Figure 1: Schematic summarising the scales of data collection (roost, subplot and tree) and density measures (A-E) with table highlighting spatial and temporal replicates of density measures within the data. Roost-level density (A) was calculated by dividing the total roost abundance by roost area. Subplot-level density was calculated either from total subplot count and subplot area (B), or from fixed-bandwidth kernel estimates (C), calculated from tree locations weighted by abundance (index values). For (C) shading showing the kernel density estimates (light shading = lower density, dark shading = higher density). Average kernel density per subplot was calculated from occupied pixels only. Tree-level 2-D density (D) was estimated from within tree abundance and canopy area (estimated by Dirichlet-Voronoi tessellations), while tree-level 3-D density (E) was calculated as the absolute count of bats in a single tree divided by the approximate volume of tree occupied (height range occupied multiplied by crown area). Tree-level 3-D density was calculated for the randomly selected subset of trees in the full dataset for which absolute count and height measures were available. The total number of datapoints for each measure can be calculated by the number of spatial replicates per roost multiplied by the number of temporal replicates. Grey colouring in measure visuals indicates where bats were absent and not included in mean calculations. Grey colouring on the map shows urban areas.