Professional Documents
Culture Documents
Abstract
Because of the ubiquitous nature of anthropogenic nitrate (NO2 3 ) in many parts of the world, determining
background concentrations of NO2 3 in shallow ground water from natural sources is probably impossible in most
environments. Present-day background must now include diffuse sources of NO2 3 such as disruption of soils and
oxidation of organic matter, and atmospheric inputs from products of combustion and evaporation of ammonia
from fertilizer and livestock waste. Anomalies can be defined as NO2 3 derived from nitrogen (N) inputs to the
environment from anthropogenic activities, including synthetic fertilizers, livestock waste, and septic effluent.
Cumulative probability graphs were used to identify threshold concentrations separating background and anoma-
lous NO3-N concentrations and to assist in the determination of sources of N contamination for 232 spring water
samples and 200 well water samples from karst aquifers. Thresholds were 0.4, 2.5, and 6.7 mg/L for spring water
samples, and 0.1, 2.1, and 17 mg/L for well water samples. The 0.4 and 0.1 mg/L values are assumed to represent
thresholds for present-day precipitation. Thresholds at 2.5 and 2.1 mg/L are interpreted to represent present-day
background concentrations of NO3-N. The population of spring water samples with concentrations between 2.5
and 6.7 mg/L represents an amalgam of all sources of NO2 3 in the ground water basins that feed each spring; con-
centrations >6.7 mg/L were typically samples collected soon after springtime application of synthetic fertilizer.
The 17 mg/L threshold (adjusted to 15 mg/L) for well water samples is interpreted as the level above which live-
stock wastes dominate the N sources.
Champaign, IL 61820; (217) 333-2167; fax (217) 244-7190; concentration in its Water Basics glossary as ‘‘a concen-
martinse@.uiuc.edu tration of a substance in a particular environment that is
4Illinois State Geological Survey, 615 East Peabody Drive, indicative of minimal influence by human (anthropo-
Champaign, IL 61820; (217) 244-2396; fax (217) 244-2785; genic) sources.’’
hackely@isgs.uiuc.edu Determining background concentrations for some
Received July 2005, accepted April 2006.
Copyright ª 2006 The Author(s) contaminants can be difficult, especially those that have
Journal compilation ª 2006 National Ground Water Association. multiple and nonpoint sources, or are reactive in the envi-
doi: 10.1111/j.1745-6584.2006.00240.x ronment. Nitrate (NO2 3 ) is an example of an ion for which
influence or at least certain types of influence. Studies problematic. Pristine areas may have substantially different
that have evaluated NO3-N data collected prior to large- physical, chemical, and biological characteristics than the
scale application of synthetic fertilizers include Hallberg potentially impacted area(s). In an example from explora-
and Keeney (1993), Katz et al. (2002), Limbrick (2003), tion geochemistry, concentrations of metals in natural
and Jones and Smart (2005) (Table 1). These data proba- water are orders of magnitude higher in mineralized areas
bly represent the best estimate of background values that than in nonmineralized areas (Runnells 1998). With
are available; however, the primary limitation in using respect to background NO3-N, if anomalies or events are
historic data is that such data sets are scarce. limited to NO2 3 derived directly from synthetic fertilizers,
Determining background NO3-N concentrations from livestock waste, and/or human sewage, background values
ground water samples collected in pristine areas is a com- will include some anthropogenically derived materials.
mon practice. In a highly cited publication, Mueller and These materials can include organic matter from crop resi-
Helsel (1996) used data from ‘‘relatively pristine’’ sites due or tilled soils, products of combustion, and atmo-
sampled as part of the USGS National Water-Quality spheric inputs due to evaporation of NH3 from fertilizer
Assessment Program to define background NO3-N in U.S. and livestock waste, as well as natural sources.
