Journal of Environmental Management 104 (2012) 152e157

Contents lists available at SciVerse ScienceDirect

Journal of Environmental Management

journal homepage: www.elsevier.com/locate/jenvman

Using electronic conductivity and hardness data for rapid assessment of stream
water quality
Michael Y. Thompson, David Brandes*, Arthur D. Kney
Department of Civil & Environmental Engineering, Acopian Engineering Center, Lafayette College, Easton, PA 18042, USA

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 22 December 2010
Received in revised form
20 February 2012
Accepted 10 March 2012
Available online 9 April 2012

A graphical screening method was previously developed by Kney and Brandes (2007) for assessing
stream water quality data using electronic conductivity (EC) and alkalinity data. The method was aimed
at providing citizen scientists involved in stream monitoring programs with a relatively simple way to
interpret EC data. The method utilizes a plot of EC against concurrent alkalinity data, and is used to
distinguish EC values for impacted or degraded streams from those that can be considered background
values in a particular geologic setting. The method performs well in areas underlain by carbonate
bedrock, as streams in those areas characteristically have EC values that are strongly correlated with
alkalinity. However, in areas of low stream alkalinity (less than approximately 50 mg/L as CaCO3), the
Kney and Brandes (2007) method was found to be much less effective in identifying impacted streams.
This paper extends the graphical screening approach to streams with low alkalinity, specically regions
underlain by clastic sedimentary or crystalline bedrock, by using the strong correlation between EC and
total hardness (TH). A baseline relationship of EC vs. TH is developed using surface water chemistry data
from Hydrologic Benchmark Network streams (deemed as having minimal anthropogenic impacts) and
regional groundwater quality data. The usefulness of the method is demonstrated by application to
publicly available stream chemistry data and to eld data collected from streams of eastern Pennsylvania
under baseow conditions. Results demonstrate that for streams with alkalinity <75 mg/L as CaCO3, the
TH-based graphical screening method should be used rather than the alkalinity-based method of Kney
and Brandes (2007).
2012 Elsevier Ltd. All rights reserved.

Keywords:
Specic conductance
Electronic conductivity
Stream monitoring
Hardness
Alkalinity
Watershed monitoring

1. Introduction
Volunteers play an integral role in monitoring water quality in
thousands of watersheds throughout the United States, and such
programs are among the best-known examples of citizen science.
State and federal regulatory agencies provide manuals, videos, and
short courses to educate watershed volunteers regarding eld
procedures and methods of analysis (USGS, 2010a). Although
a variety of biological, chemical, hydrological, and visual assessment methods are available for measuring human impacts to
streams (Karr, 1991; NRCS, 1998; Barbour et al., 1999; Gergel et al.,
2002), due to limited funding volunteer stream monitors often use
low-cost water quality test kits requiring minimal training (HACH,
2010). Typical parameters for such kits include pH, EC, turbidity,
alkalinity, hardness, dissolved oxygen, and nutrients. Interpretation
of these data is in many cases straightforward; however, EC data in
particular can be difcult to relate directly to stream impairment
because many sources of EC are naturally occurring.
* Corresponding author. Tel.: 1 610 330 5441; fax: 1 601 330 5059.
E-mail address: brandesd@lafayette.edu (D. Brandes).
0301-4797/$ e see front matter 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jenvman.2012.03.025

Several studies have demonstrated inverse relationships

between conductivity and aquatic life, starting with the classic
work of Wilhm and Dorris (1968). More recently, Pond and
McMurray (2002) found that a macroinvertebrate assessment
index was well correlated with EC in streams of eastern Kentucky,
and Potapova and Charles (2003) showed that EC and major ion
concentrations explained much of the variation in assemblage
composition of benthic diatoms for a large sample of rivers
throughout the U.S. In a study using multiple indicators within
a coastal plain watershed, Zampella et al. (2006) found high
ecological integrity scores generally represented sampling sites
with low pH and EC values. These studies suggest that EC can be
a useful indicator of stream health, provided that EC resulting from
anthropogenic sources can be distinguished from background
levels of EC.
2. Background
Stewart (2001) and Kney and Brandes (2007) have both suggested graphical methods to improve interpretation of EC data
within a citizen-based stream monitoring context. Stewart (2001)

