The diversity and severity of Nr impacts have catalyzed efforts to understand the causes and consequences of human modifications of the N cycle at local, regional, and global scales. A common approach to quantify Nr dynamics is to create a N mass balance, otherwise known as N budget, which is a simple accounting of the inputs, outputs, and accumulations of N within defined system and spatial boundaries. The first global N budget, which was developed by Delwiche (1970), noted the degree to which humans had altered the N cycle. Subsequent N mass balances have been conducted at a variety of spatial scales. The spatial extent may be individual farms (Spears et al., 2003; Oenema, 2003; van der Schans et al., 2009), watersheds (Howarth et al., 1996; Billen et al., 2011), regions (Baker et al., 2001; Gu et al., 2009), countries, or continents (Robertson, 1982; Galloway et al., 2004; Antikainen et al., 2005; Leip et al., 2011). The type of N flows (e.g. gaseous emissions, dissolved losses, or agricultural harvest) and the nature of those flows (input, output, or storage) included in any mass balance will depend on the definition of the system and study boundary.
Nitrogen flow in surface waters has been measured in rivers globally (e.g., Howarth et al., 2011) and is typically included in N budgets. In contrast, N flow to groundwater is rarely taken into account in mass balances (Gu et al., 2013). One reason for this may be that the time lags associated with transport to groundwater complicate the typical yearly budgeting process. Another reason is that depending on the study boundaries, N flow to groundwater may not be considered an output. However, NO3− is the most ubiquitous pollutant of groundwater resources. Nitrate contamination is becoming more acute and is affecting larger areas and more people (Dubrovsky et al., 2010). Besides supporting irrigated agriculture throughout the world, groundwater also serves as the primary source of drinking water for many. An estimated 32, 75, 29, and 15% of the populations of Europe, the Asia–Pacific region, Central and South America, and Australia depend on groundwater to supply their drinking water (Morris et al., 2003). In the United States, more than 33% of public drinking water comes from groundwater (Kenny et al., 2009). Locally, the reliance on groundwater may be even higher. In California, for example, groundwater provides greater than 60% of the state’s public drinking water, with 85% of the population depending on groundwater to satisfy drinking water needs at least part of the year (CDWR, 2003; SWRCB, 2013).
Elevated concentrations of NO3− in drinking water affect human health. While the most cited consequence of ingesting high NO3− levels is methemoglobinemia, or blue-baby syndrome, links have also been drawn to respiratory disease, thyroid cancer, disruption of the nervous system, and birth defects (Ward et al., 2005). The World Health Organization and the USEPA set a maximum threshold of 12.5 and 10 mg L−1 NO3–N for safe drinking water, respectively. The links between NO3− ingestion, physiological responses, and health outcomes are still an area of active research in epidemiology and not without controversy (Powlson et al., 2008).
Given the time lag between surficial activity and impact to groundwater, a historical assessment of NO3− leaching to groundwater (here referred to as loading) is critical to understand the cumulative effect that human actions have on the N cycle, generally, and specifically to protect drinking water supplies and human health. Previous N balances often lack sufficient spatial or temporal resolution to be useful in identifying measures for protection of groundwater quality (e.g., Leip et al., 2011). In this paper, we disaggregate the historical agricultural sources of NO3− to groundwater to quantify the current and potential future drivers of groundwater NO3− contamination for two groundwater basins in California. The Tulare Lake Basin (TLB) and Salinas Valley (SV) make interesting case studies for this issue because of (i) the diversity and productivity of irrigated agricultural systems, (ii) hydrological conditions that result in groundwater and not surface water being the primary NO3− sink, and (iii) the reliance of large and small communities on groundwater for drinking water. The study regions are illustrative of the global tension between intensive food, feed, fiber, and biofuel production and safe drinking water. Our objective is to identify past and present sources of N loading from agricultural and other sources to groundwater at basinwide and subbasin scales to understand the causes of current contamination and inform future corrective activities.
Materials and Methods
The study area encompasses the TLB and SV groundwater basins (Fig. 1) with semiarid, mediterranean climatic conditions. From approximately 15 April to 15 October, daytime air temperatures can exceed 35°C in the TLB but are more moderate in the SV, and there is negligible, if any, precipitation in the TLB. Dry summer conditions contrast sharply with prevailing conditions between October and April when cool, moist weather produces episodic periods of intense rainfall. The TLB groundwater basin encompasses the southernmost and driest portion of California’s Central Valley (<250 mm annual rainfall). It extends over 20,800 km2 of alluvial fans and basin deposits with minimal topographic expression and slopes generally <1% throughout. Surface water drainage is internal, while groundwater extraction has led to long-term overdraft in some portions of the TLB (Faunt, 2009). The similarly flat, nearly 1400 km2 SV groundwater basin extends to the south from Monterey Bay at the Pacific Coast. Unlike the TLB, there is surface water discharge from the SV to the ocean via the Salinas River, but flow of groundwater to the stream is negligible due to groundwater extraction (Durbin et al., 1978). For both basins, groundwater constitutes between one-third and two-thirds of irrigation water supplies depending on climate conditions. About 10% of groundwater extraction is for drinking water supplies.
Agriculture is the dominant land use in both regions (Fig. 1). The industry is among the most diverse and productive in the United States and essential to the country’s food system. Nearly 300 commodities are cultivated in the study area (CDFA, 2012). Crops ranging from lower-value cereals (e.g. wheat [Triticum aestivum L.] and oats [Avena sativa L.]) to high-value horticultural commodities (e.g., almonds, vegetables) are common. For some commodities, the study area produces a significant fraction, or in some cases essentially all, of the U.S. production. The SV is the center of lettuce and cool-season vegetable production in the United States. The TLB dairy industry accounts for more than 10% of total U.S. milk production and half of California’s milk production. The five-county study area contains the four most productive counties in the United States, in terms of cash receipts, with the fifth county nationally ranked ninth.
