Environ. Sci. Technol. 2010, 44, 4988–4997
Nitrate in Groundwater of the United States, 1991-2003 K A R E N R . B U R O W , * ,† BERNARD T. NOLAN,‡ MICHAEL G. RUPERT,§ AND NEIL M. DUBROVSKY† U.S. Geological Survey, Placer Hall, 6000 J Street, Sacramento, California 95819, U.S. Geological Survey, 413 National Center, Reston, Virginia 20192, and U.S. Geological Survey, 201 E. Ninth Street, Pueblo, Colorado 81003
Received February 17, 2010. Revised manuscript received May 7, 2010. Accepted May 25, 2010.
An assessment of nitrate concentrations in groundwater in the United States indicates that concentrations are highest in shallow, oxic groundwater beneath areas with high N inputs. During 1991-2003, 5101 wells were sampled in 51 study areas throughout the U.S. as part of the U.S. Geological Survey National Water-Quality Assessment (NAWQA) program. The well networks reflect the existing used resource represented by domestic wells in major aquifers (major aquifer studies), and recently recharged groundwater beneath dominant land-surface activities (land-use studies). Nitrate concentrations were highest in shallow groundwater beneath agricultural land use in areas with well-drained soils and oxic geochemical conditions. Nitrate concentrations were lowest in deep groundwater where groundwater is reduced, or where groundwater is older and hence concentrations reflect historically low N application rates. Classification and regression tree analysis was used to identify the relative importance of N inputs, biogeochemical processes, and physical aquifer properties in explaining nitrate concentrations in groundwater. Factors ranked by reduction in sum of squares indicate that dissolved iron concentrations explained most of the variation in groundwater nitrate concentration, followed by manganese, calcium, farm N fertilizer inputs, percent welldrained soils, and dissolved oxygen. Overall, nitrate concentrations in groundwater are most significantly affected by redox conditions, followed by nonpoint-source N inputs. Other waterquality indicators and physical variables had a secondary influence on nitrate concentrations.
Introduction The natural global nitrogen (N) cycle has been extensively altered by human activities, doubling the rate of N inputs into the terrestrial N cycle (1). The global increase in the use of N fertilizer over the last several decades has also led to increased leaching and runoff of N that threaten water quality, especially in agricultural areas where elevated nitrate concentrations are common (2-9). Although this increase in the use of fertilizer has been important for increasing crop production for food in the face of increasing population (10), * Corresponding author phone: 916-278-3087; e-mail: krburow@ usgs.gov. † U.S. Geological Survey, Sacramento, CA. ‡ U.S. Geological Survey, Reston, VA. § U.S. Geological Survey, Pueblo, CO. 4988
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 13, 2010
questions have been raised as to the sustainability of these practices without causing further impairment of water resources (11). Groundwater supplies more than 33% of the water used for public drinking-water supply in the United States (12). Contamination of groundwater by nitrate is of concern because elevated concentrations can affect human health. Additionally, many surface water bodies receive significant groundwater discharge, and excess N in groundwater can lead to ecological disturbances in receiving surface water (13-15). Elevated concentrations of nitrate in groundwater are highly variable, however, and although the overall increase in N fertilizer use may generally correspond to higher nitrate concentrations (16), it is sometimes difficult to link high nitrate concentrations in groundwater directly to overlying N inputs (17, 18). In addition to complex unsaturated zone processes that influence the N leaching to the water table (2, 19, 20), nitrate may also be affected by physical and chemical processes within the aquifer that result in some areas being more vulnerable to nitrate contamination than others (21-24). To understand where and how nitrate concentrations become elevated requires an understanding of the sources of nitrate and the factors that control how nitrate moves through the hydrologic system. This, in turn, can aid in the development of effective management practices for the most vulnerable areas. Numerous studies at regional and local scales have examined the relation between nitrate concentrations in groundwater and governing factors; however, relatively few studies have evaluated nitrate concentrations across wide regions, representing multiple aquifers and diverse hydrogeologic conditions. Since 1991, The U.S. Geological Survey (USGS) National Water-Quality Assessment (NAWQA) program has been monitoring groundwater quality in the primary aquifers of the U.S. using a common network design, sampling methods, and consistently developed