Treefall gap mapping using Sentinel-2 images

Abstract

Proper knowledge about resources in forest management is fundamental. One of the most important parameters of forests is their size or spatial extension. By determining the area of treefall gaps inside the compartments, a more accurate yield can be calculated and the scheduling of forestry operations could be planned better. Several field- and remote sensing-based approaches are in use for mapping but they provide only static measurements at high cost. The Earth Observation satellite mission Sentinel-2 was put in orbit as part of the Copernicus programme. With the 10-m resolution bands, it is possible to observe small-scale forestry operations like treefall gaps. The spatial extension of these gaps is often less than 200 m2, thus their detection can only be done on sub-pixel level. Due to the higher temporal resolution of Sentinel-2, multiple observations are available in a year; therefore, a time series evaluation is possible. The modelling of illumination can increase the accuracy of classification in mountainous areas. The method was tested on three deciduous forest sites in the Börzsöny Mountains in Hungary. The area evaluation produced less than 10% overestimation with the best possible solutions on the sites. The presented work shows a low-cost method for mapping treefall gaps which delivers annual information about the gap area in a deciduous forest.