Analyses of spatial and temporal changes of the urban environment using multi- and hyperspectral data

A relatively small part of the world is occupied by built-up area, but the spatial pattern and growth of these areas have a large impact on environmental, economical and social processes. Although settlements are an essential land use type, until this decade remote sensing methods were mainly used to map and characterize other land use and land cover categories.

With the appearance of very high spatial and spectral resolution data, and new methods, remote sensing analyses of urban changes have moved into a new direction.

These new systems and methods were used in the urban area of Szeged, which is situated in the southern part of Hungary and covers an area of 50 km2. About 160 thousand people are living in Szeged. The city has a large variety of building densities ranging from parks, high rise apartment buildings to a dense city center.

The urban land cover was investigated by satellite and aerial remote sensing methods. To analyse and map land cover and land use changes, aerial photographs, Corona, LANDSAT, SPOT and IKONOS images were used in combination with a new aerial hyperspectral imaging system called AISA Dual. These data sources enabled us to study the land use in a 35 years time frame (1972-2007), at different spatial (1-120 m) as well as spectral resolutions (up to 359 bands). The spatial and temporal changes in the city were analysed using spectral indices and multi- and hyperspectral classification.

In the AISA Dual system, for the first time two scanners are used together to acquire imagery in a spectral range of 400-2450 nm, with a programmable spectral resolution of 5 nm. Since the used dataset was acquired during one of the first flight campaigns with the instrument, quite some technical problems had to be solved and new technological skills had to be developed. Large numbers of contiguous spectral bands with a high spectral resolution allow the identification of chemical and physical material properties. This enables a more accurate classification of land cover types in urban areas. The preprocessing of the data included correction of system errors, geometric and radiometric corrections and the removal of bands with noise. The preprocessed bands were mosaicked and a spatial subset of the city was created. From the subset, a land cover classification was generated and different indices were calculated.

The poster will show the workflow to create the land cover/land use classification. The temporal changes of the natural and buildup areas in the city were analyzed using different spectral indices. The quantitative analysis of the 35 years of data indicates some important tendencies: (1) former unused green areas have been transformed to built-up areas, (2) areas with high rise apartment buildings have become greener and (3) in suburban residential regions the green and bare areas decreased in size. Our analyses show that multi- and hyperspectral methods - combined with high spatial resolution - are suitable to identify different urban surfaces.

The poster summarizes the first experiences with the new AISA Dual sensor and shows the first applications of hyperspectral analysis in urban environments in Hungary.