A multi model ensemble reveals net climate benefits from regenerative practices in US Midwest croplands

Abstract

Process-based cropping systems models (CSMs) are key components of measurement, monitoring, reporting, and verification frameworks of carbon markets, but model-specific differences limit their applicability across diverse pedo-climatic conditions and agronomic practices. Multi-model ensemble (MME) provides an opportunity to better estimate changes in soil organic carbon (SOC) and nitrous oxide (N₂O) emissions from agronomic practices at scale. We used an MME across 46 million hectares of US Midwest cropland at a resolution of 4-km² to assess the aggregate ability of different regenerative practices to sequester SOC and N₂O emissions compared to their counterfactual dynamic baselines. MME was validated against long-term trials and compared to its constituent CSMs, showing greater accuracy and lower uncertainty. The results show that adopting no-till combined with cover crops increased SOC stocks by 0.36 ± 0.12 Mg ha^-1 yr^-1, corresponding to a net regional SOC gain of 16.4 Tg C yr^-1 compared to business-as-usual baselines. These benefits are halved when each management is practiced individually, and the SOC gains are only fully realized with low initial carbon stock. By including N₂O emissions, we can assess the overall climate mitigation potential, specifically, the extent to which carbon sequestration can offset direct N₂O emissions. The magnitude of this potential varies depending on management practices and geographic location with net climate benefits on average ranging from 0 to 3 Mg CO₂-eq ha^-1 yr^-1. High-resolution MME results allow for robust estimates of climate mitigation, reducing barriers to carbon market participation and supporting regenerative agriculture initiatives at scale.

Keywords: Carbon credits; Dynamic baselines; Multi-model ensemble; N20 emissions; Regenerative agriculture; soil carbon.

Fig. 1

Representation of the dynamic baseline concept and the mean MME annual SOC change and annual N₂O emissions difference between scenarios. (A) The counterfactual scenario dynamic baselines (red), as well as the regenerative practice lines (green), account for yearly climate variability and site-specific soil properties, while the static baseline reflects the assumption of steady-state conditions. Three different generic cases are provided as examples of potential errors if changes are compared to static baseline and not the dynamic baseline. (B) MME annual soil organic carbon change (Mg ha^-1 yr^-1, 0–30 cm) and (C) annual N₂O emissions (kg ha^-1 yr^-1) as the difference between Scenario 4 (NT FN CC, no-till with full nitrogen fertilization and cover crops) and Scenario 1 (CT FN, conventional till with full nitrogen fertilization). Each UID included in the study (n = 40,000) is reported as a single-colored dot. The gray density histograms represent the distribution of UIDs values across the US Midwest cropland Maps were created using RStudio v2025.05.0.496 https://www.r-project.org.

Fig. 2

Effect of different regenerative practices on county-based MME mean annual SOC change. (A) Effect of no-till on county annual soil organic carbon (SOC) change (Mg ha^-1 yr^-1, 0–30 cm) shown as the difference between Scenario 3 (NT FN, no-till with full nitrogen fertilization) and the baseline Scenario 1 (CT FN, conventional till and full nitrogen fertilization); (B) effect of cover crop on county annual soil organic carbon (SOC) change (Mg ha^-1 yr^-1, 0–30 cm) shown as the difference between Scenario 2 (CT FN CC, conventional till with full nitrogen fertilization and cover crop) and the baseline Scenario 1 (CT FN, conventional till and full nitrogen fertilization); (C) effect of cover crop under no-till on county annual soil organic carbon (SOC) change (Mg ha^-1 yr^-1, 0–30 cm) shown as the difference between Scenario 4 (NT FN CC, no-till, full nitrogen fertilization and cover crop) and the baseline Scenario 3 (NT FN, no-till with full nitrogen fertilization). Baselines serve as counterfactual benchmarks for comparison with different scenarios. Individual counties are represented by their mean weighted values, with weights accounting for the agricultural area percentage of each county’s UID.

Fig. 3

MME mean annual SOC change rates across scenarios based on different soil textures and initial SOC stock levels across 46 M ha. Annual soil organic carbon (SOC) change (Mg ha^-1 yr^-1, 0–30 cm) across scenarios (1 to 4, from left to right), soil texture (Sandy, Loam and Clay), and initial SOC stock level (Low, < 40 Mg C ha^-1; Medium, 40–80 Mg C ha^-1; and High, > 80 Mg C ha^-1). Annual SOC change values derived from all UIDs mean across the US Midwest cropland. All scenarios use maize-soybean crop rotation.

Fig. 4

MME mean annual SOC change response to variations in initial SOC stock, clay content, and latitude across 46 M ha. Each line, represented by a unique color, corresponds to an individual model, with the black dotted line representing the MME median. They represent individual model annualSOC change (Mg ha^-1 yr^-1, 0–15 cm) response to variations in initial SOC stock (Mg ha^-1), clay content (%), and latitude (°). The different scenarios (1 to 4 from top to bottom) are displayed at the right end of each row. Scenario acronyms: CT (conventional till), NT (no-till), FN (full nitrogen), CC (cover crop).

Fig. 5

Climate mitigation (Mg CO₂-eq ha^-1 yr^-1) of Scenario 4 (NT FN CC, no-till with full nitrogen fertilization and cover crops) compared to Scenario 1 (CT FN, conventional till with full nitrogen fertilization). Climate mitigation accounts for N₂O emission and SOC change as reported in Eq. (3). Each UID included in the study (n = 40,000) is reported as a single-colored dot. The gray density histograms represent the distribution of UIDs values across the US Midwest cropland.