Science of The Total Environment

Science of the Total Environment 476–477 (2014) 144–157
Contents lists available at ScienceDirect
Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv
Pesticide Toxicity Index—A tool for assessing potential toxicity of

pesticide mixtures to freshwater aquatic organisms
Lisa H. Nowell a,⁎, Julia E. Norman b, Patrick W. Moran c, Jeffrey D. Martin d, Wesley W. Stone d
a
U.S. Geological Survey (USGS), 6000 J Street, Placer Hall, Sacramento, CA 95819, USA
b
USGS, 2130 SW 5th Avenue, Portland, OR 97201, USA
c
USGS, 934 Broadway, Suite 300, Tacoma, WA 98402, USA
d
USGS, 5957 Lakeside Boulevard, Indianapolis, IN 46278, USA
H I G H L I G H T S
• Pesticide Toxicity Index (PTI) is a tool to assess pesticide mixtures.

• PTI is an indicator of potential toxicity of pesticide mixtures to aquatic life.
• The PTI was updated and expanded to include 440 pesticides and 52 degradates.
• A new type of PTI provides a more sensitive indicator of potential toxicity.
• From published field data, daphnid mortality occurred above PTI values of 0.1 to 1.
a r t i c l e i n f o a b s t r a c t
Article history: Pesticide mixtures are common in streams with agricultural or urban influence in the watershed. The Pesticide
Received 3 July 2013 Toxicity Index (PTI) is a screening tool to assess potential aquatic toxicity of complex pesticide mixtures by com-
Received in revised form 28 November 2013 bining measures of pesticide exposure and acute toxicity in an additive toxic-unit model. The PTI is determined
Accepted 18 December 2013
separately for fish, cladocerans, and benthic invertebrates. This study expands the number of pesticides and
Available online xxxx
degradates included in previous editions of the PTI from 124 to 492 pesticides and degradates, and includes
Keywords:
two types of PTI for use in different applications, depending on study objectives. The Median-PTI was calculated
Pesticides from median toxicity values for individual pesticides, so is robust to outliers and is appropriate for comparing rel-
Mixtures ative potential toxicity among samples, sites, or pesticides. The Sensitive-PTI uses the 5th percentile of available
Aquatic toxicity toxicity values, so is a more sensitive screening-level indicator of potential toxicity. PTI predictions of toxicity in
Water quality environmental samples were tested using data aggregated from published field studies that measured pesticide
Indicators concentrations and toxicity to Ceriodaphnia dubia in ambient stream water. C. dubia survival was reduced to
≤ 50% of controls in 44% of samples with Median-PTI values of 0.1–1, and to 0% in 96% of samples with
Median-PTI values N 1. The PTI is a relative, but quantitative, indicator of potential toxicity that can be used to
evaluate relationships between pesticide exposure and biological condition.
Published by Elsevier B.V.
1. Introduction
Abbreviations: CA, concentration addition; CCWQP, Central Coast Water Quality
Preservation; CWA, Clean Water Act; DPR, California Department of Pesticide
Regulation; EC50, 50 percent effective concentration; ECOTOX, ECOTOXicology database;
A major challenge when assessing the water quality of streams is to
HCp, pth percentile hazardous concentration; Kow, octanol/water partition coefficient; link contaminant data to effects on aquatic biota, particularly for pesti-
LC50, 50 percent lethal concentration; maxTU, the maximum TU of any single component cides, which tend to occur in streams as complex mixtures with strong
in a mixture; MOA, mode of action; msPAF, multi-substance Potentially Affected Fraction; seasonal and geographic patterns. In data collected during 1992–2001
MTC, median toxicity concentration; NAWQA, National Water-Quality Assessment
by the U.S. Geological Survey's (USGS) National Water Quality Assess-
Program; NOEC, no-observed-effect concentration; O/E, observed-to-expected ratio;
OPP, Office of Pesticide Programs; PPDB, Pesticide Properties Database; PTI, Pesticide ment (NAWQA) Program, U.S. streams with substantial agricultural
Toxicity Index; RA, response addition; SMAV, species mean acute value; SSD, species sen- and(or) urban land within their watersheds had detections of 2 or
sitivity distribution; STC, sensitive toxicity concentration; TC, toxicity concentration; TU, more pesticides or degradates in water more than 90% of the time, 5
toxic unit; USEPA, U.S. Environmental Protection Agency; USGS, U.S. Geological Survey. or more about 70% of the time, and 10 or more about 20% of the time
⁎ Corresponding author. Tel.: +1 916 278 3096; fax: +1 916 278 3070.
E-mail addresses: lhnowell@usgs.gov (L.H. Nowell), jnorman@usgs.gov (J.E. Norman),
(Gilliom et al., 2006). The most common way of directly assessing po-
pwmoran@usgs.gov (P.W. Moran), jdmartin@usgs.gov (J.D. Martin), wwstone@usgs.gov tential effects of pesticides is through the use of standardized laboratory
(W.W. Stone). toxicity tests that expose a single species to a single pesticide over a
0048-9697/$ – see front matter. Published by Elsevier B.V.