ground water as being <2 mg/L. Many other researchers In this paper, we use a statistical procedure borrowed
have collected samples from areas relatively uninfluenced from exploration geochemistry to help define present-day
by human activities to estimate background NO3-N con- background concentrations for NO3-N and determine
centrations, with values varying from below detection threshold values separating sample populations having
to >4 mg/L (McKenna et al. 1988; Postma et al. 1991; different NO2 3 sources in karst aquifers of southwestern
Leenhouts et al. 1998; Cho et al. 2000; Brye et al. 2001; Illinois. We believe this technique is an improvement over
Perakis and Hedin 2002; Nolan and Hitt 2003). It should calculated means/medians and other ways by which
be noted, however, that there are areas in igneous and background values for NO3-N have been determined and
metamorphic terrains of California that are capable of applied. This procedure, in conjunction with other stat-
generating NO3-N concentrations in surface streams as istical, analytic, and land-use data, provides information
large as 6.8 mg/L (Holloway et al. 1998). In addition, N about the background/threshold concentrations and the
in sedimentary rock and sediments is associated with the nature of NO2 3 contamination in the study area.
organic fraction (e.g., kerogen in shale) that may be
released to surface water and ground water during weath-
ering (Holloway and Dahlgren 2001). Methods
Because of the long history of human impacts to the All ground water samples were collected from
global N budget, defining a pristine area can be springs and wells fed by the karst aquifers of Illinois’
S.V. Panno et al. GROUND WATER 44, no. 5: 697–709 699
sinkhole plain, southwestern Illinois. Nitrate in ground (Sinclair 1974). In this study, we used cumulative proba-
water in the study area should be associated with natural bility plots, a technique borrowed from the field of
soil processes and anthropogenic sources. The dominant geochemical mineral exploration (Sinclair 1974, 1991).
land use in the area is row-crop agriculture and small Indicators, such as metal concentrations in rock, soil, or
livestock operations. Naturally occurring sources of NO2 3 water samples, can be divided into two or more pop-
from rock-water interactions should be minimal. The ulations separated by inflection points on cumulative
karst-forming formations in southwestern Illinois are probability graphs. The populations determined in this
Mississippian-age limestones of the Salem, St. Louis, and manner are designated as background and one or more
Ste. Genevieve formations. The limestones are primarily ‘‘anomalies,’’ and the concentrations at the inflection
fine- to coarse-grained biocalcarenites with abundant fos- points are defined as ‘‘threshold’’ values. In the context of
sils (Willman et al. 1975) and contain very little organic geochemical exploration, background is used as the base-
material. Shale interbeds are generally not found in the line concentration; anything above this threshold concen-
karst region, and when they are, they are thin and organic tration is interpreted as an anomaly and indicative of the
poor (Z. Lasemi, personal communication, 2005). presence or influence of a mineral deposit (Sinclair
A total of 432 samples collected between 1994 and 1983). We reasoned that the same technique could be
2000 were compiled from several sources (Panno et al. applied to water chemistry data where it should be possible
1996, 2001, 2003; Hackley et al. in press; Panno et al. in to separate natural sources from anthropogenic sources.
press; supplementary material available as part of the Important limitations to this approach are that a minimum
on-line article at http://www.blackwell-synergy.com); all of ~100 values is needed and, as Sinclair (1991) states,
samples were collected by one or more of the authors of ‘‘the determination of thresholds must be viewed as an
this paper. Ground water samples were passed through estimation procedure in the statistical sense, subject to
0.45-lm filters during collection, and the data represent random and systematic error.’’
dissolved NO3-N. The data were divided into two groups On cumulative probability plots, maximum curvature
for statistical evaluation, one with 232 samples from of the lines reveals inflection points or threshold values
springs and seeps and the other with 200 well water between populations. This is based on the fact that the
samples. The two groups represent distinct subsurface values of a single normally or lognormally distributed
environments that are expected to influence NO3-N population will form a straight line. Two or more mixed
concentrations in different ways. Spring discharge repre- populations will result in a curved line with a pronounced
sents an amalgam of soil water and shallow and deep inflection point(s). Those inflection points are thresholds
ground water draining into a central conduit system. distinguishing two or more overlapping populations.
