Digging For Gold

"May all of you turn your eyes to us! We have been suffering along with the forest! The entire forest has! The forest has died! Now the forest is dead. It's been a long time since they killed this forest. They destroyed all the trees we used to eat fruit from! They cut down all the big trees! And who did that? The miners did! They killed them all! Our land is completely dead!" - Statement from a Yanomami leader recorded by artist Richard Mosse in Palimiu in June 2021.

The Brazilian Legal Amazon (BLA) is home to the world's largest remaining contiguous rainforest on Earth (Diniz, 2015; Rosa, 2013). Anthropogenic degradation of the Amazon is associated with different land use patterns, land cover histories, anthropogenic forces, and players. This study utilizes Geographic Information Systems (GIS) to analyze the geospatial factors that result in forest degradation and deforestation in the state of Roraima. Preliminary GIS analysis of the Brazilian Legal Amazon revealed that Roraima is an illegal mining hotspot. The study employed a predictive logistic regression model to investigate the relationship between deforestation and four geospatial inputs: slope, roads, indigenous communities, and illegal gold mining sites.

Evaluating environmental degradation in the Amazon rainforest informs a broader audience on where we stand in terms of the protection of the Amazon, which is important at both the global and the local scale; the global community relies on the Amazonian ecosystem for carbon sequestration and evapotranspiration. At the local level, the Amazon is home to various communities which rely on the Amazon basin for their subsistence. This study will primarily investigate the effects of environmental degradation on Indigenous communities in Roraima. Roraima is located in the North-Western region of the Brazilian Legal Amazon, covering approximately 223 km ²  and it is home to approximately 33 Indigenous communities. The incentive behind this research is to use GIS as a socio-environmental monitoring tool.

Artisanal gold mining emits over 200 metric tons of mercury, annually, into the Amazon basin (Crespo-Lopez, 2021). Mercury is used in gold mining because of its ability to bind with gold particles present in riverbed sediments. Miners can later separate the gold from the resulting mercury-gold amalgam by melting the metals at 90°C or 194 °F (Ahern, Adams, M. D. ) .

Mercury vapor can travel in the air over long distances and is usually deposited on trees. Trees can effectively sequester mercury from vapor fumes. Regrettably, extensive deforestation in the Amazon has reduced the air filtration capacity provided by trees. This has led to increased exposure of human populations and other forest biota to substantial amounts of this toxic metal. Logging and burning of the forest causes mercury vapor to be transported into the atmosphere at accelerated rates. Once mercury meets clouds and rain, it can be further transported along waterways and deposited into distant soils.

Another anthropogenic factor known to exacerbate mercury mobilization and biotransformation in the Amazon is hydroelectric dams. Dams create stagnant water which is ripe with bacteria. Bacteria can convert inorganic mercury (Hg ²⁺ ) into methylated mercury (MeHg) (Crespo-Lopez, 2021). Methylated mercury is the most toxic form of mercury. After being digested by bacteria, methylated mercury bioaccumulates along the riverine food chain (Crespo-Lopez, 2021). Riverine communities are largely exposed to mercury (Hg) primarily through the consumption of aquatic biota (Crespo-Lopez, 2021).

Mercury intoxication causes serious human health issues such as damage to the skin, the eyes, lungs, kidneys, the cardiovascular system, the immune system, the central nervous system, and the digestive system (Crespo-Lopez, 2021). Moreover, when food gets contaminated with mercury, it becomes a serious food security issue, especially affecting riverine communities that heavily depend on local food sources for their subsistence. According to a study conducted in 2022 by Meneses et al., riverine populations are disproportionally exposed to higher levels of this metal (Meneses, 2022). Meneses et al. demonstrated that mercury exposure was 90% higher in riverine areas in the BLA in comparison to their urban counterparts, where the exposure rate was 57.1%. (Meneses, 2022).

Preliminary site analysis of the BLA revealed a spatial cluster of illegal mining in the North-Western region of the Brazilian Legal Amazon. Hence, the scope of the project was narrowed down to focus on the state of Roraima. Roraima is a hotspot for illegal mining thus it is an optimal site to investigate the relationship between mining and deforestation in the region.

H _a : illegal mining is a statistically significant driver of deforestation in the BLA.

H _o : illegal mining is not a statistically significant driver of deforestation in the BLA.

Preliminary site analysis identified 112 illegal mining sites within Indigenous communities in Roraima. 