explanatory variables. Results of the first 20 study areas (2130 wells sampled in 1991-1995) reported by ref 5 indicated that nitrate concentrations were highest beneath shallow agricultural land, in areas characterized by high N inputs, well-drained soils, fractured bedrock, and in irrigated areas. This analysis of NAWQA results builds upon the initial assessment (5). Since this time, sampling was done in an additional 2971 wells in an additional 31 study areas, for a total of 5101 wells sampled during 1991-2003 in 51 study areas throughout the U.S. The more extensive data and expanded geographic coverage available for the current study reinforces many of the previously reported findings, allows more detailed analyses of each topic, and supports new findings. This paper describes the occurrence and distribution of nitrate in groundwater and presents statistical comparisons using a greater number of regionally available variables than reported by ref 5, including geochemical variables, groundwater age, and other aquifer and well construction characteristics that were not considered previously. Additionally, the relative importance of these variables was further evaluated. The objective was to identify the most influential factors controlling groundwater nitrate concentrations across the wide range of environmental settings in the U.S., based on an extensive data set that included environmental factors, water quality indicators from groundwater samples, and well construction data. 10.1021/es100546y
2010 American Chemical Society
Published on Web 06/11/2010
Materials and Methods The NAWQA study design for groundwater focuses on assessing the water quality conditions of shallow groundwater beneath agricultural and urban land-use settings, and in major aquifers in each study area (25). The major aquifer studies provide a broad assessment of the water-quality conditions of the used groundwater resource. During 1991-2003, networks of about 20-40 wells were sampled (Supporting Information (SI) Figure S1 and Table S1). Larger aquifer systems were divided into several subunits on the basis of hydrogeologic features. Wells were selected for sampling using a computerized, stratified, random site-selection procedure (26) to minimize spatial bias. Most of the wells sampled for major aquifer studies were domestic wells, which often tap the shallow part of the used resource and thus may not be representative of entire thickness of the aquifer. The land-use studies assess the concentrations and distribution of water-quality constituents in recently recharged groundwater (generally less than 10 years old) associated with primarily agricultural and urban land-use settings and the dominant hydrogeologic conditions in each study unit. Nitrogen inputs and water quality conditions differ substantially among different land-use settings, and sampling networks have been grouped into agricultural, urban, and undeveloped settings. Networks of about 30 wells were sampled once during 1991-2003. Within individual landuse settings, wells were selected for sampling using the same stratified, random site-selection procedure as the major aquifer wells. Most of the wells sampled were shallow monitoring wells, which were typically screened at shallower depths than the domestic wells sampled in the major aquifer studies. Although water quality in shallow wells can be affected by seasonal or annual variability, it was not feasible to collect more samples to characterize the variability at individual wells. The data set used in this analysis includes 5101 wells sampled in 51 study units: 2792 wells were sampled in 94 major aquifer studies; 1414 wells were sampled in 53 agricultural land-use studies; 861 wells were sampled in 33 urban land-use studies, and 34 wells were sampled in undeveloped areas. Wells were sampled for dissolved nitrite plus nitrate using consistent data collection protocols (27) and analyzed at the USGS National Water-Quality Laboratory using standard methods (28). Because nitrite concentrations were low, concentrations of nitrite plus nitrate are referred to as nitrate (as N) in this manuscript. Measured nitrate concentrations were compared to a background concentration of 1 mg/L. The background concentration was estimated based on data for 401 wells in undeveloped areas plus samples from 18 additional wells, using methods documented in ref 29. This value was estimated as the 75th percentile of a single representative sample from each of the 419 wells that had data. Nitrate concentrations in undeveloped areas may vary considerably among regions of the country. In particular, rangeland occurs primarily in the western U.S., whereas the forested sites occur primarily in the eastern U.S. Sites were somewhat clustered within regions, but distributed between forested and rangeland settings. Although concentrations of nitrate beneath rangeland and forest may be different, it is still valuable to compare concentrations in the nationally distributed data set to a single background value. Nitrate concentrations were statistically analyzed using data from individual wells and also using median values calculated for each network. Nonparametric methods were used because the data were not normally distributed. The Kruskal-Wallis test was used to test for differences among data groups. When comparing more than two groups, the
