No abstract

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 costefficient 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 (r s =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

No abstract

The diversity of data sources, analysis methodologies, and classification systems has led to a number of new techniques for monitoring land-cover change. However, this wide choice means that it is difficult to know which solution to choose. A system capable of integrating the results of different analyses and applying them to land-cover mapping would therefore be extremely useful. This study investigates the use of evidence pooling and neural networks in land-cover mapping. Neural networks were used to classify land-cover using evidence from spectral (Landsat-7 ETM1), textural, and topographic information. Mapping was performed using combinations of evidence source and evidence pooling techniques. The best performance was achieved using all available information with a method that summed evidence directly instead of categorizing it. While the methodology failed to reach the level of accuracy recommended elsewhere, a comparison of the number of classes used with other methods showed that the system performed better than these approaches.

No abstract