The population in the study area increased from nearly 700,000 in the mid-1940s to 2.6 million in 2005 with most living in five urban centers (Fresno, Visalia, Hanford, Bakersfield, and Salinas). About 80% of the population is sewered, and the remainder is on septic systems.
For developing an approach to estimating groundwater N loading, three major N systems are considered, each with characteristic internal N flow patterns that then lead to loading of NO3− to groundwater (Fig. 2): croplands (Flow E), livestock production facilities (Flow I), and human systems (Flow K). As the N flows in these three systems are also linked, there are indirect connections between various land uses and groundwater loading. For example, the majority of livestock manure is applied to cropland, and this land-applied manure is considered in the calculations of loading from cropland and not from livestock facilities. Methods to calculate groundwater NO3− loading differed among the crop, livestock, and human systems. Loading from livestock facilities and human systems was based on loading rates derived from census information, known excretion rates, regulatory records, and N fluxes or rates reported in the literature, while loading from cropland was calculated by closure to its system’s mass balance. To capture temporal dynamics, data and land use maps were created for five reference years 15 yr apart (1945, 1960, 1975, 1990, and 2005).
The cropland system considered here encompasses the field root zone and plant system. Groundwater loading (Flow E in Fig. 2) is computed using the difference between known inputs and outputs (closure). The independently calculated N inputs include synthetic fertilizer (Flow H), manure (Flow J), atmospheric N deposition (Flow B), organic waste from food processing and municipal wastewater treatment facilities (Flow G), atmospheric deposition (Flow B), and NO3−in irrigation water (Flow L). The independently calculated outputs in the mass balance include gaseous emissions (dinitrogen, nitric oxides, nitrous oxide, ammonia; Flow A), harvested crops (Flow F), and surface runoff (Flow D).
Alfalfa (Medicago sativa L.), a major leguminous (N-fixing) crop in the study area, was not suitable for a mass balance approach. Alfalfa is largely unfertilized and sustained by biological N fixation with groundwater leaching a small fraction (<10%) of the total biomass N produced. Groundwater N loading (Flow Ealf) was based on reported leaching rates of 30 kg ha−1 yr−1 (Letey et al., 1979).
Changes in long-term N storage within the cropland system are small and were not explicitly considered except in animal farming facilities. The average increase in soil N stocks reported for agricultural soils of California over 55 yr is 0.2% N (De Clerck et al., 2003). This is equivalent to an increase of N storage of approximately 400 kg N ha−1 or 8 kg N ha−1 yr−1 assuming bulk density was the same in the historical samples. That level of N input represents a small fraction (generally <10% and typically <3%) of the annual N fertilizer inputs to agricultural lands. Further, De Clerck et al. (2003) found large regional differences in soil N accumulation with no significant change in either region of the present study. The accumulation of N in biomass of permanent crops (e.g., trees and vines) over the 60-yr time horizon considered here is also not a significant fraction of the total crop N uptake after accounting for long-term internal N cycling by leaf fall, prunings, and reapplication of harvested trees as soil amendment (Weinbaum et al., 1994; Rosecrance et al., 1998; Weinbaum and van Kessel, 1998).
Data Sources and Calculations
The mass balance parameters that have the most leverage on the N systems are associated with the cropping system: synthetic fertilizer, manure applications, and N in harvested crops (Miller and Smith, 1976; Zhang et al., 1998). We describe the mass balance methods for these N flows in detail along with a brief description of methods used to compute others including development of historical land use maps that underlie the calculations.
The starting point for developing historic land use maps was construction of a statewide digital land use map with a 50-m raster resolution, referred to as the California Augmented Multisource Landcover (CAML; Hollander, 2010), which is an extension of an earlier, more limited effort (Hollander, 2007). The map was derived from four sources: agricultural land uses, grouped into 71 crop categories, developed for individual counties between 1997 and 2006 by the California Department of Water Resources land use maps (CDWR, 2011) and augmented by information available from pesticide use reports (CDPR, 2000); the Farmland Mapping and Monitoring Program maps (CDOC, 2011), which track urban boundaries and conversion of agricultural to urban lands over time; and the 2002 Multi-Source Land Cover map from the California Department of Forestry and Fire Protection (2002), which provides information on natural vegetation. The California Augmented Multisource Landcover (Hollander, 2010) yielded two sets of land use maps representing 1990 and 2005 conditions. For pre-1990 periods, CAML relied on a stochastic, back-casting algorithm (Hollander et al., unpublished data, 2012) to assign crops to individual 50-m pixels based on 2005 land use information and based on reported changes in the total county area reported for each crop within historic county Agricultural Commissioner Crop Reports (ACRs) (county crop reports for the years studied—1943–1947, 1958–1962, 1973–1977, 1988–1992, 2003–2007—can be found at the California Department Food and Agriculture Website, http://www.cdfa.ca.gov/exec/county/countymap/).
Alternative sources of land use information, also needed for the backcasting, are annually issued ACRs, which we digitized into an electronic database for five 5-yr periods centered around 1945, 1960, 1975, 1990, and 2005. The median annual area and production for approximately 250 different crops in the study area were calculated for each of the five time periods. For our analysis, each ACR crop was explicitly linked to one of 71 crops recognized in the CAML digital map. The backcasting produced maps with the appropriate numbers of pixels for each crop type, while location of the crops was only an approximation. Maps of corrals and lagoons on dairy facilities were digitized from aerial photography. Cropland under management by a dairy and used for dairy manure application was identified through the state dairy reporting system introduced in 2007 (CVRWQCB, 2010). The manure-treated cropland associated with each dairy is linked to the individual facilities and to their reported herd size. Due to the lack of accurate historic information, the location of dairy facilities was assumed to be fixed over time but changes in associated cropland followed reported land use changes.
