Sulfur fertiliser use in the Midwestern US increases as atmospheric sulfur deposition declines with improved air quality

The study addresses how sulfur (S) inputs to Midwestern U.S. agriculture have shifted as atmospheric S deposition has declined due to air quality regulations. Historically, crops obtained much of their S from atmospheric deposition associated with fossil fuel emissions (acid rain), but improved air quality has reduced this free S supply. Concurrently, maize and soybean acreage and yields have increased, raising S demand. The research aims to quantify trends in S fertiliser sales (as a proxy for applications) alongside atmospheric S deposition, and to compare these trends with nitrogen (N), phosphorus (P), and potassium (K) fertilisers. The broader purpose is to evaluate implications for soil S dynamics, environmental consequences, and the need for optimized S management in high-productivity agricultural systems.

Prior research has extensively examined N and P management in agriculture to maximize yields and minimize environmental impacts such as greenhouse gas emissions and eutrophication. Historical studies of acid rain established environmental harms from elevated S deposition, including soil and water acidification and metal mobilization. Recent work documents large declines in atmospheric S deposition in North America and Europe following air quality regulations, and an emerging shift from atmospheric S inputs to agricultural additions. Studies also highlight yield benefits of improved air quality (reductions in ozone, particulate matter, SO2, NO2) for maize and soybean in the Midwest. However, comparable attention to the fate, optimization, and environmental impacts of agricultural S inputs has been limited, motivating this analysis.

• Fertiliser sales data: Compiled county-level annual fertiliser sales (1985–2015) from the Association of American Plant Food Control Officials (AAPFCO). Products were grouped by nutrient content (N, P, K, S), counting S whether as a target nutrient or carrier in compound fertilisers. When nutrient percentages were not reported, a low/average/high range was estimated from public sources; results were robust, and average values were used. Sales adjustments accounted for carryover (unused product subtracted the following year), which can produce negative year-to-year values. To address potential mismatches between purchase location and application location, data were aggregated across a 12-state Midwestern region.
• Normalization and crop data: Elemental loads (N, P, K, S) were normalized by total crop acreage per state and year from USDA sources, which include major row crops and certain harvested acreage (e.g., hay). S exported in maize tissues was estimated using University of Nebraska-Lincoln tissue S content values combined with USDA NASS maize acres planted; soybean S export was not included due to data limitations on single vs double cropping. Yield trends for soybean, maize (grain), and maize (silage) were derived from USDA NASS yield and harvested area.
• Atmospheric S deposition: Total S deposition was estimated using wet-only deposition sulfate concentrations from the National Atmospheric Deposition Program (NADP), dry S deposition from the EPA CASTNET network, and PRISM precipitation data. Spatial models used Kriging to interpolate wet sulfate concentrations and dry deposition (SO2 and particulate sulfate) from monitoring sites across the 12-state region. Annual total S deposition was generated for each state for 1989–2017. Uncertainties arise from measurement errors in wet and dry components and spatial extrapolation.
• Statistical analyses: Linear regressions (R ‘lm’) were used to quantify trends over time for atmospheric S deposition, fertiliser sales (N, P, K, S), and yields. Slopes, R², and p-values were reported. Hypotheses tested included whether N, P, K trends differ from S trends (all S-containing products and sulfur-S products >85% S content) and whether slopes differ from zero (p < 0.05).

