Location and Physical Setting

Long Island is in the southeasternmost part of New York State and extends east-northeastward roughly parallel to the Connecticut coastline. It is surrounded by salt water and is bounded by the Atlantic Ocean on its eastern and southern shore, the Long Island Sound on its northern shore, and New York Bay and the East River on its western shore (fig. 1). The total length of Long Island is about 193 kilometers (km), and its maximum width is 37 km. The total land area (including smaller islands within the administrative boundaries of the island but excluding the bordering bays) is about 3,608 square kilometers (km²; Cohen and others, 1968).

Kings and Queens Counties, on the western end of Long Island, are densely populated. Kings County is the westernmost county and has a land area of about 184 km². Queens County is bordered by Kings County on the west and Nassau County on the east and has a greater land area than Kings County, about 282 km². Nassau County has about 762 km² of land area (26 km² of which is marshland) and is divided into three towns—Hempstead, North Hempstead, and Oyster Bay—and two cities—Glen Cove (within Oyster Bay Township) and Long Beach (within Hempstead Township). Suffolk County, the easternmost county, shares a border with Nassau County and has about 2,380 km² of land area (18 km² of which is marshland); Suffolk County is divided into 10 towns—Babylon, Brookhaven, East Hampton, Huntington, Islip, Riverhead, Shelter Island, Smithtown, Southampton, and Southold (fig. 1).

In 2020, Kings, Queens, and Nassau Counties had population densities that ranked in the top 20 in the entire United States (U.S. Census Bureau, 2021a). However, 100 years ago, more than 40 percent of Long Island was farms and pastures, and the urbanized areas were primarily in Kings and western Queens Counties (U.S. Department of Agriculture, 2019; National Agricultural Statistics Service [NASS], 2020).

Purpose and Scope

This report presents the data and methods used to estimate annual nitrogen loading from 1900 to 2019 from available land use and land cover data, population and agricultural census data, and sewer area information. The methods described in this report made the diverse sources of information comparable and useful to consistently estimate annual nitrogen load for the major nonpoint sources throughout the analysis. The results of this analysis include both island-wide and countywide trends and annual nonpoint source nitrogen loads from 1900 to 2019. The results also include estimates of the spatial variability of each of the six nitrogen source types across Long Island. Examples of this spatial variability for selected years are included in this report. The complete dataset of spatially variable nitrogen sources for each year are documented in a U.S. Geological Survey data release (Monti, 2024), which includes comma delimited files of gridded nitrogen data at a resolution of 500 feet (ft).

Previous Studies

Previous studies of the occurrence and concentration of nitrogen in the groundwater and surface waters of Long Island include Perlmutter and Koch (1975), Katz and others (1980), Ragone and others (1980), and Bleifuss and others (2012). The relationship between land use and water quality was assessed by Eckhardt and others (1989), LeaMond and others (1992), Eckhardt and Stackelberg (1995), Stackelberg (1995), and Pearsall (1996). Estimates of nonpoint source nitrogen loads to Long Island Sound were developed by using watershed models (Georgas and others, 2009). Scorca and Monti (2001) and Monti and Scorca (2003) estimated total nitrogen loads from groundwater and surface water entering Long Island Sound and the south shore estuaries.

Related studies in the Chesapeake Bay watershed are applicable to the Long Island aquifer system. Bachman and Phillips (1996) sampled the base flow of streams representing a variety of watershed characteristics during selected seasons and estimated nitrogen loads from the calculated median instantaneous yields. Phillips and others (1999) evaluated nitrate loads discharged from groundwater to streams in a wide range of hydrogeomorphic regions throughout the Chesapeake Bay watershed and estimated the residence time of groundwater in the aquifer system prior to discharging to streams. Bachman and others (1998) also developed relationships between groundwater discharge, base-flow nitrate yields, and hydrogeomorphic regions in the Chesapeake Bay watershed.

Nitrogen loads from agricultural fertilizer have been described in several reports that include each county for the contiguous United States: Alexander and Smith (1990) focused on the agricultural use of fertilizer from 1945 to 1985, Gronberg and Spahr (2012) estimated both agriculture and residential use of commercial fertilizer from 1987 to 2012, Cao and others (2018) provided county-level estimates of nitrogen for the contiguous United States from 1850 to 2015, and Falcone (2021) estimated nitrogen and phosphorus from fertilizer and manure from 1950 to 2017, also at the county level.