http://dx.doi.org/10.1016/j.scitotenv.2013.12.088
L.H. Nowell et al. / Science of the Total Environment 476–477 (2014) 144–157 145
range of concentrations for a specified period of time. Comparison of criteria for protection of aquatic organisms (Stephan et al., 1985).
such toxicity test results among pesticides (or for a pesticide among The median and geometric mean values are more robust to outliers,
test organisms) indicates the relative toxicity of these pesticides (or which is important because toxicity values for a given compound
the relative toxicity of that pesticide to these test organisms) under may vary over several orders of magnitude, reflecting variability
standardized test conditions, but doesn't address many factors that are both among species and among tests for the same species.
important in extrapolating to field conditions. Such factors include the • For pesticides with little or no experimental toxicity test data avail-
dose–response relationship; mode(s) of action; endpoints selected; able, previous studies either omitted such pesticides from the index
environmental factors (such as organic carbon, pH, and temperature); (Anderson, 2008; Munn et al., 2006) or estimated their toxicity values
and potential for additive or interactive effects of contaminant mixtures. from structurally similar compounds (Battaglin and Fairchild, 2002;
This paper addresses one of these factors—exposure to pesticide mix- Schäfer et al., 2011b), compounds with the same MOA (Sala et al., 2012),
tures—and presents the Pesticide Toxicity Index (PTI) as a robust and or the octanol/water partition coefficient (Kow) using quantitative-
readily applicable screening tool for interpreting the biological signifi- structure–activity relationships (Schäfer et al., 2011a, 2011b).
cance of concentration data for pesticide mixtures in hydrologic
systems. For the present study, as will be described, we selected the median
Procedures developed in previous studies to predict or evaluate the and 5th percentile of all applicable acute toxicity values (LC50/EC50s)
effects of pesticide mixtures using standardized single-species toxicity for a given compound and taxonomic group to compute TUs, in order
test data commonly share two steps: (1) environmental concentrations to maximize the number of pesticides included in the index, while
of individual pesticides are divided by compound-specific measures of using a robust and readily transparent approach based on experimental
toxicity or effect, and then (2) a model is applied to combine the contri- toxicity data.
butions of individual pesticides in the mixture. In the second step, after concentrations have been scaled to toxicity,
First, individual chemical concentrations have long been scaled to a model is applied to assess the contribution of individual components
some measure of toxicity or effect (Bliss, 1939) as part of a toxic unit of the mixture to overall toxicity. A number of approaches have been
(TU) approach (Sprague and Ramsay, 1965). Studies have differed in used in previous studies:
key details, however, such as type(s) of toxicity measures used, taxo- • The most common model for combining TUs of pesticide components
nomic groups included, procedures for selecting one toxicity value to of a mixture is concentration addition (CA), which entails summing
represent a compound when multiple values were available, and proce- TUs for individual compounds in a mixture (Bliss, 1939; Sprague and
dures for handling pesticides with no toxicity data available (examples Ramsay, 1965). CA assumes that components of the mixture behave
cited below). as concentrations or dilutions of one another, differing only in their
potencies, and do not interact. In theory, CA applies to compounds
• TUs for pesticides have been calculated using acute 50-percent lethal with a similar mode of action (MOA) and dose–response curves.
or effective concentrations (LC50 or EC50, respectively) (Battaglin and • The response addition (RA) model (also called independent ac-
Fairchild, 2002; Guy et al., 2011; Munn and Gilliom, 2001; Munn et al., tion) assumes that compounds in the mixture act independently,
2006; Schäfer et al., 2011a, 2011b; Stenström, 2013), chronic no- so the combined effect is calculated according to the probability
observed-effect concentrations (NOECs) or LC50/EC50s multiplied of non-excluding processes (Bliss, 1939; de Zwart and Posthuma,
by a safety factor to represent chronic effects (Anderson, 2008), 2005).
water-quality standards (Stenström, 2013), and hazardous concentra- • In the maximum TU (maxTU) approach, only the most toxic com-
tions (HC) derived from species sensitivity distributions (SSD) (Guy ponent of a mixture is considered, and potential contributions
et al., 2011; Maltby et al., 2005, 2009; Sala et al., 2012; Schäfer et al., from other compounds in the mixture are assumed to be negligible
2013; Whiteside et al., 2008). A SSD entails fitting a statistical distribu- (Liess and von der Ohe, 2005; Schäfer et al., 2011a, 2011b, 2012).
tion to toxicity data for certain broad taxonomic groups, and SSDs can • For complex pesticide mixtures, de Zwart and Posthuma (2005)
be constructed from acute LC50/EC50 values (Schäfer et al., 2013; combined CA (for compounds with the same MOA) and RA (for
Whiteside et al., 2008), chronic NOEC values, or any other selected compounds with different MOAs) in a two-stage model called the
toxicity criterion (Posthuma and de Zwart, 2006). From the concen- multi-substance Potentially Affected Fraction (msPAF) model.
tration–effect relationship, a specific HC (HCp) can be derived that The msPAF model has been applied to assess assemblage metrics
corresponds to effects on p% of species. for fish and invertebrates (Posthuma and de Zwart, 2006, 2012;
• TUs have been calculated for a single taxon such as Daphnia magna Schäfer et al., 2013).
(Guy et al., 2011; Liess and von der Ohe, 2005; Schäfer et al., 2011a,
2011b; Stenström, 2013), the most sensitive taxon (i.e., the taxon For the present study, as will be described, we used the CA model.
with the lowest LC50) (Schäfer et al., 2013), or taxonomic groups Studies of mixtures have shown that for compounds with a similar
such as cladocerans (Munn et al., 2006). Similarly, SSDs may represent MOA, the CA model generally provides accurate predictions and the
a fairly narrow taxonomic group such as arthropod invertebrates RA model may underestimate effects; but for compounds with different
(Maltby et al., 2005, 2009; Sala et al., 2012), crustaceans (Whiteside MOAs, the RA model provides more accurate predictions and the CA
et al., 2008), or broader groups such as freshwater invertebrates model may overestimate toxicity (Backhaus and Faust, 2012; Belden
(Schäfer et al., 2013), aquatic animals (Etterson, 2011), or a combina- et al., 2007; Faust et al., 2000, 2003; Junghans et al., 2006). However,
tion of aquatic animals, aquatic macrophytes and algae (Maltby et al., several studies of pesticide mixtures have shown that toxicity predicted
2009). using the CA model was within a factor of 2 or 3 of the observed toxicity,
• When multiple toxicity values are available for a given pesticide, dif- regardless of the MOA of the components (Belden et al., 2007; Faust
ferent studies selected the minimum value (Anderson, 2008), the meet al., 2003; Warne, 2003). Moreover, application of the RA model, or a
dian (Munn et al., 2006), the geometric mean (MacDonald et al., 2000; two-stage combination of CA and RA models, requires additional data
Sala et al., 2012), or the mean value (Battaglin and Fairchild, 2002). on the MOA for each compound (which may be unknown or may vary
The U.S. Environmental Protection Agency's (USEPA) Office of Pesti- for organisms in different trophic levels) and on dose–response curves
cide Programs (OPP) uses minimum values for ecological risk assess- for compounds with dissimilar MOAs (which are usually required to as-
ment and for calculating OPP aquatic-life benchmarks (http://www. sess the effect that each compound would have if it were applied singly
epa.gov/oppefed1/ecorisk_ders/aquatic_life_benchmark.htm). The at the concentration at which it is present in the mixture) (Backhaus
geometric mean is used by USEPA's Office of Water for computing and Faust, 2012). For field samples, which may contain a large number
Species Mean Acute Values (SMAV) in deriving USEPA water-quality of pesticides with similar, dissimilar, and unknown MOAs, CA appears to
146 L.H. Nowell et al. / Science of the Total Environment 476–477 (2014) 144–157
be a slightly conservative, but broadly applicable model with a relatively predicting whether pesticide mixtures in water samples are likely to
small likelihood of underestimating effects (Backhaus and Faust, 2012; be toxic to aquatic organisms.
Belden et al., 2007). The present paper updates and modifies the PTI, which initially was
A limitation of the CA and RA models is that they do not account for developed for 75 pesticides (Munn and Gilliom, 2001) and later ex-
potential interactive effects among pesticides in a mixture. In general, panded to 124 pesticides analyzed by the NAWQA program (Munn
interactive effects have been reported to occur in 10–30% of studies et al., 2006). The objectives of this paper are to (1) expand the target
(Deneer, 2000; Faust et al., 1994; Warne and Hawker, 1995), were fairly list of pesticides to all pesticides used in agriculture in the U.S. from
evenly divided between synergism and antagonism, and typically did 1992 to 2011; (2) update and expand experimental toxicity data using
not deviate much from additivity (Warne and Hawker, 1995). Most additional data sources and acceptable endpoints; (3) investigate the
studies of pesticide interaction have evaluated only limited endpoints, possibility of basing the PTI on more sensitive toxicity values; and (4)
such as acute mortality or immobilization for animals and simple apply the PTI method to past field studies to test how well the PTI pre-
short-term growth measures for plants (Belden et al., 2007). The dicted toxicity in these studies and to illustrate its application to inter-
incidence of interactive effects may depend on the endpoint measured pret water-quality data.
and test duration (Cedergreen and Streibig, 2005), on the test chemicals
(Deneer, 2000; Lydy et al., 2004), and possibly on test species 2. Methods
(Cedergreen et al., 2008).
Schäfer et al. (2011a) preferred the maxTU over the summed TUs The PTI methodology used in the present study follows that of
(CA model) because the sum of TUs exhibited a stronger dependency previous editions (Munn and Gilliom, 2001; Munn et al., 2006),
on the number of compounds measured and could overestimate the except for some important modifications. The target analyte list
toxicity of compounds with a dissimilar mode of action. Because the was expanded to include nearly 500 pesticides from an agricultural
maxTU approach considers only the component with the highest toxic use database for the U.S. (described below). Toxicity data sources
unit value, it does not require additional data on MOA or dose–response. and endpoints were expanded to increase the number of pesticides
The observation that RA models may underestimate toxicity for with experimental toxicity data, including the use of the LC50 as
mixtures of similarly acting compounds (Backhaus and Faust, 2012) well as the EC50 (immobilization) endpoints for cladocerans; use
suggests that the same may be the case for the maxTU approach, of non-standard data for those pesticides that have no standard
which includes only the most toxic component of a mixture. However, test data available; and addition of toxicity data from OPP risk assess-
the sum of TUs and the maxTU were highly correlated with one another, ment documents and the Pesticide Properties Database (PPDB). Fi-
and both had a similar relationship to macroinvertebrate abundance in nally, the PTI from previous editions (here called “Median-PTI”)
streams in Australia and Europe (Schäfer et al., 2011a). was supplemented with a second type of PTI (“Sensitive-PTI”) that
The PTI approach presented in this paper is a variation of a risk- was calculated using more sensitive toxicity concentrations. These
based scoring system described by Kimerle et al. (1997). The PTI was changes are described further in Sections 2.1–2.3.
first developed by Munn and Gilliom (2001) for use with data on pesti-
cides in stream water collected as part of the NAWQA Program, and was 2.1. Toxicity data sources
updated in 2006 (Munn et al., 2006). The PTI is a TU procedure that fol-
lows the CA model. This approach was selected because CA is a slightly A master list of 484 pesticides was compiled from agricultural
conservative (protective), broadly applicable model that requires only pesticide use lists for 1992 to 2011 (GfK Kynetec, 2009; personal
standard toxicity data for individual pesticides, and is suitable as a communication, Melissa Sims, Senior Business Development Manag-
robust screening tool for application to complex environmental mix- er Americas, GfK Kynetec, June 1, 2012), and was composed of 197
tures of pesticides with similar, dissimilar, or unknown MOAs. Thus, herbicides; 135 insecticides, acaricides and nematicides; 105 fungi-
the PTI approach allows us to maximize the number of pesticides in- cides; 9 fumigants; 4 adjuvants and synergists; and 34 other miscel-
cluded in the mixtures index, using only experimental data. As original- laneous pesticides. In addition, a master list of 104 pesticide
ly developed, PTI values are derived separately for fish, cladocerans, and degradates included high priority (Tier 1) degradates identified dur-
benthic invertebrates by using the median of acute LC50s and EC50s for ing a pesticide prioritization effort by the U.S. Geological Survey
appropriate test species in standardized tests. Because only experimen- (Norman et al., 2012).
tal data are used, any pesticides without standardized test data available For these 484 pesticides and 104 degradates, the following sources
for organisms in a given taxonomic group are omitted from the PTI were searched for experimental toxicity test data: USEPA's ECOTOXicol-
calculations for that taxonomic group. The PTI may under-represent ogy (ECOTOX) database (U.S. Environmental Protection Agency,
the relative acute toxicity of pesticides in a water sample because 2012a); USEPA registration and risk assessment documents cited in
PTI calculations are limited to pesticides that (i) are measured and support of the OPP aquatic-life benchmarks (U.S. Environmental
(ii) have toxicity data available. Because it is based on median toxic- Protection Agency, 2012b); and the PPDB (University of Hertfordshire,
ity values, the original PTI is relatively robust to outliers, but it would 2013). The ECOTOX database was queried for toxicity data for pesticides
not necessarily be indicative of possible effects on all species in all on July 5, 2012 and degradates on February 20, 2013. Toxicity data from
life stages that may be exposed in hydrologic systems. OPP documents and the PPDB were compiled during September–Octo-
The PTI can be applied to pesticide mixtures detected in ambient ber 2012 for pesticides and May 2013 for degradates. Details about each
water samples from hydrologic systems and used to interpret data in of these data sources are provided in Appendix A of the online supple-
a number of ways. PTI values for individual samples can be used to mentary information that accompanies this paper. Toxicity test data
rank sites or samples according to their expected relative toxicity due were selected as follows, in order of priority:
to pesticides, to assess changes in potential toxicity over time at a single
site, or to predict which of the pesticides detected may be the greatest • Standardized toxicity test data from the ECOTOX database (U.S.
contributor to potential toxicity for a site or sample. PTI values for sam- Environmental Protection Agency, 2012a). Standardized data are de-
ples, seasons, or sites have been used as explanatory variables in multi- fined as meeting the screening criteria in Table 1;
variate analyses designed to determine which environmental variables • Toxicity test data from core or supplemental studies underlying OPP
best explain spatial patterns in the structure of a biological community aquatic-life benchmarks or summarized in registration documents
(Fuhrer et al., 2004; Waite et al., 2006). As described further in cited by USEPA (2012b);
Section 3.2, the PTI originally was designed to be an indicator of relative • Non-standard toxicity test data from the ECOTOX database—i.e., tests
toxicity, and was not necessarily appropriate as a sensitive tool for that did not meet the screening criteria in Table 1, or that had
Table 1 cladocerans, the inclusion of the LC50 values is new, as only EC50s
Standard criteria for toxicity test duration, endpoint, and measured effect, by taxonomic based on immobilization were used previously (Munn and Gilliom,
group. [EC50, 50 percent effective concentration; LC50, 50 percent lethal concentration].
2001; Munn et al., 2006). Both endpoints are widely used and it is diffi-
Taxonomic group: Fish or benthic invertebrates Cladocerans cult to ascertain organism mortality with certainty without examining
Duration (hours) 96 48 the organism under a dissecting microscope for a heartbeat (Jonczyk
Endpoint LC50 LC50 or EC50 and Gilron, 2005). Standardized methods are available for both immobi-
Effect Mortality Mortality or immobilization lization (Commission of the European Communities, 1992; Organiza-
Value Discrete Discrete
tion for Economic Cooperation and Development, 2004; U.S.
Environmental Protection Agency, 1996) and mortality (U.S.
Environmental Protection Agency, 2002). Inclusion of LC50 values for
unbounded toxicity values (i.e., reported as “greater than” or “less cladocerans may increase the median toxicity value for cladocerans to
than” a specified concentration, such as N100 μg/L); some degree; however, this represents only one of many sources of un-
• Toxicity values compiled from the PPDB (University of Hertfordshire, certainty in experimental toxicity test data (Section 3.4). The added un-
2013). certainty from combining LC50 with EC50 values is countered by the
increase in the number of data records: Of the 2064 toxicity values for
Bioassay data used in this paper are for fish and invertebrates only. cladocerans in the present study, 739 (36%) were LC50 values, allowing
When querying the ECOTOX database, specific species were not queried determination of toxicity concentrations for an additional 24 pesticides
by name; the species list of aquatic animals included crustaceans, fish, for which EC50 values for immobilization were not available. The ratio-
insects/spiders, other invertebrates, and molluscs. Most species in the nale for including LC50 values for cladocerans is discussed further in Ap-
data query output are uniquely found in freshwater environments; pendix A of the supplementary information.
however, a limited number of species spend part of their life cycle in For pesticide compounds that did not have standardized test data
the freshwater environment (e.g., salmon), or are known to inhabit ma- (Table 1) available, non-standard test data were selected using the