• Atmospheric S deposition decreased markedly: from 4.7 to 1.1 kg S ha⁻¹ yr⁻¹ (1987–2019), at −0.12 kg S ha⁻¹ yr⁻¹ (R² = 0.96, p < 0.0001), approaching pre-industrial background levels.
• S fertiliser inputs increased substantially: all S-containing products rose from 1.3 to 4.9 kg S ha⁻¹ yr⁻¹ (1985–2015), at 0.10 kg S ha⁻¹ yr⁻¹ (R² = 0.82, p < 0.0001) across Midwestern croplands. Sulfur-S products (>85% S content) also increased significantly (R² = 0.71, p < 0.0001).
• Relative rates: Since 1985, S product use increased far faster than other nutrients. Percent-change slopes: N 0.7% yr⁻¹ (R² = 0.52), P 1.2% yr⁻¹ (R² = 0.62), K 0.1% yr⁻¹ (R² = 0.01; not significant, p = 0.52), all S products 7.7% yr⁻¹ (R² = 0.82), sulfur-S 6.6% yr⁻¹ (R² = 0.71). From 2009–2015, S increases accelerated: 29.6% yr⁻¹ for all S products (R² = 0.91, p < 0.006) and 21.1% yr⁻¹ for sulfur-S (R² = 0.55, p < 0.04). S increases were significantly different from N, P, K (p < 0.0001).
• Crop dynamics: Maize and soybean area expanded, especially into the Upper Midwest. Maize grain and silage yields increased significantly (slopes 0.12 and 0.53 MT ha⁻¹ yr⁻¹, respectively; p < 0.004), enhancing nutrient export and fertiliser demand.
• Mass balance context: A quasi-mass balance indicates that in recent years, atmospheric deposition plus fertiliser S may meet maize S export, but the estimate is conservative (excludes soybean export and riverine sulfate losses, which can exceed crop tissue S export). Average regional S fertiliser input is ~5 kg S ha⁻¹ yr⁻¹, with high local variability.
• Soil S pools: Estimated upper 0.9 m soil S pools range from 0.002–2.3 kg S m⁻² (median 0.23 kg S m⁻²), indicating substantial legacy S reservoirs with spatial heterogeneity likely influencing fertiliser needs and S mobility.
• Drivers of increasing S fertiliser: Declining atmospheric S deposition, rising yields and crop footprint, spatial variability in soil S cycling and availability, and management choices (e.g., fertiliser product switches such as ammonium sulfate) collectively explain observed trends.

The findings demonstrate a pronounced shift in the agricultural sulfur cycle for the Midwestern U.S.: as atmospheric S deposition declined due to improved air quality, farmers increasingly relied on S fertilisers, with growth far outpacing N, P, and K changes—especially after 2009. This directly addresses the study’s question by linking regional-scale fertiliser sales trends to atmospheric inputs and crop outputs. The implications are twofold: (1) agronomic—continuing high yields and expanded maize/soybean production will maintain or increase pressure for S applications; and (2) environmental—enhanced S additions may affect soils, waters, and biogeochemical processes, reminiscent of historical concerns with N and P. Although cleaner air also benefits crop yields through reduced pollutants, the transition from diffuse S inputs (deposition) to targeted inputs (fertilisers) necessitates new management strategies. Spatial heterogeneity in soil S pools and cycling likely modulates local S needs and losses, suggesting that one-size-fits-all recommendations are suboptimal. The observed mass balance, along with evidence that riverine sulfate export can exceed crop S export, underscores the need to quantify internal soil processes and off-field fluxes to inform sustainable S management.

This study compiles and analyzes three decades of fertiliser sales data alongside atmospheric deposition and crop statistics to reveal a major transition in the Midwestern U.S. sulfur cycle: declining atmospheric deposition concurrent with rapidly increasing S fertiliser use, which outpaces changes in N, P, and K. The work highlights the agronomic necessity and environmental risks of growing S inputs, and calls for development of sulfur management tools that optimize application timing and rates to sustain yields while minimizing environmental impacts. Future research priorities include: generating temporal and spatial datasets of soil S storage and process rates; resolving crop-specific and management-induced variability in S needs; integrating soybean and other crop exports in regional S balances; quantifying riverine sulfate fluxes and their drivers; and assessing long-term ecosystem consequences of elevated agricultural S loading.

• Fertiliser sales as proxy for application: AAPFCO sales are reported by county of purchase, which may not match application location; some records lack county identifiers. Data aggregated to the regional scale mitigates but does not eliminate this limitation.
• Product composition gaps: Some products lacked reported N, P, K, S percentages; estimated ranges were used, with trends robust to these assumptions. Trace nutrient-containing products may have been omitted but likely contribute minimally to regional totals.
• Accounting adjustments: Year-to-year negative values can occur due to subtracting unused product carried over, affecting annual variability.
• Crop export estimates: Maize S export estimates use fixed tissue S content without a time series; soybean S export not included due to cropping complexity, making mass balance conservative. Riverine sulfate export, which can exceed crop tissue S export, was not included.
• Atmospheric deposition modeling: Uncertainties arise from wet/dry measurement errors, modeled deposition velocities, and spatial interpolation from limited monitoring sites (especially for dry deposition).
• Spatial heterogeneity: Soil S pool estimates are uncertain and spatially variable; no temporally or spatially explicit datasets exist for regional croplands, limiting process-based interpretation.