Files
Abstract
Forest stands are fundamental management units, consisting of trees of uniform composition and structure. Creating stand maps is time-consuming requires expertise and is subjective. The dynamic nature of forests necessitates regular updates and revisions to stand maps. This thesis aims to provide automated methods of tackling this problem, using both aerial imagery and LiDAR. We first examine the effectiveness of traditionally employed metrics to evaluate stand maps over three forest landscapes of varying complexity and degrees of management. We present a fast workflow within ArcGIS Pro that is aimed at maximizing homogeneity of stands while allowing flexibility in terms of the feasibility of stand sizes. To address the issue of subjectivity between experts and the requirements of the land, we also propose an alternative, hands-on approach for refinement using random forests. This significantly reduces human effort while maintaining the required precision.