Journal Article

Scaling High‐Resolution Soil Organic Matter Composition to Improve Predictions of Potential Soil Respiration Across the Continental United States

Bowman, Maggie, Karra, Satish, Qafoku, Odeta, Zhao, Qian, Shi, Cheng, Toyoda, Jason, Kew, Will, Graham, Emily B., Mudunuru, Maruti, Bargar, John R., Corilo, Yuri

2025

Overview

Despite the importance of microbial soil organic matter (SOM) respiration in regulating the flux of carbon between soils and the atmosphere, soil carbon cycling models remain primarily based on climate and soil properties, leading to large uncertainty in predictions. To address this knowledge gap, we analyzed high‐resolution water‐extractable SOM profiles from soil cores collected across the United States by the 1,000 Soils Pilot of the Molecular Observation Network. Our innovation lies in using machine learning to distill thousands of SOM formula into tractable units; and it enables integrating data from molecular measurements into soil respiration models. In surface soils, SOM chemistry provided better estimates of potential soil respiration than soil physicochemistry, and using them combined yielded the best prediction. Overall, we identify specific subsets of organic molecules that may improve predictions of global soil respiration and create a strong basis for developing new representations in process‐based models. Plain Language Summary Soil organic carbon is one of the largest and most active pools in the global carbon cycle. Microbial decomposition of soil organic matter (SOM) releases large amounts of carbon dioxide (CO2) to the atmosphere. This process is soil microbial respiration. To evaluate if SOM composition can improve predictions of soil respiration, we collected soils from across the continental US, and analyzed both soil physicochemistry and molecular SOM composition, as part of the Molecular Observation Network. We developed machine learning based workflow to extract key SOM signatures and used the SOM signatures to predict potential rate of soil respiration, compared to standard soil physicochemistry. The results suggested that SOM composition improved the prediction of potential soil respiration in surface soils, where most soil carbon is actively cycled. In deeper soils, model performance was not improved by SOM chemistry, possibly due to the greater importance of mineral‐associated SOM. Our results identified key SOM molecules in predicting potential soil respiration and supported the significance of SOM dynamics in future development of soil carbon models. Key Points Molecular measurements of dissolved soil organic matter (SOM) are critical to accurately predict soil respiration at the continental scale Machine learning extracts key molecules from complex high‐resolution SOM profiles to explain differences in potential soil respiration Dissolved SOM profiles provide more power in predicting potential respiration in surface soils than subsoils

• Publisher: John Wiley & Sons, Inc,American Geophysical Union (AGU),Wiley
• Subject: Atmosphere / Carbon / Carbon cycle / Carbon dioxide / climate change / Composition / Cores / ENVIRONMENTAL SCIENCES / Learning algorithms / Machine learning / Microorganisms / Organic carbon / Organic chemistry / Organic matter / Physicochemical properties / Predictions / Respiration / Signatures / Soil analysis / soil carbon / Soil chemistry / Soil organic matter / Soil properties / Soil respiration / Soil surfaces / Soils / Workflow