Integrated studies on the landscape have sought to contemplate the heterogeneity of factors and processes to the analysis of the rhythm, form and direction in which anthropic features are established due to the change of use and land cover. The complexity in determining the response variables and the poorly detailed scale in which mappings of natural resources in the Amazon region are produced have been a major constraint to models that seek to represent the dynamics of land use change in small regions. The objective of this study was to model the evolution of the landscape and a prediction scenario for year 2047 of a forest environment with open savannas and campinarana formations in the central part of Roraima, northern Amazonia, investigating the hypothesis that changes in the use of land occurred between 1984 and 2017 determined changes in soil organic carbon stocks. Seven land use classes were quantified and spatialized, and transitional matrices, at 10-year intervals, identified the pace and magnitude of landscape anthropization. The conversion of forested and ecotoneous areas by agricultural activity was evidenced and, added to the recurrent regime of fire, especially in high altitude forests, revealed that deforestation and land transitions for non-conservation uses did not lead to good productivity farming practices, but favored impoverished and degraded subsistence, whose main reason is the inadequacy of family agriculture to the acid / dystrophic soils destined to the settlements, the timid presence of technology. For the modeling of organic carbon stocks in the soil, the Random Forest algorithm, 141 pedological, climatic predictors and data derived from remote sensors, with the 30 most important variables being predicted. Prediction maps of COT stocks and statistics were evaluated for four layers up to the first 0.3 m of soil, with spatial resolution of 30 m, and COT stocks were estimated for different land uses and geoenvironments. The results showed that areas of strong undulating forest and candidates for organic carbon sinks in the region are being particularly impacted by fire and erosion, making them less resilient to environmental disturbances. Three anthropogenic processes relevant to land use changes were modeled in the Land Change Modeler (LCM) through binary logistic regression with fit indexes of the satisfactory models. Transition potecncials maps and change trends were produced, and a prognosis scenario for the year 2047 for each of the anthropic processes was generated using the Markov Chains. The highest losses (17%) were estimated for the forested areas until 2047, in the absence of environmental governance actions. Approximately 3.1 Tg of Carbon (25.3%) were lost considering the estimated COT stock in 2017 and the remaining areas of the main uses in 1984, confirming the hypothesis established for the study. In general, the "cattle-raising" and the primitive management of the lands in this Amazonian environment, left as a legacy the precarious subsistence of small producers, who still experience the misadventures of agricultural colonization projects conducted in the past, in addition to the inestimable losses to the local biodiversity.
|
