Mapping of submerged aquatic vegetation using hyperspectral airborne data and a physically based process chain

Thomas Heege, Sabine Miksa
DLR Oberpfaffenhofen, Münchner Str. 20
D-82230 Weßling, Germany
thomas.heege@dlr.de

Nicole Pinnel
Technical University of Munich, Limnological Station
Hofmark 3, D-82393 Iffeldorf, Germany

Viatcheslav Kisselev
Institute for Informatics and Automation of the RAS
St. Petersburg, 14 line, 39, Russia 199178

Peter Hausknecht
HyVista Corporation
22 Delhi Road, N. Ryde, 2113, NSW, Australia

ABSTRACT

Mapping the littoral zone and the submerged vegetation is of prime importance for the ecological evaluation of natural waters. Remote sensing techniques are efficient for such mapping tasks, if the retrieval algorithms and processing methods are robust, transferable and independent from additional ground truth measurements. The Modular Inversion Program (MIP) follows this concept. It is a processing tool designed for the recovery of hydro-biological parameters from multi- and hyperspectral remote sensing data. The architecture of the program consists of physical inversion schemes that derive bio-physical parameters from the measured radiance signal at the sensor. Therefore, the coherence and validation of each processing level is traceable by optical measurements. Program modules exist for the retrieval of aerosols, sun glitter correction, atmospheric and water surface corrections, retrieval of water constituents among others. Mapping the bottom coverage in optically shallow waters is performed using program modules, which calculate the bottom depth, the bottom coverage and the macrophyte species probabilities in an iterative manner.

The processing system has been tested using multi-temporal data collected with the hyperspectral airborne scanner HyMap at Lake Constance, Lake Starnberg and Rottnest Island/Australia. For the purposes of validation, subsurface reflectance spectra measured with Ramses radiometers, in situ probes of water constituent concentrations and extensive in situ observations are used.

Last Update: 2005-05-2