M.Y. Thompson et al. / Journal of Environmental Management 104 (2012) 152e157

developed the concept of a chemical perturbation index (CPI) using

alkalinity, hardness and EC data. The CPI subtracts the pairwise
Spearman (rank) correlation coefcients between the three
parameters from the maximum value of 3.0; impacted streams tend
to have a high CPI value while unimpacted have a low CPI value. The
method was applied to several streams in eastern Tennessee, and
the author indicates that stream-specic chemistry data are needed
for the CPI method to be used effectively. Stewart (2001) suggested
that a graphical/visual screening method has the advantage of
being easily understood by citizen scientists, and that reliance on
more advanced statistical methods is likely to discourage use in
stream monitoring programs.
Kney and Brandes (2007) (henceforth designated KB07) used
a conceptually similar framework to Stewart (2001), based on the
idea that water quality data from watersheds with carbonate
bedrock will typically show a strong correlation between EC and
alkalinity, and additional non-carbonate ions will tend to produce
anomalously high EC/alkalinity ratios. The baseline relationship
between EC and alkalinity is developed using regional groundwater
or stream chemistry data rather than stream-specic data. Unlike
the Stewart (2001) CPI method, the KB07 graphical screening
method involves plotting sample chemistry data directly against
the baseline relationship. If the data consistently plot above the
upper bound estimate of the baseline EC/alkalinity relationship, the
stream is deemed to be anthropogenically impacted. KB07
demonstrated the method with data from various streams of the
Delaware River basin. However, in areas of low EC and alkalinity, EC
is not well correlated with alkalinity, and this screening method
was found to be ineffective. An alternative method using total
hardness (TH) as a normalizing parameter rather than alkalinity
was suggested for such cases (KB07).
To investigate the hypothesis that EC would be more strongly
correlated with TH than with alkalinity in areas of low alkalinity, we
rst collected data from four streams in the lower Delaware River
and Lehigh River basins in eastern Pennsylvania underlain by shale
bedrock and with minimal anthropogenic impact. Baseow
conditions were required for samples to be included, so that the
relationships between EC, alkalinity, and TH are not affected by
mixing and dilution with rainfall and surface runoff (KB07). Fig. 1a
shows the weak correlation between EC and alkalinity for these

data, plotted with the baseline EC/alkalinity relationship as established by KB07. However, Fig. 1b shows a strong linear relationship
between EC and TH is shown for the same samples. This provides
initial support to the hypothesis that for low alkalinity streams, an
improved screening method can be developed using TH rather than
alkalinity.
In this paper we use TH as the EC-normalizing parameter to
extend the graphical screening approach to low alkalinity (<75 mg/
L as CaCO3) streams, which are typical of areas underlain with
clastic sedimentary (sandstone, mudstone, shale) and crystalline
bedrock. The method is tested using publicly available data (USEPA
Storage and Retrieval (STORET), the Delaware River Basin
Commission (DRBC), and the United States Geological Survey
(USGS)), as well as using eld samples collected from streams in the
lower Delaware River and Lehigh River basins in eastern
Pennsylvania.
3. Materials and methods
3.1. Development of the graphical screening method
The extended graphical screening method uses the same
general approach as KB07; however, in addition to groundwater
quality data we also use stream water quality data from streams
within the Hydrologic Benchmark Network (HBN) to establish the
baseline EC/TH relationship for unimpacted conditions. The HBN
includes only streams of minimal anthropogenic inuence and high
water quality (Mast and Turk, 1999). Table S-1 [supplementary
material] lists the streams used to develop the baseline EC/TH
relationship; the high percentage of the watershed area covered by
forest is an indicator of the high quality of these streams. To avoid
using stream chemistry data from high ow periods when dilution
reduces stream concentrations, only data from baseow periods
was used. This was dened operationally by only using data from
sampling dates when the stream was at a ow rate less than the
25th percentile ow based on the historical streamow data for
that particular date. Two sources of groundwater data were
combined with the stream data: Nystrom (2007) data from monitoring wells in the Catskill State Park, and Reese and Lee (1998)
regional data from clastic sedimentary formations in Pennsylvania.

