Compositions of urban surface materials and their use for deriving urban structure types in 30m hyperspectral remote sensing data

Kurzfassung

Future spaceborne imaging spectrometers such as EnMAP allow for the monitoring the Earth’s surface and overcome the time- and cost critical limits of airborne based hyperspectral sensor systems. However, the spatial resolution of EnMAP is 30 m, which is in many cases not sufficient to provide essential information on urban areas that are characterized by a dense network of spectrally different objects. Given the spectral resolution of the EnMAP, this leads to complex mixtures. At the same time, the spectral resolution is an advantage and opens up new opportunities for urban analyses. Despite the spectral and spatial complexity of urban settlements, characteristic materials can be found in the composition of similar urban structures such as industrial areas, residential areas and open spaces. This study presents a new concept of spaceborne imaging spectroscopy data analyses. The presented strategy presumes that the composition of material mixtures changes gradually across urban spaces and thus, gradient analyses applied to urban areas could be a suitable methodology to derive surface material compositions. This information is important for urban weather and climate modeling and specifies the requirements to better describe urban canopy models. This study focuses on 1) the adaptation of gradient analysis from vegetation ecology to analyze gradual material transitions in the composition of urban surface materials and 2) the mapping of these material gradients in simulated EnMAP data of Munich, Germany, to identify characteristic material compositions. The Detrended Correspondence Analysis was used to analyze the similarities of material composition for a number of systematically distributed samples in a detailed material map. Results demonstrate the cooccurrences of urban materials along two urban gradients. The determined gradient scores serve as an indicator for the composition of urban surface materials in the respective sample. The resulting models were subsequently applied on the image data to map the gradients. These maps reveal a rural-to-urban transition and allow for a differentiation of urban structures with regard to the degree of imperviousness from the first gradient, as well as the differentiation of subtleties of relatively similar urban structures based on the patterns resulting from the second gradient. The results demonstrate that even in spaceborne hyperspectral remote sensing data detailed structural information is retained, which for the first time enables the identification of characteristic material compositions in very complex spectral mixtures.