High resolution imagery form early Corona missions – an attractive data source for the assessment of settlement dynamics?

As data from the American espionage satellite program Corona, declassified in the year 1995 by President Clinton, covers a wide range of the earth’s surface with high resolution images of a ground resolution between 25 ft (KH 1-4) and 6 ft (KH 4b) it seems to be attractive source in the assessment of landscape dynamics. Reaching back to the early 1960s, Corona images seem to offer the opportunity to bridge more than 40 years of land cover changes and to allow the recognition of even slow processes or small scale and one-time events. Additional, since one scene covers more than 1600 km² this source promises an efficient geometrical correction of the images compared to aerial photography.
Our study was performed on a north-west to south-east orientated transect at western Germany and Luxemburg low mountain range regions. Test sites in Luxembourg, France and different federal states in Germany are selected. The study was conducted employing a set of images taken on September, 30th 1962 by the Corona system KH4 with a resolution of approximately 8 meters and nearly unimpaired by clouds. To assess the dynamics of detected landuse changes and to identify regional driving forces behind them a comparison between the mentioned Corona data set and current images of SPOT 5 HRG and ASTER sensors was performed.
Due to the use of panchromatic images, different unsupervised and supervised classification methods added by segmentation and object-orientated approaches had to be investigated. To balance different illumination of certain regions within one image, caused by clouds, cloud shadows or terrain, images had to be segmented before classifying. As one experience from the comparison of different sites, optimal regional recording conditions for both images play a crucial role for achieved quality of results. To reach a satisfying accuracy level, in some cases only a pre- classifying, followed by a visual post-classification, was feasible. Though in most cases results had to be manually improved they are usable for the detection of urban development and the classification of the extent and spatial shape of the growth in settlements.