Water from point sources, such as septic effluent, mixes The NO3-N concentrations from spring (n ¼ 232)
with water from nonpoint sources, such as synthetic fertil- and well samples (n ¼ 200) were log transformed prior to
izers, and water from forested areas. Conversely, the plotting on separate cumulative probability plots. Class
ground water that is sampled in the wells, while intersect- intervals of 0.25 mg/L for springs and 0.20 mg/L for
ing crevices and conduits, may undergo mixing of con- wells were selected by trial and error in order to create
taminated, near-surface sources with ground water from smooth curves on the probability plots, resulting in ~20
deep (uncontaminated) sources. classes for each data set. The probability of each class
The spring water samples were collected from 16 was calculated by consecutively adding the number of
karst springs and 16 cave seeps and the well water samples samples in each set, in order, from highest concentration
from 58 wells completed in the karst aquifer (Figure 1). to lowest. The cumulative number in each class was then
Well depths ranged from 6 to 190 m. One of the springs divided by n in order to calculate cumulative percent
and two of the wells were sampled 10 or more times; the probability. The midpoints between data classes were
number of samples collected from the other springs, plotted against cumulative percentages, and a relatively
seeps, and wells ranges from one to seven. In order to smooth line was constructed connecting the data points.
assess the significance of the effects of (1) individual Thresholds were chosen at the intersection of two ‘‘lin-
springs or wells and (2) the date of sampling on NO3-N ear’’ segments of the probability curves. Because of the
levels, analysis of variance (ANOVA) was conducted on paucity of data points at the extremes of the data set,
two-way layout models with individual springs or wells abrupt changes in slope may be present that reflect only
as one factor and date of sampling as the second factor. the incompleteness of the data set and not an inflection
point or threshold.
Cumulative Probability Plots In addition to NO3-N concentrations, NO2 3 stable iso-
Many data sets of the chemical composition of tope values were available for 28 well samples and atrazine
ground water can contain one or more populations (e.g., concentrations were available for 206 and 172 spring and
background and anthropogenic contributions from one or well samples, respectively (Panno et al. 1996, 2001, 2003;
more sources). Consequently, the determination of a back- Hackley et al. in press). Nitrate isotope data can be a useful
ground concentration for a particular analyte can be diffi- tool for identifying sources of NO2 3 in the environment, as
cult. The mean and standard deviation of a data set have well as determining if denitrification has occurred (Heaton
been used to estimate a background value using the mean 1986; Aravena and Robertson 1998; Panno et al. 2001).
plus 2 standard deviations. This technique is arbitrary (in Chloride (Cl2) and bromide (Br2) data (Panno et al. in
the selection of the statistics) and does not consider press, supplementary material), land-use information, and
the possibility of the presence of multiple populations sample date were also used to help interpret NO2 3 sources
and the results from the NO3-N probability graphs. Plotting the spring water samples. Nitrate was detected in all the
Cl/Br vs. Cl2 and vs. total N (Panno et al. 2006) is a new spring water samples, while NO3-N was below detection
technique that can separate different sources of Cl2. All (0.01 mg/L) in 15% of the well water samples.
the data used in this paper, sample locations, land use for In the two-way ANOVA model with individual
well water samples, and additional plots discussed herein springs and sampling dates as factors, the individual
may be found in an Illinois State Geological Survey Open spring was highly significant (P < 0.001), while the date
File Series paper (Panno et al. in press) and in the journal’s of sampling was not significant at all (P ¼ 0.912). In the
supplementary material. two-way ANOVA model with individual wells and dates
of sampling as factors, individual wells were significant
at the 0.1 level (P ¼ 0.077); date of sampling was again
Results not significant (P ¼ 0.160).
Figure 2. Cumulative probability graph for 232 spring water samples (32 springs and seeps) showing threshold values.