350

153

350

300

300

250

250

Electrical Conductivity (S/cm)

Electrical Conductivity (S/cm)

EC = 2.579 * TH + 9.2689
R = 0.9933

200

150

100

50

200

150

100

50

Eastern Pennsylvania Samples

Kney and Brandes (2007)
Upper Bound
Lower Bound

0
0

15

30

45

Alkalinity (mg/L as CaCO3)

60

75

20

40

60

80

100

120

Hardness (mg/L as CaCO3)

Fig. 1. (a) Relationship between EC and alkalinity for samples from unimpacted shale regions of the lower Delaware River and Lehigh River basins in eastern Pennsylvania. The
dashed line represents the relationship between conductivity and alkalinity as established by Kney and Brandes (2007). (b) Relationship between EC and TH for the same data set.

154

M.Y. Thompson et al. / Journal of Environmental Management 104 (2012) 152e157

Fig. 2 shows the EC vs. TH data for all of the streams in Table S-1,
as well as the groundwater data, separated by data source. Because
of the high quality of the streams, the majority of the stream data
have very low EC; however, the groundwater chemistry data
extend to >200 mS/cm. It is apparent from Fig. 2 that a single linear
EC vs. TH relationship describes both stream and groundwater
chemistry over a large geographic region. Upper and lower bounds
on Fig. 2 were dened by lines offset from the baseline relationship
by a value equal to twice the standard error of the estimate. These
lines represent conservative upper and lower bounds for unimpacted streams during baseow conditions.
Using this plot, it is now possible to plot a particular set of
stream data (EC vs TH) and determine whether or not it is likely to
be impacted by anthropogenic sources. However, for the EC/TH
plot, unlike the KB07 EC/alkalinity plot where impacted stream
data would consistently plot above the baseline, data from
impacted streams may fall on both sides of the EC/TH baseline
relationship. This is because anthropogenic discharges often
contain ions contributing to both EC and TH. Data that plot above
the upper bound are most likely due to discharges whose component ions contribute minimally to TH, whereas data points that fall
below the lower bound may be the result of discharges that contain
TH-contributing ions. This is illustrated with a simple mixing model
in the following section.
3.2. Mixing model example
We now demonstrate with a simple mixing model how stream
data may plot under two different discharge scenarios: a road deicing
salt (CaCl2) with hardness ions, and a fertilizer source (KNO3) without
hardness ions. Upstream of the discharge, a theoretical stream has
the chemistry shown in Table S-2 [supplementary material]. The
steady-state mixing model is given by CMIx (QSTRCSTR QDCD)/QMIx,
where Q and C represent ow rate and concentration, respectively.
Subscripts STR, D and MIX correspond to stream upstream of
discharge, the discharge, and the stream below the discharge,

200

In this section, the graphical screening method is tested with

two demonstrations. The rst uses publically available water
quality data from watersheds in New England, the mid-Atlantic
region, and the Appalachian Mountains. The second utilizes eld
data collected during the summers of 2009 and 2010 from
numerous sampling points across multiple watersheds in eastern
Pennsylvania, all underlain by clastic sedimentary and/or crystalline bedrock.
4.1. Application using existing water quality data
Data from the USEPA STORET (USEPA, 2010), USGS (USGS,
2010b) and the DRBC (DRBC, 2004) databases were compiled for

200
Electrical Conductivity (S/cm)

Electrical Conductivity (S/cm)

250

4. Results and discussion

250

Esopus Creek, NY
Cataloochee Creek, NC
Holiday Creek, VA
Young Woman's Creek, PA
Reese and Lee (1998)
Nystrom, E.A. (2007)
Trend-Line
Upper Bound
Lower Bound

300

respectively. The electrical conductivity of each mixture was calculated using a semi-empirical procedure for naturally occurring
waters, utilizing the specic conductances, absolute value of the
charge, ionic strengths and millimolar concentrations of each ion
(Clesceri et al., 2005).
In Fig. 3, each of the four scenarios starts with the point labeled
upstream of discharge and after mixing, ends with the point
labeled for each specic case. As shown in Fig. 3, the discharges
containing hardness-contributing ions tend to result in data points
below the lower bound. Discharges with non-hardness ions simply
raise the EC value of the stream sample until it falls above the upper
bound. Note that it is possible that in certain cases where the ratio
of increased EC to TH is approximately the same as the slope of the
baseline EC/TH relationship, the method could fail to identify
a particular sample as impacted; however, in such cases it is likely
that repeated sampling events would reveal points lying outside
the upper and lower bounds. In summary, in applying the EC/TH
graphical screening method for low alkalinity streams, anthropogenic inuences can be indicated by stream water chemistry data
plotting either above the upper bound or below the lower bound in
Fig. 2.

