Analysis of deforestation and protected area effectiveness in Indonesia: A comparison of Bayesian spatial models

Abstract Tropical deforestation in Southeast Asia is one of the leading causes of carbon emissions and reductions of biodiversity. Spatially explicit analyses of the dynamics of deforestation in Indonesia are needed to support sustainable land use planning but the value of such analyses has so far been limited by data availability and geographical scope. We use remote sensing maps of land use change from 2000 to 2010 to compare Bayesian computational models: autologistic and von Thunen spatial-autoregressive models. We use the models to analyze deforestation patterns in Indonesia and the effectiveness of protected areas. Cross-validation indicated that models had an accuracy of 70–85%. We find that the spatial pattern of deforestation is explained by transport cost, agricultural rent and history of nearby illegal logging. The effectiveness of protected areas presented mixed results. After controlling for multiple confounders, protected areas of category Ia , exclusively managed for biodiversity conservation, were shown to be ineffective at slowing down deforestation. Our results suggest that monitoring and prevention of road construction within protected areas, using logging concessions as buffers of protected areas and geographical prioritization of control measures in illegal logging hotspots would be more effective for conservation than reliance on protected areas alone, especially under food price increasing scenarios.