multiple-stage Kruskal-Wallis test was used to determine which groups were different. Spearman’s rho was used to evaluate the correlation of two variables. Test results were evaluated at the 0.05 level of significance. The Partition platform in the software JMP (SAS Institute, Inc.) was used to analyze a subset of the groundwater data set by Classification and Regression Tree (CART) analysis. The CART analysis performs recursive, binary partitions of a continuous dependent variable to identify factors that best predict membership in comparatively homogeneous data clusters called nodes. The partition at a particular parent node is determined by maximizing the Log Worth statistic, which is related to the p-value corresponding to the reduction in sum of squares realized by the child nodes (30). The fitted value at a node is the mean of the observations at the node. Interactions between variables are indicated by variables at the same level on opposing branches of the tree. The data subset (2,257 observations) consisted of wells in agricultural and urban land-use studies and was restricted to sites that had major and minor element data. Wells from the land-use studies were selected because these wells were expected to most closely correlate with N inputs. N input data for 500 m buffers around groundwater sampling sites were estimated from data on fertilizer (1987-2001), manure (1982, 1987, 1992, and 1997) and atmospheric deposition (1985-2001); data were modified from that in 31. Because networks were sampled over a 10year span, 4 years of N input data bracketing the sampling period for each network were averaged to arrive at one number to represent N inputs for each site. The reduction/oxidation (redox) conditions of groundwater at the time of sampling were evaluated using field measurements of dissolved oxygen (DO) and laboratory measurements of Mn and Fe using a redox classification methodology similar to that described by ref 32. Because of the potential bias in using nitrate as a redox indicator for a statistical analysis using nitrate concentrations, the classification scheme outlined by (32) was modified to exclude nitrate. Groundwater was classified as oxic if DO was >0.5 mg/L, Mn concentrations e50 µg/L, and Fe concentrations e100 µg/L. Groundwater was classified as reduced if DO concentrations were e0.5 mg/L, Mn concentrations were >50 µg/L, and Fe concentrations were >100 µg/L. Groundwater was assigned a mixed classification if concentrations of multiple constituents met the criteria for both oxic and reduced classifications. Soil drainage characteristics were determined from soil hydrologic group data in the State Soil Geographic (STATSGO) database (33) and were compiled in 1:250 000 scale GIS maps. The soil hydrologic group variable has four major categories ranging from soils with high to moderate infiltration rates (groups A and B, respectively) to soils with slow and very slow infiltration rates (groups C and D, respectively). The categorical variable was converted to a continuous variable by using a GIS to compile the percent area of each soil hydrologic group within 500 m buffers around each sampled site. Similar to soil infiltration properties, a variable for recharge rate was determined by combining an estimate of mean annual recharge from base flow analysis (34) with total groundwater withdrawals for irrigation (35) in 500 m buffers around each sampled site. The type of aquifer in which wells are screened can affect travel times and redox conditions in the aquifer and thus also affect nitrate concentrations in groundwater. Wells screened in unconfined aquifers are more directly connected to inputs from land surface, whereas confined aquifers are separated from the water table by a low permeability layer or layers that impede the movement of land-surface-derived contaminants to the aquifer. A subset of data (996 wells) was identified to evaluate the importance of aquifer type (unVOL. 44, NO. 13, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4989
FIGURE 1. Distribution of nitrate concentrations in groundwater for each type of well network. Box plots labeled with different letters (A, B, C) indicate that differences in nitrate concentrations among each group are significant. confined or confined). To eliminate spatial bias in the data set, only wells from major aquifer networks were used. Networks were selected for inclusion in the data set if they included wells screened in both the unconfined and confined parts of the aquifer.
