Vital for sustainable soils: Revealing microbial enzyme hotspots across Brazilian farmlands

The extracellular enzymatic activities of soil microorganisms are critical in driving carbon, phosphorus, and nitrogen cycles, which have a significant impact on environmental, agricultural, and climatic conditions. Leveraging proximal and remote sensing technologies, coupled with machine learning techniques, we predicted and mapped soil enzyme activities across 3.4 million km 2 of Brazilian agricultural landscapes. By using the Recursive Feature Elimination algorithm, we identified specific spectral bands in the Vis-NIR-SWIR and Mid-IR ranges that correlate with three key enzymes: beta-glucosidase, acid phosphatase, and urease. Using these bands, we formulated transfer functions for the enzyme activities that allow us to mine the Brazilian Soil Spectral Library, and upscale it to produce national maps at 30 m spatial resolution. These maps showed strong correlations with point observations of enzyme activity. We observed significant associations between these enzymes and Fe and Al oxides, enzyme patterns aligned with Brazil's soil mineralogy, particularly hematite and kaolinite distributions, and a direct link with soil carbon. Land use also influenced enzyme activities; notably, areas under forest, soybean, and sugarcane use showed higher activity levels. However, urease activity decreased in areas treated with nitrogen fertilizers. Our findings demonstrate how satellite-based enzyme monitoring can transform national-scale soil health monitoring and management.