SOILS-AI

Campaign name: Soil Organic Indicators at Large Scale for Artificial Intelligence (SOILS-AI)

A key barrier to scaling up soil organic carbon (SOC) dynamics is the inherent spatial variability of soils, along with the lack of standardized molecular-level data for Earth system model prediction from local to continental scale. Accurately representing SOC processes at such scales requires integrating meter-scale variability into model parameterization. Pedotransfer functions provide a systematic framework for linking fine-scale observational data to model parameters, enabling the upscaling of biogeochemical processes and SOC dynamics.  

The Environmental Molecular Sciences Laboratory (EMSL) is leading a campaign, which supports the recent presidential memo on FY 2027 national research and development priorities and demonstrates how AI accelerates scientific discovery within Earth sciences. The "Soil Organic Indicators at Large Scale for Artificial Intelligence (SOILS-AI)" campaign is a sampling effort across the continental United States designed to improve the representation of high-value soil taxa in the Molecular Observation Network (MONet) database, creating and publicizing a massive structured scientific dataset of soils and subsurface molecular measurements for AI model training. For example, AI models trained on preliminary MONet data have already identified critical data gaps in existing predictive models for SOC, particularly in estimating carbon use efficiency and soil respiration at regional scales.  

Instruments and resources 

Soil cores will be collected using the standard MONet collection methods and will be analyzed using MONet workflows. Using EMSL computing resources, the campaign will develop new soil-based metrics. 

Contact 

If you have questions or are interested in learning more about how you can participate, contact Emily Graham.