Linking remote sensing and soil properties to model mercury spatial patterns in a natural reserve in the Amazon rainforest
摘要
Mercury (Hg) accumulation in Amazonian soils is influenced (governed) by atmospheric deposition, hydrological processes, and the rapid turnover of organic matter. Understanding its spatial distribution is critical for assessing ecosystem vulnerability and informing conservation strategies in tropical forests. We collected thirty surface soil samples within a 5 × 5 km grid in a protected area of the Amazon rainforest to investigate spatial relationships of total mercury (THg) concentrations. Variogram analysis and geostatistical interpolation—Ordinary Kriging (OK) and Co-Kriging (CK)—were applied to characterize spatial autocorrelation. Remotely sensed variables, including Gross Primary Productivity (GPP) and Height Above Nearest Drainage (HAND), were integrated alongside soil classification and soil organic matter (SOM) to improve predictive accuracy. THg concentrations, determined by cold vapor generation atomic absorption spectrophotometry, averaged 60.51 ± 18.16 μg kg⁻1, with Oxisols and Ultisols showing comparable means (~ 62 μg kg⁻1). CK models incorporating GPP and HAND achieved the lowest mean absolute error (MAE = 0.027), with similar spatial ranges for THg, GPP, and HAND supporting their joint predictive capability. GPP alone was a significant predictor (MAE = 0.04), consistent with links to atmospheric Hg deposition, terrain morphology, and sediment redistribution. Spatial patterns revealed Hg hotspots in intermediate topographic zones where organic and inorganic leaching (podzolization) and high humic activity occur, while Hg losses were associated with transport from upland plateaus and seasonally flooded areas. These results demonstrate that coupling geostatistical modeling with remote sensing products provides a robust approach for predicting Hg distribution in Amazonian soils and can inform environmental monitoring and management efforts in sensitive tropical ecosystems.