150

100

150

100
Upstream of discharge
Case 1
50

Case 2

50

Case 3

EC = 2.7185 * TH + 4.8467
R2 = 0.9564

Case 4
0

0
0

20

40
60
80
Hardness (mg/L as CaCO3)

100

120

Fig. 2. Relationship between EC and hardness for the streams listed in Table S-1 (open
symbols) and groundwater data (lled symbols). The bounds are offset by twice the
standard error of the mean, and enclose 96% of all points.

20

40
60
Hardness (mg/L as CaCO3)

80

100

Fig. 3. EC vs. TH plot of a hypothetical stream for four different discharge scenarios.
The discharge constituents, concentrations, and ow rates for each case are shown in
Table S-2.

M.Y. Thompson et al. / Journal of Environmental Management 104 (2012) 152e157

400

Ammonoosuc
French
West Stony
Brodhead
Martins
Tohickon

350

Dyberry
Hardwood
W.B. Ausable
Catawba
Tobyhanna

400

350

300

300

Electrical Conducivity (S/cm)

Electrical Conductivity (S/cm)

155

250

200

150

250

200

150

100

100

50

50

0
0

15

30

45

60

75

20

Alkalinity (mg/L as CaCO3)

40

60

80

100

120

Hardness (mg/L as CaCO3)

Fig. 4. (a) Stream EC and alkalinity data plotted with the Kney and Brandes (2007) graphical screening method. (b) The same datapoints analyzed using the TH-based method. The
bounds are the same as in Fig. 2 (note the difference in y-axis scales between plots). For both (a) and (b), lled symbols represent impacted streams, open symbols indicate
unimpacted streams. Data from the USEPA STORET, USGS, and DRBC.

11 streams as summarized in Table S-3 [supplementary material].

Three of these streams were analyzed previously by KB07 and all of
them are underlain by non-carbonate bedrock. Both methods are
used to screen the same dataset to illustrate that the TH-based
method may reveal that streams are in fact impacted, when the
alkalinity-based KB07 method does not. We use stream data with
alkalinity up to 75 mg/L as CaCO3; this allows us to compare
performance of the two methods near the lower limit alkalinity of

b 400

400
Jeans Run
Mud Run

350

350

Sober's Run
W.B. Bushkill
Pine Run

300

Electrical Conductivity (S/cm)

300

Electrical Conductivity (S/cm)

50 mg/L as CaCO3 suggested by KB07 for the alkalinity-based

screening method.
In Fig. 4a, the stream data is plotted using the KB07 screening
method. Because only four points plots above the upper
bound, one might conclude that all 11 streams are unimpacted.
However, the Pennsylvania Department of Environmental
Protections (PADEP) 303(d) water quality report lists ve of the
streams as being impaired (see Table S-3). This provides evidence

250

200

150

250

200

150

100

100

50

50

0
0

15

30

45

Alkalinity (mg/L as CaCO3)

60

75

20

40

60

80

100

120

Hardness (mg/L as CaCO3)

Fig. 5. (a) Eastern Pennsylvania stream EC and alkalinity data plotted with the Kney and Brandes (2007) graphical screening method. (b) The same datapoints analyzed using the
TH-based method. The bounds are the same as in Fig. 2. For both (a) and (b), lled symbols represent impacted streams, open symbols indicate unimpacted streams.