Synthetic Fertilizer Use
Neither current nor historic synthetic fertilizer application rates (kg ha−1) are well documented, especially for the large variety of crops grown in the area. We estimated crop-specific synthetic fertilizer inputs for the 71 CAML crops based on surveys from 1999 to 2010 (UC–D, 2013; USDA 2013a) according to the method developed by Rosenstock et al. (2013). For 1990, N fertilizer application rates reported in USDA (2013a), which lists growers’ self-reported fertilizer use, were not significantly different from 2005. Also, estimates of California fertilizer-use sales remained relatively unchanged over the 15-yr period (Rosenstock et al., 2013). Hence, fertilizer rates for 1990 were assumed to be equal to those in 2005.
Rauschkolb and Mikkelsen (1978) conducted a statewide expert survey of fertilizer use in 1973. The authors listed common rates for 48 different crops, which are used here for the 1975 period. The authors also listed average fertilizer use by crop group (e.g., agronomic, fruits and nuts, and vegetables) for 1950 (used for 1945), 1960, and 1973 (Rauschkolb and Mikkelsen, 1978). We calculated the percentage change for each crop group between two sequential periods (e.g., 1990 vs. 1975, 1975 vs. 1960, and 1960 vs. 1945) and computed fertilizer use for the crops not included in the survey in 1973 and for all 1945 and 1960 crops based on 1990 estimates and the percentage change for the respective crop group.
Manure for Cropland Application
Livestock manure N applications to cropland were calculated by animal-system size and excretion rates. Manure N production at each facility was calculated on the basis of dairy herd size reported since 2007 to the regional water quality regulatory agency (e.g., CVRWQCB, 2010). The fate of this dairy manure was partitioned into atmospheric losses and permanent soil N storage during collection and storage, on-farm application on forage crops (areas reported in permit records), and exported application outside of dairy farms (specific areas unknown).
We calculated manure production to quantify the transfer of manure N to cropland. The USDA National Agricultural Statistics Service’s Census of Agriculture and annual surveys for 1945, 1960, 1967, 1975, 1990, 2003–2007, and 2010 were used to measure state animal herd size and milk production changes over time (USDA, 2013b). Current excretion rates from lactating cows and dry cows are 462 and 195 g N d−1, respectively, or 153 kg N yr−1 adult cow−1 (Harter et al., 2007). We assumed 1.17 calves and heifers (support stock) per adult cow excreting an additional 45 kg N yr−1 adult cow−1 (Pettygrove et al., 2009). Historic excretion rates were obtained by considering relative changes in USDA-reported milk production, dairy herd size, and milk N use efficiency in California. Comparing 1973 with 2005 excretion rates (Harter et al., 2007) and USDA reported milk production, milk N use efficiency was estimated to have linearly increased from 21% in 1945 to 25% in 2005. Relative excretion rates per cow, adjusted for increases in milk N use efficiency and milk production, were estimated to be 42, 54, 71, and 91% of 2005 levels for 1945, 1960, 1975, and 1990, respectively. Due to lack of historic reporting data, we approximated historic farm size by assuming that 2005 farms grew at the same rate as the state herd. Through the 1960s, cows in the study area generally grazed openly in irrigated pastures or on public lands, and manure was not intensively managed. Nitrogen excretion and deposition in pastures likely did not exceed pasture-buffering capacity and thus presented no significant leaching to groundwater in 1960 or prior. Manure in excess of pasture N capacity was assumed to have been applied to grain crops. Transition to confined dry-lot and freestall-based dairy farming with irrigated forage crops instead of pasture occurred predominantly during the 1960s and 1970s. Land application of manure to forage crops was assumed to have occurred since the 1975 reference period. Beginning around 1980, dairy facilities exported (mostly dry) manure to neighbors and elsewhere, typically within county boundaries. Not much data exists on the amount and fate of exported manure. We assumed a linear increase in exports from none before 1981 to 19% in 2005 (and later), which is based on ACR data. In 2005, manure N applications to forage crops within individual dairy facilities amount to 24% of the facility-specific excretion rate after accounting for export and atmospheric losses. Atmospheric losses are 38% (USEPA, 2004) before land application and an additional 10% after land application (see below). We conservatively assume that after 1975, 19% of the excreted manure N accumulated in soil storage (recovery loss, Kellogg et al., 2000) and at least 50% of the nutrient needs in forage crops within dairy facilities were met by synthetic fertilizer (van der Schans et al., 2009), with the remainder managed with available manure N. Exported manure N was simulated as being applied to all crops within county boundaries (outside of dairies) as an organic amendment at rates proportional but in addition to a crop’s synthetic fertilizer use. While an oversimplification, no data exist to support a better spatial resolution of this exported manure N flux.
Nitrogen Export in Agricultural Product
Harvested N for each of the 71 CAML crops was computed from ACR production data using the moisture and N content reported in the USDA crop nutrient tool (USDA, 2013c). Reported production, and hence harvested N, showed significant interannual variations. We therefore compiled a 5-yr data series centered on each of the five reference years and computed the median to represent each period.
Human Systems Contribution to Cropland Systems
Municipal and food processing waste applications and their application areas were based on discharge rates and land application areas reported to regulatory agencies. Nitrogen loading for wastewater treatment plants and food processors was determined based on (i) the waste discharge requirement reports provided by the respective environmental regulatory agency, (ii) the spatial distribution of land applications and percolation basins extracted from the California Integrated Water Quality System Project and USEPA Facilities Registry System, (iii) surveyed facilities for biosolid information, (iv) census data (U.S. Census Bureau, 2001; Forstall, 2011), and (v) the Hilmar database, which combines data on waste discharge requirements and monitoring data (Rubin et al., 2007; Sunding and Berkman, 2007; Sunding et al., 2007). Information collected includes sewered population, flow rates of wastewater treatment plants, flow to recharge basins, quantity of discharge to surface water and to irrigated agriculture, seasonal variation in flows, nitrogen concentrations, area or receiving lands (agriculture and percolation basins), and volume of biosolids. The resulting data from 40 wastewater treatment plants and 132 food processing facilities accounted for 90% of wastewater treatment design flow and 63% of food processors. Surface water discharge other than reported amounts was assumed to be negligible. Nitrogen loading to groundwater from nonsurveyed facilities was estimated based on the following rules: If N concentration of waste discharge from the facility was not reported, it was estimated based on the measured correlation between discharge N concentration and total flow of known wastewater treatment plants or food processors. A 50–50 split between receiving land units was assumed when the relative split of discharge between recharge basins (direct groundwater loading) and agricultural land application was unknown. Biosolids application from nonreported facilities was not included. In the case of unknown land area receiving discharge, we used the correlation between flow and area at known facilities to estimate land area for agricultural land application and the average across known facilities to estimate land area of recharge basins. Historical N loading was backcasted based on changes in county-level human populations in reference years as reported by the U.S. Census Bureau.
