Soil erosion in agricultural landscapes reduces crop yields, leads to loss of ecosystem services, and influences the global carbon cycle. Despite decades of soil erosion research, the magnitude of historical soil loss remains poorly quantified across large agricultural regions because preagricultural soil data are rare, and it is challenging to extrapolate local-scale erosion observations across time and space. Here we focus on the Corn Belt of the midwestern United States and use a remote-sensing method to map areas in agricultural fields that have no remaining organic carbon-rich A-horizon. We use satellite and LiDAR data to develop a relationship between A-horizon loss and topographic curvature and then use topographic data to scale-up soil loss predictions across 3.9 × 10⁵ km² of the Corn Belt. Our results indicate that 35 ± 11% of the cultivated area has lost A-horizon soil and that prior estimates of soil degradation from soil survey-based methods have significantly underestimated A-horizon soil loss. Convex hilltops throughout the region are often completely denuded of A-horizon soil. The association between soil loss and convex topography indicates that tillage-induced erosion is an important driver of soil loss, yet tillage erosion is not simulated in models used to assess nationwide soil loss trends in the United States. We estimate that A-horizon loss decreases crop yields by 6 ± 2%, causing $2.8 ± $0.9 billion in annual economic losses. Regionally, we estimate 1.4 ± 0.5 Pg of carbon have been removed from hillslopes by erosion of the A-horizon, much of which likely remains buried in depositional areas within the fields. Significance Conventional agricultural practices erode carbon-rich soils that are the foundation of agriculture. However, the magnitude of A-horizon soil loss across agricultural regions is poorly constrained, hindering the ability to assess soil degradation. Using a remote-sensing method for quantifying the absence of A-horizon soils and the relationship between soil loss and topography, we find that A-horizon soil has been eroded from roughly one-third of the midwestern US Corn Belt, whereas prior estimates indicated none of the Corn Belt region has lost A-horizon soils. The loss of A-horizon soil has removed 1.4 ± 0.5 Pg of carbon from hillslopes, reducing crop yields in the study area by ∼6% and resulting in $2.8 ± $0.9 billion in annual economic losses.

Erosion degrades soils, reduces crop yields, and diminishes ecosystem services. The total amount of soil that has been eroded since the initiation of farming is unknown in most agricultural landscapes, which hinders assessment of soil erosion trends. In the Midwestern U.S., erosion has caused native prairie remnants to become perched above surrounding farmland, providing an opportunity to measure historical soil erosion rates. We use high‐resolution topographic surveys conducted across erosional escarpments at the boundary between 20 prairies and adjacent agricultural fields and show the median reduction in soil thickness ranges from 0.04 to 0.69 m, corresponding to erosion rates of 0.2–4.3 mm year⁻¹_, with a median value of 1.9 mm year⁻¹. We used an association between the measured reduction in soil thickness and topographic curvature to predict regional soil erosion integrated since the beginning of farming to the present. We estimate a median historical erosion rate of 1.8 ± 1.2 mm year⁻¹, which is nearly double the rate considered tolerable by the U.S. Department of Agriculture (USDA). Current soil loss predictions from the USDA National Resources Inventory (NRI) and the Daily Erosion Project (DEP) are lower than our historically averaged erosion rate by a factor of 3 and 8, respectively. We suggest that the NRI and the DEP underpredict soil loss rates because they do not include tillage erosion, a process shown to be important throughout the Midwestern U.S. Our findings indicate that further implementation of conservation practices is needed to reduce the high centennial‐averaged soil erosion rates that we measure to sustainable levels.

Permafrost soils are a critical component of the global carbon cycle and are locally important because they regulate the hydrologic flux from uplands to rivers. Furthermore, degradation of permafrost soils causes land surface subsidence, damaging infrastructure that is crucial for local communities. Regional and hemispherical maps of permafrost are too coarse to resolve distributions at a scale relevant to assessments of infrastructure stability or to illuminate geomorphic impacts of permafrost thaw. Here we train machine learning models to generate meter‐scale maps of near‐surface permafrost for three watersheds in the discontinuous permafrost region. The models were trained using ground truth determinations of near‐surface permafrost presence from measurements of soil temperature and electrical resistivity. We trained three classifiers: extremely randomized trees (ERTr), support vector machines (SVM), and an artificial neural network (ANN). Model uncertainty was determined using k‐fold cross validation, and the modeled extents of near‐surface permafrost were compared to the observed extents at each site. At‐a‐site near‐surface permafrost distributions predicted by the ERTr produced the highest accuracy (70%–90%). However, the transferability of the ERTr to the sites outside of the training data set was poor, with accuracies ranging from 50% to 77%. The SVM and ANN models had lower accuracies for at‐a‐site prediction (70%–83%), yet they had greater accuracy when transferred to the non‐training site (62%–78%). These models demonstrate the potential for integrating high‐resolution spatial data and machine learning models to develop maps of near‐surface permafrost extent at resolutions fine enough to assess infrastructure vulnerability and landscape morphology influenced by permafrost thaw.