rine or estuarine habitats but were tested under freshwater conditions. screening procedure below. Non-standard toxicity values were charac-
Bioassay data records were removed for sponges, microcrustaceans terized by the direction of bias where possible.
such as rotifers and ostracods, worms such as turbellarians and oligo-
chaetes, protozoans, and ciliates, which were not retained with the in- • Concentration: If non-standard tests with discrete values were avail-
vertebrates. In addition, bioassays were limited to laboratory tests able, then only these tests were used, and any tests with results re-
conducted in freshwater, which the ECOTOX database defines as having ported as unbounded or as concentration ranges were omitted. If no
salinity of 4 parts per thousand or below (U.S. Environmental Protection available studies reported discrete values, then toxicity values that
Agency, 2011). Subsequently, ECOTOX test data were subdivided by were unbounded or ranges were retained, subject to further screening
taxonomic group, and classified as standard (Table 1) or non-standard below. A toxicity value expressed as a greater-than concentration
as described in Section 2.2. (e.g., N 100 μg/L) was considered to overestimate toxicity, and a result
For pesticides without standardized test data in ECOTOX, but that expressed as a less-than concentration to underestimate toxicity. For
have acute OPP aquatic-life benchmarks for fish and(or) invertebrates, test results expressed as a concentration range (e.g., N 100 μg/L but
toxicity values associated with those benchmark values (U.S. b1000 μg/L), the unbounded low end of the range was retained
Environmental Protection Agency, 2012b) were compiled from the (this only applied to one test).
cited source document. Toxicity values obtained from OPP documents • Effect: For fish studies, the only acceptable endpoint was mortality.
were the cited LC50 or EC50 values, and were not equivalent to the For cladocerans and benthic invertebrates, if one or more studies
acute OPP benchmarks themselves, which include an additional safety had immobilization or mortality measured, all tests with other ef-
factor of 2. Some OPP benchmark values are unbounded because the fects measured (such as population growth or hatching success)
EC50 or LC50 was greater than the highest concentration tested or less were omitted; if all invertebrate tests had non-standard effects
than the lowest concentration tested. Unbounded LC50 or EC50 values measured, then these tests were retained, subject to screening for
were considered non-standard data (Table 1; Section 2.2). duration. Typically the direction of bias in such cases was rated as
Aquatic toxicity values for pesticide compounds that did not have unknown.
data available from ECOTOX or OPP documents were obtained from • Duration: If all available studies had non-standard exposure pe-
the PPDB (University of Hertfordshire, 2013), when available. Data in- riods, then the tests with the duration closest to the standard dura-
cluded experimental 96-h LC50 values for fish (usually rainbow trout), tion were selected; if test durations were both shorter and longer
48-h EC50 values for aquatic invertebrates (usually D. magna), and 96- (to a comparable degree) than the standard test duration, then
h LC50 values for sediment dwelling organisms (usually the midge, the longer period was selected to be more protective. For example,
Chironomus riparius). In the present study, toxicity values from the if the standard duration is 4 days, and there were data for 1, 2, 5,
PPDB were limited to those with (1) standard test durations and (2) and 21 days, then the 5-day study was selected. In occasional
confidence scores of 5 (verified with regulatory purpose), 4 (verified), cases where the longer exposure periods did not have the lowest
or 3 (unverified but known source). For LC50 values, the effect (mortal- (most sensitive) toxicity values, then all discrete toxicity values
ity) can be inferred, but for cladoceran EC50 values, the effect was not associated with all acute test durations were included. A study
reported in the PPDB but was accepted for the present paper as likely re- with long exposure duration was considered to overestimate tox-
ferring to immobilization, and therefore standard. Toxicity values from icity, and a study with short exposure duration to underestimate
the PPDB were considered standard if the test duration, endpoint, and toxicity.
species were appropriate and the values were discrete; if the reported
toxicity value was unbounded, the endpoint was non-standard, or the 2.3. Development of the Pesticide Toxicity Index
species was not reported, then the toxicity value was considered to be
non-standard. The PTI is determined separately for three taxonomic groups: fish,
cladocerans, and benthic invertebrates. Cladocerans are treated sepa-
2.2. Data screening process rately from benthic invertebrates because of differences in physiology,
habitat and functional group. In addition, their long history of use in tox-
The criteria for standardized test data in Table 1 are identical to those icity testing has resulted in a large cladoceran dataset with a relatively
used in previous editions of the PTI, with one exception—for small number of genera.
The PTIt is the sum of toxicity quotients for each pesticide compound The toxicity value percentiles were determined for each pesticide
(i) measured in ambient water from a hydrologic system, for each taxo- compound within a taxonomic group using the Cunnane (1978) plot-
nomic group (t): ting position of 0.4, as recommended by Helsel and Hirsch (1992), and
linear extrapolation. The quantile.default function in Spotfire S + 8.1
Xn (TIBCO Software Inc., 2008) was used to perform the computations.
PTIt ¼ i¼1
Ei =TCi;t ð1Þ For pesticide compounds with 20 or more toxicity values available,
the 5th percentile was used as the STC. For pesticide compounds with
13 to 19 toxicity values available, the STC was the 5th percentile as de-
where Ei is the concentration of pesticide i, n is the number of detected termined using linear extrapolation, with the two lowest ranked toxic-
pesticides in an environmental sample, and TCit is the toxicity concen- ity values used to extrapolate near or below the minimum value. One
tration for the pesticide i for the taxonomic group t. E and TC are consequence of this extrapolation method is that if the lowest two tox-
expressed in the same units. icity values were a tie, then the 5th percentile value was equal to the
The PTI of Munn and Gilliom (2001) and Munn et al. (2006) was minimum. For pesticide compounds with 12 or fewer toxicity values,
computed from Eq. (1) using the median of toxicity values available the minimum value was used as the STC in lieu of the 5th percentile.
for a given pesticide. The median was selected because, for pesticides, Monte Carlo simulations (detail in Supplemental Appendix A) showed
there is a large difference in sensitivities of different taxonomic groups that, for pesticides with 12 or fewer toxicity values available, the prob-
(Maltby et al., 2005, 2009; Sala et al., 2012; van den Brink et al., 2006). ability was less than 50% that the minimum toxicity value was a statis-
Even under standardized test conditions within a taxonomic group, tical outlier, defined as within the lowest 5th percentile of a
there is considerable variability in toxicity values for many pesticides. hypothetical large population of toxicity values.
In data compiled for the present study, the distribution of standard tox- As described in Supplemental Appendix A, an alternative approach
icity values for a given pesticide within a taxonomic group spanned up was evaluated in which toxicity tests were first aggregated by species
to 3–4 orders of magnitude for pesticides with 20 or more toxicity test and SMAVs were computed as the geometric mean of all tests for a
values available, and extreme values—defined as more than three orders given species and compound; the STC then was determined to be either
of magnitude from the median—occurred for 17 to 41% of these pesti- the lowest SMAV or (for compounds with more than 12 species) the
cides, depending on the taxonomic group. Factors that may contribute 5th percentile of the SMAVs. The resulting species-weighted STCs were
to variability include differences between test organisms, test condi- not only highly correlated with the unweighted STCs, although species
tions, and pesticide test material (see Section 3.4). weighting raised STC values for many pesticides, but also (for those pesti-
On the other hand, the use of median toxicity values in the PTI cides whose lowest SMAV was based on a single test) increased the like-
means that the concentration selected for a given pesticide may lihood of selecting a possible low outlier as the STC (see Supplemental
not be representative of some sensitive species or life stages. For cer- Appendix A.3 for further discussion and examples). For compounds
tain applications of the PTI, a more sensitive indicator of toxicity may with multiple tests available for only a single species, selecting the lowest
be preferable. Therefore, for the present study, the PTI method was SMAV as the STC resulted in a value that was similar to (or even greater
modified to include two types of PTI values for a given taxonomic than) the MTC, rather than a more conservative, sensitive value. In theory,
group, which are computed by using two different types of TC values it would be preferable to weight species equally when determining the
in Eq. (1): STC; however, in practice this was problematic because of the unevenness
and scarcity of the toxicity test dataset (Supplemental Appendix A.3). Al-
• The Median-PTI is calculated from Eq. (1) using the median toxicity
though the unweighted STC is more influenced by results for some spe-
concentration (MTC), which is defined as the median of the toxicity
cies than for others, a species-weighted STC would be more influenced
values available for each compound towards the taxonomic group.
by results of some individual tests than of others.
The Median-PTI is equivalent to the original PTI of Munn et al.
(2006), except for some modifications to the kinds of toxicity data
2.5. Application of the Pesticide Toxicity Index to interpret water-quality
that are acceptable (Sections 2.1 and 2.2).
data
• The Sensitive-PTI is calculated from Eq. (1) using the sensitive toxicity
concentration (STC). The STC is either the 5th percentile or (if fewer
To test the PTI model and to illustrate its application to water-quality
than 13 values were available) the minimum toxicity value for each
data, published studies were compiled that reported concurrent data on
compound towards the taxonomic group (Section 2.4).
pesticide concentrations and aquatic toxicity. These studies reported
MTC values are robust to outliers and provide a more consistent raw data for both (1) concentrations of a substantial number of pesti-
representation of relative toxicity values among pesticides, whereas cides in ambient stream water and (2) organism survival in toxicity
STC values provide a more sensitive indication of potential toxicity to tests conducted in the laboratory with undiluted ambient water. Studies
aquatic organisms. Because extreme values were fairly prevalent in included a series of reports and memoranda on pesticides and toxicity in
the toxicity test dataset, care was taken to minimize the likelihood California streams published by the California Department of Pesticide
of selecting an outlier as the STC (Section 2.4 and Supplemental Regulation (DPR) (Ensminger et al., 2009; Ganapathy, 1999a, b; Gill,
Appendix A). 2002; Jones, 2000; Kim et al., 1999a, b; Nordmark, 1999, 2000;
Walters et al., 2000a, 2000b, 2000c, 2000d, 2000e, 2000f, 2000g,
2.4. Determination of STC values 2000h, 2000i, 2000j, 2000k, 2000l, 2000m, 2001a, 2001b, 2001c,
2001d, 2001e, 2001f, 2001g, 2001h, 2001i, 2001j) and the Central
The 5th percentile of the available toxicity values was selected as the Coast Water Quality Preservation (CCWQP) (2008, 2009). Some inde-
STC because it is a sensitive indicator of relative toxicity, yet it is more pendent studies also provided suitable data (Kuivila and Foe, 1995;
robust to outliers than the minimum value is. Species that have a greater Phillips et al., 2012).
number of toxicity tests available will have more influence on the STC These studies measured acute toxicity to Ceriodaphnia dubia in 4 to
than other species. The most frequently tested species tended to be 8-day laboratory exposures to undiluted ambient stream water, so re-
among the more sensitive species within this dataset (Supplemental sults were compared to the PTI for cladocerans (Section 3.3.1). When
Appendix A); therefore, STCs tended to be biased low (i.e., to overesti- data were aggregated from these studies, the total number of individual
mate toxicity), which was considered acceptable because the objective samples was 479 (Table 3). Samples from CCWQP (2008, 2009) and
of the STC was to establish a sensitive indicator of potential toxicity to Kuivila and Foe (1995) were exposed for 6–8 days (total n = 174 sam-
aquatic life. ples), whereas the remaining samples in the aggregated dataset were
exposed for 4 days (n = 305 samples). Each study analyzed between 8 three taxonomic groups. The number of unique species was 228 for
and 27 parent pesticides, and collected water samples between 1993 fish, 47 for cladocerans, and 284 for benthic invertebrates. Table 2 sum-
and 2009 (Table 3). All aggregated samples were from California and marizes the number of compounds with toxicity values and the number
most studies focused on organophosphate insecticides, which were of underlying bioassays for pesticides and degradates, by taxonomic
responsible for placement of some California streams on the Clean Water group. MTC and STC values for individual pesticides and degradates in
Act (CWA) §303(d) list (Central Coast Water Quality Preservation Inc., each taxonomic group are provided in Supplemental Appendix B for
2008, 2009; Ensminger et al., 2009). use in calculating PTI values from pesticide concentrations measured
The results from one study in Table 3 that analyzed multiple classes in environmental samples. MTC and STC values are listed for each taxo-
of pesticides—a study of dormant spray pesticides by the California DPR nomic group separately—for fish (Table B.1), cladocerans (Table B.2),
(Gill, 2002; Nordmark, 1999, 2000)—are described in more detail to and benthic invertebrates (Table B.3)—including the number of toxicity
illustrate application of the PTI to interpret water-quality data tests, number of species, type and source of toxicity data, and any appli-
(Section 3.3.2). For the purposes of discussion, the term “toxic” is de- cable bias for pesticides and degradates with non-standard toxicity data.
fined in this paper as reduction of C. dubia survival to ≤50% of that in
control samples. Although not statistically based, the 50% value in this
3.1.1. Fish
definition corresponds to the EC50/LC50 values from which PTI values
MTC and STC values for fish were determined for a total of 435 pes-
are calculated. Spearman rank correlations and least-square regressions
ticides and 48 degradates. Many more bioassays were available for pes-
were performed using DataDesk 6.3.1.
ticides (6685) than for degradates (173; Table 2). Of the total bioassays,
Invertebrate community condition in streams across the U.S. was
most (98%; Table 2) met the definition of standard data from Table 1.
assessed as part of the USGS NAWQA Program during 1993–2005
The median number of standardized bioassays per compound was
(Bryant and Carlisle, 2012; Carlisle and Meador, 2007; Carlisle et al.,
four for pesticides and one for degradates; the median number for
2013; Yuan et al., 2009). Invertebrate communities were characterized
non-standard bioassays was one for both pesticides and degradates.
by quantifying the deviation of a site's measured macroinvertebrate
The use of non-standard data for pesticide compounds that did not
composition from the expected composition, expressed as the ratio of
have standard data available increased the total number of compounds
observed/expected (O/E) taxa. The O/E index provides a way to com-
with MTC and STC values by 78 (22% increase) for pesticides and 18
pare across sites because it is standardized by each site's potential for
(60% increase) for pesticide degradates (Tables 2 and B.1). Just over
species composition and richness. Expected invertebrate assemblage
half (53%) of the pesticides and most (83%) of the degradates with
compositions (E) for each site were the sum of predicted probabilities
non-standard data for fish have only a single toxicity value, which
of capturing common species at a given site, derived from predictive
means that the MTC and STC are the same for these pesticides. Overall,
models developed from regional reference sites (Carlisle and Hawkins,
the standard dataset for fish includes toxicity data from 6692 bioassays
2008; Carlisle and Meador, 2007; Yuan et al., 2008).
representing 228 different species, including warm and cold water fish,
Based on data compiled for Carlisle et al. (2013), NAWQA sites were
but about half of the bioassays are for one of three species: rainbow
identified that had both (1) an invertebrate community sample with an
trout (23%), bluegill (18%), and fathead minnow (11%) (Appendix C,
O/E ratio and (2) data for 26 pesticides analyzed in water at all sites at
Table C.1).
least 3 times during the 90 days prior to the invertebrate biological sam-
ple, with at least 2 samples occurring in different calendar months. A total
of 128 stream sites sampled between 1993 and 2004 met these condi- 3.1.2. Cladocerans
tions. The potential toxicity of pesticides at these sites was characterized MTC and STC values for cladocerans were determined for a total of
using the maximum Median-PTI for benthic invertebrates that occurred 423 pesticides and 47 degradates; a total of 2063 bioassays were avail-
in samples within the 90-day time period. The relationship between O/E able for pesticides and 116 for degradates (Tables 2 and B.2). The medi-
ratio and the Median-PTI was evaluated using least-squares regression. an number of bioassays per compound was one, for both standard and
nonstandard data. Addition of the non-standard dataset for cladocerans
3. Results and discussion increased the number of compounds with MTC and STC values by 68
(19%) for pesticides, and 19 (68%) for degradates (Tables 2 and B.1).
3.1. Toxicity concentrations — MTC and STC There are 17 genera and 47 species represented in the standard dataset
(Tables C.1 and C.2); however, most tests used D. magna (63%), C. dubia
For pesticides and degradates combined, MTCs and STCs were based (12%), Daphnia pulex (5%), Daphnia longispina or Simocephalus serrulatus
on toxicity values from a total of 10,837 bioassays for 559 species in (3% each). The non-standard cladoceran data set includes only 5 species,
Table 2
Summary of the number of compounds and bioassays in the standard and non-standard datasets for the three taxonomic groups: fish, cladocerans, and benthic invertebrates. [b, number of
bioassays].
Bioassay data by taxonomic group Pesticides Pesticide degradates
Fish Cladocerans Benthic invertebrates Fish Cladocerans Benthic invertebrates
No. of total compounds 435 423 205 48 47 15

