Agricultural expansion is a major threat to global biodiversity, with profound impacts in tropical regions targeted for commodity goods. Due to a growing demand for food and fiber consumption, privately-owned lands are undergoing rapid transformation from wild ecosystems to commodity production areas. These landscapes can still play a critical conservation role by providing connectivity between large protected areas while also retaining remnant essential habitat for some species. Brazil is a megadiverse country that, despite having one of the highest agricultural expansion rates in the world, has a well-established policy that regulates land use in rural private lands. This policy, known as the Forest Code, offers great opportunities to protect biodiversity in agricultural regions such as the last agricultural frontier located in the Cerrado biome. In this study, we quantify the effectiveness of the Forest Code for achieving a biodiversity-representative network of land-use restricted sites in a commodity-driven agricultural landscape. We develop a fine-scale, spatially explicit gap analysis to estimate conservation gaps from policy-protected sites, using vegetation physiognomic types and a set of bird species as biodiversity surrogates. In addition, we investigate how rural properties can help address conservation gaps by analyzing their policy compliance status. 

Spatially-explicit information featuring a detailed vegetation types are necessary to support conservation and ecological analyses. Accurately mapping such categories at scales compatible with land management is still a challenge for Neotropical savannas. For this project, we developed a supervised systematic classification framework combining machine learning and Geographic Object-Based Image Analysis (GEOBIA) to map vegetation structural types at 5-m resolution in the Cerrado biome. Our resulting maps feature 13 land cover classes with 87.6% overall accuracy, of which all 11 major vegetation classes were identified.