Lidar-based MaxEnt models to support conservation planning for endangered Red-cockaded Woodpeckers in urbanizing environments

Abstract

Sensitive and intensively managed species require carefully thought-out management plans to promote and maintain specific habitat conditions. Urban features and land-use change must be assimilated into these habitat management plans, as they become increasingly present globally. As a case study, several MaxEnt species distribution models were developed that could enable habitat management efforts for the endangered Red-cockaded Woodpecker (RCW) in moderate to increasingly urbanized environments. Model development began with a classification of fine-scale, lidar-based habitat indicators with the area normalized at the stand level and developed around known habitat characteristics of RCW. Other explanatory rasters included distance to different urban features, and experimentation with spectral layers outside the visible light spectrum. Models were trained using presence data from a relatively small but comprehensively surveyed population in Montgomery County, Texas, and three compartments that were recently pedestrian surveyed for RCWs on the Sam Houston National Forest. The former is experiencing moderate levels of urbanization, and the latter is in earlier stages. The best performing model predicted RCW presence 94% of the time at a 0.4 probability threshold and resulted in an area under the curve (AUC) of 0.88. Successful model development required a specific combination of steps and data processing, including the use of lidar-based habitat indicators created using data fusion and machine learning classification, land-use features, and non-visible spectra. These methods can provide valuable insights into strategic habitat planning for the RCW and other sensitive species in urbanizing landscapes. This study reinforces what habitat characteristics promote RCW success, while providing valuable insights to guide management activities around urbanization. These could include mapping suitable recruitment areas that remain unoccupied, spatially identifying where habitat quality was lacking or sufficient, and predicting the impact of future land-use change. This case study demonstrates that species distribution modelling can be successfully applied at subpopulation and fine scales, and for the practical purpose of enabling habitat and conservation planning where anthropogenic activities are adding challenging complexities.