Other Cropland Fluxes
There were several smaller N inputs to and losses from cropland that we took into account for the cropland mass balance. We estimated 2005 atmospheric N deposition to cropland based on the modeled results reported in Fenn et al. (2010). Historic N deposition for 1945 to 1990 was scaled using N emissions. For 1975 to 2005, we used reported oxides of N (NOx) and ammonia (NH3) emissions (CARB, 2009). Estimates of emissions for 1960 and 1945 were derived from an assumed linear increase in NOx from 1900 (zero emissions) to 1975. Similarly, we assumed zero NH3 emissions for 1900 and a linear increase until 2005 (the only reported period). Nitrogen inputs to cropland from irrigation water in 2005 were computed from long-term average use of groundwater for irrigation, which were 600 mm yr−1 in the SV (Durbin et al., 1978) and 450 mm yr−1 in the TLB (Faunt, 2009) and median nitrate concentrations, which were computed separately for each of 16 groundwater basins from data reported for the period from 2000 to 2010 in public databases (Boyle et al., 2012). Estimated 2005 rates for irrigation water inputs ranged from 1 kg N ha−1 yr−1 in basins pumping deep groundwater to 39 kg N ha−1 yr−1 in an unconfined aquifer subbasin with relatively high nitrate concentrations. Historic groundwater NO3− concentrations in irrigation water are not well documented. Hence, a linear increase was assumed over the historic period, from negligible irrigation water contributions in 1945 to those estimated for 2005. Surface water used for irrigation, which originates exclusively in the granitic Sierran uplands of the TLB watershed, contributes negligible amounts of N to cropland. We conservatively assumed that N loss in surface runoff was high (14 kg N ha−1 yr−1) for mixed agricultural areas (Beaulac and Reckhow 1982).
An N emission factor is derived from available data and reported as percentage of N applied:
∙N2O: 1%. This is the default emissions factor of direct field emissions used by the IPCC (De Klein et al., 2006).
∙N2: 1.8%. This emissions factor is based on the average N2:N2O ratio reported for agricultural sites (Schlesinger 2009).
∙NH3: 4%. This is the typical emissions factor for California fertilizers, using values provided by Battye et al. (2003); however, Krauter et al. (2006) suggests 3.2% based on the average emissions factor measured from 14 California fields with varying types of applied fertilizer.
∙NO: 2.1%. This is the average emissions across 8 crops and 20 sites in California’s Central Valley (Matson et al., 1997).
These four fluxes suggest a total of 9% of applied N is emitted to the atmosphere as gas. Here, because of the relatively high inputs of other sources of N, gaseous losses were calculated conservatively as 10% of all input N instead of just from synthetic fertilizer or manure N (after land application). Rates of runoff and relative gaseous losses were assumed constant across the period of study.
Direct Groundwater Loading from Human and Animal Systems
Direct N loading to groundwater from human systems (Flow K) includes leaching of fertilizers applied to turfgrass in lawns and golf courses, leaching from septic systems, leaky sewer pipes, and treated wastewater that is percolated to groundwater via recharge basins. In livestock systems, direct N contributions to groundwater (Flow I) considered here include manure N leaching to groundwater from the surface of production facilities. Dairies account for over 99% of N fluxes associated with animal systems in the TLB. Modern dairy production in the region takes place exclusively in concentrated animal feeding operations (CAFOs). Before the 1970s, pasture farming dominated the dairy industry. We considered three separate potential dairy sources of groundwater NO3− after 1970: manure-treated croplands via Flow J, open corrals and feedlots, and manure storage lagoons.
Wastewater discharge permit records were used to separate land-applied wastewater N (input to the cropland mass balance, flow G) from direct percolation N (Flow K). Estimates of incidental leakage from urban sewer system before treatment were obtained by surveys of city engineering staff in the five major urban areas and combined with reported sewer flow rates and population density. Estimated losses vary widely and are highly uncertain. Based on interviews with local operators, we used a net urban leakage rate of 10 kg N ha−1 yr−1, which represents approximately 10% of total sewer N. Losses from lawns and golf courses were assumed to be small (Petrovic, 1990, Raciti et al., 2008) and added an additional 10 kg N ha−1 yr−1 to urban groundwater loading. Historic changes were assumed to be tracked by the changes in mapped extent of urban areas. Septic system losses outside urban areas were based on population density (Forstall, 2011), U.S. census data on septic systems (U.S. Census Bureau, 1990), per capita excretion rates of 13.3 g d−1 (Crites and Tchobanoglous, 1998), and volatile losses assumed at 20% of excreted N. Historic septic losses were scaled relative to historic county population. Groundwater NO3− loading from corrals and lagoons (Flow I) from 1975 onward was based on unit area groundwater recharge rates and NO3− concentrations found in previous field studies, 183 kg N ha−1 yr−1 (BVA, 2003; Ham, 2002; Harter et al., 2002; Harter et al., 2007; Miller et al., 2008, Vaillant et al., 2009; van der Schans et al., 2009).