Land Use and Land Cover

Land use classifications designate the human activities or decisions determining how the land is used. Land cover describes the physical land type, such as developed (urban), forested, agricultural, or wetland areas. Because the data for land use and land cover (often found in tandem) were derived from multiple sources, the resolutions varied, and the land use categories changed over time (tables 1 and 2).

Land use and land cover data used in this analysis were obtained from several sources to enable analysis for the period from 1900 to 2019. The primary datasets (Sohl and others, 2014, 2016, 2018a, b) consisted of 13 land use categories available as annual raster grids for 1938–2005 at a 250-meter (m) resolution that were derived from modeled historical land use and land cover for the conterminous United. These datasets are thematically and spatially consistent with the land cover projections produced as part of the National Assessment of Ecosystem Carbon and Greenhouse Gas Fluxes project (Zhu and Reed, 2014).

Although several land use models are available for the years prior to 1938, the resolutions were too large to be appropriate for this analysis. These land use models are Steyaert and Knox (2008) at 10-km resolution, University of Illinois at Urbana-Champaign (2012) at 0.5-degree resolution, and Goldewijk and others (2007) at a 0.5-degree latitude and 1-degree longitude resolution. Therefore, the 1938 land use and land cover raster grid from Sohl and others (2016, 2018a) was used to represent land use conditions from 1900 to 1938 to maintain a consistent resolution throughout the 1900–2005 period. The use of the data from 1938 to represent all the prior years of the analysis does introduce a level of uncertainty in the estimates. Although the changes in land use categories prior to 1938 were primarily in Kings and Queens Counties, where populations increased during this period, more detailed spatial information about the extent of those changes was limited.

Two additional sources were used for land use and land cover categories from 2006 to 2019. The National Land Cover Database (NLCD) provided data for 2006 and 2007 (Fry and others, 2011). The NLCD uses a raster grid with a 30-m resolution with 15 land use and land cover categories for Long Island. The USDA cropland data layer was used for land use from 2008 to 2019 (NASS, 2019). The datasets were downloaded from the interactive CropScape mapping application.

The various land use and land cover categories from these sources were reclassified into five categories for the purposes of this analysis: developed, crop, pasture, undeveloped, and water (fig. 3; table 2). This reclassification simplified the land use and land cover categories in each raster grid and provided a means to consistently calculate the nitrogen load from year to year. The spatial distribution of the five categories and the shift from agricultural and undeveloped lands in 1938 to mostly developed lands in 2019 are illustrated in figures 3A and 3B respectively. Land use changes from 1900 to 2019 represent a transition from extensive agricultural use in 1900, when 40 percent of the land area was used for crops, to predominantly developed (residential) land use. Developed areas increased from less than 27 percent in 1938 to more than 68 percent of the total land mass in 2019.

Groundwater Flow Model Polygon Grid

A polygon grid was developed with 348 rows and 1,309 columns with each 500- by 500-ft cell representing a 250,000-ft² area. This polygon grid was chosen because it is the same grid used in a regional groundwater flow model for Long Island described in Walter and others (2020). Although the flow model grid extends beyond the land area of Long Island and into the surrounding saline water bodies, nitrogen loads were considered to be zero in those areas. Conditions at the center point of each land cell polygon were used to estimate nitrogen loads for each 500- by 500-ft cell and nonpoint source from 1900 to 2019 (Monti, 2024).

Building Footprints

Building footprints were downloaded from the New York State Building Footprints website (State of New York, 2023). Although these polygons represent the outlines of all the structures in New York, only those for the four counties on Long Island were used in this study. The building footprints were used to distribute population and household estimates for the developed land use areas. Building footprints with an area less than 500 ft², such as garages and sheds, were omitted to avoid assigning population values to those structures. This data source and others were used to estimate the septic and pet waste nitrogen loads.

Golf Courses

Golf courses were delineated from tax parcel boundaries or drawn from Streets and Imagery Hybrid maps (Esri, 2013, 2015). The golf course areas were applied in this analysis to calculate the nitrogen fertilizer use from golf courses. There were approximately 51 golf course polygon cells in Nassau, and 78 in Suffolk County that were potential sources of nitrogen over the course of the analysis period. The total golf course area in Nassau and Suffolk Counties was 75 km², with 33 km² in Nassau County and 42 km² in Suffolk County, representing about 4 and 2 percent of the total land area in the two counties, respectively.