A histogram and box and whisker plot of the data are also shown.
thresholds established by the graphs and are thus properly Threshold NO3-N Concentrations
accounted for in determining background levels. The spring water population with NO3-N concen-
The successful application of the cumulative proba- trations <0.4 mg/L may represent near-pristine or the low
bility graph method does not require normally or lognor- end of present-day background conditions in the conduit-
mally distributed data (Sinclair 1991). For springs, it was dominated ground water in the study area; the average
clear from the histogram of the log NO3-N values N concentration of precipitation in the area is ~0.6 mg/L
(Figure 2) that the overall distribution was not lognormal. for 2003 (National Atmospheric Deposition Program
The inflection points on the probability graph indicated [NADP] 2004). There were only five samples in this pop-
an interpretable breakdown of the distribution of the log- ulation, thus defining them as a separate population may
ged values, as follows: (1) above a logged value of 0.83 not be warranted. All these samples were collected from
(original NO3-N value of 6.7 mg/L), the distribution was seeps in a cave draining small nonagricultural watersheds
skewed to the left with a relatively long right-hand tail, during periods of high rainfall. It is likely that this was
similar to an exponential distribution; (2) between the runoff that passed rapidly through the soil zone via sink-
logged values of 0.4 and 0.83 (2.5 and 6.7 mg/L), the holes or associated macropores. The threshold at 2.5 mg/
distribution was skewed in the opposite direction, to the L was interpreted to be the upper bound of present-day
right; and (3) between the logged values of 20.4 and 0.4 background, and concentrations above this value were
(0.4 and 2.5 mg/L), the distribution was fairly uniform. considered to be events, samples for which there were
For wells, there was an analogous breakdown (Figure 3). land-based anthropogenic sources. The background pop-
Above a logged value of 1.2 (17 mg/L), the distribution ulation (n ¼ 33) would likely include NO2 3 from some
was again skewed to the left; between the logged values anthropogenic sources, such as soil organic matter from
of 0.3 and 1.2 (2.1 and 17 mg/L), the distribution was crop residues, combustion products (St. Louis is <50 km
fairly uniform; between the logged values of 21 and 0.3 away and generally upwind of the study area), and evapo-
(0.1 and 2.1 mg/L), the distribution was somewhat ration of NH3 derived from synthetic fertilizer. For the
skewed to the left. Fundamentally, the probability graph population with NO3-N concentrations between 2.5 and
approach split the overall distribution into distinct compo- 6.7 mg/L, the anthropogenic sources of NO2 3 contamina-
nents and, in so doing, identified threshold values. tion were an amalgam of N fertilizer, animal waste, and
S.V. Panno et al. GROUND WATER 44, no. 5: 697–709 703
septic effluent. This was by far the largest population of
spring water samples (n ¼ 165), highlighting the vulnera-
bility of karst systems to surficial contamination. The
land use in the ground water basins from which samples
with NO3-N concentrations >6.7 mg/L came was domi-
nated by row-crop agriculture. More than half of these
samples were collected in late May or June and thus may
have contained NO2 3 from N fertilizer applied in the
springtime (typically late April or early May) that had
been flushed out by heavy rains.
There were many more samples in the population
with the lowest NO3-N concentration for the well water
samples compared to the spring water samples (n ¼ 38),
even though the threshold concentration was slightly
lower (0.1 vs. 0.4 mg/L). This population is primarily
made up of wells located in areas where land use was
primarily nonagricultural (Figure 4) or very deep wells
where conditions can be anoxic (Panno et al. in press,
supplementary material). Well water samples are more
likely to have lower NO3-N concentrations than spring Figure 4. Box and whisker plot of NO3-N concentrations in
water because they are relatively more protected from well samples based on predominant land use in the area. The
surficial contamination (longer flow paths), and denitrifi- Livestock category generally includes wells in areas with
cation is likely to be more substantial due to longer resi- both row crop and livestock land use.
dence times and a potentially greater component of soil
water and matrix flow. Further, most of the ground water
samples collected showed evidence of denitrification were probably dominated by animal waste; these samples
(Figure 5; Panno et al. in press, supplementary material); were collected from relatively shallow wells that were in
it is likely that some of the samples with NO3-N concen- very close proximity to livestock (Panno et al. in press,
trations <0.1 mg/L were affected by denitrification within supplementary material).
the soil zone and/or localized areas within the karst aqui- Even though spring water samples were more likely
fer. Defining 0.1 mg/L as the natural background in well to exhibit NO2 3 contamination than well water samples,
water samples, however, is not warranted because of the their concentration range was much smaller (Panno et al.