County-level, ACR-based, crop-specific N fluxes were estimated by using an electronic spreadsheet aggregated to crop groups, county, and study area. A spatially distributed N mass balance with a 50-m resolution was computed by developing a Matlab (MathWorks) script to link the various data sources, to distribute place-based N fluxes spatially, and to assemble all other land use-specific and temporally varying N fluxes at the period and pixel level. Each 50-m pixel is characterized by its land use, census block, and, where applicable, attributed to an individual sewered municipality, dairy, wastewater treatment plant, or food processor facility.
Statistical error analysis of cropland groundwater N loading was implemented by performing a Monte Carlo simulation of the study area cropland mass balance with 10,000 random realizations. Uncertainty about the individual cropland N mass balance components was expressed as a normal distribution with means equal to the aggregated results from the spatially distributed, simulated mass balance analysis and by assuming that the 95% confidence interval for total study area N flux is ±20% for each of the largest N flux variables (synthetic N, manure N applied, harvested N) and ±40% for each of the other, smaller variables in the cropland N mass balance (e.g., human systems contributions to cropland, atmospheric losses and deposition).
Results and Discussion
Areal Extent of Agriculture
Agriculture expanded significantly over the past six decades in the study area. In total, cropland, excluding alfalfa, nearly doubled in less than 30 yr, from 0.7 million ha in the mid-1940s to nearly 1.0 million ha in the early 1960s and to 1.3 million ha in the mid-1970s (Fig. 1 and Fig. 3). Mirroring national trends, it has remained at that level for the past four decades. Changes in the extent of agriculture have differed by county and by crop (Table 1). Only Monterey County has remained constant in area harvested over the entire historic period. Recent increases in reported harvested land area are predominantly due to more intensive cropping practices with the same area providing double and triple harvest of vegetables. In TLB, the area dedicated to alfalfa and small grain and hay crops expanded after 1945 but has leveled off or decreased in the last 40 yr, along with the area dedicated to field crops. These decreases are due to significant expansion of specialty crops including grapes, citrus, nuts, tree fruits, and vegetables and berries.
|Crop group or county||1945||1960||1975||1990||2005|
|Grain and hay||210,651||353,793||304,459||161,263||223,468|
|Vegetables and berries||75,809||90,490||132,626||209,524||280,433|
Livestock is overwhelmingly dominated by dairy cattle in the four-county TLB. Unlike cropland area, the number of dairy cattle has increased exponentially throughout the past six decades. In 1950 (first reporting census), TLB had 108,000 milk cows, which quadrupled to 409,000 by the 1992 census and more than doubled again over 15 yr to 878,000 milk cows (1,050,000 adult cows) in the 2007 census. The dairy herd in SV decreased from 10,000 in 1950 to 2,000 in 2007. In 2007, in addition to the dairy herds, two large beef cattle CAFOs (up to 130,000 head in 2007) and a few poultry (15 million broilers and turkeys) and swine facilities (10,000 head) were located in the study region.
Increases in overall livestock production have cascading effects on cropland, particularly in the TLB. Between 1945 and 2005, the area planted to potential animal feed (grain and hay, alfalfa and pasture, and corn [Zea mays L.]) increased first in the 1950s before decreasing in the 1980s, then increased again sharply after 1990. Corn, mostly for feed and double-cropped with winter grains, steadily increased in acreage and nearly quadrupled between the mid-1970s and 2005 (Table 1).
Intensity of Agriculture
Both the unit area inputs of N (fertilizer and feed) and outputs of N in harvested crop and livestock products and manure have increased in the agricultural system in the study area, but with distinctly different long-term dynamics that have significant implications for groundwater loading.
Nitrogen application rates have increased over time for most crops (Table 2), but the amount of applied N typically varies among crops (Rosenstock et al., 2013). An approximate gradient of 2005 typical N applied rates among major crop groups is as follows: multicropped vegetables (≥200 kg ha−1 season−1) > nuts (∼200 kg ha−1 season−1) > field crops without alfalfa (120 to 200 kg ha−1 season−1) > tree fruits (∼110 kg ha−1 season−1) > grapes (20–70 kg ha−1 season−1) > alfalfa (0–30 kg ha−1 season−1). As a result of changes in fertilization rate and crop areas over time, the partitioning of the total fertilizer use among crop groups has significantly shifted (Table 2). Early on (1945), grain and hay crops, cotton (Gossypium hirsutum L.), and subtropical crops were the primary users of N fertilizer. In 2005, vegetables and berries accounted for more than 25% of typical fertilizer N applied and nuts for about 10% of the total. Crops that dominate elsewhere in the United States—cotton, field crops (dominated by corn), and grain and hay crops—use just over half of all typical fertilizer applied in this study area (∼120 Gg N yr−1).
|Crop group||1945†||1960||1975||1990||2005||2005 avg. N input reductions needed‡||2005 avg. county exported (excess) manure N rate§|
|Gg N yr−1||%||kg ha−1 yr−1|
|Grain and hay||11.0||26.0||30.8||25.2||42.1||8||20–100|
|Vegetables and berries ¶||6.50||11.7||22.1||44.7||59.2||38||20–110|
|All crops except alfalfa||37.0||81.6||140.6||201.2||225.1|
Currently, the typical N applied as synthetic or manure fertilizer (not including excess N, defined as applications of manure or organic waste in excess of typically applied or recommended rates) is 242 Gg N yr−1 in the study area, which is a sixfold increase since 1945. Only 5% is contributed by manure N (Table 3). Synthetic N inputs to cropland quintupled between 1945 and the late 1980s (Fig. 3). Since then, synthetic fertilizer use has remained relatively stable.