Impervious Surfaces

Estimates of impervious surfaces on Long Island were used to eliminate areas from the calculations of residential fertilizer application. Raster grids are available for 1974, 1982, 1992, 2002, and 2012 (Falcone, 2017) and provided the percent of impervious land cover for each raster cell. The mapping of these surfaces was derived by comparing anthropogenic impervious surfaces from the 2011 version of the NLCD (Fry and others, 2011) with the national land use dataset developed from an assessment of human-influenced land use trends for the conterminous United States in 2012 (Falcone, 2015). The raster grids provide an estimated imperviousness (humanmade surfaces as a percentage of the total land use) at a 60-m resolution. The impervious surfaces available for 1974, 1982, 1992, 2002, and 2012 were used to represent the imperviousness of the developed land use category for 1900–74, 1975–82, 1983–92, 1993–2002, and 2003–19, respectively. Using the 1974 impervious surface data for a period as long as 1900 to 1974 was not ideal; however, impervious surface information did not exist before 1974. It is possible, therefore, that the imperviousness during the earlier part of the study period has been overestimated and, in turn, residential fertilizer use, which was calculated from the developed land use data, has been underestimated.

Census of Population

The U.S. Census Bureau has enumerated the country’s population every decade since 1790. These data are available at different levels of geographical detail (U.S. Census Bureau, 2002); however, for this study, the town and city population data were useable because the administrative boundaries on Long Island did not change between 1900 and 2019. Town and city populations from 1790 to 1980 were inventoried by Shupe and others (1982) using data from the U.S. decennial census collections. This dataset, combined with the U.S. Census Bureau decennial collections for 1990, 2000, 2010, and 2020 (U.S. Census Bureau, 2021b), were merged into a single table of decennial populations at the town and city level for 1790 to 2020. The population estimates for the years in between the decennial collections were calculated by linear interpolation (fig. 4A).

In 1900, more than 1 million people resided in Kings County. By 1930, Kings County had 2.5 million residents and Queens County exceeded 1 million residents (fig. 4B). Even as population densities were increasing in the westernmost parts of Long Island, population densities in Nassau and Suffolk County remained low until 1945, when suburban communities began to develop, and populations increased eastward (fig. 4B). The populations in Nassau and Suffolk Counties exceeded 1 million residents in 1955 and 1968, respectively (fig. 4B). By 2020, the estimated population of Long Island exceeded 8 million people (fig. 4A), representing a more than fivefold increase since 1900.

Census of Agriculture

Agricultural information, including farm acreage, livestock count, and crop type, is collected by the USDA as part of the census of agriculture (USDA, 2019) and is provided at both State and county levels. The census of agriculture originated as part of the 1820 decennial census, when agricultural pursuits within each household were added to the census questionnaire. A separate agricultural census began in 1840 on the same schedule as the decennial census of population until legislation in 1919 (The Act Providing for the Fourteenth Census, Public Law 65–325, 40 Stat. 1291) approved an additional mid-decennial census of agriculture taken in 1925 and every 10 years afterwards until 1950. Between 1954 and 1974, the U.S. Census Bureau conducted the census of agriculture in years ending in “4” and “9.” In 1997, the census of agriculture responsibility was transferred to the NASS, which began conducting this census in years ending in “2” and “7” (USDA, 2019).

Historical census of agriculture publications used in this study are available for 1900, 1910, 1920, 1925, 1930, 1935, 1940, 1945, 1950, 1954, 1959, 1964, 1974, 1978, 1982, 1987, 1992, 1997, 2002, 2007, 2012, and 2017 (NASS, 2020). The census of agriculture provides the best available data on total farm acreage, crop acreage, and livestock enumerations for each county on Long Island. These data were tabulated by county. Estimates for the years between census collections were calculated by using linear interpolation between collection dates to generate a table of estimated annual farm acreage (fig. 5A), crop acreage (fig. 5B), and livestock counts in each county. The 2017 census was the last census available at the time of this analysis, and those values were used for the years 2017–19.

Farmed lands in the four counties exceeded 391,000 acres in 1900, or about 40 percent of the total land area on Long Island. Farm acreage decreased substantially from 1900 to 2019, when less than 3 percent of Long Island was used for agriculture. Most of this remaining agricultural land was on the peninsulas in eastern Suffolk County (fig. 3B).