156

M.Y. Thompson et al. / Journal of Environmental Management 104 (2012) 152e157

that the KB07 method should not be used for low alkalinity
streams.
In Fig. 4b, the same data points are plotted using the TH-based
screening method. Tobyhanna, Brodhead, Tohickon and Catawba
now plot with <20% of their points within the upper and lower
bounds, indicating that there is anthropogenic inuence. This is
consistent with the streams being listed on the 303(d) list. Martins
Creek is interesting in that it shows rather low alkalinity but high
hardness values. Using the KB07 alkalinity-based method, the
majority of the Martins Creek data points fall within the upper and
lower bounds; however when using the TH-based method, most of
the points are outside (below) the bounds, which is consistent with
its 303(d) listing. For the example shown in Fig. 4a and b, the data
were not ltered for baseow conditions, so some of the low outlier
data points may be due to dilution that occurs at high ow conditions. A few high outliers may indicate errors in measurement, or
possibly episodic point source contamination.

publically available data shown in Fig. 4. Similarly to the KB07, the

method itself is transferable but the baseline relationship used here
(Fig. 2) is empirical and users need to be cautious about applying it
too far aeld. It cannot for example be applied to arid or semi-arid
regions where high evapotranspiration rates yield high dissolved
ion concentrations.
The graphical screening method is intended primarily as a tool
for citizen scientists to assist in interpreting their stream chemistry
data. Its advantages are its simplicity and rapidity, and its potential
for identifying episodic discharges. Ideally it would be used in
combination with other water quality metrics, such as macroinvertebrate testing, as part of a comprehensive stream monitoring
program. As in the KB07 method, the screening method does not
identify the specic contaminant(s) causing degradation, and it
cannot identify nonionic constituents that may be degrading
stream quality. The method is intended only to provide a rapid
screening evaluation of whether a stream is impacted, which could
then be followed up with more specic analysis as necessary.

4.2. Application using eld data from lower Delaware River basin
To further evaluate the TH-based screening method, eld
sampling was conducted in the Bushkill Creek, Jeans Run, Sobers
Run, Mud Run and Pine Run watersheds (see Table S-4
[supplementary material]) in eastern Pennsylvania during baseow conditions in summer 2009 and 2010. Electrical conductance
was measured using a WTW Multi 3500i meter with a TetraCon
325 EC probe. Alkalinity was measured by a titration with .005 N
sulfuric acid, and was expressed in mg/L as CaCO3. Hardness was
measured by titration with .0025 M Etheylenediaminetetraacetic
Acid (EDTA), using Eriochrome Black T indicator, and was expressed
in mg/L as CaCO3 (Clesceri et al., 2005).
As in the previous section, the eld data were analyzed using
both the KB07 alkalinity-based screening method and the TH-based
method (see Fig. 5a and b). The two methods are consistent in
showing that the four streams not listed on the PADEP 303(d) list
are unimpacted, with all points except one (a Mud Run datapoint
near the lower bound) plotting within the bounds. However, the
Pine Run results again demonstrate a case where only the TH-based
method gives a result consistent with the 303(d) listing. In this case,
all points when plotted with the KB07 alkalinity-based method are
within the upper and lower bounds, suggesting that the stream is
unimpacted; however, when plotted using the TH-based method,
all points plot outside the bounds, which is supported by its 303(d)
listing. This particular case shows that streams with alkalinity up to
75 mg/L should be analyzed with the TH-based method rather than
the method of KB07.
5. Conclusions and Limitations
This paper extends the previous graphical assessment method
(KB07) to streams with low alkalinity using the observed strong
correlation between EC and TH. A baseline relationship of EC vs. TH
was created using stream chemistry and groundwater data from
relatively undisturbed watersheds in New England, the MidAtlantic States and the Appalachian Highlands, which are dominated by clastic sedimentary and crystalline bedrock. Streams with
EC and TH data that plot outside these bounds are deemed to be
impacted by anthropogenic sources. We recommend that for
streams with alkalinity <75 mg/L as CaCO3 and TH <120 mg/L as
CaCO3, the TH-based graphical screening method be used rather
than the alkalinity-based method of KB07.
Although the eld study was applied locally (Delaware River
basin in eastern Pennsylvania) for demonstration, the baseline EC/
TH relationship (Fig. 2) can be applied regionally to streams
underlain by similar types of bedrock, as demonstrated with