|Nonalfalfa cropland mass balance component||1945||1960||1975||1990||2005|
|Gg N yr−1|
|H. Synthetic fertilizer N applied on cropland after accounting for manure use as fertilizer||43.7||93.8||155.2||210.7||230.1|
|J. Land applied manure|
|On dairy (only after mid-1960s)||–||–||32.0||42.2||48.0|
|Off dairy (only after 1980)||–||–||–||6.9||38.7|
|Long-term farm soil N storage (recovery loss after 1980)||6.9||38.7|
|G. Biosolids and wastewater effluent N||2.0||3.1||2.5||2.9||5.7|
|B. Atmospheric N deposition (cropland without alfalfa)||5.0||10.2||15.6||15.7||13.5|
|L. Irrigation water N (groundwater nitrate)||0||4.3||10.6||15.4||21.4|
|Nitrogen outputs [Gg N yr−1]|
|F. Harvested N from cropland||25.0||65.0||94.2||110.4||139.6|
|D. Surface runoff N from cropland||9.9||15.4||19.4||18.1||19.3|
|A. Atmospheric N losses from cropland||5.1||11.1||21.6||29.4||35.8|
|E. From nonalfalfa cropland||10.7||19.9||80.7||136.0||162.7|
|Ealf. From alfalfa cropland||3.7||6.6||4.5||3.8||4.5|
|I. From dairy facilities (corrals and lagoons)||–||–||1.7||1.7||1.7|
|K. From human systems|
|Urban lawns and golf courses||0.2||0.4||1.0||1.6||1.8|
|Other relevant fluxes|
|Synthetic N applied on nonalfalfa cropland, unadjusted‡||44.2||94.8||160.7||221.7||242.2|
|Synthetic N applied on alfalfa||1.5||3.7||3.3||1.5||1.8|
|Dairy manure N flux to cropland before 1970s||7.7||15.8||–||–||–|
|Incidental to grazed pasture at 135 kg N ha−1 yr−1||4.2||9.9||–||–||–|
|Applied to grain crops||3.5||5.9||–||–||–|
Total N harvested in agricultural products has been progressively increasing in the study area. Over the past 60 yr, all crops have seen dramatic and steady increases in the rate of harvest removal despite little increase in farmed land area since the 1970s (intensification) and despite largely unchanged commercial fertilizer N application rates (higher N use efficiency) since the late 1980s. In 2005, the harvested N from cropland was 140 Gg N yr−1 (Table 3), not including alfalfa. Other cropland outputs are smaller: atmospheric losses were assumed to be one-tenth of the inputs (36 Gg N yr−1), and runoff was estimated to be 19 Gg N yr−1 from cropland (about 5% of inputs).
Nitrogen removal in harvest is greatest per hectare in alfalfa due to its high rates of biological N fixation. Yields removing more than 400 kg N ha−1 yr−1 are common. In total, more than one-third (74 Gg N yr−1) of N exported off-farm each year in 2005 resulted from alfalfa production. Production of corn and other field crops removes about 200 kg N ha−1 yr−1. Where summer corn and winter grain is double-cropped (typically in the vicinity of dairies), annual harvest N removal also may exceed 400 kg N ha−1 yr−1. Cumulatively, field, grain, and hay crops (without alfalfa) account for nearly another third of the harvested N. Vegetables take up 90 to 120 kg N ha−1 yr−1 per harvest, but many are double (and sometimes triple) cropped each year, averaging 1.7 crops per year in 2005 (1.6 in 1990 and single-cropping in 1975 and earlier). Vegetables account for approximately 15% of the total N harvested and nuts account for approximately 10% of the total N harvested, about 100 kg N ha−1 yr−1 for average yields. The least amount of N is taken up by average harvests of subtropical fruit (e.g. citrus, avocado, olive, kiwi, persimmon), stone fruit, and grapes: at rates of 50, 25, and 17 kg N ha−1 yr−1, respectively.
The intensity of milk production and milk N yield has also increased significantly. In 1945, a cow could be expected to produce 3240 kg milk yr−1 (17 kg N yr−1). Today, an average cow produces 9710 kg milk yr−1 (50 kg N yr−1), a 300% increase. At the same time, about 19% more feed N was converted to milk N in 2005 compared with 1945 (higher milk N use efficiency). Multiplying the milk N production by herd size suggests that 41 Gg milk N was produced in 2005 compared with 1.9 Gg N in 1945. Additionally, the total N excretion, after accounting for increases in milk N use efficiency, milk production, and herd size, was 6, 12.4, 25.5, and 44.6% of 2005 levels in 1945, 1960, 1975, and 1990, respectively, an unabated exponential increase of manure N (doubling every 15 yr). This has created a large pool of N inputs to cropland not matched by concurrent decreases in commercial fertilizer N use. In the last 30 yr, with animals mostly confined rather than on pasture and with manure applied to croplands, the total manure that is land applied has increased nearly threefold from 32 Gg N yr−1 in 1975 to 87 Gg N yr−1 in 2005. Due to the limited area of forage cropland controlled by dairies and used for manure N applications (∼60,000 ha), only 14% of land-applied manure N is estimated to replace synthetic fertilizer N with the remainder being applied in excess of typical synthetic fertilizer application rates (excess N).
Simulated, mass balance–derived groundwater loading from croplands (other than alfalfa) has increased from 11 to 163 Gg N yr−1 over the 60-yr study period and is now the largest N flux out of cropland (Fig. 1, Table 3). Between 1975 and 2005, the contribution to cropland groundwater N loading due to manure excess N increased from 26 Gg N yr−1 (33% of all N loading) to 75 Gg N yr−1 (46% of all N loading). By 2005, an additional 39 Gg N yr−1 is added to long-term soil N storage in dairies with risk of future leaching (e.g., Vaillant et al., 2009).
Significant differences exist in groundwater loading intensity between major crop groups. Vineyards are the least intensive crop group, with loading rates less than 35 kg N ha−1 yr−1, followed by rice (Oryza L.) and subtropical tree crops (about 60 kg N ha−1 yr−1), then tree fruits, nuts, and cotton (90–100 kg N ha−1 yr−1). Crops with the most groundwater N loading are those that consist of multiple crops harvested sequentially on the same land within a single year including vegetables and some berry crops (over 150 kg N ha−1 yr−1). With typical synthetic fertilizer applications alone, field crops (other than cotton), grain, and hay crops (not including alfalfa), even if double-cropped, have relatively low groundwater N loading (<35 kg N ha−1 yr−1).
