Downscaling European species atlas distributions to a finer resolution: implications for conservation planning

摘要: Aim One of the limitations to using species' distribution atlases in conservation planning is their coarse resolution relative needs local planners. In this study, a simple approach downscale original species atlas distributions finer outlined. If such procedure yielded accurate downscaled predictions, then it could be an aid available real-world decisions. Location Europe. Methods An iterative based on generalized additive modelling used European 50 x km 2189 plant and terrestrial vertebrate c. 10 grid resolution. Models are trained 70% data evaluated remaining 30%, receiver operating characteristic (ROC) procedure. Fitted models interpolated A British dataset comprising 81 passerine-bird as test bed assess accuracy predictions. European-wide, predictions further terms ability reproduce: (1) spatial patterns coincidence richness scores among different groups; (2) richness, rarity complementarity hotspots. Results There was generally good agreement between observed fine-resolution for passerine Britain (median Jaccard similarity = 70%; lower quartile 36%; upper 88%). contrast, correlation relatively low (rho 0.31) indicating pattern error propagation through process overlaying many species. It also found that measures model fitting were poor predictor models' interpolate at fine resolutions -0.10). Although hotspots not fully coincident modelled coarse-resolution data, or there evidence able maintain cross-taxon species-richness scores, least groups. Downscaled uncover important environmental gradients otherwise blurred by data. Main conclusions Despite uncertainties, downscaling procedures may prove useful identify reserves more meaningfully related variation. Potential errors arising from presence false positives reduced if downscaled-distribution records projected occur outside range excluded. However, usefulness limited data-rich regions. applied data-poor regions, need undertake research understand structure models. particular, would investigate which poorly modelled, where why. Without assessment difficult support unsupervised use most situations.