Abstract
The use of satellite remote sensing for operationally assessing urban land cover changes is generally constrained by the limited spatial resolution of the image series which cover sufficiently long time periods. Only SPOT HRV‐Panchromatic (PAN) images have in fact a sufficient spatial resolution (10 × 10 m) and are now available for more than 12 years. In most cases, however, the utility of these data for land cover change assessment has been limited by their poor spectral information content. A new multi‐step methodology is currently developed to extract the information related to urban land cover changes from SPOT HRV‐PAN multitemporal images. The methodology is divided in two main phases. During the first phase (change assessment) directional Principal Component Analysis (PCA) is applied to each SPOT HRV‐PAN image pair in order to identify the areas where change had occurred. The second phase (change interpretation) is performed by applying the following steps:
-
Multi‐scale extraction of textural information from each SPOT HRV‐PAN image by the computation of parametric and nonparametric statistics within moving windows of different sizes;
-
Fuzzy classification of the multi‐scale textural images produced;
-
Maximisation of the multi‐scale fuzzy membership grade images to produce multitemporal land cover maps.
-
Change interpretation by comparison of the multitemporal land cover maps for the areas identified as changed in step 1.
The methodology was applied to three SPOT HRV‐PAN images taken in 1988,1994 and 1998 over the Chinese area around Xiamen, where strong urban expansion processes had taken place. The results obtained testify to the efficiency of the method, which correctly assessed and interpreted the main land cover changes occurred in the area during the study decade.
Notes
Corresponding author.