Acknowledgements
Author MYT was funded by a grant from the Andrew W. Mellon
Foundation to Lafayette College during summer 2009 and 2010. We
thank an anonymous reviewer for helpful comments on the
manuscript.
Appendix A. Supplementary material
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.jenvman.2012.03.025.
References
Barbour, M.T., Gerritsen, J., Snyder, B.D., Stribling, J.B., 1999. Rapid Bioassessment
Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates and Fish. EPA 841-B-99-002, second ed. U.S. Environmental
Protection Agency; Ofce of Water, Washington, D.C.
Clesceri, Lenore S., et al. (Eds.), 2005. Standard Methods for the Examination of
Water and Wastewater, twenty rst ed. United Book Press, Baltimore, MD.
Delaware River Basin Commission (DRBC), 2004. Lower Delaware monitoring
program: 2000e2003 results and water quality management recommendations. Report and data available at: www.state.nj.us/drbc/LD/index.htm.
Gergel, S.E., Turner, M.G., Miller, J.R., Melack, J.M., Stanley, E.H., 2002. Landscape
indicators of human impacts to riverine systems. Aquatic Sciences 64, 118e128.
Hach Company (HACH), 2010. [Hach2O] Your formula for water analysis. http://
www.hach.com.
Karr, J.R., 1991. Biological integrity: a long neglected aspect of water resource
management. Ecological Applications 1, 66e84.
Kney, A.D., Brandes, D., 2007. A graphical screening method for assessing stream
water quality using specic conductivity and alkalinity data. Journal of Environmental Management 82, 519e528.
Mast, M.A., Turk, J.T., 1999. Environmental Characteristics and Water Quality of
Hydrologic Benchmark Network Stations in the Eastern United States, 1963-95:
U.S. Geological Survey Circular 1173-A.
Nystrom, E.A., 2007. Ground-Water Quality in the Delaware River Basin, New York,
2001 and 2005-2006: USGS Open File Report 2007-1098, 37 pp.
Pond, G.J., McMurray, S.E., 2002. A Macroinvertebrate Bioassessment Index for
Headwater Streams in the Eastern Coaleld Region, Kentucky. Kentucky
Department for Environmental Protection, DOW, Water Quality Branch,
Frankfort, KY.
Potapova, M., Charles, D.F., 2003. Distribution of benthic diatoms in U.S. rivers in
relation to conductivity and ionic composition. Freshwater Biology 48,
1311e1328.
Reese, S.O., Lee, J.J., 1998. Summary of Groundwater Quality Monitoring Data
(19851997) from Pennsylvanias Ambient and Fixed Station Network (FSN)
Monitoring Program. http://www.dep.state.pa.us/dep/deputate/watermgt/wc/
subjects/srceprot/ground/sympos/ground_mont_rpt.htm. Retrieved June 17,
2005.
Stewart, A.J., 2001. A simple stream monitoring technique based on measurements
of semi-conservative properties of water. Journal of Environmental Management 27, 37e46.
United States Department of Agriculture Natural Resources Conservation Services
(NRCS), 1998. National Water and Climate Center Technical Note 99e1, Stream
Visual Assessment Protocol.

M.Y. Thompson et al. / Journal of Environmental Management 104 (2012) 152e157

United States Environmental Protection Agency (USEPA), 2010. The storage and
retrieval data warehouse (STORET). Retrieved June 14, 2010. http://www.epa.
gov/storet/.
United States Geological Survey (USGS), 2010a. National Field Manual for the
Collection of Water-quality Data: U.S. Geological Survey Techniques of WaterResources Investigations. book 9, chaps. A1eA9. Retrieved July 12, 2010.
http://water.usgs.gov/owq/FieldManual.

157

United States Geological Survey (USGS), 2010b. USGS water data for USA. Retrieved
Jun 1, 2009eJun 16, 2010. http://waterdata.usgs.gov/nwis.
Wilhm, J.L., Dorris, T.C., 1968. Biological parameters for water quality criteria.
BioScience 18, 477e481.
Zampella, R.A., Bunnell, J.F., Laidig, K.J., Procopio, N.A., 2006. Using multiple indicators to evaluate the ecological integrity of a coastal plain stream system.
Ecological Indicators 6, 644e663.