The Amazon rainforest significantly influences rainfall patterns, a crucial ecosystem service supporting agriculture in the region and beyond, particularly in the Southern Brazilian Amazon (SBA), a major producer of soy and beef. Despite this, deforestation in the SBA is increasing, reaching 30% forest loss by 2020. This raises concerns about the impact on rainfall and agriculture. While prior research suggests a link between deforestation and altered rainfall, the specific threshold of forest loss beyond which rainfall decreases remains uncertain. This study aims to empirically determine this threshold at various geographical scales and evaluate the resulting economic consequences for agriculture under different governance scenarios. The study's significance lies in its potential to inform policies aimed at balancing agricultural expansion with environmental conservation in the Amazon.

Existing literature indicates that the Amazon rainforest plays a crucial role in regulating regional rainfall. Studies suggest both linear and non-linear relationships between deforestation and rainfall. Some studies propose that small-scale deforestation can initially enhance local rainfall due to changes in mesoscale moisture circulation. However, as deforestation progresses, it's hypothesized that a critical threshold is reached, beyond which further forest loss leads to a sharp decline in rainfall. Regional-scale modeling studies have estimated this threshold to be between 30% and 50% forest loss, depending on deforestation patterns. However, prior to this study, there was a lack of empirical evidence using observed rainfall data to confirm this threshold and its consequences.

This study analyzed the relationship between annual rainfall and Amazon forest loss from 1999 to 2019 using a multiscale approach. Rainfall data were obtained from the Tropical Rainfall Measuring Mission (TRMM) satellite, while deforestation data came from the Program to Calculate Deforestation in the Amazon (PRODES). To account for spatial and temporal variability in rainfall, the researchers calculated rainfall anomalies by detrending the effects of geographic location, elevation, and interannual variability. These anomalies were then correlated with forest loss at different grid cell sizes (28 km, 56 km, 112 km, and 224 km). A multivariate adaptive regression spline (MARS) model was used to identify the nonlinear relationship between forest loss and rainfall. Future deforestation scenarios were simulated using the Otimizagro model under weak and strong environmental governance scenarios. Finally, the economic impact of these scenarios on agricultural revenue (soybean and beef) was assessed by considering projected yield losses due to rainfall reduction.

The analysis revealed a non-linear relationship between forest loss and rainfall, varying with geographical scale. At the 28 km scale, rainfall initially increased with forest loss up to ~58%, but then decreased sharply beyond this threshold. The threshold was lower at larger scales (48% at 56 km and 23% at 112 km). At the 224 km scale, a linear decrease in rainfall with forest loss was observed. This dual response is consistent with the theory of deforestation breeze effects at smaller scales and albedo changes/reduced evapotranspiration at larger scales. The study also showed that widespread deforestation leads to a negative-sum game, with total rainfall reduction outweighing local gains. The simulations predicted that under a weak governance scenario, 55% of 28x28 km grid cells in SBA could reach the critical deforestation threshold by 2050. This could lead to significant economic losses in agriculture, with estimated annual losses of up to US$1 billion from reduced soybean and beef productivity. Stronger governance could significantly mitigate these losses.

The findings provide strong empirical evidence supporting the hypothesis that extensive deforestation in the Amazon leads to a decline in rainfall and reduced agricultural productivity. The identification of critical deforestation thresholds at different scales is a crucial contribution, highlighting the importance of considering spatial context in assessing the impacts of deforestation. The economic analysis demonstrates that the ecological consequences of deforestation translate into substantial economic losses, suggesting that current agricultural expansion practices are unsustainable. The results underscore the need for effective environmental governance to balance economic development with environmental conservation.

This study empirically demonstrates the negative impact of deforestation on rainfall and agricultural revenue in the Southern Brazilian Amazon. The identification of critical deforestation thresholds at various scales provides valuable insights for policy-making. The significant economic losses projected under a weak governance scenario highlight the urgent need for stronger environmental regulations and sustainable land management practices. Future research could focus on refining the models used, exploring the impacts of climate change in conjunction with deforestation, and investigating adaptation strategies for agriculture in a changing Amazon.

While the study utilizes high-quality data and sophisticated models, some limitations exist. The analysis relies on correlations, and while efforts were made to address potential confounding factors, it's difficult to completely rule out other influences on rainfall patterns. The economic assessment is based on current prices and productivity levels and may not fully capture future uncertainties. The model's accuracy depends on the accuracy of the underlying deforestation and land use projections.