Individual crop types used in this analysis included corn, potato, various vegetables, orchards, vineyards, sod, hay, and grains. The crop names were not consistent across census years. For instance, potatoes were sometimes called “potatoes” and sometimes “Irish potatoes.” These crop types with multiple names were grouped together to create consistency across census collections. When summing the acres of individual crop types, the total acreage of specified crops was less than the total acreage classified as farmed. This difference was noted, and the acreage was assigned as an unreported crop category in this analysis (fig. 5B).

Livestock enumerations from agricultural collections used in this analysis included horses, cattle, hogs, pigs, swine, sheep, goats, ducks, geese, chickens, and turkeys. To account for inconsistencies in how these counts were taken from census to census, this analysis grouped livestock by type, creating categories such as pigs and hogs, or goats and sheep.

Sewer System Areas

Human waste enters the environment through either an onsite septic system or a sewer system. Most areas begin with onsite septic systems and transition to a sewer system as the area develops. There are three principal types of sewer systems on Long Island—sanitary, storm, and combined sewers. Construction of sewer systems in New York City began in the mid-1800s (Veatch and others, 1906), and the New York State legislature established the Brooklyn Board of Sewer Commissioners in 1857. These early sewer systems were combined sewer systems that collected rainwater and human waste. Storm sewers primarily collect rainwater, but not human waste, and are not represented as a sewer system in this report. For this analysis, sanitary and combined system areas together are referred to as sewer system areas. It was assumed in this analysis that the entire populations in Kings and Queens Counties used sewer systems that discharge treated wastewater directly into the surrounding surface-water bodies for the whole study period.

For Nassau and Suffolk Counties, sewer data were obtained from the Long Island Index (2020), which provided maps of sewer districts. These maps were then georeferenced to create digital sewer system outlines for this analysis. By 2019, the cumulative area serviced by sewer systems was approximately 1,254 km², or about 35 percent of the entire Long Island area (fig. 6). The year that an area switched from onsite septic systems to a centralized sewer system was estimated and assigned to each digitized outline by using previously published reports (Suter, 1937; Reilly and others, 1983; Monti and Scorca, 2003; Busciolano, 2005), descriptive data of municipal wastewater treatment plants (State of New York, 2020), and a timeline of estimated community connections (Peter E. Pyne, Jr., Nassau County Division of Environmental Health, written commun., 2017). A start year could not be determined for approximately 47 km² of sewer system areas, in which cases a year was estimated from the “developed” reclassification code of the land use and land cover grid (table 2). The assigned start year for these areas was the year when 75 percent of the sewer system area overlapped the aggregate of the developed grid cell area (fig. 6).

Atmospheric Deposition of Nitrogen

Data sources for both wet and dry atmospheric annual deposition of nitrogen are available as raster grids for 1900 to 2019 from two sources. The first data source is the 1-km-resolution North American Climate Integration and Diagnostics (NACID) Nitrogen Deposition Version 1 database (Hember, 2018a, b). These raster grid units are in kilograms nitrogen per hectare per year and are available for each year from 1900 to 2013. The second data source is from the National Atmospheric Deposition Program (NADP) National Trends Network (Schwede and Lear, 2014; NADP, 2021). The NADP Total Deposition Science Committee developed total (wet plus dry) deposition raster grids at about a 4-km resolution, also in kilograms nitrogen per hectare per year, for each year from 2000 to 2019. Combining these two datasets into one required additional processing steps because neither dataset extent completely covered the land area of Long Island. These additional steps are discussed further in the “Estimating Nitrogen Loads” section.

Annual Nitrogen Loading

Annual nitrogen loading is the mass of nitrogen, in kilograms per head count of humans and livestock per year, per hectare per year for agriculture or residential fertilization, or per household per year for pets. Loadings used for this analysis are from previous publications, as described in the following subsections and summarized in table 3 and were held constant throughout the 120-year period, with the exception of atmospheric deposition and agricultural fertilization rates, which varied during the analysis period.

