall, for the western side of Andres Island, Great Bahama Bank, the bottom depth ranged from~2.0 to 8.0 m. This is consistent with our general understanding and depth retrievals from other observations and methods [61,62,67]. On average, the difference is~28.0% when compared with that derived from MERIS [62], which is consistent with those reported in the literature [26,28].…”

“…Overall, for the western side of Andres Island, Great Bahama Bank, the bottom depth ranged from~2.0 to 8.0 m. This is consistent with our general understanding and depth retrievals from other observations and methods [61,62,67]. On average, the difference is~28.0% when compared with that derived from MERIS [62], which is consistent with those reported in the literature [26,28]. However, obviously, there are erroneous outputs, where the bathymetry is~15.0 m for the Tongue of the Ocean (TOTO), which is known to be~2000 meters deep.…”

“…Such a result is expected because, as shown by Equation ( 9), R rs of shallow water is governed by at least 4-5 variables; thus, a ratio of R rs at two bands cannot resolve all unknowns, unless some of them are nearly constant or covarying with each other for a region of interest. However, even for this region in the Bahamas, as shown by Barnes et al [61] and Garcia et al [62], their H, IOPs, and bottom substrates vary spatially. This is further evidenced by the spatial variation of að440Þ (see Figure 3) derived from HOPE for the matchup data in Figure 2, where að440Þ varied from~0.03 to 0.09 m -1 , showing limited correlation (R 2 as 0.18 and an inverse relationship) with H Lidar .…”

“…Such high predictability is not always the case [28] (also see Table 2). Figure 2 shows matchup measurements (>3500 pairs) over the Great Bahama Bank, an environment with generally clear water and shallow depths [43,61,62] These values indicate that such a ratio at most explained <40% of the variance for this dataset, although the radiometric measurements came from the same image, and that the distance of the points (see the red dashed line in Figure 2(a)) spans~110 km. Most of the remaining variances (>60%) are likely from the water column and bottom properties (i.e., assuming that uncertainties from sun-sensor geometry and atmospheric properties can be omitted), and these variations could not be resolved from Z 2−3 .…”

Branching Algorithm to Identify Bottom Habitat in the Optically Complex Coastal Waters of Atlantic Canada Using Sentinel-2 Satellite Imagery

“…Inversion methods solve simultaneously for depth, bottom reflectance, and water IOPs but are highly sensitive to atmospheric correction errors and were developed for hyperspectral data (Lee et al, 1999). To date, they have had limited applications for multispectral data due to the high number of unknown parameters (absorption, scattering, depth, and bottom reflectance) relative to the number of spectral bands (Garcia et al, 2020). In the case of Sentinel-2, only three water-penetrating spectral bands are available at a 10-m spatial resolution, which is inadequate to characterize both the water column, depth, and bottom reflectance (Dierssen et al, 2019).…”