Investigation of the potential of hyperspectral sensors for bathymetry applications using airborne HyMap data from Lake Constance

Abstract

The potential of hyperspectral sensors in monitoring the bottom depth in shallow waters is investigated by numerical simulations and by analysis of hyperspectral data. The theoretically expected accuracy of bottom depth determination is studied by combining forward and inverse calculations for different conditions of the water medium and the bottom type. The parameters defining these conditions are changed in value starting from a clear-water scenario to various other optical situations. The spectral range of all calculations is 400 to 900 nm and the spectral sampling interval is 1 nm. All simulations are done with the Water Colour Simulator (WASI) software which is optimised to modelling and analyzing optical in situ measurements in aquatic environments. A physically based method for the inversion of hyperspectral remote sensing data is applied to data of the airborne sensor HyMap. The data were acquired during the HyEurope flight campaign in June 2004. HyMap spectra are inverted by adjusting modelled top-of-atmosphere radiance spectra to the measured ones by fitting the concentrations of 3 aerosol types (rural, maritime, urban) and 3 water constituents (suspended matter, chlorophyll, Gelbstoff). The determined concentrations of aerosols and water constituents are used to correct the HyMap images for atmospheric and air/water interface effects and for the angular dependency of the water body reflectance. In the final step bottom depth is determined for all pixels, with fixed concentrations of all water constituents at the deep-water average values. All data inversion and image processing is performed by use of the Modular Inversion and Processing System (MIP). These results are compared with bottom depth maps derived from echosounding measurements and aerial photography, and an assessment of error sources is given. WASI simulations proved that an accuracy of bottom depth better than 10 % is theoretically possible for low concentrations of water constituents between 2 and 8 m, independent from the bottom type. Errors in the concentration of the water constituents Gelbstoff and suspended matter may have a significant negative effect on the accuracy of bottom depth (ZB) determination if considered constant during the inversion. The results from the inversion technique by MIP showed that it is capable to estimate the bottom depth values from the HyMap image with a root mean square error of 20 cm after correction of systematic errors. The results on error propagation from the WASI simulations suggest that fixing of the water constituents during the inversion with MIP may be the main error source for the systematic error observed in the determination of bottom depth.