However, these leaching rates do not account for excess manure N applied (Table 2). Most of the excess N is thought to become additional groundwater N loading, although high dissolved organic carbon associated with manure leachate (Chomycia et al., 2008) may lead to atmospheric losses greater than the 10% assumed here (Singleton et al., 2007). In 2005, countywide average excess manure N applied on field (predominantly corn), grain, and hay crops (not including alfalfa) that are managed by dairies ranged from 400 to 1000 kg N ha−1 yr−1. The highest countywide average rates, exceeding 700 to 1000 kg N ha−1 yr−1, occurred on dairies in Kern, Kings, and Tulare Counties. In addition, assuming that exported manure is distributed as excess N in proportion to typically applied synthetic fertilizer N, average excess N on most crops range from 0 to 20 kg N ha−1 yr−1 in Fresno County (lowest manure excess N besides Monterey County) to 10 to 110 kg N ha−1 yr−1 in Tulare County (highest manure excess N). Large uncertainty exists about the amount and fate of manure N exported from dairies, the recovery losses due to long-term soil N storage within animal facilities, and the atmospheric losses of N after excretion. Actual excess N rates and resulting groundwater N loading on individual facilities will vary widely. Total manure N applications across the study area, however, represent a reasonable estimate and are useful to illustrate the magnitude of impact from manure N.
Direct leaching of manure N to groundwater from animal corrals and manure lagoons are small when considering the scale of the entire study region but are locally intensive. These sources account for an estimated 1.5 and 0.2 Gg N yr−1, respectively (<1% of total excreted N). However, there is significant uncertainty about the overall magnitude of corrals and lagoons as groundwater NO3− sources. Actual loading may range somewhere between 0.5 and 8 Gg N yr−1 for corrals and between 0.2 and 2.0 Gg N yr−1 for lagoons given the range of reported leaching rates (Harter et al., 2002; van der Schans et al., 2009). Beef, poultry, and swine facilities generate a total of about 0.9 Gg N yr−1 that is land applied as manure or compost.
Our historical reconstruction of agricultural land use indicates that agriculture has contributed on the order of 5000 Gg N to the aquifers beneath the SV and TLB since 1945. Likely, very little of the NO3−is denitrified during transport (Green et al., 2008). Assuming transit times of over 20 yr for NO3− transmission through the vadose zone and to domestic wells (Burow et al., 2007), which are typical of even the shallowest wells, groundwater concentration today is mostly the result of activities before 1990. Before this time, much of the N inputs were the result of synthetic fertilizer use, and thus, lands receiving synthetic N inputs bear much of the burden for past NO3− contamination. Past estimates of field-scale N loading from that period range between 25 and 912 kg ha−1 yr−1 (median of 190 kg ha−1 yr−1 for 25 sites) depending on cropping system and soil type (Adriano et al., 1972; Pratt and Adriano, 1973), which is consistent with results of the cropland N mass balance.
Croplands may be the primary historical delivery mechanism of NO3− loading from agricultural systems, but animal agriculture, particularly dairy production, is a dominant driver today. In the 1980s, a shift occurred in the importance of N inputs from fertilizer to manure when the proportion of manure N applied increased relative to synthetic fertilizer N, which has leveled off. Clearly, the recent increases in manure N applied to cropland are the result of increasing herd size. Changes in husbandry practices have also translated to increased milk production, which in turn increased feed demand and manure N availability. Given the time lag between surficial activities and changes in groundwater quality, the current concentrations of NO3− in groundwater, and the trends in land use documented in this study for 1975 to 2005, chronic groundwater NO3− contamination in the study area is likely to become more acute and widespread in the future before improving due to recent regulatory actions or those currently under development. Even if N loading to groundwater were dramatically curtailed today, it would be decades before concentrations broadly began to decline. Denitrification in the aquifers can reduce NO3− concentrations over time but, in many areas, would likely take several decades to remove just half of the N (Green et al., 2008). The dynamics of historic groundwater loading indicate that NO3− contamination is an agricultural legacy that the TLB and SV will not soon forget.
The steady increase in groundwater N loading from agriculture has been the net result of three factors: changes in the extent of agriculture (until the 1970s), changes in the intensity of agriculture (ongoing), and changes in the efficiency of agriculture (ongoing). Not counting N applications in excess of typical rates, 58% of N applied to croplands was harvested in 2005 compared with 50% in 1990, a 16% improvement. The total amount of surplus N (inputs, not including excess manure applications minus harvest, Table 3) was similar between 1990 and 2005 (about 145 Gg N) despite 26% higher N harvested over the time period.
Though promising, upward trends in efficiency of N use appear insufficient to offset the rapid increase in manure N applied, with commensurate increases in excess N applied. Actual total simulated surplus N (inputs, including excess N minus harvest) increased from 184 Gg N yr−1 in 1990 to nearly 220 Gg N yr−1 in 2005 due to the increase in land-applied manure N (Table 3).
Improvements in N efficiency can be associated with other worthwhile goals (e.g., higher profits and less water use with drip irrigation). Yet, the triple transformation of more area in cropland (until the 1970s), more area planted with N intensive crops (since the 1970s), and more dairy manure (unabated exponential increase) cannot be offset by efficiency increases alone. The net effect of agricultural expansion and intensification has been a significantly greater N groundwater loading over the extent of the study area despite concurrent increases in efficiency.