Results and Discussion Nitrate Occurrence and N Input. Nitrate was detected at concentrations above 1.0 mg/L in 50% of the wells sampled. Shallow groundwater beneath agricultural land had the highest median concentration of nitrate (3.1 mg/L). Concentrations in shallow groundwater beneath urban land (1.4 mg/L) were lower than beneath agricultural land, but were higher than groundwater sampled in major aquifers (0.56 mg/L) (Figure 1). The land-use studies sampled shallow, recently recharged groundwater and are likely to reflect recent N inputs at the land surface. Nitrate concentrations were higher than the USEPA maximum contaminant level (MCL) of 10 mg/L in 437 wells (8%); concentrations were above the MCL in 20% of wells in the agricultural land-use setting, 3% of wells in the urban land-use setting, and 4% of wells in major aquifers. These results are similar to that in ref 5; however, in this study, concentrations were significantly different among all three network types: agricultural landuse, urban land-use, and major aquifer wells (p < 0.001). In contrast to spatial patterns noted in the previous study using fewer networks (5) in which the highest nitrate concentrations were clustered in the mid-Atlantic and western U.S., high median concentrations (the highest 25%) are widely distributed across the U.S. (Figure 2). The highest concentrations are likely the result of high N inputs and conditions favorable to nitrate transport in groundwater. Nitrate concentrations in most major aquifer studies (Figure 2c) are low or medium (the lowest 75%), with the highest concentrations in parts of the Northeast, the Central Plains, and the Southwest, and the lowest concentrations in the Midwest along the Mississippi River to Gulf Coast aquifers. High concentrations in the Central Plains and Southwest could be influenced by irrigation practices and the limited ability for natural attenuation of nitrate in these aquifers, which may accelerate the downward movement of nitrate in groundwater (36, 22). Wells in the Northeast aquifers are shallower than wells in most of the other major aquifer studies, and may represent younger groundwater that has moved rapidly through the system. Low concentrations in the major aquifers from the Midwest to Gulf Coast are likely attributed to a combination of physical and chemical properties that inhibit rapid nitrate transport to the deeper aquifers. 4990
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 13, 2010
Spatial variations in the occurrence and distribution of nitrate concentrations in groundwater are due, in part, to differences in N inputs. In the U.S., the amount of N entering the environment from the application of N fertilizer has increased 20-fold since 1945, with about 20 million metric tons of N introduced in 1997 (31). A very small portion, about 2-4%, of the total N fertilizer is used in nonagricultural settings such as city parks and residential lawns. The mass of N in manure produced by livestock is about one-half as much as that applied in fertilizer, and atmospheric deposition contributes about one-half as much N as manure does, and less than a quarter as much N as does fertilizer application. The proportion of each type of N input, the rate of fertilizer use, and the mode of application varies widely throughout the U.S. Nitrate concentrations in oxic wells in shallow agricultural and urban land-use studies were weakly but significantly correlated to total N input for the 500 m area around each well (p ) 0.01, rho ) 0.28) (SI Figure S2). Similar to results shown in the previous study (5), some of the networks with high N input have low nitrate concentrations and some of the networks with low N input have high nitrate concentrations. These results highlight that other factors besides N input influence nitrate concentrations in groundwater. Age of Groundwater and Relation to Depth and Well Type. Because of the geometry of groundwater flow systems and because of higher N input in more recently recharged groundwater, the age of groundwater and corresponding nitrate concentrations are expected to become stratified with depth in the subsurface (e.g., refs 4, 5, 37). A subset of the wells had tritium (3H) data, an environmental tracer that can be used to trace the flow of water recharged within the past 50 years (38). Elevated 3H concentrations indicate that groundwater was recharged after the early 1950s thermonuclear device testing. 3H concentrations were inversely correlated to well depth below the water table (p < 0.001, rho ) -0.39) indicating that shallow groundwater is generally younger than deep groundwater. Similarly, nitrate concentrations were also inversely correlated to well depth below water table (p < 0.001, rho ) -0.14), as noted by ref 5. Nitrate concentrations were significantly correlated with 3H concentrations (p < 0.001, rho ) 0.27), consistent with higher nitrate in more recently recharged groundwater. 