Classification of 'potential' forests based on remote sensing data

The aim of this study is to estimate the area with forest vegetation that does not yet meet the criteria formulated in the FAO/UN definition (minimum height 5 m, minimum canopy cover 10%, minimum area 0.5 ha), but will potentially meet them in the future (5 years or more, depending on the individual site conditions), which means that (according to the definition) they also represent forest areas. The study was conducted in the Bia??owie??a Glade. Tree species were classified individually and then divided into two groups: those that will reach a height of 5 m in the future and those that will not (grey willow, hawthorn). Hyperspectral (reduced with MNF transformation) and ALS???based features were used for classification with the SVM algorithm. Classification accuracy based on ALS data was better than that of hyperspectral data for indi??? vidual species but similar for the two species groups ??? 95.5% (Kappa 87.5%). Information about species and height was used to perform the classification of a fishnet layer into ???forests???, ???potential forests??? and ???non???forests???, with an accuracy of 96% (Kappa 87.7%). A map of forests and potential forest vegetation was created in the form of a thematic map, taking into account height, canopy cover, area of the complex and land use. This study provides new solutions in the context of cli??? mate change, deforestation and the need for reporting the forest area by individual countries (including Poland) to the FAO/UN.