Application of the Winter and Early-Spring Satellite Images for Assessment of the Birch Forest Coverage on the Abandoned Agricultural Lands

Abstract

More than 40 million hectares of agricultural land were abandoned after the collapse of the Soviet Union. A significant part of the land is covered by spontaneously regenerating woody and shrubby vegetation. When identifying the forest regeneration, the stands with a tree cover of more than 50% are accurately identified. It is difficult to identify the initial stages of forest regeneration on the abandoned agricultural lands using summer satellite images because of little difference between the young trees and saplings due to their low height and low density on the one hand, and herbaceous vegetation on the other. The purpose of this work was to apply winter and early-spring satellite images for assessments of the tree cover of birch-dominated stands (Betula pen-dula Roth.) formed on the abandoned agricultural lands (See Fig. 1). We used 189 releves of birch forests on the abandoned agricultural lands in the broad-leaved forest zone of the Republic of Bashkortostan. A regression analysis of the evaluation of the tree cover was carried out using the values of the spectral reflectance of the RED, NIR, SWIR11, and SWIR12 bands, as well as the values of the NDFSI snow index from seven cloudless Sentinel-2 images taken between 04.11.2020 and 13.05.2021 (See Fig. 2, 3). When selecting optimal regression models, the values of correlation coefficients (R) and determination coefficients (R2) were used to assess the model quality. To test the possibility of using the obtained models for assessing the tree cover of the stand at earlier succession stages, we involved the data on the tree cover from 36 geobotanical releves, where the crown density of the stand was visually evaluated in July 2013. Then, the described procedure was applied to calculate the tree cover using the Landsat-8 image taken on 25.03.2014. When creating regression models to calculate the tree cover, the best results were obtained using the red band of early spring images during the period when snowpack is still solid (from mid-March to the first half of April) (See Table 1). The correlation between the tree cover and the spectral reflectance of the red band was -0.90. The model allowed us to determine accurately the tree cover of birch forests aged from 18 to 20 years which prevail in the zone of broad-leaved forests in the Republic of Bashkortostan. The accuracy of the model for determining the tree cover according to the obtained regression models for other dates is unstable and highly likely influenced by the snow depth and the seasonal dynamics of changes in the radiation intensity of the red and infrared bands (See Table 2, 3). To conclude, the equations calculated from modern satellite images can be used to assess the tree cover using retrospective images at earlier succession stages of the abandoned field recovery. When using early-spring images, the snow depth should be taken into account since the snowpack melting dates can vary greatly from year to year. The paper contains 3 Figures, 3 Tables, and 41 References. The Authors declare no conflict of interest.