METHOD OF STEPWISE CLASSIFICATION OF SATELLITE IMAGES FOR THE THEMATIC MAPPING OF FOREST COVER

Abstract

Introduction. Question of unification and automation of the forest cover decoding are of great importance in connection to increasing number of used satellites. Regional estimates of forest cover are significant for solving problems of monitoring and validation of thematic maps by alignment o satellite images of middle and high spatial resolutions. The goal of research is to develop method of stepwise classification of satellite images of middle resolution for thematic mapping of forest cover with the user of classification systems of FAO, NASA and Russian state forest inventory. Objects and methods. Approbation of the methodology was carries out on the territory of Middle Povolzhje including Republics of Mari El, Chuvashia, Tatarstan, Kirov and Nizhny Novgorod region. In the research we used archival Landsat images of 2011 and 2014 years. Spectral separability of forest classes on the satellite images, transformed by the “Tasseled cap” procedure to the images of Brightness, Greenness, and Wetness, was carried out in the “feature space”. For determination of the age groups (young, middle-aged, mature and over mature stands) of the decoding forest class each thematic layer was subject to detailed classification with scale of specified gradations and iterations. Results and discussion. After separation of all classes under unsupervised classification a detailed thematic map on 14 and 5 classes of land cover was derived. Similar values of spectral signatures showed mature and middle aged forest stands, which brings to the mixture of classes in the process of satellite images classification. Maximum spectral separability revealed main forest classes TNEC_MaO (Tree Needleleaved Evergreen Closed_Mature and Overmature), TBDC_MaO (Tree Broadleaved Deciduous Closed_Mature and Overmature), and TMC_MaO (Tree mixed closed _Mature and Overmature). The analyses of newly obtained multi temporal thematic maps on the territory of Middle Povozhje show a decrease of TNEC_MaO class area on the 2,8%. At the same time on the base of thematic maps of 2001 and 2014 there is a slight increase in TNEC_M class (Tree Needleleaved Evergreen Closed_ Middle aged) on 2,3%, and TNEO_W (Tree Needleleaved Evergreen Closed_Wetland) on 0,6%. The research results could be used for the implementation of forest monitoring with the use of remote sensing techniques. Conclusions. We suggest improved methodology of stepwise classification of forest cover on satellite images of middle spatial resolution on the base of international systems of land cover classifications. The research results could be used during implementation of forest monitoring by the methods of remote sensing, carrying out forest activities on increasing productivity, conservation of biodiversity and sustainable use of forest resources.