Many nature conservation organizations focus on creating protected areas that they acquire and then manage themselves. By doing so, they have to sustain a wide variety of costs. A critical part of conservation planning and spatial prioritization is accurately accounting for spatial heterogeneity of costs, especially acquisition costs. Studies have revealed that this could lead to large efficiency savings, but they have based this claim on methods of estimation prone to affect the spatial pattern of variation in costs. Due to the lack of data regarding actual conservation costs faced by conservation organizations, a common approach is to rely on more readily available proxies, such as GDP per capita or agricultural land value. While these proxies can seem reasonable, the lack of mechanistic understanding of what determines actual conservation costs means that there is a risk that such proxies do not preserve the underlying spatial signal of actual costs. For example, the nature of land parcels targeted for conservation or the dynamics of conservation land transactions can be very different compared to agricultural lands. The spatial pattern and the variation of one might not accurately reflect those of the other.
We use cost data from actual land acquisition deals made since 1990 by The Nature Conservancy, the world's largest land trust. We describe the nationwide pattern of actual acquisition costs, as faced by a conservation organization, and we identify the factors explaining their spatial distribution. We use a model averaging approach by building a set of generalized linear models, fitting acquisition costs to a variety of ecological, socio-economic and geographical parameters at the county level. We then retain best models, using AIC-based selection, and average them to produce a new, accurate and updated cost prediction model.
We create the first nationwide map of acquisition costs, and use it to explain their spatial pattern across the continental U.S. Costs per hectares of protected areas show patterns of spatial aggregation and tend to be highest in coastal regions and metropolitan areas, ranging in size across several orders of magnitude, in our case study. We show that agricultural land value used as sole proxy for acquisition costs returns spatially biased estimates; however it remains a useful predictor within our model, significantly explaining part of the observed spatial heterogeneity of costs. This work reveals how better land cost prediction may skew conservation practice through spatial constraint.