Of urban mice and men: How habitat structure, grassland management, and landscape configuration shape urban small mammal communities in Stoke-on-Trent, UK.

Urban green spaces are increasingly expected to support biodiversity, yet they are often treated as ecologically equivalent despite variation in habitat structure, management and landscape context. Small mammals remain underrepresented in multiscale research within post-industrial cities. This thesis investigated how patchscale habitat conditions and landscape configuration shape small mammal communities across Stoke-on-Trent and Newcastle-under-Lyme, Staffordshire, United Kingdom, and whether these relationships could support spatially explicit assessment of management scenarios.

Repeated Longworth trapping was undertaken across 21 green space patches over five survey periods, generating 7,200 nominal trap nights, equivalent to 6,966 adjusted trap nights after accounting for unavailable traps, and 765 captures. Responses were assessed using species richness, catch per unit effort (CPUE), Simpson’s reciprocal diversity and species-specific CPUE for wood mouse (Apodemus sylvaticus), field vole (Microtus agrestis), bank vole (Myodes glareolus) and common shrew (Sorex araneus). GIS quantified habitat composition, woody linear vegetation, paths, watercourses, patch area and grassland management within patches, together with green space amount, green space count, nearest green space distance, hedgerows, watercourses and roads at 250 m and 500 m extents. Patch- and landscape-scale relationships were modelled separately before supported predictors were integrated into a spatial CPUE model evaluated using site-blocked cross-validation.

Urban green spaces were not ecologically equivalent. Woody linear density was the most consistent positive within-patch predictor, while path density and management intensity were negatively associated with CPUE. Patch area had limited explanatory value, and focal species showed contrasting responses to grassland, woody structure, watercourses and disturbance. At the landscape scale, subdivision and isolation were the most consistent negative predictors, while road density was negatively associated with CPUE. The combined model retained a positive association with woody linear density and negative associations with management intensity, green patch count and road density. Site-blocked cross-validation produced a Pearson correlation of 0.719 between observed and predicted CPUE. Across 598 environmentally comparable unsurveyed patches, mean predicted CPUE declined from 9.83 under low mowing to 6.85 under moderate mowing and 4.77 under intensive mowing. Urban habitat value therefore depends on internal structural quality, grassland management and matrix permeability rather than area or designation alone.