Określanie lesistości Polesia Ukraińskiego na podstawie wyników klasyfikacji sezonowych obrazów kompozytowych Landsat 8 OLI

Training dataset for modelling of forest cover was created after classification of multispectral satel− lite imagery IKONOS−2 with spatial resolution 3.2 m (acquisition date – 12.08.2011). As a result, we created binary forest cover map with 2 categories: ‘forest’ and ‘not−forest’. That allowed us to compute the tree canopy cover for each pixel of Landsat 8 OLI, using vector grid with cell size of 30×30 m. Classification model was developed using training dataset that included 17,000 observations, 10,000 of them represented results of IKONOS−2 classification. Aiming to avoid errors of agricultural lands inclusion into forest mask because of lack of data, additionally we collected about 7000 random observations with canopy cover 0% that had been evenly distributed within unforested area. Random Forest (RF) model we developed allowed us to create continuous map of forests within study area that represents in each pixel value of tree canopy closeness (0−100%). To convert it into a discrete map, we recoded all values less than 30% as ‘no data’ and values from 30 to 100% as 1. Forest mask for two selected administrative districts of Chernihiv region (NE Ukraine) was created after screening map from small pixel groups that covered area less than 0.5 ha. Obtained results were compared with Global Forest Change (GFC) map and proved that GFC data can be used for forest mapping with tree canopy closeness threshold 40%. On considerable areas of abandoned agricultural lands in the analysed regions of Ukraine, forest stands are formed by Scots pine, silver birch, black alder and aspen. Existence of such forests substantially increases (on 6−8%) the forested area of Gorodnya and Snovsk districts of Chernihiv region – comparing to official forest inventory data. However, such stands are not protected and have high risks to be severed by wildfires, illegal cuttings with aim to renew the agricultural production, by diseases, insects and other natural disturbances.