Modelagem espacial do carbono em florestas monodominantes de Peltogyne gracilipes no norte de Roraima

Abstract

Species distribution maps and carbon stocks based on the correlation of field samples with environmental constraints are relevant inputs for adaptation and mitigation actions in the context of regional and global climate change. The aim of this study was to model the carbon stocks of monodominant Peltogyne gracilipes Ducke (Leguminosae) forests in northern Roraima based on environmental conditions. Carbon data from the forest inventory of 129 plots (0.05 ha = 6.45 ha) were used in Maracá Island. Predictor variables were altitude, slope, drainage distance and Normalized Difference Vegetation Index (NDVI) for the sample plots. Multiple Linear Regression (RLM) techniques were adopted to spatialize carbon stocks and map P. gracilipes conglomerates in Maracá and surrounding areas (~ 450,000 ha). The Maximum Entropy Method (MaxEnt) based on P. gracilipes occurrence points (57 plots) and environmental variables was also used to evaluate the performance of the RLM model. The carbon-based regression model (Mg ha-1) indicated that environmental variables are significantly associated with P. gracilipes monodominance (R2 = 0.35; p <0.0000). The RLM method based on carbon percentage was more accurate (Global Accuracy = 0.55) than MaxEnt (Global Accuracy = 0.41) to model the areas potentially with presence of the studied species, allowing to verify that these areas have average altitude. (147.5 ± 44.3 m) significantly lower (t = 6.78; p = 2.71e-11) compared to the areas without (186.1 ± 58.6 m). P. gracilipes monodominant forests and their carbon stocks are spatially conditioned by topographic and NDVI variables. These results are similar to field studies undertaken at the microscale, indicating that similar methodological approaches can be taken to reduce uncertainties about the spatialization of arboreal carbon stocks in the spatial macroscale of this region of the Brazilian Amazon.