3H concentrations were also used to approximate whether groundwater is predominantly old (recharged in 1952 or earlier) or predominantly young (recharged after 1952). Groundwater with 3H concentrations greater than 2.5 pCi/L was classified as young and groundwater with 3H less than or equal to 2.5 pCi/L was classified as old. Nitrate concentrations were significantly higher in young groundwater than old groundwater (p < 0.001) (Figure 3). Most of the old groundwater had nitrate concentrations less than or equal to 1.0 mg/L (62%), which is consistent with low N inputs before the early 1950s. Most of the young groundwater had nitrate concentrations greater than 1.0 mg/L (59%), reflecting more recent N input rates. However, the age of groundwater is not definitive, as a few wells (1%) with old groundwater had nitrate concentrations higher than the MCL, and 41% of the wells with young groundwater had nitrate concentrations less than or equal to 1.0 mg/L. Another factor important to consider is well type because of potential differences in nitrate concentrations due to well depth, screen length, and resulting groundwater age. Nitrate concentrations were detected above 1.0 mg/L most frequently in monitoring wells, followed by domestic and then publicsupply wells (Table 1). In this study, sampled monitoring wells were typically screened near the water table (median well depth below water table of 3.8 m), whereas sampled domestic wells were screened at deeper intervals in the aquifer (median well depth below water table of 26 m). A subset of the major aquifer
FIGURE 2. Median nitrate concentration in (a) agricultural land use, (b) urban land use, and (c) major aquifer studies sampled by the NAWQA program during 1991-2003. Low, medium, and high concentrations correspond to the 0-25th, 25th-75th, and 75th-100th percentiles, respectively.
FIGURE 3. Distribution of nitrate concentrations by age of groundwater. Age of groundwater was determined by tritium concentrations, in which old groundwater had 3H concentrations e2.5 pCi/L and young groundwater had 3H concentrations >2.5 pCi/L. wells sampled were public-supply wells, which were screened at deeper intervals in the aquifer (median well depth below water table of 75 m) than the domestic wells. High nitrate concentrations were most prevalent overall in shallow monitoring wells, with 12% above the MCL (Table 1). High nitrate was less common in deeper domestic wells, with 7% above the MCL. High nitrate concentrations were the least common in deep public-supply wells, with only 2% above the MCL. These wells were sampled prior to treatment
and/or blending and therefore do not represent concentrations delivered to consumers. Although it is useful to be aware of bias introduced by analyzing data with a combination of well types, it is difficult to separate differences in concentration that are due to depth in the aquifer from differences due to attenuation (such as denitrification), mixing, or dispersion. The monitoring wells sampled typically had short screened intervals, whereas the domestic wells had longer screened intervals. In addition to being deeper in the aquifer, public-supply wells had the longest screened intervals and samples likely represent a mixture of a wide range of ages of groundwater. Geochemical Conditions. The redox conditions of groundwater are an important geochemical factor explaining nitrate concentrations in groundwater because reduced geochemical conditions promote attenuation of nitrate through denitrification. Reduced conditions predominate in groundwater in areas where large amounts of organic carbon are present, and where oxygen is depleted, such as along long groundwater flow paths or in low permeability, waterlogged soils. Previous studies (5, 21) noted correlations between nitrate concentrations and indirect, nonchemical measures of denitrification, such as poorly drained soils or histosols. In this study, analysis of additional variables, including use of a redox classification scheme using chemical indicators in sampled groundwater, provide additional insight into the relations among nitrate concentrations, redox conditions, soil properties, and groundwater age. VOL. 44, NO. 13, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4991
8 20 3 12 major aquifers agricultural land use urban land use all networks
mg/L, Milligrams per liter; %, percent; AG LUS, agricultural land use study; URB LUS, urban land use study; na, not applicable.