No. of total bioassays (b) 6685 2063 1734 173 116 66
Overall median b per compound 3 1 3 1 1 2
Standard data
No. of compounds 357 355 139 30 28 10
No. of bioassays 6540 1978 1571 152 97 53
Median b per compound 4 1 4 1 1 2
Non-standard data
No. of compounds 78 68 66 18 19 5
No. of bioassays 145 85 162 21 19 13
Median b per compound 1 1 1 1 1 2
Increase in no. of compounds by adding non-standard data 22% 19% 47% 60% 68% 50%
Table 3
Studies included in the aggregated dataset used to plot survival from Ceriodaphnia dubia toxicity tests as a function of Median-PTI and Sensitive-PTI (Fig. 1) and to determine the optimum
Median-PTI and Sensitive-PTI thresholds (Fig. 2). All toxicity tests were conducted with undiluted ambient stream water. [CA DPR, California Department of Pesticide Regulation; Carb,
carbamate; CCWQP, Central Coast Water Quality Preservation; d, days; Herb, herbicide; I, insecticides; No., number; OP, organophosphate; Pyr, pyrethroid; Ref, reference; Tox, toxicity].
Author or Study No. No. Sampling Number of parent pesticides analyzed Tox test General sampling design Ref
agency area sites samples dates duration
OP-I Carb-I Pyr-I Other-I Herb Total
(d)
CA DPR Sacramento River watershed (Wadsworth Canal)