proximity of agricultural contamination to even poten- in press, supplementary material). This is probably due to
tially pristine wooded areas within the study area. For the greater mixing of water from different areas within
example, total N in precipitation from pristine forests of conduits; consequently, point sources of contamination
the Pacific Northwest of the United States is as low as can go unrecognized in spring water samples. Conversely,
0.02 mg/L (NADP 2004). Given that NO3-N concen- a contaminated well would be diluted or flushed out at
trations in spring water in the study area during and fol- a somewhat slower rate than a contaminated spring. Well
lowing rainfall events reflect the NO3-N concentration water compositions can reflect the effects of both point
in precipitation, it follows that natural background for and nonpoint sources. An additional factor is that soil and
NO3-N within the karst aquifer was probably <0.1 mg/L. sediment stored in conduits may be a fairly continuous
There is evidence that shallow and deep ground source of NO2 3 to the springs, through dissolution and
water mix in some wells because many wells are uncased oxidation of organic matter (Panno and Kelly 2004).
below the soil-bedrock interface in the study area (Panno Thus, there would be elevated NO3-N concentrations in
et al. 1996). Thus, NO3-N concentrations can be elevated springs even during periods when one would expect low
(up to 30 mg/L) in even the deepest wells of the study concentrations, such as the late summer.
area. The threshold value at 2.1 mg/L was interpreted as The present-day threshold NO3-N values of between
being the upper limit of present-day background, compa- 2.1 and 2.5 mg/L in this karst environment are similar to
rable to the 2.5 mg/L value found for the spring water the most commonly cited values for ground water in the
data. The slightly lower concentration for the well water United States; for example, the threshold value used by
samples suggests that the well water samples had under- the USGS for shallow ground water is 2.0 mg/L (USGS
gone more denitrification than the spring water samples. 2002). However, our technique identified two additional
The far greater percentage of spring water samples above thresholds, one for present-day precipitation at ~0.4 mg/L
present-day background also indicates that the well water (similar to the threshold value for surface streams of 0.6
samples were less likely to be impacted and more likely mg/L [USGS 2002]), and one for ground water containing
to have undergone denitrification. The NO2 3 in samples animal waste at ~17 mg/L.
with NO3-N concentrations between 2.1 and 17 mg/L was
primarily from N fertilizer and septic effluent; these sam- Other Chemical Data
ples were collected from areas dominated by row-crop The NO2 3 isotope and pesticide data and land-use
agriculture and in the vicinity of a private septic system. information and Cl/Br plots (this investigation; Panno et al.
Well water samples with concentrations exceeding 17 mg/L in press, supplementary material) support the threshold
704 S.V. Panno et al. GROUND WATER 44, no. 5: 697–709
Figure 5. d15 NO23 vs. NO3-N for 28 well water samples (data from Hackley et al. in press). Wells contaminated with animal
waste have high NO3-N concentrations (>15 mg/L). Arrow indicates increasing d15NO2 3 values as NO3-N concentrations
decrease due to denitrification.
values and interpretations of sources determined from the likely from synthetic fertilizer and soil organic carbon;
probability graphs. A plot of d15N values vs. NO3-N con- the decreasing NO3-N concentrations with increasing
centrations for 28 of the well water samples (data from d15N values is suggestive of denitrification (Hackley et al.
Hackley et al. in press) shows a sample population with in press). The two samples with NO3-N concentrations
elevated NO3-N concentrations (>15 mg/L) and d15N val- <2.1 mg/L and d15N values >20& had probably under-
ues between 8& and 13& (Figure 5). All these samples gone considerable denitrification, although the original
were collected at farms from wells located in close prox- source for these samples was no longer determinable. The
imity to livestock and were determined, by Hackley et al. other samples with NO3-N concentrations <2.1 mg/L may
(in press), to have been affected by animal waste. Nitrate represent present-day background conditions or minor
from soil or synthetic fertilizer sources tends to have rela- amounts of denitrification.