374 1 9 384
type of network
a
183 938 781 1,902 4 22 6 7 42 78 94 48 1965 406 17 2,388 2 na 0 2
number sampled
40 na 33 40
number sampled wells with nitrate greater than 10 mg/L (%)
public-supply wells
38 59 53 54
wells with nitrate greater than 10 mg/L (%) number sampled
observation wells
wells with nitrate greater than 1.0 mg/L (%) wells with nitrate greater than 10 mg/L (%) wells with nitrate greater than 1.0 mg/L (%)
domestic wells 9
wells with nitrate greater than 1.0 mg/L (%)
TABLE 1. Summary of Nitrate Occurrence by Well Type and Percent above Thresholda 4992
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 13, 2010
FIGURE 4. Median nitrate concentrations and percent of oxic wells grouped by percent of well drained soils in 500 m buffers around the well. The depth to seasonally high water table characteristic in the soils data reflects saturated near-surface soils resulting from poor drainage or other hydrologic factors. Areas with a shallow seasonal water table correspond to areas with reduced conditions in shallow groundwater (SI Figure S3). This soil variable showed a distinct spatial pattern: areas with a shallow depth to seasonally high water table and reduced groundwater tend to occur in the eastern half of the U.S. where the climate is humid and soils have a high organic content. In contrast, oxic conditions tend to occur in the western half of the U.S. where the climate is dry, soils have a low organic content, and irrigation pumping often increases the depth to water. These results are consistent with a recent study by ref 22, where regional redox patterns were evaluated in the context of principal aquifer lithologies. Another soil characteristic, the percent of well-drained soils (here quantified as the sum of percent of Soil Hydrologic Groups A and B in a 500 m well buffer), show a similar relation among soil properties, redox conditions, and corresponding nitrate concentrations in groundwater. Median nitrate concentrations and the percent of oxic wells are the lowest for wells located in areas with less than or equal to 25% welldrained soils. The nitrate concentrations and percent of oxic wells generally increase as the percent of well-drained soils increases (Figure 4). It is noteworthy that the major aquifer wells show the same relation among nitrate concentration, percent oxic wells and percent well-drained soils as the land use wells, but the nitrate concentrations tend to be lower overall for a similar value of percent of oxic wells. The lower overall nitrate concentrations in the major aquifer wells may reflect the greater depth and/or age of groundwater in the major aquifer wells compared to the land use wells, but suggests that these deeper wells are affected by the same processes as the shallower wells even though most of the concentrations are below 1.0 mg/L. Previous research (5, 21) indicated the importance of the correlation between welldrained soils and nitrate concentrations. This study further links soil drainage characteristics with oxic geochemical conditions, suggesting that well-drained soils influence both transport and attenuation factors that control nitrate concentrations in groundwater. Young groundwater tends to be more oxic because dissolved oxygen may be progressively depleted along
FIGURE 7. Nitrate concentrations by aquifer type for a subset of wells from major aquifer networks. Networks were included if both unconfined and confined wells were sampled in the same study area. FIGURE 5. Median nitrate concentration grouped by groundwater age and redox classification.
FIGURE 6. Median nitrate concentration and nitrogen input grouped by redox classification for agricultural and urban land use studies. groundwater flow paths with time. The highest nitrate concentrations occur in young, oxic groundwater: 66% of the wells with nitrate concentrations above 1.0 mg/L and 78% with nitrate above the MCL occur in young, oxic water (Figure 5). Reduced groundwater rarely had concentrations over the MCL, and then only in young groundwater. As another indicator of the importance of redox conditions on nitrate concentrations, nitrate concentrations are highest in young, oxic groundwater regardless of N inputs (Figure 6). For example, N input in agricultural areas is the same for wells across the three redox classes, but median nitrate is 5.6 mg/L in oxic wells and
