Cropland expansion in the United States produces marginal yields at high costs to wildlife

The study addresses the resurgence of cropland expansion in the United States since the mid-to-late 2000s and examines its persistence and consequences for agricultural production and wildlife habitat. Despite initial drivers such as high commodity prices, biofuel industry growth, and reductions in conservation program extent, key uncertainties persist regarding ongoing expansion rates, the productivity of newly cultivated lands, and impacts on biodiversity. The authors hypothesize that contemporary cropland expansion is increasingly occurring on marginal land, resulting in diminishing production gains alongside significant ecological costs. To test this, they map annual cropland expansion and abandonment across the US from 2008 to 2016, evaluate expected yields on new versus existing croplands, and assess impacts on habitats critical to Monarch butterflies, nesting waterfowl in the Prairie Pothole Region, and long-term grasslands harboring native plant communities.

Prior work documents renewed conversion of grasslands and natural areas to cropland in the US beginning in the late 2000s, coinciding with market and policy shifts. Studies show US cropland expansion can cause significant carbon emissions, reducing climate benefits compared to expansions in some other regions. Grassland conversion is linked to losses in pollinators and birds, with grasslands supporting substantially more Monarch milkweed pods and serving as prime breeding habitat for North American ducks. Remaining unplowed grasslands contain high native plant diversity. A UN assessment identifies agricultural land use change as a major driver of global biodiversity loss. Earlier analyses have identified regional hotspots of conversion and noted that lands moving into or out of the Conservation Reserve Program tend to have lower-than-average yields. However, detailed, high-resolution assessments of the persistence of expansion, relative productivity, and explicit wildlife habitat impacts across the US have been lacking.

Study design focused on three questions: the annual spatiotemporal patterns of cropland expansion post-2008; the yield performance of new croplands relative to existing fields at national and local scales; and the impacts on selected wildlife habitats (Monarch milkweed, duck nesting accessibility in the Prairie Pothole Region, and long-term grasslands). Conversion detection: The authors used the USDA Cropland Data Layer (CDL, 2008–2017) and NLCD (2001, 2006, 2011) to create binary cropland/noncropland time series, applying majority spatiotemporal filtering and class-specific rules to identify annual conversions at 30 m resolution. Conversions required evidence of noncropland for 6–10 years prior and cropland for two years post-conversion; intermittent cropland was separately classified. A five-acre minimum mapping unit was applied to aggregated transition classes to reduce noise. Confused classes (e.g., fallow/idle cropland versus hay/pasture) and unlikely transitions (developed to cropland) were excluded; special treatments addressed perennials and rice. Yield modeling: Crop-specific random forest models (for corn, soybeans, wheat) predicted representative yields based on county-level yield data (2008–2017) and gridded biophysical covariates, including TerraClimate monthly climate and water balance metrics, soils productivity indices (NCCPI), and topography (elevation, slope, aspect). Models were trained (~13,000 records per crop) with 30% held out for validation; tuned parameters were 250 trees, mtry=21, node size=5. Predictions produced expected average yields for new and existing croplands. National yield differential was computed as deviation from the modeled national average across existing croplands for each crop; local differentials compared new cropland yields to those of existing croplands within 10 km × 10 km neighborhoods. Converted land characteristics were assessed using slope (NED), land capability class and hydric soils (gSSURGO), and climate water deficit (TerraClimate). Wildlife habitat impacts: Milkweed stem losses were estimated following Pleasants (2016), with explicit accounting for conversion of CRP grasslands to better approximate pre-conversion milkweed densities and uncertainties. Duck breeding pair nesting accessibility was estimated using USFWS thunderstorm maps (ca. 2012), assigning midpoint values for category ranges and 110 pairs/mi² for >100 category; accessibility was averaged over the PPR by transition class. Long-term grasslands (unplanted, uncultivated for ≥25 years) were identified using NLCD (1992, 2001, 2006, 2011) by excluding pixels ever classified as cropland (82) or pasture/hay (81). Accuracy estimation multiplied state- and class-specific superclass accuracies for pre- and post-conversion years; expected user’s and producer’s accuracies for detected changes were 71.0–86.9%. Results were compared with USDA NRI (2007–2015), USGS NLCD (2008–2016), and USDA Census of Agriculture (2007–2017).

