Parameterizing biochar effect on climate-smart agriculture using artificial intelligence and land surface model
Sabrina Wilson1, Sagar Gautam2, Umakant Mishra2, Yang Song1
1 Department of Hydrology and Atmospheric Science, the University of Arizona
2 Sandia National Lab
Agricultural land expansion has significant biogeophysical and biogeochemical implications for climate change, calling for developing climate-smart agricultural practices. One of the broad applied soil management practices, biochar addition, has the potential to sequestrate carbon, hold water, and increase nutrient availability. To date, increased model efforts have outlined long-term single characterized biochar addition to a few sites, whereas there is still a lack of clear understanding about the effect of biochar addition on soil chemical and physical processes and the consequent impact on greenhouse gas (GHG) emission, carbon sequestration, and crop productivity, much less a comprehensive upscaling impact over diverse environments. To address this knowledge gap, we explored the heterogeneity of biochar’s effect by integrating global-scale biochar addition experiments with process-based and artificial intelligence (AI) models. We incorporated biochar decomposition dynamics and adsorption capacity into the Community Land Model (CLM5.0). We integrated a total of 359 biochar addition experiments from 69 diverse locations across the globe to train an AI-based surrogate model for parameterizing the biochar effect on the decomposition of natural soil organic matter (SOM), GHG emissions, and crop yields and applied the coupled CLM-AI model to investigate the heterogeneity of biochar’s effect. We found that the carbon benefit of biochar addition would be more obvious in nutrient-deficient and water-limited soils due to increased SOM and inorganic nutrient adsorption and water-holding capacity by biochar, additional carbon input, and improved nutrient availability resulting from biochar-derived nitrogen mineralization.