1 42 1998–2000 8 2 – – 9 19 4 Baseline sample in Dec, then sample twice 1
1 24 2000–2001 8 2 – – 9 19 4 weekly through dormant spray season 2
CA DPR San Joaquin River watershed (Orestimba Creek)
1 18 1997–1998 8 2 – – 9 19 4 Baseline sample in Dec, then sample twice 3
1 20 1998–1999 8 2 – – 9 19 4 weekly through dormant spray season 4
1 20 1999–2000 8 2 – – 9 19 4 5
CA DPR Del Puerto and Orestimba Creeks
2 21 2007–2008 10 – 8 – – 18 4 Monthly samples from Dec–June; 3 storm 6
samples in Dec, Jan, Feb
CA DPR Orange County
3 8 1999 8 – – – – 8 4 Rainfall runoff sampling 7
12 90 1999–2000 8 – 1 3 – 12 4 Monthly sampling; Feb 2000 samples 8
coincided with rainfall
8 38 2000 8 – 1 3 – 12 4 Rainfall runoff sampling 9
2 14 2000–2001 5 – 1 3 – 9 4 Monthly sampling; no rainfall runoff 10
CCWQP Lower Salinas and Santa Maria watersheds
18 18 2007 19 – – – – 19 6–8 d Sample during dry season 11
4 6 2008 14 6 – – 7 27 6–8 d Sample during dry season 11
21 21 2008 19 – – – – 19 6–8 d Sample during dry season 11
23 94 2006–2007 19 – – – – 19 6–8 d Sample twice per year in dry 12
season and twice in winter/wet season
Kuivila San Francisco Estuary
a a a
2 35 1993 4 1 – – 3 8 7d Daily samples for 3 weeks; 13
included some rainfall events
Phillips Santa Maria River and Oso Flaco Creek watersheds
10 10 2009 6 – 9 – – 15 4 Bimonthly for 1 year; full pesticide analysis 14
on 1 sample per site
a
19 pesticides analyzed but only those detected were identified
1. Nordmark (1999, 2000) 8. Kim et al. (1999b); Walters et al. (2000c); Walters et al. (2000e, f, g, h, i, j, k, m)
2. Gill (2002) 9. Walters et al. (2000l)
3. Ganapathy (1999a) 10. Walters et al. (2000a, b, d, 2001a, b, c, d, e, f, g, h, i, j)
4. Ganapathy (1999b) 11. Central Coast Water Quality Preservation Inc. (2009)
5. Jones (2000) 12. Central Coast Water Quality Preservation Inc. (2008)
6. Ensminger et al. (2009) 13. Kuivila and Foe (1995)
7. Kim et al. (1999a) 14. Phillips et al. (2012)
and D. magna accounts for 90% of the non-standard bioassays 3.2. PTI values for a water sample
(Table C.1).
MTC and STC values for individual pesticide compounds in Supple-
mental Appendix Tables B.1 to B.3 can be used with Eq. (1) to calculate
3.1.3. Benthic invertebrates Median-PTI and Sensitive-PTI values, respectively, for a sample. Pesti-
MTC and STC values for benthic invertebrates were determined for cide compounds with non-standard MTC and STC values have greater
205 pesticides and 15 degradates from a total of 1734 bioassays for pes- uncertainty compared to compounds for which standardized toxicity
ticides and 66 bioassays for degradates (Tables 2 and B.3). Addition of test data are available, but the decision to exclude that pesticide from
the non-standard data for benthic invertebrates increased the number the PTI because it has a non-standard MTC or STC is equivalent
of compounds with MTC and STC values by 66 (47%) for pesticides to treating it as a non-detection in the sample. For some applications,
and 5 (50%) for degradates. The standard dataset had 3 or more bioas- it may be preferable to include an uncertain or biased toxicity value
says for 59% of compounds and represented 259 species, with amphi- (i.e., non-standard MTC or STC value) for a given pesticide, in lieu of
pods of the genus Gammarus making up nearly 20% of the standard leaving that pesticide out of the PTI. However, potential bias of pesti-
bioassays (Tables B.3 and C.1). The non-standard dataset for benthic in- cides with non-standard data should be considered when applying the
vertebrates included toxicity data from 44 different species (Table C.1), PTI to interpret water-quality data, especially if a pesticide with a non-
with the greatest number of bioassays for the midge, C. riparius (16% of standard MTC that is biased low (i.e., overestimates toxicity) tends to
the total), followed by the mosquito, Aedes aegypti (9%) and crayfish, dominate the PTI of a water sample.
Procambarus clarkii (8%). Collectively, the standard and non-standard MTC and STC values are significantly correlated with one another
benthic invertebrate datasets include data for 284 species and 172 (p b 0.001) within a taxonomic group (R2 of 0.84 to 0.93). The ratio of
genera. Benthic crustaceans account for 25% of the genera and 33% MTC to STC values for a given pesticide reflects the degree of variability
of the species; insects/spiders account for 47% and 38%, and molluscs in the toxicity data available for that pesticide. The MTC/STC ratios vary
account for 28% and 28% of the genera and species, respectively the least for cladocerans (coefficient of variation, CV = 400%), the
(Table C.2). group containing the smallest number of unique species. Fish are
intermediate (CV = 500%) and benthic invertebrates, which is the (A)

most diverse taxonomic group, have the most variable MTC/STC ratios 120
(CV = 1,100%). The MTC/STC ratio is 10 or higher for 9% of pesticides
for cladocerans, 15% of pesticides for fish, and 22% of pesticides for ben-
thic invertebrates. The Median-PTI dampens differences in toxicity
Ceriodaphnia dubia survival (percent of control)

100
values among taxa, tests, and test conditions, and provides a measure
of central tendency for comparing the toxicity of one pesticide to
another towards a given taxonomic group.
When MTCs (medians) are used, the purpose of the index is to rep- 80
resent the relative toxicity of sites, samples, or individual pesticides—

