Mapping the dynamics of forests in the north of the Moscow Region based on satellite images of the Landsat series

Abstract

Solving the problem of identifying areal and formational changes in forests is an urgent problem in studying the dynamics of forest cover [Coppin, 2004; Senf et al., 2017]. The presented study is devoted to identifying the spatiotemporal variability of forest cover in the Moscow Region using the example of the Klinskiy district (Fig. 1). As part of this work, the problem of a methodological approach to identifying changes in the composition of forests was solved and changes in the formational composition of forests in the study area were established. The work used 153 geobotanical relevés made between 2013 and 2021, satellite images of Landsat-8 (2014–2021) and Landsat-5 (1985–1986), a digital elevation model and forest plans. The work scheme includes 5 stages (Fig. 2): 1) data collection in GIS; 2) classification of communities, development of legends for maps of the forest’s formational composition; 3) processing of satellite images; 4) automated classification of multi-temporal composites with the formation of a training sample; 5) analysis of changes. For the analyzed periods 1985–1986 and 2014–2021 cloudless satellite images were selected for the beginning, middle and end of the growing season, as well as the end of winter and beginning of spring (Table 1). For them, the vegetation indices NDVI, EVI and NDMI were calculated, pruning was carried out according to the forest mask, and subsequently the materials were combined into composites. Classification of prepared composite images for 1985–1986 and 2014–2021 was carried out using discriminant analysis method with a training sample prepared according to available relevés and forest plans. This made it possible to establish with sufficiently high accuracy the formation composition of forests at the dates of the period under study: 1985–1986 and 2014–2021, the accuracy of the determination was 79.3 % and 78.8 % (Table 2). A similar technique for selecting satellite images was used previously for mapping the current state of forests in the Moscow region [Chernenkova et al., 2019; Kotlov, Chernenkova, 2020]. As a result, maps of the formational composition of forests for the indicated periods were obtained, and an analysis of changes was carried out. The identified changes show a relatively small decrease in the total forested area from 55.1 % to 52.1 %. The composition of forests has changed significantly; only 26 % of forests have not changed their formational affiliation, of which the pine formation is the most stable (82.6 % have not changed their formational affiliation). A general increase in the share of birch and aspen formations was revealed at the present stage compared to 1985–1986, and it can be assumed that, despite a slight change in the total forested area of the region (from 55.1 % to 52.1 %), the composition of forests is becoming more disturbed (Table 4). As a result of the work, a methodology was tested for identifying formational changes in the forests based on the use of multi-temporal satellite images of the Landsat series. The territorial and qualitative changes in the forest cover of the study area have been established. The presented methodology has prospects for testing not only in the Moscow Region, but also in other regions.