Introduction
Estimates of biodiversity loss from land-use change overwhelmingly derive from local-scale studies, but disturbance often spans vast spatial scales transcending abiotic gradients and biogeographic barriers. Because land-use change might differentially impacts range-restricted versus widespread species, a key challenge for large-scale conservation interventions is to understand how large-scale biodiversity change is related to local-scale impacts.

Objectives
We aimed to quantify biodiversity change across the forest/pasture transition in Colombia at scales ranging from local to near national, and then to tease apart the key abiotic gradients that structure the spatial scaling of biodiveristy loss.

Methods
We assembled an unprecedented avian point-count dataset in forests and pastures across Colombia, recording 969 bird species across the Amazon, Llanos, Magdalena, Santa Marta, and all three Andean Cordilleras. We then applied a novel addition to the community occupancy model to derive detailed trait-based predictions of occupancy for more than 1600 bird species in forest and pasture across Colombia and assess the scaling relationships apparent in the results. Finally, we applied generalized dissimilarity models to the forest and pasture data separately in order to understand how abiotic gradients in elevation and precipitation, together with biogeographic barriers imposed by mountains and valleys structure turnover in forest and pasture separately.

Results
Regional-scale biodiversity losses exceed mean local-scale losses by up to 75 percent. Excess regional biodiversity losses accumulate rapidly with a region's spatial extent, but they are even more closely related to the region's compositional heterogeneity, and Whittaker's beta-diveristy (expressed as a multiplicative partition) yields an approximately scale-free predictor of excess regional-over-local losses. Elevation gradients play a paramount role in structuring turnover biotic homogenization in the system.

Conclusions
Effect sizes measured at local scales severely underestimate the large-scale biodiveristy impacts of land-use change in heterogeneous regions, and particularly on tropical mountins.