MTC values may be higher than effect concentrations for some species
and life stages in a given hydrologic system, but they are relatively 60
robust to outliers. When the STC values are used, the index is better suit-
ed for use as a screening level, because it is a more conservative (protec-
tive) indicator of the potential for toxicity. Although the 5th percentile
40
of the available toxicity values was selected as the basis for the
Sensitive-PTI (as discussed in Section 2.4), it does not necessarily pro-
tect 95% of species and it is not equivalent to a 5th-percentile hazardous
concentration (HC5) from a SSD (Etterson, 2011). The PTI methodology 20
computes the 5th (and 50th) percentiles of all suitable independent
toxicity tests, whether tests are based on the same species or on differ-
ent species. It is intended to prevent low outlier values, such as from 0
singularly sensitive species, statistically anomalous results, or unusual 0.0000001 0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10 100 1,000
test conditions, from skewing the results. Median-PTI

(B)
3.3. Example applications of the PTI to interpret water-quality data 120
3.3.1. Validation test of the PTI using aggregated data from published
studies
100
Data aggregated from published studies were used to test how
well the PTI approach predicted the incidence of observed toxicity,

(i.e., reduction in C. dubia survival). The available studies (Table 3),
which analyzed different numbers and classes of pesticides at different 80
analytical reporting levels, are not considered to represent a robust test

of the PTI model. However, these results provide some indication of PTI
values at which toxicity has been observed in field-collected water 60
samples.
Based on aggregated data from published studies (Table 3 and
Supplemental Appendix D), Fig. 1 shows survival of C. dubia in
40
4–8 day toxicity tests in ambient stream water as a function of the
Median-PTI (Fig. 1A) and the Sensitive-PTI (Fig. 1B). With some excep-
tions (those symbols shown on y-axis), survival remained near 100%
until the Median-PTI reached about 0.1. Survival rates decreased sharply 20
at Median-PTI values between about 0.1 and 0.5, with highly variable %
survival within this range. Of samples with Median-PTI values N 1, 97%
had C. dubia survival reduced to ≤ 50% relative to controls, and 96% 0
had 0% survival. Percent survival declined to 50% well below the theo- 0.0000001 0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10 100 1,000
retical value of 1 for the Median-PTI. The same studies plotted in rela- Sensitive-PTI
tion to the Sensitive-PTI look very similar (Fig. 1B), except that most
results are shifted to the right, so that C. dubia survival appears to EXPLANATION
drop at Sensitive-PTI values between 0.5 and 1. Results were similar Ganapathy (1999a, b) Kim et al. (1999b), Walters et al. (2000c, e–m)
among studies and no single study dominated the pattern shown. Nordmark (1999, 2000) Walters et al. (2000a, b, d, 2001a-m)
The pattern shown in Fig. 1 was the same for studies that exposed Gill (2002) CCWQP (2008)
Jones (2000) CCWQP (2009)
C. dubia to ambient water for 4 days vs. 6–8 days and also for studies
Ensminger et al. (2009) Phillips et al. (2009)
that measured only organophosphate insecticides vs. multiple classes
Kim et al. (1999a) Kuivila and Foe (1995)
of pesticides (not shown). Organophosphates accounted for over half
of the Median-PTI in 71% of samples overall, and in 86% of samples
Fig. 1. Ceriodaphnia dubia survival on 4–8 day exposure to ambient water, as a function of
with C. dubia survival reduced to ≤50%. This occurred in part because Median-Pesticide Toxicity Index (PTI) (A) and Sensitive-PTI for cladocerans (B), computed
organophosphates are highly toxic, but also because most of the pub- for pesticides detected in the sample. Each symbol represents a single sample (n = 479
lished studies had focused on organophosphate insecticides because of samples). Results are aggregated from published studies in Table 3.
CWA §303(d) listing of streams. Where organophosphates were the
only pesticides analyzed (141 samples) or the only pesticides detected
(110 additional samples), the PTI model is indistinguishable from an
organophosphate TU model, in which only pesticides with this common C. dubia survival reduced to ≤50%. In these 81 “toxic” samples, organo-
mode of action (acetylcholinesterase inhibition) are included. However, phosphates accounted for less than half of the Median-PTI in 32%
of 228 samples with multiple pesticide classes detected, 81 samples had of samples, and less than one quarter of the Median-PTI in 17% of
Samples classified correctly by threshold, in percent

samples—indicating that other pesticides besides organophosphates 100
may have contributed to the observed toxicity in these samples.

The incidence of toxicity in samples with Median-PTI values b1 is
80
expected because the lowest toxicity values were excluded by use of
the median, pesticides may be present at concentrations below the
analytical detection level, and contaminants may have been present
60
that were not analyzed in the sample or that were not included (be-
cause toxicity data were not available) in the PTI. In addition, variability
is expected because of uncertainty in MTCs used to compute PTI values, 40
deviation from the CA model that underlies the PTI approach, and envi-
ronmental factors affecting toxicity in the sample. Lack of toxicity at EXPLANATION
Median-PTI values N1 is more difficult to explain, but this occurred in 20 Median-PTI

