Computing maps of forest structural diversity: Aggregate late
Example data: https://github.com/antiphon/point-pattern-examples
Abstract
Local forest biodiversity hotspots are small areas within a landscape or a single stand, characterized by a high variability in e.g. species composition, size distribution or spatial pattern, in comparison to the surrounding areas. Their identification is important, e.g. for planning routes of harvesters, for selecting retention tree groups from the cutting area, and for selecting set-aside areas at landscape level. Traditional optical remote sensing enables prediction of forest attributes at large areas, but is typically restricted to a fixed spatial resolution. The fixed resolution is problematic especially for diversity indices as it contradicts the ecological meaning of local diversity which varies as a function of scale. While traditionally diversity predictions were produced with area-based approaches, combining 3D point-cloud-data-based single tree detection with field data enables the production of tree-level data, creating new opportunities for forest structure quantification. Particularly, at the single-tree level the ecological scale can be separated from the technical resolution. We demonstrate the importance of distinguishing scales when producing forest diversity maps. Furthermore, local diversity indices are typically computed at systematically or randomly selected locations in the landscape. We present new, alternative indices, defined through individual trees’ neighbourhoods, and show via simulated examples how the new indices greatly improve detection of local diversity. We also compare data from Panama and Finland at a shared ecological scale. We conclude that a tree-level data should not be aggregated to any technical scale before computing indicators. The separation of scales also helps produce indicator maps comparable across different studies. We recommend conditional indicators of local diversity over unconditional ones.
