Open access publications in the SLU publication database

Abstract

Wet soil in forested landscapes of the boreal zone is often associated with large open wetlands or peatlands, but it may also be hidden beneath closed forest canopies on drained wetlands or wet strips beside streams known as riparian zones. Since these wet areas are shrouded by canopy they have traditionally been difficult to map from aerial photographs. Consequently, most of them are not marked on current maps. Movements of heavy forestry machines can severely disturb the soil and impair water quality in these wet and moist environments, so driving should be avoided in them. However, to avoid driving and plan forestry operations in these areas it is important to know where they are. Therefore, the aim of the studies this thesis is based upon was to evaluate and develop methods to map these hidden streams and wet areas so they can be considered in forestry planning to minimize negative effects of forestry on soil and water quality. High resolution digital elevation models (DEMs) are becoming more accessible as more countries are conducting laser scanning campaigns, and hydrological features such as streams and wet areas can now be topographically modelled on large scales. Stream networks extracted from these high resolution models are more accurate than current stream networks, provided that appropriate methods are used to pre-process the DEM. Forest-covered wet areas can also be mapped using DEMs, but no currently available method is universally applicable. Therefore my colleagues and I used machine learning to combine stream networks acquired using multiple methods with other soil and climate maps to more accurately predict wet areas. The new maps generated in this manner are relatively cheap to produce and can be used to plan forestry operations in great detail. Protective buffer zones can be placed around previously unknown headwater streams and these maps can be used to avoid driving in sensitive areas with heavy forestry machines.