Automated segmentation of vegetation structure units in a Mediterranean landscape

In Mediterranean regions, the combination of disturbances, life histories, plant regeneration traits, and microhabitat variability form highly heterogeneous vegetation mosaics which shift in space and time. Consequently, structure-based forest management is emerging as a superior alternative to management of vegetation formations in such areas. Delineation of management units in these areas is often based on manual interpretation of aerial imagery coupled with field surveys. Here, we propose an alternative approach that is based on segmentation of remotely sensed height and cover maps derived from light detection and ranging (LiDAR) imagery. A large suite of alternative segmentation maps was generated using multiresolution segmentation (MS) with different parameters, and an area-fit approach used to select the map that most successfully captured a reference set of structural units delineated manually. We assessed the feasibility of this approach in a nature reserve in northern Israel, compared the resulting map with a traditional vegetation formations map, and explored the performance of the segmentation algorithm under various parameter combinations. Pronounced differences between the structure and formation maps highlight the suitability of this approach as an alternative to the existing methods of delineating vegetation units in Mediterranean systems, and possibly in other systems as well.