Reducing Uncertainties in Net Carbon Capture to Advance Wetlands as Natural Climate Solutions

Abstract Wetlands play a crucial role as natural climate solutions (NCS) by sequestering atmospheric carbon dioxide (CO 2 ) in the form of organic carbon (OC) in soils. However, spatial heterogeneity and temporal variability in OC sequestration rates introduce uncertainties that must be addressed to inform climate policy and meet national climate targets. This study integrates expert knowledge with statistical learning techniques to develop localized estimates of OC sequestration rates in wetlands within agricultural landscapes. Experts identified direct process controls—including carbon quantity and quality, cation exchange capacity, aggregate reactivity, redox potential, and air temperature—along with human activities that influence these controls. Using geospatial proxies for these variables, alongside field‐based OC sequestration measurements, we compared and trained statistical learning models that achieved high predictive accuracy (adjusted R 2 ≥ 0.70, MAE ≤ 0.15 Mg C ha −1 yr −1 , RMSE ≤ 0.19 Mg C ha −1 yr −1 ). Variable importance analysis identified wetland inundation probability (a proxy for wetland redox potential), Human Impact Index (a proxy for wetland carbon quantity and quality), and landscape soil properties (proxies for wetland soil texture, cation exchange capacity, and aggregate reactivity) as the most influential predictors explaining the spatiotemporal variation in OC sequestration rates. This study demonstrates that statistical learning models, informed by expert knowledge of process controls, can estimate OC sequestration rates within wetlands, providing potentially critical data to guide policy development and wetland management as effective NCS strategies.