Fertilizer

Koppelman (1978) estimated an annual rate of lawn fertilizer application of 49 kilograms nitrogen per hectare (kg N/ha) on Long Island. Valiela and others (1997) provided annual rates of fertilizer application to lawns and golf courses that range from 49 to 227 kg N/ha on Cape Cod. Lloyd and others (2016) applied an annual rate of 100 kg N/ha in a nitrogen land use model on the north shore of Long Island to represent the residential fertilizer application rate. In this report, annual rates of 100 and 146 kg N/ha were used to represent the residential fertilizer and golf course application rates, respectively, which represent values typical of those found in the literature.

Valiela and others (1997) estimated an annual agricultural fertilizer application rate used on different crop types of 136 kg N/ha which was derived by averaging a variety of crop types using information provided by Loehr (1974), Stanley (1988), Hayes and others (1990), and Correll and others (1992). This rate overestimates agricultural fertilization for the early part of the 20th century, when organic fertilizers predominated. Organic fertilizers, such as animal manure, generally have a lower nitrogen per ton than modern synthetic fertilizers (Ross and Mehring, 1938), which were not available in the early 1900s.

Synthetic inorganic fertilizers originated in the early part of the 20th century by using the Haber-Bosch chemical-processing method developed in 1909 at the University of Karlsruhe in Germany (Ross and Mehring, 1938). As more processing facilities became active and farmers realized the potential benefits, the use of this type of fertilizer predominated by the middle part of the 20th century (Mehring and others, 1957). The USDA (1966) reported the average nutrient content of fertilizer in the United States from 1850 to 1964, showing it increased substantially between 1930 and 1950. The fertilizer application rates used in this analysis reflect this transition in agricultural areas from organic to inorganic nitrogen and the corresponding increase in nitrogen content, which was estimated to occur in the early 1940s to mid-1950s.

A set of curves were constructed to represent a nitrogen application rate for agricultural areas that varies over time to reflect the change from an organic to an inorganic fertilizer with increasing nutrient content. The curves were structured by using four intervals that represent the estimated changes. The first interval, from 1900 to 1919, represents an organic source of fertilizer with an average increase of less than 0.6 percent in nitrogen per unit volume applied during these 20 years. The second interval, from 1920 to 1930, represents the period when the invention of the Haber-Bosch chemical-processing method to make synthetic fertilizer and its initial adoption provided an average increase of 1.8 percent of nitrogen per unit volume applied from fertilizer above the previous interval. During the third period, from 1931 to 1957, the rate of transition from organic to synthetic fertilizer accelerated as both the popularity of synthetic fertilizers and their content of nitrogen increased, providing an average increase of 3.6 percent of nitrogen per unit of fertilizer applied. The period from 1958 to 2019 represents a plateau in the nitrogen content of synthetic fertilizers as they increased less than 0.01 percent above the previous interval.

These varying application rate curves were applied to three demand categories representing crops with low, medium, or high nitrogen demands (fig. 7). A fourth category was computed as the average of the three demand categories. A current range of nitrogen needs for specific crops was determined from a series of guidelines published by Cornell University (Reiners and others, 2019). With the assistance of the regional Cornell Cooperative Extension specialist (Nora Catlin, Agriculture Program Director/Floriculture Specialist, written commun., 2021), the guidelines provided estimated ranges for the current nitrogen application rates for each crop type. The dashed lines in figure 7 represent the minimum and maximum guideline of the crops in each demand category. High-demand crop types include potatoes and sod; medium-demand crop types include sweet corn and other corn varieties; low-demand crop types include vegetables, grains, and vineyards. The curves were constructed such that they plateau within the minimum and maximum guidelines published between 1960 and 2019; the high and medium demand curves are closer to the minimum guideline rates, and the low demand curve is about midway between the minimum and maximum.

Septic System Nitrogen Source

Septic system nitrogen loading from 1900 to 2019 was calculated annually using a subset of building footprints that excluded garages and sheds (buildings less than 500 ft²). This dataset does not include the years buildings were built in Nassau and Suffolk Counties and therefore required correlating the building data with the developed land use category. The first year a building footprint overlapped with a developed category of the land use and land cover raster grid was the year assigned to that building footprint. In some instances, this method did not yield a result because the land use was never categorized as developed. In these cases, a year was assigned on the basis of proximity to other buildings to which a year had been assigned.