• Extent and rates: From 2008–2016, 10.09 million acres were converted to cropland; 3.52 million acres of cropland were abandoned/converted to noncrop. Gross expansion peaked in 2011 at 1.94 million acres and stabilized near ~1 million acres/year during 2013–2015. Net annual cropland change ranged from 0.38 to 1.33 million acres, indicating persistent expansion.
• Sources and geography: Grasslands (including pasture and hay) accounted for 88% of land converted to crops. Hotspots included the Prairie Pothole Region (ND/SD), Dissected Till Plains (IA/MO), High Plains (KS/OK/TX), with emerging activity along the Canadian border of the Northern Great Plains and the Interior Low Plateau (KY/TN). Abandonment was greatest along the Mid-Atlantic, Gulf Coast, and parts of the Pacific Northwest.
• Yield performance (national): New croplands had lower expected yields than existing croplands nationally for corn (-10.9%, SDspatial 13.8%) and soybeans (-8.4%, SDspatial 14.9%), and slightly higher for wheat (+1.3%, SDspatial 21.1%). Overall, 69.5% of new cropland areas produced yields below national averages for the respective crops.
• Yield performance (local): Compared to nearby existing croplands within 10 km × 10 km neighborhoods, new croplands averaged lower yields: corn -1.1% (SDspatial 1.8%), soybeans -0.6% (SDspatial 1.1%), wheat -0.7% (SDspatial 2.9%). Local deficits were larger in highly cultivated areas with little remaining natural land.
• Biophysical marginality: New croplands had steeper slopes (mean 3.35% vs 2.00%), experienced 3.3% higher climate water deficit, and were less frequently on hydric soils (8.10% vs 19.19%) than existing croplands, indicating more erosive and water-limited conditions.
• Monarch habitat impacts (Midwest): Approximately 220 million (SE ±189 million) milkweed stems were lost due to conversion of grasslands, wetlands, and shrublands to corn/soy across 2008–2016, equating to 8.5% of the 2008 regional total. Converted lands had an average of 53.7 (±46.0) stems/acre prior to conversion—3.4× higher than the 15.6 (±10.4) stems/acre on remaining existing natural lands.
• Waterfowl habitat impacts (PPR): Converted areas had estimated accessibility of 42.7 breeding pairs/mi² (SDspatial 30.6), nearly double that of existing croplands (22.9; SDspatial 24.7) and 37% greater than other unconverted habitat (31.2; SDspatial 30.2). Losses amounted to 138,000 nesting opportunities (2.8% of the regional total).
• Long-term habitat loss: 2.8 million acres (28% of new cropland) came from land not cultivated or planted to pasture/hay for at least 25 years; 2.3 million acres (81% of that long-term habitat) were unimproved grasslands. Relative shares of converted long-term habitat: 26% of grasslands, 29% of wetlands, 44% of forest, and 52% of shrublands.
• Cross-dataset corroboration: Spatial patterns and overall trends of expansion align with independent estimates from USDA NRI, USGS NLCD, and USDA Census of Agriculture, though this study’s gross conversion estimates are generally more conservative.

The findings confirm persistent cropland expansion across the US and demonstrate that new cultivation generally occurs on biophysically marginal lands, leading to lower yield returns compared to existing croplands, both nationally and locally. This dual marginality—poorer land quality in less suitable regions—dampens production gains from extensification. Areas with little remaining uncultivated land show more pronounced local yield deficits, reflecting land scarcity and competition, while regions with more natural cover exhibit yields far below national averages despite less local competition. The biophysical traits of newly cultivated land (steeper slopes, higher water deficits, fewer hydric soils) imply greater agronomic risks, potential cultivation constraints, and elevated environmental costs (e.g., erosion, water quality impacts, drought vulnerability or irrigation pressure). Ecologically, expansion disproportionately removes high-quality habitats: milkweed-rich grasslands for Monarchs, high-accessibility nesting habitats for ducks in the PPR, and longstanding native grasslands, shrublands, wetlands, and forests, implying outsized biodiversity impacts. These trade-offs support conservation strategies that prioritize at-risk, high-value habitats and suggest that projections of future supply and area–yield relationships should incorporate diminishing yield returns on newly converted land. More broadly, the results strengthen arguments for emphasizing intensification and demand-side measures over further cropland expansion to meet agricultural demand while safeguarding ecosystem services.

This study provides a high-resolution, annual national assessment showing that US cropland expansion since 2008 has continued at more than one million acres per year, predominantly converting grasslands and other natural habitats. Newly cultivated lands generally yield less than existing croplands and are characterized by steeper slopes and greater water limitations, while conversion disproportionately eliminates high-quality wildlife habitats, including milkweed-rich Monarch breeding areas, prime duck nesting landscapes in the PPR, and long-term native grasslands. These findings imply that continued extensification will likely exacerbate agronomic and ecological costs, bolstering the case for alternative pathways to meet demand: improving agricultural efficiency, closing yield gaps, reducing waste, and shifting diets, alongside conservation policies and supply-chain interventions that protect remaining intact habitats. Future research should expand to additional taxa and assess cumulative biodiversity, climate, water use, and water quality trade-offs of land conversion versus intensification at multiple ecological and spatial scales.

Results depend on the classification accuracy of satellite-derived land cover datasets (CDL and NLCD) and the change-detection algorithm. Challenges include distinguishing active fallow from abandonment and differentiating short-term idling from long-term abandonment; class confusions (e.g., hay/pasture vs. fallow/alfalfa) required exclusions. A five-acre minimum mapping unit and conservative rules may undercount some conversions. Expected user’s and producer’s accuracies for change detection ranged from 71.0% to 86.9%. Yield models predict average expected yields for 2008–2017 conditions and do not capture year-to-year weather variability, specific management practices (e.g., irrigation distinctions), or genetic advances; wheat varieties were not differentiated. Milkweed loss estimates carry substantial uncertainty (large SE), and actual densities on some grassland types may differ from assumptions. Comparisons with other datasets involve differing definitions (gross vs. net change, crop class coverage), which may affect magnitude comparisons.