Cost estimates are a key input for spatial conservation planning. Because accurate estimates of cost tend to be difficult to obtain, modelers are often forced to resort to single imperfect proxies (e.g. accessibility, land cover, or county-level cost averages). Here I present a methodology to estimate the cost of conservation easements – permanent transfers of partial rights – from large-scale data on parcels and real estate transactions. I use PLACES (the Private Land Conservation Evidence System developed at Boston University), to extract a number of predictors of land value at the parcel level (e.g. sale, terrain, accessibility, water access, land cover, nearby protection, income, population density, etc.). I let machine learning algorithms (ensemble methods) compete for the best prediction of sales prices of encumbered and unencumbered properties. Predictions are compared with data from 150 state-sponsored easements and appraisals in Colorado.
Results/Conclusions
Results indicate that automated appraisals of conservation easements can be reasonably accurate, and in some cases even outperform official appraisers. The approach scales well, and can be applied across large landscapes. Resulting cost estimates can be used to incentivize landowner participation, estimate budgets for conservation interventions, and identify tax fraud in easement donations (which has been suggested to amount to billions of dollars). With further improvements in datasets and prediction algorithms, access to high-resolution cost estimates for conservation planners might become available within the foreseeable future.