The differing resolutions of the raster grids used by the historical land use and land cover, NLCD, and cropland layers were transferred to the building footprints to do the overlay analysis. A second overlay using the centers of the building footprints and the polygon grid assigned a row and column from the nitrogen loading grid for each building. The two overlays result in the estimated year a building was constructed and a consistent 250,000-ft² area from the grid cells, minimizing resolution differences in land use for developed areas.

The annual town or city population counts were used to distribute population among the building footprints in order to calculate an average population per building. Summarizing the building footprint row and column of each polygon grid resulted in a population per grid cell from 1900 to 2019. This combination of data—the number of buildings and the number of people within them for each grid cell—allowed for a consistent and detailed representation of the annual shifts in population densities in the developed areas, including residential, commercial, and industrial areas. This approach assumed that most of the town and city residents were near the commercial and industrial areas where they worked.

For each year of the study, 4.8 kg N/person (table 3) was multiplied by the aggregated population per developed polygon cell, which resulted in a human waste nitrogen load for each cell. This approach distributed the load equally within the census town, tract, or block geographic area. When a building on a developed land use and land cover cell was contained within an active sewer system extent area (fig. 6), the waste load of nitrogen for that cell was set to zero and only the developed cells outside the sewer system extent were used to estimate nitrogen from septic systems for that particular year.

Residential Fertilization Nitrogen Source

The mass of nitrogen from residential fertilization was calculated from the land use and land cover raster cells that were categorized as developed from 1938 to 2019. This calculation required converting the raster cell area to hectares because estimated residential fertilization application rates in Lloyd and others (2016) are reported as kilograms nitrogen per hectare per year. Multiplying the estimated fertilizer application rate (table 3) by the raster cell area provided the residential nitrogen load. This approach alone likely overestimated residential loading because each developed land use and land cover cell contained some impervious areas that were not fertilized and not every developed area or homeowner applies fertilizer. To adjust for this overestimation, the impervious surfaces within the developed raster cells were used to calculate a mean percent imperviousness for each developed land use and land cover category and the pervious surface percentage was then used to calculate the residential fertilization nitrogen load for only the pervious areas of the developed cells. Another adjustment was based on an estimated percentage of homeowners who use fertilizer. Giblin and Gaines (1990) estimated that the percentage of households on Cape Cod that used fertilizers was about 34 percent. In this analysis, we rounded the Giblin and Gaines (1990) estimate to 30 percent, assuming that household fertilizer may be used more frequently in some communities than in others, so 30 percent seemed a reasonable estimate for Long Island. Multiplying the residential fertilization nitrogen load for only the pervious areas by 0.30 resulted in the final residential fertilizer nitrogen load for each year analyzed.

A different approach was used to calculate the residential fertilizer nitrogen load prior to 1938 because the developed area category of the land use and land cover data was held constant during this period and therefore the residential load also remained constant, which was considered unrealistic. An average ratio of annual residential nitrogen load to septic system load of 0.53, computed based on the period 1938 to 2019, was multiplied by the human waste nitrogen load at each developed cell to compute annual residential fertilizer nitrogen load for the years prior to 1938.

Golf course areas were included in the residential fertilizer estimates. Each golf course was assigned a start and end year, and an annual application rate of 146 (kg N/ha)/yr was applied to the entire golf course area during the years it was active (table 3).

Agricultural Fertilization Nitrogen Source

The land use and land cover cells categorized as cropland were used to calculate the distribution and annual nitrogen load from agricultural fertilization. The most refined geographic level of data collected by the USDA is at the county scale; therefore, a countywide agricultural fertilizer nitrogen load was calculated from the USDA total farm and reported crop acreage (NASS, 2020), converted into hectares, and then multiplied by a variable application rate that represents the transition from organic to inorganic fertilizer and the nitrogen demand category by crop type (fig. 7).

The most commonly cultivated crop on Long Island was the potato. In 1945, potato farms covered more than 65,000 acres and represented about 42 percent of inventoried crops. In 1969, about 34,000 acres were used for potatoes, constituting the greatest proportion of all the inventoried crops at almost 53 percent of the total farmed acres, primarily in Suffolk County (figs. 5A and 5B). Corn farming used more than 28,000 acres in 1900, accounting for about 7 percent of the total farmed acres. In 1969, the acres of corn had fallen to just under 9,000 acres, but still accounted for about 13 percent of the total farmed acres. Other crops, including vegetables and grains, used about 98,000 acres in 1900, or 25 percent of farmed acres; however, by 2002, only about 12,000 acres were used, accounting for about 33 percent of all the farmed acres on Long Island. The total farmed acres recorded as individual crops was less than the total reported farm acreage in the agricultural census data. This difference was categorized as unreported crops and accounted for more than 240,000 acres in 1900, about 62 percent of the total farmed acres. In 2012, about 26,000 acres were unreported, which was about 66 percent of the total farmed acres (fig. 5B).

