Accurate estimates of the cost of land protection are useful for understanding where biodiversity conservation goals can be achieved at the lowest cost to society. However, because good cost maps are rare in tropical countries, conservation planning studies often ignore cost or rely on untested proxies, such as agricultural rent or land use intensity. Here, we show how analysts can estimate land protection costs using original data of public land acquisitions, global predictor datasets, and simple machine-learning models. For the Colombian Andes, a global biodiversity hotspot, we find that the principal driver of land protection cost is urban proximity, not agricultural rent. We derive cost estimates that predict public land protection costs more accurately than available cost proxies and identify new protection priorities for 143 threatened bird species. A more systematic collection of cost records of land protection will help inform public decisions on national and global biodiversity protection priorities.

See manuscript.