Sensitive-PTI
only 3 samples from 2 studies, out of a total of 479 samples. Possible
explanations include the fact that dose–response curves vary for differ-
ent pesticides and mortality occurs at concentrations below the EC50 or 0
0 0.5 1 1.5 2 2.5 3
LC50 endpoints used to derive MTCs; use of the MTC to represent a Possible 50%-mortality threshold for cladocerans
given pesticide, which may underestimate toxicity to some organisms
and overestimate toxicity to others; the assumption that toxicity is Fig. 2. Effectiveness of various 50%-mortality thresholds for the Pesticide Toxicity Index
(PTI) in correctly classifying samples in the aggregated dataset as nontoxic or toxic. Each
additive with no chemical interaction for pesticides from multiple clas-
symbol shows the percentage of samples in the entire dataset (n = 479) that was correct-
ses with different modes of action; and the fact that the PTI does not ly classified using the threshold value on the x-axis. The gray-shaded areas indicate the
consider environmental factors (e.g., organic carbon, temperature) optimum 50%-mortality threshold range (x-axis value corresponding to the maximum
that may affect bioavailability and(or) toxicity. y-axis value) for the Sensitive-PTI and Median-PTI.
In summary, although data available to test the PTI model are limit-
ed, C. dubia survival was reduced to ≤50% of controls in 44% of samples on the x-axis. The x-axis represents the possible 50%-mortality thresh-
with Median-PTI values in the range of 0.1 to 1, and to 0% in 96% of sam- old values ranging from 0.01 to 3, and the y-axis indicates how well
ples with Median-PTI values N1. For the Sensitive-PTI, C. dubia survival each threshold value performed in explaining C. dubia mortality in the
was reduced to ≤50% of controls in 81% of samples in the range of 0.1 aggregated dataset. At low 50%-mortality threshold values, the correct
to 1, and in 89% of samples with Sensitive-PTI values N 1. classification rate is low because a large proportion of samples above
In theory, for TUs that are based on acute LC50/EC50 data, the CA these low thresholds were nontoxic. As increasingly higher values for
model predicts 50% mortality at a theoretical threshold of 1, and theo- the 50%-mortality threshold values are considered, fewer of the samples
retically based thresholds were proposed in previous studies by above the threshold are nontoxic, so the percentage of samples that
Battaglin and Fairchild (2002) and Anderson (2008) for classifying sam- were correctly classified increases to a maximum. As potential 50%-
ples according to the probability of toxicity from pesticide mixtures. In mortality threshold values continue to increase above the level associat-
contrast, an empirical threshold was previously developed for contam- ed with this maximum, the number of toxic samples below the thresh-
inant mixtures in sediment (Ingersoll et al., 2001; MacDonald et al., old begins to increase, causing the frequency of correct classification to
2000) on the basis of validation data from field studies, which indicated decline. The optimum 50%-mortality threshold for correct classification
a high probability of toxicity at values N 0.5 for an additive sediment- of samples in this dataset—that is, the value resulting in the maximum
quality index called the Mean Probable Effect Concentration Quotient. percentage of correctly classified samples—is 0.3 for the Median-PTI
Similarly, in the present study, the validation dataset consisting of and 0.9–1.1 for the Sensitive-PTI (Fig. 2).
field data on pesticides and C. dubia survival (Table 3 and Supplemental The 50%-mortality thresholds of 0.3 and about 1 for Median-PTI and
Appendix D) can be used to determine an apparent empirical threshold Sensitive-PTI, respectively, are empirical thresholds that apply only to
at which the PTI model best explains the incidence of toxicity observed the studies on which they were based. Although in this analysis, toxicity
in the reviewed studies, but in this case the “threshold” is actually based was defined as reduction of C. dubia survival to ≤50% of controls during
on a 50%-mortality response. An empirical threshold of toxicity should a 4–8 day exposure, the results were similar when toxicity was defined
result in the correct classification of as many samples as possible—so as reducing survival to ≤70% of that of controls (the optimum 70%-mor-
that all or most of the samples above the empirical threshold are toxic, tality threshold for the Median-PTI was still 0.3) or ≤40% of that of con-
and all or most of the samples below the threshold are nontoxic— trols (the optimum 40%-mortality threshold for the Median-PTI was
however “toxic” may be defined. 0.35). This indicates robustness in the analysis, in that an empirical
Fig. 2 shows the percentage of samples in the aggregated dataset “mid-level” mortality threshold was not strongly influenced by the
that were correctly classified as “toxic” or “nontoxic,” as a function of exact percentage survival used to define a “toxic” sample. Because the
various possible 50%-mortality threshold values for Median-PTI and empirical 50%-mortality thresholds for the Median-PTI and Sensitive-
Sensitive-PTI. In the analysis shown in Fig. 2, toxicity was defined as re- PTI predict about 50% mortality, it is likely that some toxicity—some
ducing C. dubia survival to ≤50% of controls, or 50% mortality, because degree of reduced survival, sublethal effects, and(or) adverse effects
this corresponds to the definition of LC50 and EC50 on which the cla- on invertebrate communities—will occur at lower index levels.
doceran Median-PTI and Sensitive-PTI are based. Correspondingly, Whether the empirical 50%-mortality thresholds derived here
“nontoxic” is used to indicate survival N50% relative to control. Correct would successfully predict mortality in future studies of large numbers
classification is defined as the sum of (i) the number of samples with of pesticides from multiple classes is uncertain because (1) this analysis
PTI above a potential 50%-mortality threshold and also toxic (as defined is based on a limited number of samples (479); (2) the studies on which
above), plus (ii) the number of samples with PTI below the threshold this analysis is based each determined a relatively small number of pes-
and nontoxic, as a percentage of (iii) the total number of samples. It is ticides (8 to 27 per study); (3) many of these studies were focused on
important to acknowledge that the potential 50%-mortality thresholds organophosphate insecticides, which share a common mode of action
under discussion designate index levels at which C. dubia survival was and have been shown to exhibit additive toxicity to cladocerans
reduced to ≤50%, whereas in actuality, mortality and sublethal toxicity (Bailey et al., 1997); and (4) concentrations of organophosphate insec-
will occur at levels below LC50 and EC50 values. ticides have decreased in streams (Ryberg et al., 2010; Sullivan et al.,
In Fig. 2, each symbol represents all the data (n = 479 samples) 2009) since USEPA restricted their uses in both agricultural and urban
assessed relative to the corresponding possible 50%-mortality threshold applications (U.S. Environmental Protection Agency, 2000, 2001),
whereas concentrations of insecticides in other classes, including carba- addition to pesticide contamination, including other chemical stressors
ryl, pyrethroids, and fipronil, have increased (Ryberg et al., 2010). (e.g., nutrients, salinity, acidification, hydrophobic contaminants in
sediment), physical stressors (e.g. sediment, temperature, flow alter-
3.3.2. Determination of potentially toxic pesticides in a watershed ation, habitat degradation) and biological stressors (e.g. invasive or
Results from one of the studies in Table 3 (and included in the aggre- non-native species), as frequently noted in the literature (e.g., Bryant
gated dataset in Section 3.3.1 and Appendix D) are used as an example and Carlisle, 2012; Carlisle et al., 2013; Tate and Heiny, 1995; U.S.
to illustrate the application of the PTI to interpret water quality data. In Environmental Protection Agency, 2006).
this study of dormant-spray pesticides and associated water toxicity
conducted by the California DPR, water samples were collected from 3.4. Limitations
the Sacramento River watershed at Wadsworth Canal, a tributary of
the Sutter Bypass, which receives inflow from dormant spray areas The PTI has several limitations, which must be carefully considered
and has no major inputs from municipal or industrial sources (Gill, in applications:
2002; Nordmark, 1999, 2000). Samples were collected in December,
prior to the onset of dormant spray season, and again during January • The PTI is a relative ranking system that indicates that one sample is
to March, continuing until dormant spray had ceased, for three consec- likely to be more or less toxic than another sample, but does not indi-
utive years (1998–2001). The parent pesticides analyzed were 8 organ- cate that toxicity will necessarily occur.
ophosphate insecticides, 2 carbamate insecticides, and 9 herbicides. • Toxicity values are based on short-term laboratory experiments with
Acute toxicity to C. dubia was measured in 4-day static renewal tests. EC50 (nonlethal response) or LC50 (mortality) endpoints; the PTI
Median-PTI and Sensitive PTI values for cladocerans were calculated does not reflect long-term/chronic exposure or (except for immobili-
using Eq. (1) and the pesticide concentrations reported in the California zation in cladocerans) incorporate sublethal endpoints.
DPR study. Data from the first two years (the first 42 of the 66 total sam- • The PTI does not account for environmental factors, such as dissolved
ples) of the Wadsworth Canal study (Nordmark, 1999, 2000) are shown organic carbon, particulates, pH, and temperature, which can affect
in Fig. 3. Total pesticide concentrations (Fig. 3A) tended to be low in the toxicity and bioavailability of pesticides.
December, and the herbicide diuron was the only pesticide detected. • The PTI assumes that pesticide toxicity is additive and there is no
The number of unique pesticides detected increased to seven during chemical interaction (synergism or antagonism), which may not be
the spray season, with total pesticide concentrations peaking in late the case for complex mixtures of pesticides from different chemical
January and February (Fig. 3A). Median-PTI values also peaked in late classes and with different MOAs across all taxonomic groups and life
January and February (Fig. 3B) and were dominated by diazinon, with stages.
minor contributions from methidathion in some samples. The same pat- • The PTI does not take into account the dose–response curves of either
tern was observed for Sensitive-PTI (Fig. 3C), except that the scale of PTI single-chemical or mixtures exposures.
values was higher and the contribution of methidathion was barely • The PTI is limited to pesticides measured in the water column; hydro-
visible. Although the herbicides simazine, hexazinone, diuron and phobic pesticides may be underrepresented in terms of potential tox-
bromacil combined to account for a third to half of the summed concen- icity, especially to benthic organisms.
trations of pesticides detected during the dormant spray season, organ- • Uncertainty in the relative toxicity of compounds is high for pesticides
ophosphates accounted for over 99% of the Median-PTI for samples in with relatively few bioassays available. The 10,837 bioassays in this
which C. dubia survival was ≤ 50%. Toxicity was observed in samples data set are divided among 440 pesticides and 52 degradates, 559 dif-
with Median-PTI greater than about 0.3 and Sensitive-PTI greater than ferent species, and three taxonomic groups, making the number in
about 0.8. The percent survival of C. dubia at the Wadsworth Canal site each group relatively small. Although this does not preclude the use
over the 3-year study (Fig. 3D) was significantly correlated with both of the data as the best available, it demonstrates the sparseness of
the Median-PTI and the Sensitive-PTI (p b 0.0001, Spearman rho = available data on the toxicity of many currently used pesticides.
−0.67), which each explained 43% of the variability in % survival in • Most data are for standard test species (rainbow trout, bluegill, and
separate least-squares regressions. The relationship also was signifi- fathead minnow for fish, and daphnids for invertebrates). Analysis
cant for the summed TUs for organophosphate insecticides only of SSDs for pesticides showed that values for the three typical OPP
(p b 0.0001, Spearman rho = − 0.68), which explained 37% of the test species (rainbow trout, bluegill, and D. magna) typically resulted
variability in % survival. PTI analysis, such as that shown here, can in smaller HC5 values (i.e., indicating greater toxicity) compared to
aid in determining which sites, samples, or pesticides are the most random subsets of three toxicity values for the available test species
important in terms of potential toxicity to aquatic organisms. (Etterson, 2011).
• Many factors contribute to the high variability observed in laboratory
3.3.3. Invertebrate community condition in streams across the U.S. toxicity tests, including formulation of the pesticide; test species, age
Median-PTI values for benthic invertebrates were calculated for and condition of individual organisms used; water conditions (pH,
pesticide concentrations detected in water at 128 stream sites across temperature, concentration and type of dissolved organic carbon);
the U.S. sampled during 1993–2004 by the USGS NAWQA Program. testing environment (flow-through or static); and whether test or-
MTC values were available for 20 of the 26 pesticides analyzed in ganisms are fed during exposure (Jonczyk and Gilron, 2005;
these samples. Macroinvertebrate community condition, characterized Persoone et al., 2009). For applications in which certain individual
by the O/E ratio at the same 128 stream sites, was analyzed relative to compounds are particularly important, special attention should be
the maximum Median-PTI for benthic invertebrates occurring within given to the variability in toxicity test results for those compounds.
the 90-days prior to the date of the invertebrate sample, as described
in Section 2.5. The O/E ratio is a measure of biodiversity. An O/E ratio 4. Conclusions
of 1 indicates that the measured invertebrate composition is equal to
the expected composition (U.S. Environmental Protection Agency, The PTI is a relative, but quantitative, indicator of potential tox-
2006), with ratios b0.8 indicating biological degradation (Carlisle and icity that can be used in study design or to interpret water quality
Meador, 2007). There was a significant inverse relationship between data, relate pesticide exposure to biological condition, and priori-
O/E and the Median-PTI (p = 0.01, not shown), but the Median-PTI tize future assessments. The PTI uses toxicity concentrations avail-
explained only 5% of the variability in O/E ratios (biodiversity). This able for 440 pesticides (out of a total of 484 pesticides with
small amount of variability explained is not unexpected because inver- reported agricultural use during 1992 to 2011) and 52 pesticide
tebrate community condition in streams is affected by many factors in degradates. Combined, the toxicity data for pesticides and
6
(A)
5
Concentration (µg/L)
4
4.5
4
(B)
3.5
3
Median-PTI
2.5
2
EXPLANATION
1.5
Methidathion
1
Diazinon
0.5 Carbaryl
0 Simazine
Hexazinone
12
Diuron
(C) Bromacil
10
Ceriodaphnia dubia
survival (part D only)
Sensitive-PTI
8 50% Survival
0
120
(D)
100
80
60
40
20
0
12/7/98
12/9
1/4/99
1/6
1/11
1/13
1/18
1/20
1/25
1/27
2/1
2/3
2/8
2/10
2/15
2/17
2/22
2/24
3/1
3/3
12/6
12/8
1/3/00
1/5
1/10
1/12
1/17
1/29
1/24
1/26
1/31
2/2
2/7
2/9
2/14
2/16
2/21
2/23
2/28
3/1
3/6
3/8
Sampling date
Fig. 3. Summed pesticide concentrations (A), Median-Pesticide Toxicity Index (PTI) (B), Sensitive-PTI (C), and Ceriodaphnia dubia survival as percentage of control for samples from the
Wadsworth Canal during the 1999 and 2000 dormant spray seasons (D). Pesticide concentrations and C. dubia survival are from Nordmark (1999, 2000). The height of each segment in the
stacked bar indicates the corresponding pesticide's contribution to the total. Circles above bars indicate that these samples were toxic to C. dubia, defined as having survival reduced to
≤50% relative to control after 4-day exposure.
degradates include 10,837 bioassays representing 559 different Two types of PTI, as described herein, can be calculated for a field sam-
species in three taxonomic groups. Toxicity endpoints include ple. The Median-PTI is calculated using median toxicity concentrations
LC50 for fish, EC50 or LC50 for cladocerans, and LC50 or EC50 for for each pesticide in a sample mixture. The Sensitive-PTI uses the 5th per-
benthic invertebrates. centile of toxicity concentrations (except that the minimum is
substituted for pesticides with small data sets). The Median-PTI approach Belden JB, Gilliom RJ, Lydy MJ. How well can we predict the toxicity of pesticide mixtures
to aquatic life? Integr Environ Assess Manag 2007;3:364–72.
is more robust to outliers, whereas the Sensitive-PTI may be appropriate Bliss CI. The toxicity of poisons applied jointly. Ann Appl Biol 1939;26:585–615.
for applications in which a conservative comparison to a screening level is Bryant WL, Carlisle DM. The relative importance of physicochemical factors to stream
needed. biological condition in urbanizing basins: evidence from multimodel inference.
Freshw Sci 2012;31:154–66.
To test the PTI model, data on concurrent pesticide concentrations Carlisle DM, Hawkins CP. Land use and the structure of western US stream invertebrate
and toxicity to C. dubia in ambient stream water were aggregated assemblages: predictive models and ecological traits. JN Am Benthol Soc 2008;27:
from published field studies, PTI values were calculated, and empirical 986–99.
Carlisle DM, Meador MR. A biological assessment of streams in the eastern united states
50%-mortality thresholds for the Median-PTI and Sensitive-PTI were
using a predictive model for macroinvertebrate assemblages. J Am Water Resour
determined to be about 0.3 and 1, respectively. These 50%-mortality Assoc 2007;43:1197–207.
thresholds correctly predicted toxicity (defined as ≥50% mortality) or Carlisle DM, Meador MR, Short TM, Tate CM, Gurtz ME, Bryant WL, et al. The quality of
our Nation's waters—ecological health in the Nation's streams, 1993–2005. Circular.
nontoxicity (b50% mortality) in about 90% of samples in the reviewed
Reston, VA: U.S. Geological Survey; 20131391 [120 pp.].
studies. Although field data are not available to validate the PTI model Cedergreen N, Streibig JC. Can the choice of endpoint lead to contradictory results of
for a comprehensive range of pesticides, the analysis of available data mixture-toxicity experiments? Environ Toxicol Chem 2005;24:1676–83.
described here indicates that substantial mortality may occur at Cedergreen N, Christensen AM, Kamper A, Kudsk P, Mathiassen SK, Streibig JC, et al. A
review of independent action compared to concentration addition as reference
Median-PTI values in the range of 0.1 to 1—based on the observation models for mixtures of compounds with different molecular target sites. Environ
that ≥ 50% mortality to C. dubia occurred in only 6% of samples from Toxicol Chem 2008;27:1621–32.
the aggregated dataset with Median-PTI values b 0.1, but in 44% of Central Coast Water Quality Preservation Inc.. Phase I follow-up water quality monitoring:
organophosphate pesticide sampling final data report. Central Coast Region Condi-
samples in the 0.1 to 1 range. Mortality likely would be high at tional Ag Waiver Cooperative Monitoring Program; 2008 [37 pp.].
Median-PTI values N1, where 96% of samples in the aggregated valida- Central Coast Water Quality Preservation Inc.. Supplemental water quality monitoring for
tion data set showed 100% mortality to C. dubia. For the Sensitive-PTI, organophosphate pesticides and aquatic toxicity. Central Coast Region Conditional Ag
Waiver Cooperative Monitoring Program; 2009 [24 pp.].
the percentage of samples with ≥ 50% mortality was 5% in samples Commission of the European Communities. Commission Directive 92/69/EEC of 31 July
with Sensitive-PTI values b 0.1, 19% in the 0.1 to 1 range, and 89% at 1992 adapting to technical progress for the 17th time Council Directive 67/548/EEC
Sensitive-PTI values N1. on the approximation of laws, regulations and administrative provisions relating to
the classification, packaging and labelling of dangerous substances. C.2. Acute toxicity
The PTI can be applied to pesticide mixtures detected in ambient
test Daphnia. Commission of the European Communities; 1992.
water samples, and PTI values for individual water samples then can Cunnane C. Unbiased plotting positions—a review. J Hydrol 1978;37:205–22.
be used to rank sites according to their expected relative potential tox- de Zwart D, Posthuma L. Complex mixture toxicity for single and multiple species:
proposed methodologies. Environ Toxicol Chem 2005;24:2665–76.
