Search

The detectability of adjacency effects (AE) in ocean color remote sensing by SeaWiFS, MODIS-A, MERIS, OLCI, OLI and MSI is theoretically assessed for typical observation conditions up to 36 km offshore (20 km for MSI). The methodology detailed in Bulgarelli et al. (2014) is applied to expand previous investigations to the wide range of terrestrial land covers and water types usually encountered in mid-latitude coastal environments. Simulations fully account for multiple scattering within a stratified atmosphere bounded by a non-uniform reflecting surface, sea surface roughness, sun position and off-nadir sensor view. A harmonized comparison of AE is ensured by adjusting the radiometric sensitivity of each sensor to the same input radiance. Results show that average AE in data from MODIS-A, and from MERIS and OLCI in reduced spatial resolution, are still above the sensor noise level (NL) at 36 km offshore, except for AE caused by green vegetation at the red wavelengths. Conversely, in data from the less sensitive SeaWiFS, OLI and MSI sensors, as well as from MERIS and OLCI in full spatial resolution, sole AE caused by highly reflecting land covers (such as snow, dry vegetation, white sand and concrete) are above the sensor NL throughout the transect, while AE originated from green vegetation and bare soil at visible wavelengths may become lower than NL at close distance from the coast. Such a distance increases with the radiometric resolution of the sensor. It is finally observed that AE are slightly sensitive to the water type only at the blue wavelengths. Notably, for an atmospheric correction scheme inferring the aerosol properties from NIR data, perturbations induced by AE at NIR and visible wavelengths might compensate each other. As a consequence, biases induced by AE on radiometric products (e.g., the water-leaving radiance) are not strictly correlated to the intensity of the reflectance of the nearby land., Highlights • Quantification of adjacency effects for a wide range of coastal environments • Comparison of adjacency effects for data from various ocean color sensors • Relevance of adjacency effects as a function of the SNR of satellite sensors

The presented adjacency correction algorithm is based on the use of the point spread function (PSF) which allows calculating the contribution of reflections from the nearby pixels to the apparent radiance of the target. The analytical expression of the PSF for an arbitrary stratified atmosphere is obtained in the approximation of primary scattering, whereas the full equation of radiative transfer is used for the estimation of the radiance reflected from the surface. The algorithm is sensor independent and can be applied for processing images of water basins with arbitrary shape of the shore line and under different geometries of observation. The program using this algorithm is included in Modular Inversion Program — MIP (Heege et al., 2014) for processing of satellite images on a routine basis. Examples of processing results are presented in the paper.

An extensive comparison between radiative transfer codes in the atmosphere–ocean system is proposed and performed. The full angular radiance distribution at several optical depths in atmosphere and water is intercompared on a set of idealized problems designed to study codes’ accuracy in modeling separate, specific system features. In-water profiles of upwelling nadir radiance, upwelling and downwelling irradiance are intercompared for a realistic case extracted from an experimental data set. Two models are involved: the FEM numerical algorithm, based on the finite element method, and the P HO-TRAN 3D backward Monte Carlo code. The results show an optimal agreement between the codes under any condition. Codes’ relative differences are always lower than the estimated statistical error on the P HO-TRAN results.

The well-known expression of Cox and Munk for the bi-directional reflectivity of a rough water surface tends to infinity for reflection angles close to the horizontal direction. This creates evident difficulties in applying it in numerical radiative transfer schemes. Detailed analysis of the process of reflection from a non-flat surface shows that the singularity arises from some contradictory implicit assumptions adopted in the theoretical derivation of the bi-directional reflectivity expression. In the present paper, these assumptions are modified in order to provide for consistency and a bi-directional reflectivity expression applicable for all possible reflection angles is derived. An analytical form of Fourier expansion of bi-directional reflectivity in azimuthal variables is also obtained.