Spatial airborne laser scanning features for predicting forest biodiversity indices
With recent developments in remote sensing technologies, plot-level forest resources can be predicted utilizing airborne laser scanning (ALS) data. The prediction is most often assisted by ALS feature variables which summarize the vertical distribution of the pulse returns. We introduce new ALS feature variables that focus instead on the spatial or horizontal distribution of the pulse returns. These spatial features are based on the patterns of patches of vegetation at a number of height levels, which are extracted from the ALS data using thresholded canopy height models. We propose to use spatial summary characteristics, most importantly the Euler number and the empty-space function, to capture the structure of the patches and the empty space around them. We illustrate usefulness of the proposed spatial features for predicting different forest biodiversity indices that summarize the spatial structure of trees, species diversity, or their breast height diameter distribution. We employ the proposed spatial features, in addition to commonly used features from literature, in the well-known k-nn estimation method to predict the biodiversity indices. We present the methodology on the example of a study site in Central Finland.
READ FULL TEXT