Red squirrel habitat mapping using remote sensing

The native Eurasian red squirrel is considered endangered in the UK and is under strict legal protection. Long-term management of its habitat is a key goal of the UK conservation strategy. Current selection criteria of reserves and subsequent management mainly consider species composition and food availability. However, there exists a critical gap in understanding and quantifying the relationship between squirrel abundance, their habitat use and forest structural characteristics. This has partly resulted from the limited availability of structural data along with cost-efficient data collection methods. This study investigated the relationship between squirrel feeding activity and structural characteristics of Scots pine forests. Field data were collected from two study areas: Abernethy and Aberfoyle Forests. Canopy closure, diameter at breast height, height and number of trees were measured in 56 plots. Abundance of squirrel feeding signs was used as an index of habitat use. A GLM was used to model the response of cones stripped by squirrels in relation to the field collected structural variables. Results show that forest structural characteristics are significant predictors of feeding sign presence, with canopy closure, number of trees and tree height explaining 43% of the variation in stripped cones. The GLM was also implemented using LiDAR data to assess at wider scales the number of cones stripped by squirrels. The use of remote sensing -in particular Light Detection and Ranging (LiDAR) - enables cost efficient assessments of forest structure at large scales and can be used to retrieve the three variables explored in this study; canopy cover, tree height and number of trees, that relate to red squirrel feeding behaviour. Correlation between field-predicted and LiDAR-predicted number of stripped cones was performed to assess LiDAR-based model performance. LiDAR data acquired at Aberfoyle and Abernethy Forests had different characteristics (in particular pulse density), which influences the accuracy of LiDAR derived metrics. Therefore correlations between field predicted and LiDAR predicted number of cones (LSC) were assessed for each study area separately. Strong correlations (rₛ=0.59 for Abernethy and 0.54 for Aberfoyle) suggest that LiDAR-based model performed relatively well over the study areas. The LiDAR-based model was not expected to provide absolute numbers of cones stripped by squirrels but a relative measure of habitat use. This can be interpreted as different levels of habitat suitability for red squirrels. LiDAR-based GLM maps were classified into three levels of suitability: unsuitable (LSC = 0), Low (LSC < 10) and Medium to High Suitability (LSC >=10). These thresholds were defined based on expert knowledge. Such a classification of habitat suitability allows for further differentiation of habitat quality for red squirrels and therefore for a refined estimation of the carrying capacity that was used to inform population viability analysis (PVA) at Abernethy Forest. PVA assists the evaluation of the probability of a species population to become extinct over a specified period of time, given a set of data on environmental conditions and species characteristics. In this study, two scenarios were modelled in a PVA package (VORTEX). For the first scenario (Basic) carrying capacity was calculated for the whole forest, while for the second scenario (LiDAR) only Medium-to-High suitable patches were considered. Results suggest a higher probability of extinction for the LiDAR scenario (74%) than for the Basic scenario (55%). Overall the findings of this study highlight 1) the importance of considering forest structure when managing habitat for squirrel conservation and 2) the usefulness of LiDAR remote sensing as a tool to assist red squirrel, and potentially other species, habitat management.