The USDA agricultural census collection reports of crop and farmed acreage were inventoried by Brakebill and Gronberg (2017) and, in this report, were reclassified into four categories according to the low, medium, high, or average nitrogen demand of the crop (fig. 7). The high-demand application rates were applied to the acres of potatoes and sod. The medium-demand application rates were applied to the acres of corn and the low-demand application rates were applied to all other reported crop acres. The unreported acres of crops received an application rate that is the average of the low, medium, and high rates.

The specific locations of crop cultivation could not be determined; therefore, for this analysis, the combination of crops and their nitrogen application rates were aggregated to yield a county load for each year. This county load was uniformly distributed across the crop land use and land cover cells within the county boundaries. This provided the annual load of nitrogen from agricultural fertilization at each crop land use and land cover cell. Loads at each land use and land cover raster cell were then overlaid onto the polygon grid to apportion the load to each polygon cell.

Livestock Waste Nitrogen Source

Livestock waste was also considered as a source of agricultural nitrogen in this analysis. The nitrogen load from livestock waste, particularly if animals grazed on pastureland or were fed from local food sources, may be overestimated in this analysis because the nitrogen in their feces would have been derived from the atmosphere or from fertilizers applied to these pastures (Valiela and others, 1997), not from external sources. Quantifying this potential overestimate of nitrogen loading from livestock waste would require a detailed time-varying, nitrogen-source analysis for the pastures throughout Long Island, for which data are unavailable. Instead, a nitrogen load from livestock waste was calculated by using a livestock type count (NASS, 2020) and corresponding nitrogen loading (table 3) which were summed at the county level. Except for waterfowl, annual livestock nitrogen loads were reduced by 50 percent to account for the assumption that half of the year their food is locally grown, and the nitrogen content of the feed is already accounted for in other source categories. The calculated waterfowl load was reduced by 75 percent as the waterfowl headcount is based on the number brought to market annually and not the number on farms at any point in time. Other potential overestimates of nitrogen loading from livestock are discussed in the “Uncertainties and Limitations” section.

Estimates of nitrogen load from livestock varied based on the loading of nitrogen per head for each type of animal (table 3). The land use and land cover grid cells categorized as pasture (table 2; fig. 3) were overlaid with county boundaries to determine the number of pastureland grid cells in each county, the assumption being that the livestock contributions should be applied only to pasture lands. The county livestock nitrogen load, not including waterfowl, was then distributed uniformly to the appropriate land use and land cover cells by dividing this county nitrogen load by the number of pasture-land grid cells in each county for each year of the study. Waterfowl loads from ducks and geese were similarly distributed equally but only to locations along the shoreline and streams, where duck farming was known to have occurred.

Pet Waste Nitrogen Source

Annual nitrogen loads from pet waste were calculated by using the same nitrogen loading of 1.95 kg N/household/yr and 1.46 kg N/household/yr (table 3) for dogs and cats, respectively, as reported in the Suffolk County subwatersheds wastewater plan (CDM Smith, 2020). In New York, in 2018, the average numbers of dogs and cats per household were 0.43 and 0.38, respectively (American Veterinary Medical Association, 2018); these numbers were assumed to be constant throughout the study period. The number of households per 500-ft polygon grid cell was determined from U.S. Census Bureau (2022) estimates of the national average number of people per household from 1960 to 2019 and the 1900–2019 population estimates per grid cell. The average number of people per household in 1960 was used to estimate the number of households from 1900 to 1959. Multiplying the pet waste nitrogen loading per household (table 3) by the number of households in each grid cell resulted in an annual pet waste nitrogen load from 1900 to 2019, for each grid cell.

