Stefanie Adam

Quantification of bio-physico-chemical properties of intertidal sediment at different spatial scales using remote sensing

Coastal-erosion models require accurate and detailed information on bio-physico-chemical properties of intertidal sediments. In this article, the potential of remotely sensed data at different spatial scales for the quantification of sediment properties was assessed. The focus on specific absorption features of water at 1450 nm, clay at 2206 nm and chlorophyll-a at 673 nm provided a physically based approach and interpretable results. The moisture content in unsaturated sediment could be accurately related to absorption (coefficient of determination r 2 = 0.89 and root-mean-square error of prediction (RMSEP) = 3% relative gravimetric moisture content) in laboratory and field measurements. The empirical model relating chlorophyll-a content with its field absorption (r 2 = 0.6) was successful on airborne sensor measurements. Though suitable in laboratory conditions, the absorption of shortwave infrared light by clay was not observed in field or airborne spectra. The empirical models showed clear dependency on the study area and sensor, suggesting more research on intercalibration of spectrometers and on factors influencing the interaction of light with sediment properties.

Chapter 9 sediment characterization of intertidal mudflats using remote sensing

Abstract In this paper an automated method for hyperspectral image classification is proposed. The method is based on a linear transformation of each spectrum in the hyperspectral cube. Different sediment types and land covers were classified using two dimensions of the transformed data space. The methodology is applied to hyperspectral images of the IJzermonding mudflat, acquired by the Compact Airborne Spectrographic Imager (CASI) in 2001 and 2003. Comparable classification results were obtained using a standard classification method employed in hyperspectral image processing. The similarity between classification results was 82 and 65% for the images of 2001 and 2003, respectively. The superiority of the proposed user-friendly method lies in its autonomy, reliability and objectivity. The proposed method uses the underlying statistical information of the data set, while the standard method is mainly based on expert knowledge.