A method for integrating the Breeding Bird Survey and Forest Inventory and Analysis databases to evaluate forest bird–habitat relationships at multiple spatial scales

The limited spatial scales of many bird–habitat studies restrict inference regarding large scale bird–habitat relationships. A potential solution to this challenge is integrating the USFS Forest Inventory and Analysis (FIA) and USGS Breeding Bird Survey (BBS) databases. We describe a methodology for integrating these databases into a uniform dataset for modeling bird–habitat relationships at multiple spatial scales. We accumulated route-level BBS data for four species guilds (canopy nesting, ground-shrub nesting, cavity nesting, early successional), each containing a minimum of five bird species. We developed 43 forest variables at the county level using FIA data from the 2000 inventory cycle within 5 physiographic regions in 14 states. We examined spatial relationships between the BBS and FIA data at three hierarchical scales: (1) individual BBS routes, (2) FIA units, and (3) physiographic sections. At the BBS route scale, we buffered routes at 100 m, 1 km, and 10 km radii, intersected these buffers with county boundaries, and developed weighted averages for each forest variable within each buffer width. Weights were a function of the percent of area each county had within a buffer. We calculated 29 landscape structure variables from 1992 National Land Cover Data (NLCD) imagery using Fragstats within each buffer width. At the BBS route scale, we developed models relating variations in bird occupancy and abundance to forest and landscape structure within each buffer width using classification and regression trees (CART). We aggregated the FIA variables to the FIA unit and physiographic section scales and recalculated the landscape variables within each unit and section using NCLD imagery resampled to a 400 m pixel size. We used regression trees (FIA unit scale) and general linear models (GLM, physiographic section scale) to relate variations in bird abundance to the forest and landscape variables. At the BBS route scale, 80% of the best CART models accounted for >50% of the variation in bird occupancy and abundance. Among FIA units and physiographic sections, the regression trees accounted for an average of 54.1% and the GLMs accounted for an average of 66.3% of the variability in bird abundance, respectively. This methodology shows promise for integrating independent databases for evaluating bird–habitat relationships across broad spatial extents, and the hierarchical nature of these models provides a potentially consistent means of evaluating management options at varying spatial scales.