A comprehensive model of predictive deforestation was developed by referencing similar studies which made use of ArcGIS Pro and the statistical modeling software R, to predict future deforestation patterns with spatially explicit inputs (Bavaghar, 2015) (Rosa, 2013) (Ornat, Smith , & Zajaczkiwsky, 2020). Model results were used to identify at-risk areas or hot spots of environmental degradation to inform future sites of environmental protection. Selected shapefiles were reprojected to the Albers Equal Area Conic Continental projection with a South American 1969 datum enabling seamless data processing across several UTM zones (Griffiths, 2018). The PRODES deforestation data from 2008-2022 was selected as the main raster processing environment, roughly 30 x 30 meters, cell centroid and nearest neighbor interpolation.

The probability model engendered in this study is a business-as-usual scenario, assuming existing deforestation patterns and variable inputs remain constant. This model differs from existing models by accounting for illegal mining and Indigenous territories as predictive variables. These variables were selected based on data availability as well as preliminary research.

Beta coefficients derived from this model resulted in a probability deforestation raster which elucidates future deforestation hotspots in Roraima.

Project results failed to reject the null hypothesis that mines were not statistically significant predictors of deforestation. However, the model revealed that roads (p value = 7.025e-05) and indigenous communities (p value = 0.0002653) were the most statically significant predictors of deforestation in Roraima at the 95% confidence interval level.

A model based on Euclidean distance may not be the most optimal model for assessing the complexity of all the inputs that lead to environmental degradation and subsequently full forest conversion into deforested land. Future studies should consider the effects of certain factors, such as the relationship between waterways, mining sites, forest degradation and deforestation.

Future research parameters could focus on techniques which are better suited for analyzing higher dimensional models, for example deep convolutional neural networks (CNNs) have been shown to be good predictors of deforestation on pixel level data (Ball, 2022)). The future scope of this analysis should also focus on toxin dispersal modeling of Hg and methylmercury (MeHg), as shown in this research mercury poses a serious risk to human health at both the local and the global level.

Ball, J. G. (2022). Using deep convolutional neural networks to forecast spatial patterns of Amazonian deforestation. Methods in Ecology and Evolution, 13(11), 2622-2634.

Bavaghar, M. P. (2015). Deforestation modelling using logistic regression and GIS. Journal of Forest Science, 61(5), 193-199.

Bonilla, E. X. (2023). Health impacts of Smoke Exposure in South America: Increased Risk for Populations in the Amazonian Indigenous territories.  Environmental Research: Health, 1(2), 021007. .

Crespo-Lopez, M. E.-O.-A.-S. (2021). Mercury: What can we learn from the Amazon? Environment international, 146, 106223.

Diniz, C. G. (2015). DETER-B: The New Amazon Near Real-time Deforestation Detection System. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 8(7), 3619-3628. .

Griffiths, P. J. (2018). Reconstructing Long Term Annual Deforestation Dynamics in Pará and Mato Grosso Using the Landsat Archive. Remote sensing of environment, 216, 497-513. .

Lapola, D. M. (2023 ). The Drivers and Impacts of Amazon Forest Degradation. Science, 379(6630).

Meneses, H. D.-d.-C. (2022). Mercury contamination: A growing threat to riverine and urban communities in the Brazilian Amazon. International Journal of Environmental Research and Public Health, 19(5), 2816.

Ornat, M., Smith , K., & Zajaczkiwsky, S. (2020). Analyzing the Factors of Deforestation in the Amazon Using GIS and Logistic Regression. Ontario, Canada: University of Guelph.

Rajão, R. (2013). Representations and discourses: the role of local accounts and remote sensing in the formulation of Amazonia's environmental policy. Environmental Science & Policy, 30, 60-71.

Rosa, I. M. (2013). Predictive modelling of contagious deforestation in the Brazilian Amazon. PloS one, 8(10), e77231.

Spiegel, S. J. (2012). Mapping spaces of environmental dispute: GIS, mining, and surveillance in the Amazon. Annals of the Association of American Geographers, 102(2), 320-349.

T. O. Assis, A. P. (2022). Projections of Future Forest Degradation and CO2 Emissions for the Brazilian Amazon. Sci. Adv. 8.

Ye’kwana, H. A. (2022). Yanomami under attack: illegal mining on Yanomami Indigenous Land and proposals to combat it. Instituto Socioambiental.