Compilation and analysis of fertilizer sales data
We compiled fertilizer sales data from the Association of American Plant Food Control Officials (AAPFCO) for all years currently available (1985-2015) (1). Sales data (in pounds of each product purchased) are reported annually by farmers and aggregated at the county level. We separated all fertilizers into groups containing nitrogen (N), phosphorus (P), potassium (K), and sulfur (S), regardless of whether each nutrient was the target addition (e.g., N in ammonium sulfate) or a carrier (e.g., S in ammonium sulfate); we used this approach to calculate a total load of each nutrient for our study region. In some cases, the percentage N, P, K, and/or S was provided, while in others it was not. In the absence of reported nutrient content data, we calculated a range of possible values (low, average, and high) based on publicly available product information and/or published literature. By multiplying the elemental content of a product by its total product mass sold per county, we calculated a mass of N, P, K, and S for each product per county per reporting period. It is possible that some products containing trace amounts of N, P, K, and S were not included in our analysis. However, these trace amounts would be a small amount of the total load, which is normalized over all cropland in the Midwestern U.S. region (Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, Ohio, North Dakota, South Dakota, and Wisconsin).
Ultimately, products for which N, P, K, and/or S content was not reported in the AAPFCO dataset did not have appreciable differences in the calculated loads across the range of possible stoichiometric values and the trends were the same. Thus, we reported the loads based on the average reported content of each element in each fertilizer product. In addition, we found that the county-level data reporting was imperfect in some instances. For example, sometimes the data were reported for a state, but the county (FIPS) identifier was absent. We also recognized that large producers might purchase fertilizers in a different location (e.g., county) from where it was applied. Thus, we aggregated the data to evaluate trends across a 12-state region in the Midwest encompassing much of the maize and soybean cultivated in the U.S.
Note that in cases when more fertilizer product was purchased than was used in a particular location over the course of the year, the unused mass was subtracted from the value in the following year. This resulted in reporting of some negative values year to year, reflecting a more accurate accounting of fertilizer sales over the entire period.
To estimate area-normalized S loads, we divided the total weight of each fertilizer (N, P, K, S) by the total crop acreage data reported for each state by year from the USDA Economics, Statistics, and Market Information System (2). The total crop acreage includes area planted for maize, sorghum, oats, barley, rye, winter wheat, Durum wheat, other spring wheat, rice, soybean, peanuts, sunflower, cotton, dry edible beans, potatoes, sugar beets, canola, and proso millet. Hay, tobacco, and sugarcane are included as harvested acreage. These totals include double-cropped acres and unharvested small grains planted as cover crops. Again, crops other than maize and soybean were the minority (< 25%) of the total acreage in the study region. To estimate S exported in maize tissues, we used tissue data reported by University of Nebraska-Lincoln Extension (3) and maize acres planted from the USDA NASS QuickStats database (4). It is important to recognize that this estimate is conservative; we do not have a time series of S content of maize tissues, nor do we include soybean, as we could not parse areas single- or double-cropped with maize. Finally, we estimated yield trends in soybean, maize (grain), and maize (silage) for the Midwest region using a combination of yield (quantity per acre) and acres harvested reported in the USDA NASS QuickStats database (4).
Estimation of atmospheric S deposition fluxes
We estimated total atmospheric S deposition for the study region using annual volume-weighted sulfate concentrations in wet-only deposition measurements from the National Atmospheric Deposition Program (NADP) (5), estimates of dry S deposition from the U.S. Environmental Protection Agency Clean Air Status and Trends Network (CASTNET) (6) and precipitation quantity data from the PRISM spatial climate datasets (7). We interpolated total deposition for unmonitored regions using point estimates. This analysis was accomplished using a spatial model that incorporates precipitation quantity, annual volume-weighted mean concentrations of S in precipitation, and the dry deposition data for particulate sulfate and sulfur dioxide. The model then uses a Kriging approach to determine the spatial pattern of S concentration in precipitation from the network of NADP stations within the 12 Midwestern U.S. states included in this analysis. Similarly, we used Kriging to generate spatial patterns of dry S deposition using point data obtained from 14 sites monitored as part of the CASTNET program. The annual total S deposition was generated for each of 12 states for all individual years between 1989—the year when adequate point data were available through the networks to develop the kriging models—and 2017.
There are multiple sources of uncertainty associated with the components of annual total S deposition that contribute to the overall uncertainty in flux estimates. For wet S deposition there is uncertainty in the weekly measurements of precipitation volume, as well as sulfate concentrations; weekly measurements are summed to give annual fluxes of wet sulfate deposition. The quality assurance procedures for the NADP are summarized by the laboratory (5). For dry S deposition, there are uncertainties associated with the measurements of gaseous sulfur dioxide and particulate sulfate concentrations, as well as the modeled deposition velocity values to estimate deposition flux. Finally, there is uncertainty associated with the spatial extrapolation of point measurements from the deposition networks to the state scales. We estimated areal-normalized atmospheric S deposition using state areas for the Midwest study region.
References
1. Association of American Plant Food Control Officials (Online: https://www.aapfco.org/).
2. USDA Economics, Statistics, and Market Information System. https://usda.library.cornell.edu/concern/publications/j098zb09z (Accessed July 2021)
3. University of Nebraska-Lincoln. Farm and acreage—sulfur deficiency in corn. https://newsroom.unl.edu/announce/lancasterextension/9974/59326. (URL accessed June 2022).
4. USDA National Agricultural Statistics Service. https://www.nass.usda.gov/Data_and_Statistics/ (Accessed July 2021)
5. National Atmospheric Deposition Program. http://nadp.slh.wisc.edu/ (Accessed July 2021)
6. Clean Air Status and Trends Network. https://www.epa.gov/castnet (Accessed July 2021)
7. PRISM Climate Group. https://prism.oregonstate.edu/ (Accessed July 2021)
