In the field of optical remote sensing (400 to 800 nm wavelength) in natural waters, physically based inversion methods can currently be applied only when bottom effects are negligible. In shallow water areas, the remotely sensed signal in the visible range is strongly influenced by the bottom albedo, which must be taken into account for the development of remote sensing algorithms. The aim of this study was to investigate the variability of the remotely sensed signal and the feasibility and uniqueness of inversion model solutions when bottom albedo is not negligible. A well-established and validated forward model (Hydrolight 3.1) was applied to inland waters using measurements of water constituents and light field parameters in deep and shallow water areas from the test site Lake Constance. The shallow water measurements were also used for the derivation of specific bottom reflectance spectra. By using this adjusted forward model, a large number of spectral measurements was simulated, where the optical properties of the water and the bottom were varied within the range of the in-situ values at Lake Constance, but also below and above this range to cover a more general range of concentrations and to extend the study to a wide number of case-2 waters. The dependence of the under water light field on the concentration of phytoplankton and suspended matter, gelbstoff absorption, bottom reflectance, bottom depth, surface wind speed, solar zenith angle, and viewing angle was parameterised. Further parameters including phase functions and specific optical properties of water constituents were kept constant at values typical for Lake Constance. A set of analytical equations for calculating water and bottom properties was developed and implemented in a public-domain software (WASI) to provide a fast and user-friendly tool of forward and inverse modelling of optical data. A new inversion technique based on the analytical parameterisations was developed to estimate the concentrations of the water constituents, the bottom depth, and the coverage of bottom types in shallow water. The errors of the model were studied depending on the water constituent concentrations, bottom depth, and bottom reflectance. The effect of multi-parameter inversion on error propagation was also investigated as well as the influence of sensor characteristics like signal noise, radiometric, and spectral resolution. The new methodology was validated using in-situ data measured in Lake Constance.
More information is available at http://www.sub.uni-hamburg.de/opus/volltexte/2005/2325/