Remanescentes florestais da Mata Atlântica: por que existem?

Abstract

Tropical forests harbor more than half of the planet’s species. They play a fundamental role in climate regulation, biodiversity conservation, and the provision of ecosystem services. The Brazilian Atlantic Forest, one of the most biodiverse and threatened biomes in the world, has already lost approximately 80% of its original cover and currently exhibits a fragmented landscape, predominantly composed of small forest remnants smaller than 50 hectares. This results from intense anthropogenic pressure over the centuries, including urban, agricultural, and silvicultural expansion, particularly in more accessible areas. In this context, the present study aimed to identify the environmental and socioeconomic factors that influence the persistence of forest remnants in the Atlantic Forest and to assess whether there are distinct spatial patterns across the biome. We tested 48 predictor variables related to forest occurrence, maintenance, regeneration, and deforestation, using Principal Component Analysis (PCA) to reduce dimensionality. The analyses were conducted using three global models: Ordinary Least Squares (OLS), Spatial Lag Model (SLM), and Spatial Error Model (SEM), as well as one local model: Geographically Weighted Regression (GWR). The SEM showed the best performance among the global models (lowest AIC and low residual autocorrelation), while the GWR was the most effective overall (AICc = 363149; Adjusted R² = 0.32), as it captured non-stationary spatial variations. Both models revealed consistent patterns: physical variables such as elevation, slope, and soil attributes (organic carbon, nitrogen, clay content) were positively associated with forest persistence. Additionally, the presence of Protected Areas was also positively related to forest maintenance, highlighting the critical role of governance. Conversely, variables associated with accessibility (roads, highways, railways), agriculture, and silviculture were negatively associated with forest persistence. We conclude that although GWR provides greater local accuracy, SEM and GWR are complementary models.