Constraints on Remedies
Because current and historical practices have led to high leaching losses from cropland soils, reductions will be needed in order for agriculture to comply with California law, which stipulates that groundwater cannot be degraded. To provide a broad reference point of what the source loading numbers mean with respect to potential groundwater pollution, it is useful to introduce operational benchmarks that indicate whether NO3− leached in recharge to groundwater exceeds the NO3− drinking water standard at the field and study area scales. Our benchmark for low intensity vs. high intensity of NO3− leaching is 35 kg N ha−1 yr−1 and is based on average groundwater recharge rates (∼300 mm yr−1), the maximum drinking water level for NO3− (45 mg L−1), and evidence that there is only limited denitrification occurring in the vadose zone of many California regions (Green et al., 2008). Few cropping systems (e.g., alfalfa and grapes) in the study area leach less than 35 kg N ha−1 yr−1. Even where no excess manure N is applied, significant synthetic fertilizer N reductions would be required to lower groundwater N loading to 35 kg N ha−1 yr−1 while maintaining yields: in cotton, nuts, tree fruit, subtropical, and vegetable crops, necessary reductions range from 30 to 60% (Table 2). Aggregated across the 1.4 million ha of CAML cropland (not including alfalfa), the benchmark for total annual NO3− loading in the study area is 50 Gg N yr−1, 30% of simulated current loading rates. The large total NO3− loading to groundwater relative to this benchmark indicates a high potential for past and current regional groundwater degradation.
In some systems, especially high-intensity, double-cropped systems with N inputs typically exceeding 400 kg N ha−1 yr−1, it is difficult to develop management practices that could meet this threshold, as shown in a meta-analysis by Zhou and Butterbach-Bahl (2013). These cropping systems would need to reach over 90% partial nitrogen balance (N inputs–N exports). Currently, few California cropping systems exceed 50% partial nitrogen balance on average (Rosenstock et al., 2013). While theoretically plausible and even practically sustainable, it would mean that many production systems operate near N equilibrium. Achieving this level of efficiency would require transformative changes in how these systems were managed, which is often at great cost to the farmer (Medellín-Azuara et al., 2013).
In 2007, California adopted regulatory measures for Central Valley dairies requiring that all N inputs total at most 140 to 165% of harvest. If all cropland in the study area were under such regulations, the total allowable N application to cropland, at today’s crop harvest output, would be on the order of 195 to 230 Gg N yr−1, which is 55 to 64% of the current total N inputs to cropland (360 Gg N yr−1). Total groundwater NO3− loading (after accounting for atmospheric losses) would then indeed be on the same order as the above benchmark if yields were sustained at the current level, which indicates the potential value of such guidelines.
Considering the cropland N mass balance, another underlying challenge is that many inputs cannot be reduced, such as atmospheric deposition, irrigation water NO3−, application of urban and industrial organic waste, and the manure generated by the native dairy herd. These now total 127 Gg N yr−1, which is over half of the amount of synthetic fertilizer N applied (230 Gg N yr−1; Table 3). The issue is similar California-wide, as irrigated cropland is about 2.5 times larger and the statewide dairy herd is about two times larger than the study area. Therefore, addressing this issue requires a fundamental shift in crop nutrient and waste management practices. Nutrient management practices need to be developed that account for and use a large amount of N inputs that come from nonsynthetic fertilizer sources. Additionally, waste management practices are needed that convert dairy, municipal, and food processing organic wastes into more effective, transportable, and accountable fertilizer.
Finally, our analyses provide average estimates of the magnitude and intensity of NO3− loading from a wide range of land use categories. Average annual NO3− loading for specific categories are based on simplified assumptions and on limited data with varying degrees of accuracy. The numbers given represent a best, albeit uncertain, approximation of the actual NO3− loading from specific sources. Furthermore, agricultural discharges of NO3− to groundwater may vary widely between individual fields, farms, or facilities of the same category due to differences in operations, management practices, and environmental conditions. We currently lack data to take such variability into account. Since our estimates do not account for edaphic differences in production, we caution that actual local groundwater NO3− loading at any location within the study area is likely to vary from those projected. With such inherent heterogeneity, complexity, and uncertainty, and given the additional uncertainty of the groundwater system itself, it is tempting to lose sight of the overarching effect of intensive agriculture on groundwater quality.
The estimated amount of NO3− loading in 2005 (163 Gg N yr−1) is double the estimated amount of groundwater NO3− loading from cropland in the mid-1970s (81 Gg N yr−1). Assuming recharge rates are similar over the historic period, concentrations of NO3− in recharge have likely doubled over the past 30 yr. For the study area, statistical error analysis furthermore indicates that the 95% confidence interval for 2005 cropland N loading to groundwater (not including alfalfa) ranges from 100 to 220 Gg N yr−1, which is well above the benchmark of 50 Gg N yr−1. The amount of groundwater NO3− loading is therefore of such magnitude that no matter the uncertainty the overarching finding is that agriculture has and continues to significantly degrade groundwater quality in the study area.
Significance and Implications
Intensive N use has boosted crop yields in the past and is a part of the solution to feeding over 9 billion people by the middle of the 21st century (Erisman et al., 2008; Foley et al., 2011). However, the environmental burden of this intensification is now clearly apparent (Galloway et al., 2008; Vitousek et al., 2009). Though not without uncertainty, our estimates provide insight into both the variability and the cumulative magnitude of groundwater NO3− loading from diverse sources. In this way, the N mass balance delivers historical understanding and, perhaps more important, critical information to target remedial action, be it investments in research, technological change, farmer and consumer outreach, or policy interventions. Unequivocally, California agriculture cannot continue along its current N trajectory and still preserve groundwater quality, thus placing Californians and the agricultural industry in precarious positions. The burden and opportunity to curb NO3− loading does not rest with producers alone. National and global consumer food choices and global markets will continue to incentivize N intensification of crop and animal production systems in California, the United States, and agricultural regions around the world. A combination of government incentives, regulations, industry initiatives, and consumer choices informed by better knowledge of the environmental costs associated with nonpoint sources of N pollution is needed to address the issue of NO3− loading to groundwater over the long term.