Atmospheric Deposition Nitrogen Source

An annual atmospheric nitrogen deposition rate was applied to all land area grid cells, regardless of land use and land cover category. Deposition rates are from the North American Climate Integration and Diagnostics (NACID) Nitrogen Deposition Version 1 database (NDEP1) for the years 1900–2013 (Hember 2018a, b) and the NADP (2021) National Trends Network for the years 2000–19 (Schwede and Lear, 2014); both raster grids are reported in kilograms nitrogen per hectare per year. These data were multiplied by the cell area to generate the atmospheric deposition nitrogen loading at each cell. Extrapolations of the raster grids were necessary because atmospheric deposition data from the two sources did not completely overlap the land area across Long Island. The extrapolation process used a mean 3 by 3 focal statistic GIS tool that targeted grid cells with no defined data. A value was calculated by using the surrounding grid cells to produce raster grids that completely covered the land area of Long Island. For the years when the two data sources overlapped (2000–13), a ratio of NADP values to NACID–NDEP1 values was calculated and was used to adjust the NACID–NDEP1 data.

Calculating a ratio for the overlapping period 2000–13 was necessary because the NADP data are updated annually and Schwede and Lear (2014) included estimates for several nitrogen species that were not considered by Hember (2018a, b) in the NACID–NDEP1 database, thereby providing a more accurate representation of the atmospheric load. Using the extrapolated raster grids allowed for a total island-wide load to be calculated annually from 2000 to 2013 from both data sources. The ratio calculated for these 14 years was 0.58, indicating that, on average, the NACID–NDEP1 data for island-wide load estimates were 73 percent higher than the NADP data because of differences in the approaches used to create the raster grids. The datasets used different interpolation models, but the model used for NADP is the more current one. The ratio (0.58) was used to adjust the extrapolated 1900–99 NACID–NDEP1 raster grids and, together with the extrapolated 2000–19 NADP raster grids, provided a continuous annual representation of the atmospheric deposition nitrogen load from 1900 to 2019.

Nitrogen Loads, Island Wide

For the period 1900 to 2019, the average annual island-wide nitrogen load from the six nonpoint source components was about 14.92 million kilograms (Mkg N; table 4). This load is an aggregate of 4.15 Mkg N from septic systems, 3.28 Mkg N from residential fertilizer, 3.19 Mkg N from agricultural fertilizer, 1.22 Mkg N from livestock waste, 0.98 Mkg N from pet waste, and 2.10 Mkg N from atmospheric deposition (fig. 8A; table 4).

Area-normalizing the average annual island-wide nitrogen load for the 1900–2019 period results in an average annual nitrogen load of about 4,100 kilograms of nitrogen per square kilometer (kg N/km²; table 5). The average annual area-normalized nitrogen load for septic systems, residential fertilizer, agricultural fertilizer, livestock waste, pet waste, and atmospheric deposition are 1,100, 910, 880, 340, and 270, and 580 kg N/km², respectively (table 5).

All six nonpoint source contributions (fig. 8B) varied over time (fig. 8A). The three largest island-wide average annual nitrogen loads are septic systems, residential fertilizer, and agricultural fertilizer, which contributed 28, 22, and 21 percent, respectively, of the island-wide total. The peak annual loads for these three sources were at different times because of land use changes between 1900 and 2019. Island-wide, agricultural fertilizer was the largest contributing source of nitrogen from 1900 to 1958, at which time septic systems became the largest contributing source. The load of nitrogen from agricultural fertilizer peaked in 1950 with 6.13 Mkg N, about 39 percent of the island-wide total for that year; Suffolk County accounted for 83 percent of this amount. The island-wide septic system nitrogen load peaked in 1971 at 7.78 Mkg N, or about 38 percent of the island-wide total for that year. About 67 percent of the total load from septic systems was in Suffolk County, and 33 percent was in Nassau County. In 2006, residential fertilizer surpassed septic systems as the largest nitrogen source, island wide. The conversion from agricultural, pastural, and undeveloped land into developed residential and industrial communities from 1900 to 2019 and large-scale sewer projects starting in the early 1970s are reflected in these changes in contributions from individual sources.

Nitrogen Loads, by County

The temporal and spatial patterns of nitrogen loading from 1900 to 2019 varies by county and is a function of historical changes in land use and in changing human and animal populations within each county. The nitrogen loads at the island-wide scale were compared at the county scale to understand the specific changes that occurred within each county. The nitrogen loads from the nonpoint sources from each county were compared by normalizing the nitrogen load by the county land area. The county and area normalized loads were compared to the island-wide loads and graphed to provide a visual of the comparison.