Epsilon Archive for Student Projects

Abstract

Management of young forest stands is important for the future economical outcome. Cleaning is a way to control the competition between plants and a total of 1 443 000 ha are in need of cleaning in Sweden. The cleaning is usually performed when the trees are 2 – 6 m high and the most common is to remove deciduous trees to favor coniferous species. To handle the amount of forests in need of cleaning there is a need for efficient inventory and planning of these areas. Remotely sensed data can be used to make these processes more efficient. Earlier studies have shown that variables that are commonly used for forest management planning can be accurately estimated when using photogrammetry with aerial images. In this study the use of 3D point clouds as an aid for field inventory when planning for cleaning has been evaluated. This was done by studying field inventoried sample plots and a 3D point cloud derived from high-resolution images collected with a drone in the county of Västerbotten. The need of cleaning was predicted with logistic regression with an overall accuracy of 82 %. Field attributes; average height, stem number and ΣH2 was predicted using linear regression with a relative RMSE of 43.9 %, 44.4 % and 76.2 %, respectively. The results show that it is possible to estimate cleaning need and field attributes using 3D point clouds derived from high-resolution images collected with a drone. Cleaning need can be predicted with high accuracy. The method is time consuming, hence evaluations regarding costs and time compared to manual field inventory is required if the method is to be implemented in practical use.