W artykule zostały przedstawione wyniki prac nad utworzeniem metody określania zmian użytkowania ziemi z wykorzystaniem cyfrowej analizy wysokorozdzielczych zdjęć satelitarnych. Przebadano różne typy danych satelitarnych – Landsat, SPOT, IRS, jak również różne sposoby tworzenia map różnicowych, bazujące na odejmowaniu wartości spektralnych oraz na klasyfikacji danych. W wyniku tych prac sformułowano wnioski dotyczące optymalnej metody określania zmian użytkowania ziemi na przykładzie fragmentu aglomeracji warszawskiej.
Results of complex studies aimed at preparation of optimal method for detection of land use changes on the basis of digital analysis of multitemporal satellite data were presented in the article. Various types of satellite images, characterized by different ground and spectral resolutions – Landsat TM, SPOT HRV, IRS 1C and QuickBird – were used in these studies. At the preparatory phase methods of geometric and radiometric correction, as well as normalization methods, were analyzed. As a result of these works normalization method, optimal for comparing multispectral, multitemporal images, based on stable in time land cover types, was determined. At the second stage of the works methods of creating change detection maps on the basis of multitemporal satellite data were thoroughly analyzed. The following approaches were studied: Method of subtracting radiometric values in particular spectral channels Method of subtracting normalized difference vegetation index Method of comparing principal component images Classification of multitemporal dataset Method of comparing two classification images Method of analysis of change vectors As a result of the performed research works usefulness of the above mentioned methods was assessed, considering precision of distinguishing changes of land use within urban areas. In case of the methods based on subtraction of original and transformed images the work was concentrated on finding optimal threshold levels for separating changed and non-changed areas. It was found, that values of threshold level depends on type of satellite data and on spectral band. In case of the methods based on classifications the research works were aimed at determination of the method, which enables to distinguish the detailed land use/land cover classes, especially within urban areas, with the adequate accuracy. As a result of comparison of various methods it was found, that the method of independent classification of satellite data collected at different dates and their comparison is the most effective method for creating change detection maps. For urban areas, characterized by very complex land use pattern, method of object-oriented classification, which allows to take into consideration spectral and non-spectral image features, proved to be optimal. Application of this method for Warsaw area enabled to distinguish four levels of urban density, characterized by various contribution of anthropogenic objects and green areas. Comparison of two classification maps, prepared with the use of this method, allowed for preparation of change detection map, presenting character of changes of land use and land cover within Warsaw area. This map represents changes concerning transition of non-urban areas to built-up land and transformations related to urban density.