tively low d15N values, <10& and typically <5& in Illi- About 19% of the well samples had detectable atra-
nois, while NO2 3 from livestock waste or septic systems zine. Those with detectable atrazine had significantly
has somewhat larger d15N values, generally between 8& greater NO3-N concentrations than samples without
and 20& (Hackley et al. in press). The picture is compli- (median ¼ 6.4 mg/L vs. 0.21 mg/L), and more than 90%
cated by denitrification; as NO2 3 is degraded, the d N
15 of the samples with detectable atrazine had NO3-N con-
value of the NO2 3 remaining in solution increases. We centrations greater than the estimated present-day back-
conclude from these and land-use data that the NO3-N ground of 2.1 mg/L (Figure 6). Atrazine is obviously an
from the samples with concentrations >15 mg/L were pri- anthropogenic contaminant, and its absence from samples
marily from livestock waste, while the NO3-N in samples with NO3-N concentrations <2.1 mg/L is additional evi-
with lower concentrations and similar or lower d15N val- dence that 2.1 mg/L is an appropriate present-day thresh-
ues were primarily from the degradation of soil organic old concentration in this area. In contrast, ~70% of spring
matter and synthetic fertilizer sources. The data points samples had detectable atrazine, and the difference in
clustered along and to the left of the denitrification vector NO3-N concentrations between samples with and without
in Figure 5 were all from wells located in cropped areas. atrazine was not significant (Figure 6). This is additional
The 15 mg/L concentration separating livestock waste from evidence that there is considerable mixing of water from
soil/synthetic fertilizer sources is commensurate with the different areas within conduits that discharge at springs.
threshold values determined from the well water probabil- The separation of well water samples affected by ani-
ity graphs (17 mg/L). It appears that for wells with NO3- mal waste, septic effluent, and row-crop agriculture was
N concentrations <15 mg/L, the shallowest wells (<50 m), also accomplished through the use of a recently developed
presumably with the youngest water, have undergone less technique using Cl/Br vs. Cl2 and other related plots
denitrification than most of the deeper wells. (Panno et al. 2006, in press, supplementary material).
Most of the samples with concentrations between the
present-day threshold value of 2.1 and 15 mg/L had d15N Natural Background vs. Present-Day Background
values between 2& and 10& (Figure 5). Land use in the Most published NO3-N concentrations identified as
vicinity of these well water samples was predominately indicating anthropogenic contamination range from 0.08 to
row-crop agriculture. Consequently, the NO2 3 was most 4.5 mg/L (Table 1). Several factors contribute to this wide
S.V. Panno et al. GROUND WATER 44, no. 5: 697–709 705
range, including that estimates were made using different
techniques and for diverse environments. An important con-
sideration is that investigators have defined two separate
background values, one for natural and one for present-day
conditions. Natural background NO3-N concentrations in
shallow ground water should by definition include only
NO2 3 from dissolved geological N or in recharge water in
areas unaffected by humans. This would be limited to natu-
ral atmospheric inputs, natural organic matter degradation
in soils and the saturated zone, wildlife waste, and dissolu-
tion of N-containing minerals (Figure 7). Ideally, natural re-
fers to presettlement conditions, but because there are
essentially no presettlement data, natural background values
are generally determined for samples collected in natural
areas, such as forests and prairies. Unfortunately, in much
of the world, especially in intensively farmed areas such as
the Midwestern United States, there are very few natural
areas and the ones that do exist are not particularly pristine.
For example, humans are responsible for nearly doubling
Figure 6. NO3-N concentrations as a function of atrazine the amount of fixed N entering land surface on a global
detection in 172 well water samples and 206 spring water scale (Tilman et al. 2001). Combustion products from coal
samples. Threshold values of 2.1 and 15 mg/L determined burning and vehicle exhaust and NH3 volatilized from syn-
from cumulative probability graph for well water samples
thetic fertilizers contribute to atmospheric inputs over large
are shown. Of the well water samples with nondetectable
atrazine, 28 had NO3-N concentrations below the detection regions. In addition, most forests and prairies have not con-
limit (0.01 mg/L). tinuously existed since presettlement times but represent
new growth.
Figure 7. Nitrogen sources for the different sample populations. Sources for a particular population include those for
populations with lower NO3-N concentrations. NO3-N concentrations determined for spring and well water samples from
Figures 2 and 3.