Colloquium

Optimizing species distribution models for woodpeckers using LiDAR-derived vegetation metrics

Organisator Laboratorium voor Geo-informatiekunde en Remote Sensing
Datum

di 24 juni 2025 09:00 tot 09:30

Locatie Gaia, gebouwnummer 101
Droevendaalsesteeg 3
101
6708 PB Wageningen
+31 (0) 317 - 48 17 00
Zaal/kamer 2

by Wouter de Vries

Abstract
Biodiversity has been increasingly threatened by environmental change and habitat loss, making the need for monitoring biodiversity clear. Species distribution models (SDMs) are an important method for understanding and predicting the distribution and the change in numbers. Recent advances in SDMs allows the inclusion of remote sensing data, such as LiDAR derived vegetation data. These metrics capture structural vegetation data, covering vegetation cover, density and structural complexity.

This research investigates whether incorporating LiDAR derived vegetation metrics improves the predictive performance and ecological value of SDMs for five woodpecker species in the Netherlands. A total of 25 LiDAR variables (as derived from Kissling et al., 2023) were added to an existing set of environmental variables provided by Sovon (Dutch Centre for Field Ornithology). Three sets of input variables were tested and evaluated, LiDAR only, environmental only and a combination of both. A random forest (RF) model was used and the performance was evaluated using adjusted R-Squared values.

The results show that the models using only LiDAR variables had the overall lowers predictive performance, while combining LiDAR with environmental variables led to a very slight improvement over environmental-only models. Although the overall effect on model performance was limited, certain LiDAR metrics consistently ranked among the top predictor variables, suggesting their ecological importance for the woodpecker species.

These findings suggest that while LiDAR variables do not substantially improve the predictive performance under the current SDM setup, they can contribute valuable insights into the species’ habitat preferences not represented by the environmental variables. Future research could dive more into predictor variable selection based on the ecology of the species and incorporate more commonly used SDM performance metrics to better compare SDM results.