Estimating forest biodiversity in airborne laser scanning assisted inventories using spatial measures
share
With recent developments in remote sensing technologies, plot-level forest resources can be estimated using airborne laser scanning (ALS) data. We present how spatial summary characteristics can be utilized for detecting the spatial structure of trees and other biodiversity variables on top of that. In particular, we summarize the spatial structure of trees using the so-called L- and empty-space function from spatial statistics. On plot-level, these functions are used for classifying forests into regular, random, or clustered patterns of tree locations. From the ALS data, we extract patches of vegetation at a number of height layers using thresholded canopy height models. The structure of the patches and the empty space around them is captured by spatial ALS feature variables of which some are new in this context and also based on the empty-space function. We estimate forest biodiversity employing, in addition to commonly used ALS feature variables, the proposed new spatial features in the well-known k-nn estimation method. We present the methodology on the example of a study site in Central Finland. We found that the new spatial ALS features are useful for remotely estimating the spatial structure of trees, their breast height diameter distribution, and even the development class of the field plots in a simple, but yet efficient approach.
READ FULL TEXT