Porównanie klasyfikacji wieloczasowych zdjęć satelitarnych MODIS

Since the 1970’s remote sensing enable constant monitoring of land cover and land use, which are considered as the most crucial environmental data. Obtaining this information at global, regional and local scales, becomes the goal of many research and application programs and has allowed for the deeper understanding of the entire Earth system. In December 1999 NASA launched the EOS Terra Satellite, followed in 2002 by Aqua, both equipped with MODIS (Moderate Resolution Imaging Spectroradiometer). Thanks to its technical specification and free distribution of the majority of its products, MODIS is considered to be the most important sensor for the global vegetation mapping. Although has been originally designed for large-scale analysis, MODIS is also used in many regional research programs. This paper presents results of two approaches of multitemporal land cover classification of MODIS data. For the study polygon of 22100 square kilometres, situated in the western Poland, four single day surface reflectance data sets, of spatial resolution of 250m and 500m, were collected for the year 2007 – one for spring and autumn, and two for summer. In the first approach supervised land cover classification was conducted for each set of single day data separately. On the basis of obtained results, final classification was elaborated as an effect of analyse of the sequence of changes for each pixel. In second approach, independent classification was conducted for the aggregation of all possessed data. The accuracy of all classifications’ results was checked against Corine Land Cover 2000 database using 4200 randomly distributed points. Obtained statistics show that comparing with single-day classifications as well as with classification of aggregated data, multi-temporal approach based on the analysis of sequence, enabled crucial improvement of accuracy of the classification.