icity caused by pesticides, to assess changes in potential toxicity over
Deneer JW. Toxicity of mixtures of pesticides in aquatic systems. Pest Manag Sci 2000;56:
time at a single site, or to predict which of the pesticides detected may 516–20.
be the biggest contributor to potential toxicity for the sample or site. Ensminger M, Bergin R, Spurlock F. Pesticide concentrations in water and sediment and
PTI values for samples, seasons, or sites also can be used as explanatory associated invertebrate toxicity in Del Puerto and Orestimba Creeks, California. Sacra-
mento, CA: California Environmental Protection Agency, Department of Pesticide
variables in multivariate analysis designed to determine which environ- Regulation; 2009 [47 pp.].
mental variables best explain spatial patterns in the structure of a bio- Etterson M. Analyses of sensitivity distributions for estimation of acute hazard concentra-
logical community. A more rigorous test of the PTI model is needed, tions to aquatic animals. Appendix C of USEPA, review of methods for characterizing
effects of pesticides and other chemical stressors to aquatic organisms. U.S. Environ-
but this will require the availability of data for pesticides from multiple mental Protection Agency; 2011.
classes and modes of action, concurrent with data on aquatic toxicity Faust M, Altenburger R, Boedeker W, Grimme LH. Algal toxicity of binary combinations of
and(or) ecological condition. pesticides. Bull Environ Contam Toxicol 1994;53:134–41.
Faust M, Altenburger R, Backhaus T, Bodeker W, Scholze M, Grimme LH. Predictive assessment
Supplementary data to this article can be found online at http://dx. of the aquatic toxicity of multiple chemical mixtures. J Environ Qual 2000;29:1063–8.
doi.org/10.1016/j.scitotenv.2013.12.088. Faust M, Altenburger R, Backhaus T, Blanck H, Boedeker W, Gramatica P, et al. Joint algal
toxicity of 16 dissimilarly acting chemicals is predictable by the concept of indepen-
dent action. Aquat Toxicol 2003;63:43–63.
Disclosure statement Fuhrer GJ, Morace JL, Johnson HM, Rinella JF, Ebbert JC, Embrey SS, et al. Water quality in
the Yakima River basin, Washington, 1999–2000. Circular. Reston, VA: U.S. Geological
Survey; 20041237 [34 pp.].
None of the authors have any actual or potential conflict of interest Ganapathy C. Preliminary results of acute chronic toxicity testing of surface water
including any financial, personal or other relationships with other peo- monitored in the San Joaquin River watershed, winter 1997–1998. Memorandum
ple or organizations within three years of beginning the submitted work from Carissa Ganapathy to Donald Weaver, January 5, 1999. Sacramento, CA: California
Department of Pesticide Regulation; 1999a [27 pp.].
that could inappropriately influence, or be perceived to influence, their
Ganapathy C. Preliminary results of acute chronic toxicity testing of surface water monitored
work. in the San Joaquin River watershed, winter 1998–1999. Memorandum from Carissa
Ganapathy to Donald Weaver, July 20, 1999. Sacramento, CA: California Department
of Pesticide Regulation; 1999b [22 pp.].
Acknowledgments GfK Kynetec. Unpublished proprietary agricultural pesticide use data for 2009, 2009.
Gill S. Preliminary results of pesticide residue analysis and acute and chronic toxicity
testing of surface water monitored in the Sacramento River watershed, winter
The authors thank USGS personnel for collecting and analyzing
2000–2001 (study number 199). Memorandum from Sheryl Gill to Kean Goh, Jan-
stream water samples and biological samples. We also thank Jerry uary 22, 2002. Sacramento, CA: California Department of Pesticide Regulation;
Tran (USGS) for his assistance with graphics, Daren Carlisle (USGS) for 2002 [25 pp.].
providing NAWQA biological data, and Robert Gilliom (USGS), Wade Gilliom RJ, Barbash JE, Crawford CG, Hamilton PA, Martin JD, Nakagaki N, et al. The quality
of our Nation's waters—pesticides in the Nation's streams and ground water,
Bryant (USGS), and two anonymous reviewers for their helpful insights 1992–2001. Circular. Reston, VA: U.S. Geological Survey; 20061291 [172 pp.].
and reviews. Guy M, Singh L, Mineau P. Using field data to assess the effects of pesticides on crustacea
in freshwater aquatic ecosystems and verifying the level of protection provided by
water quality guidelines. Integr Environ Assess Manag 2011;7:426–36.
References Helsel DR, Hirsch RM. Statistical methods in water resources. New York: Elsevier Science;
1992.
Anderson A-M. Development of an aquatic pesticide toxicity index for use in Alberta. Ingersoll C, MacDonald D, Wang N, Crane J, Field L, Haverland P, et al. Predictions of sed-
Edmonton, Alberta: Alberta Environment, Environmental Assurance; 2008W0803 iment toxicity using consensus-based freshwater sediment quality guidelines. Arch
[30 pp.]. Environ Contam Toxicol 2001;41:8–21.
Backhaus T, Faust M. Predictive environmental risk assessment of chemical mixtures: a Jonczyk E, Gilron G. Acute and chronic toxicity testing with Daphnia sp. In: Blaise C, Férard
conceptual framework. Environ Sci Technol 2012;46:2564–73. J-F, editors. Small-scale Freshwater Toxicity Testing for freshwater Environments.
Bailey H, Miller J, Miller M, Wiborg L, Deanovic L, Shed T. Joint acute toxicity of diazinon Small-scale Freshwater Toxicity Investigations. Toxicity Test MethodsSpringer;
and chlorpyrofios to Ceriodaphnia dubia. Environ Toxicol Chem 1997;16:2304–8. 2005. p. 337–93.
Battaglin W, Fairchild J. Potential toxicity of pesticides measured in Midwestern streams Jones D. Preliminary results of acute and chronic toxicity testing of surface water moni-
to aquatic organisms. Water Sci Technol 2002;45:95–103. tored in the San Joaquin River watershed, winter 1999–2000. Memorandum from
DeeAn Jones to Patricia Dunn, October 3, 2000. Sacramento, CA: California Depart- Sprague JB, Ramsay BA. Lethal levels of mixed copper–zinc solutions for juvenile salmon.
ment of Pesticide Regulation; 2000 [23 pp.]. J Fish Res Board Can 1965;22:425–32.
Junghans M, Backhaus T, Faust M, Scholze M, Grimme LH. Application and validation of Stenström JR. Mixture toxicity of pesticides and biological effects in agricultural streams.
approaches for the predictive hazard assessment of realistic pesticide mixtures. LicentiateUppsala, Sweden: Faculty of Natural Resources and Agricultural Sciences,
Aquat Toxicol 2006;76:93–110. Department of Aquatic Sciences and Assessment, Swedish University of Agricultural
Kim D, Walters J, Sava R. Preliminary background results of organophosphate analysis and Sciences; 201343.
acute toxicity testing of surface water monitored for the red imported fire ant project Stephan CE, Mount DI, Hanson DJ, Gentile JH, Chapman GA, Brungs WA. Guidelines for
in Orange County, March and April, 1999. Memorandum from Dave Kim, Johanna deriving numerical National Water Quality Criteria for the protection of aquatic
Walters, and Roger Sava to Kean S Goh, November 29, 1999. Sacramento, CA: California organisms and their uses. EPA PB85-227049. Duluth, MN: U.S. Environmental Protec-
Department of Pesticide Regulation; 1999a [7 pp.]. tion Agency, Environmental Research Laboratories; 1985.
Kim D, Walters J, Sava R. Preliminary results of pesticide analysis and acute toxicity test- Sullivan DJ, Vecchia AV, Lorenz DL, Gilliom RJ, Martin JD. Trends in pesticide concentra-
ing of monthly surface water monitoring for the red imported fire ant project in tions in corn-belt streams, 1996–2006. Scientific Investigations Report 2009–5132.
Orange County, May and June, 1999. Memorandum from Dave Kim, Johanna Walters, Reston, VA: U.S. Geological Survey; 2009. [75 pp.].
and Roger Sava to Kean S Goh, December 2, 1999. Sacramento, CA: California Department Tate CM, Heiny JS. The ordination of benthic invertebrate communities in the South Platte
of Pesticide Regulation; 1999b [8 pp.]. River Basin in relation to environmental factors. Freshw Biol 1995;33:439–54.
Kimerle R, Barnthouse L, Brown R, de Beyssac BC, Gilbertson M, Monk K, et al. Ecological TIBCO Software Inc.. TIBCO Spotfire S + 8.1 for Windows, Revision Date 11/4/2008;
effects, Chap. 4 in Chemical ranking and scoring—guidelines for relative assessments 2008.
of chemicals. In: Swanson MB, Socha AC, editors. Proceedings of the Pellston Work- University of Hertfordshire. The Pesticide Properties DataBase (PPDB) developed by
shop on Chemical Ranking and Scoring, 12–16 February, 1995, Sandestin, FL. SETAC the Agriculture & Environment Research Unit (AERU) at the University of Hert-
special publication seriesPensacola, FL: SETAC Press; 1997. p. 89–111. fordshire, 2006–2013. Accessed September–October, 2012, and May 2013, at
Kuivila KM, Foe CG. Concentrations, transport and biological effects of dormant spray pesti- http://sitem.herts.ac.uk/aeru/ppdb/en/index.htm.
cides in the San Francisco Estuary, California. Environ Toxicol Chem 1995;14:1141–50. U.S. Environmental Protection Agency. Ecological effects test guidelines. OPPTS 850.1010
Liess M, von der Ohe PC. Analyzing effects of pesticides on invertebrate communities in aquatic invertebrate acute toxicity test, freshwater daphnids, public draft. EPA
streams. Environ Toxicol Chem 2005;24:954–65. 712-C-96-114. Washington, DC: U.S. Environmental Protection Agency, Office of
Lydy M, Belden J, Wheelock C, Hammock B, Denton D. Challenges in regulating pesticide Prevention, Pesticides and Toxic Substances; 1996 [10 pp.].
mixtures. Ecol Soc 2004;9:1. [online, http://www.ecologyandsociety.org/vol9/iss6/ U.S. Environmental Protection Agency. Chlorpyrifos revised risk assessment and agree-
art1/]. ment with registrants. Fact sheet. Washington, DC: U.S. Environmental Protection
MacDonald DD, Ingersoll CG, Berger TA. Development and evaluation of consensus-based Agency, Prevention, Pesticides and Toxic Substances; 2000 [4 pp.].
sediment quality guidelines for freshwater ecosystems. Arch Environ Contam Toxicol U.S. Environmental Protection Agency. Diazinon revised risk assessment and agreement
2000;39:20–31. with registrants. Fact sheet. Washington, DC: U.S. Environmental Protection Agency,
Maltby L, Blake N, Brock TCM, van den Brink PJ. Insecticide species sensitivity distributions: Prevention, Pesticides and Toxic Substances; 2001 [4 pp.].
importance of test species selection and relevance to aquatic ecosystems. Environ U.S. Environmental Protection Agency. Methods for measuring the acute toxicity of efflu-
Toxicol Chem 2005;24:379–88. ents and receiving waters to freshwater and marine organisms. EPA 821-R-02-0125th
Maltby L, Brock TCM, van den Brink PJ. Fungicide risk assessment for aquatic ecosystems: ed. . Washington, DC: U.S. Environmental Protection Agency, Office of Water; 2002
importance of interspecific variation, toxic mode of action, and exposure regime. [266 pp.].
Environ Sci Technol 2009;43:7556–63. U.S. Environmental Protection Agency. Wadeable streams assessment — a collaborative
Munn MD, Gilliom RJ. Pesticide Toxicity Index for freshwater aquatic organisms. survey of the Nations's streams. EPA 841-B-06-002. Washington, DC: U.S. Environ-
Water-Resources Investigations Report 01–4077. Reston, VA: U.S. Geological Survey; mental Protection Agency, Office of Research and Development, Office of Water;
2001. [55 pp.]. 2006 [113 pp.].
Munn MD, Gilliom RJ, Moran PW, Nowell LH. Pesticide Toxicity Index for freshwater U.S. Environmental Protection Agency. ECOTOX user guide: ECOTOXicology Database
aquatic organisms. Scientific Investigations Report 2006–5148. 2nd ed. Reston, VA: System, Version 4.0. Duluth, MN: U.S. Environmental Protection Agency, Office
U.S. Geological Survey; 2006. [81 pp.]. of Research and Development and the National Health and Environmental Effects
Nordmark C. Preliminary results of acute and chronic toxicity testing of surface water Research Laboratory's Mid-Continent Ecology Division; 2011 [96 pp.].
monitored in the Sacramento River watershed, winter 1998–1999. Memorandum U.S. Environmental Protection Agency. ECOTOXicology Database System, Version 4.0. U.S.
from Craig Nordmark to Don Weaver, May 26, 1999. Sacramento, CA: California Environmental Protection Agency, Office of Research and Development and the
Department of Pesticide Regulation; 1999 [32 pp.]. National Health and Environmental Effects Research Laboratory's Mid-Continent
Nordmark C. Preliminary results of acute and chronic toxicity testing of surface water Ecology Division; 2012a [Updated June 15, 2012, Accessed July 5, 2012, at http://
monitored in the Sacramento River watershed, winter 1999–2000. Memorandum www.epa.gov/ecotox/].
from Craig Nordmark to Don Weaver, September 15, 2000. Sacramento, CA: California U.S. Environmental Protection Agency. Office of Pesticide Programs' aquatic life bench-
Department of Pesticide Regulation; 2000 [29 pp.]. marks (freshwater). U.S. Environmental Protection Agency, Office of Pesticide
Norman JE, Kuivila KM, Nowell LH. Prioritizing pesticide compounds for analytical Programs; 2012b [Updated May 30, 2102, Accessed September-October, 2012,
methods development. Scientific Investigations Report 2012–5045. Reston, VA: U.S. at http://www.epa.gov/oppefed1/ecorisk_ders/aquatic_life_benchmark.htm].
Geological Survey; 2012. [206 pp.]. van den Brink PJ, Blake N, Brock TCM, Maltby L. Predictive value of species sensitivity
Organization for Economic Cooperation and Development. Test No. 202: Daphnia sp. distributions for effects of herbicides in freshwater ecosystems. Hum Ecol Risk Assess
Acute immobilisation test, OECD guidelines for testing of chemicals, Section 2. 2006;12:645–74.
OECD; 2004 [12 pp.]. Waite IR, Sobieszcyk S, Carpenter KD, Arnsberg AJ, Johnson HM, Hughes CA, et al. Effects
Persoone G, Baudo R, Cotman M, Blaise C, Thompson KC, Moreira-Santos M, et al. Review of urbanization on stream ecosystems in the Willamette River basin and surrounding
on the acute Daphnia magna toxicity test — evaluation of the sensitivity and the pre- area, Oregon and Washington. Scientific Investigations Report 2006-5101-D. Reston,
cision of assays performed with organisms from laboratory cultures or hatched from VA: U.S. Geological Survey; 2006. [63 pp.].
dormant eggs. Knowl Manag Aquat Ecosyst 2009;393:29. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
Phillips BM, Anderson BS, Hunt JW, Siegler K, Voorhees JP, Tjeerdema RS, et al. Pyre- ing of monthly surface water monitoring for the red imported fire ant project in
throid and organophosphate pesticide-associated toxicity in two coastal water- Orange County, August 2000 (Study 183). Memorandum from Johana Walters,
sheds (California, USA). Environ Toxicol Chem 2012;31:1595–603. Dave Kim, and Kean S. Goh to John S. Sanders, October 5, 2000. Sacramento, CA:
Posthuma L, de Zwart D. Predicted effects of toxicant mixtures are confirmed by changes California Department of Pesticide Regulation; 2000a [10 pp.].
in fish species assemblages in Ohio, USA, rivers. Environ Toxicol Chem 2006;25: Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
1094–105. ing of monthly surface water monitoring for the red imported fire ant project in
Posthuma L, de Zwart D. Predicted mixture toxic pressure relates to observed fraction of Orange County, July 2000 (Study 183). Memorandum from Johana Walters, Dave
benthic macrofauna species impacted by contaminant mixtures. Environ Toxicol Kim, and Kean S. Goh to John S. Sanders, October 5, 2000. Sacramento, CA: California
Chem 2012;31:2175–88. Department of Pesticide Regulation; 2000b [10 pp.].
Ryberg KR, Vecchia AV, Martin JD, Gilliom RJ. Trends in pesticide concentrations in urban Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
streams in the United States, 1992–2008. Scientific Investigations Report 2010–5139 ing of monthly surface water monitoring for the red imported fire ant project in
2010–5139. Reston, VA: U.S. Geological Survey; 2010. [101 pp.]. Orange County, June 2000 (Study 183). Memorandum from Johana Walters, Dave
Sala S, Migliorati S, Monti GS, Vighi M. SSD-based rating system for the classification of Kim, and Kean S. Goh to John S. Sanders, September 1, 2000. Sacramento, CA: California
pesticide risk on biodiversity. Ecotoxicology 2012;21:1050–62. Department of Pesticide Regulation; 2000c [9 pp.].
Schäfer RB, Pettigrove V, Rose G, Allinson G, Wightwick A, von der Ohe PC, et al. Effects of Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
pesticides monitored with three sampling methods in 24 sites on macroinvertebrates ing of monthly surface water monitoring for the red imported fire ant project in
and microorganisms. Environ Sci Technol 2011a;45:1665–72. Orange County, September 2000 (Study 183). Memorandum from Johana Walters,
Schäfer RB, von der Ohe PC, Kühne R, Schüürmann G, Liess M. Occurrence and toxicity of Dave Kim, and Kean S. Goh to John S. Sanders, December 22, 2000. Sacramento, CA:
331 organic pollutants in large rivers of north Germany over a decade (1994 to 2004). California Department of Pesticide Regulation; 2000d [12 pp.].
Environ Sci Technol 2011b;45:6167–74. Walters J, Kim D, Sava R, Goh KS. Preliminary results of pesticide analysis and acute
Schäfer RB, von der Ohe PC, Rasmussen J, Kefford BJ, Beketov MA, Schulz R, et al. Thresh- toxicity testing of monthly surface water monitoring for the red imported fire ant
olds for the effects of pesticides on invertebrate communities and leaf breakdown in project in Orange County, April 2000 (Study 183). Memorandum from Johana
stream ecosystems. Environ Sci Technol 2012;45:5134–42. Walters, Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, July 6, 2000.
Schäfer RB, Gerner N, Kefford BJ, Rasmussen JJ, Beketov MA, de Zwart D, et al. How to Sacramento, CA: California Department of Pesticide Regulation; 2000e [10 pp.].
characterize chemical exposure to predict ecologic effects on aquatic communities? Walters J, Kim D, Sava R, Goh KS. Preliminary results of pesticide analysis and acute tox-
Environ Sci Technol 2013;47:7996–8004. icity testing of monthly surface water monitoring for the red imported fire ant project
in Orange County, December 1999 (Study 183). Memorandum from Johana Walters, Dave Kim, and Kean S. Goh to John S. Sanders, May 3, 2001. Sacramento, CA: California
Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, April 12, 2000. Sacramento, Department of Pesticide Regulation; 2001c [9 pp.].
CA: California Department of Pesticide Regulation; 2000f [8 pp.]. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity
Walters J, Kim D, Sava R, Goh KS. Preliminary results of pesticide analysis and acute toxic- testing of monthly surface water monitoring for the red imported fire ant project in
ity testing of monthly surface water monitoring for the red imported fire ant project Orange County, January 2001 (Study 183). Memorandum from Johana Walters,
in Orange County, February 2000 (Study 183). Memorandum from Johana Walters, Dave Kim, and Kean S. Goh to John S. Sanders, April 27, 2001. Sacramento, CA: California
Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, May 17, 2000. Sacramento, Department of Pesticide Regulation; 2001d [10 pp.].
CA: California Department of Pesticide Regulation; 2000g [9 pp.]. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
Walters J, Kim D, Sava R, Goh KS. Preliminary results of pesticide analysis and acute tox- ing of monthly surface water monitoring for the red imported fire ant project in
icity testing of monthly surface water monitoring for the red imported fire ant project Orange County, July 2001 (Study 183). Memorandum from Johana Walters, Dave
in Orange County, January 2000 (Study 183). Memorandum from Johana Walters, Kim, and Kean S. Goh to John S. Sanders, November 1, 2001. Sacramento, CA: California
Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, April 7, 2000. Sacramento, Department of Pesticide Regulation; 2001e [9 pp.].
CA: California Department of Pesticide Regulation; 2000h [8 pp.]. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
icity testing of monthly surface water monitoring for the red imported fire ant project Orange County, June 2001 (Study 183). Memorandum from Johana Walters, Dave
in Orange County, March 2000 (Study 183). Memorandum from Johana Walters, Kim, and Kean S. Goh to John S. Sanders, September 7, 2001. Sacramento, CA: California
Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, July 5, 2000. Sacramento, Department of Pesticide Regulation; 2001f [10 pp.].
CA: California Department of Pesticide Regulation; 2000i [10 pp.]. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
icity testing of monthly surface water monitoring for the red imported fire ant project Orange County, March 2001 (Study 183). Memorandum from Johana Walters, Dave
in Orange County, October 1999 (Study 183). Memorandum from Johana Walters, Kim, and Kean S. Goh to John S. Sanders, July 13, 2001. Sacramento, CA: California
Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, February 4, 2000. Sacra- Department of Pesticide Regulation; 2001g [10 pp.].
mento, CA: California Department of Pesticide Regulation; 2000j [8 pp.]. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
icity testing of monthly surface water monitoring for the red imported fire ant project Orange County, May 2001 (Study 183). Memorandum from Johana Walters, