Contribution of small agricultural reservoirs to changing carbon cycles in the Amazon-Cerrado region.

Damming small tropical streams has created hundreds of thousands of artificial reservoirs across Brazil's agricultural frontier, a lasting legacy of the country's long history of cattle ranching. The low-oxygen conditions in these water bodies favors production of methane - a potent greenhouse gas twenty-eight times more powerful than carbon dioxide. Although they are widespread and can persist for decades, we know little about the factors controlling methane production in reservoirs, or their collective contribution to the global carbon budget. This study combines field measurements, remote sensing, and modeling to quantify annual CH₄ and CO₂ emissions from small agricultural reservoirs; evaluate the spatial-temporal dynamics of reservoirs in agricultural landscapes; and estimate the individual and cumulative contributions of agricultural reservoirs to Brazil's carbon budget.

To estimate average CH₄ emissions at the reservoir scale, we sampled monthly diffusive fluxes, surface water concentrations, and ebullition from six representative reservoirs over the course of a year (March 2018-May 2019). Sampling took place at Tanguro Ranch, a large soybean farm in Mato Grosso, Brazil. Results showed strong seasonal trends, with the highest diffusive fluxes (~50 mg-C m^-2 d^-1) observed during the onset of the rainy season (November) and the lowest (~4 mg-C m^-2 d^-1) at the peak of the dry season (June-July). Ebullition may account for as much as 50% of total emissions, but fluxes were highly variable over time and space. Drone surveys of over 25 nearby reservoirs suggested that reservoir volume is linearly correlated with area, which ranged from 800-100,000 m² (median ~9,000 m²). To evaluate cumulative regional CH₄ emissions from reservoirs, we developed an automated deep learning algorithm, using Sentinel 1 and 2 satellite data to map over 54,000 man-made reservoirs across Mato Grosso. Applying this approach to the satellite time series for the Amazon-Cerrado region will allow us to estimate cumulative emissions and track changes in the number, size, and distribution of man-made reservoirs. Ultimately, this research could improve understanding of which pathways dominate methane fluxes, and help identify potential management strategies to reduce methane emissions in agricultural landscapes.