The sediment dynamics along the Belgian shoreline, studied with airborne imaging spectroscopy and LIDAR

Recent indications and model predictions leave little room for argument that the sea level worldwide is rising and that the rising will continue and will even be accelerated in the next decades. Many factors have been identified in causing the sea level rise, with warming of the oceans and melting of land ice as the dominant players. The anthropogenic influence in this has recently been confirmed by the Intergovernmental Panel on Climate Change (IPCC). The rising of the sea level goes hand in hand with climate changes that are expected to cause more storminess along the western European coastline, as well as along many other coastlines. Even without an intensified storminess and an accelerated sea level rise, the western European coastline and more in particular the Belgian shoreline, needs permanent attention as this shoreline is highly prone to erosion. The huge socio-economic value of the Belgian shoreline, which asks for a stable and save shoreface, is not compatible with the dynamic nature of the system. Therefore man has strived to keep the shoreline at its position for many decades. As a result of all this, the regular monitoring of the Belgian shoreline is absolutely necessary. In this thesis, two state-of-the-art remote sensing techniques have been explored to monitor the Belgian shoreline in the period 2000 - 2006. Airborne LIDAR or laserscanning is a well-known technique that allows making accurate Digital Terrain Models (DTMs) of the shoreface. Successive DTMs were used to calculate the amount of sediment that was eroded or deposited. As a novel technique, airborne hyperspectral remote sensing or airborne imaging spectroscopy, was applied to classify the sediment of the beach and foredunes in seven sand type classes. Several classification strategies were tried out; a comparison was made between the non-statistical Spectral Angle Mapper (SAM) and a statistical classifier based on Linear Discriminant Analysis (LOA), making use of AISA-Eagle imagery. The best classification results were obtained applying LOA in combination with a feature selection based on Sequential Floating Forward Search (SFFS). The latter is a band selection technique that chooses the sub-optimal combination of spectral bands from the hyperspectral data in order to obtain the best classification accuracy. Classifications were performed with the original bands as well as with wavelet coefficients. The latter significantly enhanced the classification accuracy obtained. The combination of LOA with SFFS resulted in an overall classification accuracy of 82%, using three wavelet coefficients. Replacing the LOA with the nonstatistical SAM algorithm reduced the overall classification accuracy to 74%, using all bands or wavelet coefficients. Tests pointed out that, as a result of the limited number of training samples available, using more than two to three bands or wavelet coefficients did not result in higher classification accuracies. HyMap data, featuring 126 bands in the Visible till Shortwave Infrared part of the spectrum, were used to demonstrate that the Visible and the Near-Infrared part of the spectrum outperform the Shortwave Infrared part for classification purposes in this application. As a side-step, airborne hyperspectral data acquired over the Molenplaat in the Westerschelde, were classified applying the LOA and SFFS algorithms. The aim of this study was to investigate whether this classification methodology could be used to identify sediment habitat types on tidal sand banks. High classification accuracies were obtained for the water content (88%), the median grain size (88%) and the chlorophyll-a concentration (84%). The organic matter content scored somewhat lower but still reached 80% overall accuracy. The four parameters classified allowed to define the general sediment habitat types which could consequently be used in more detailed biological or sedimental studies.However, the main goal of this thesis was the applica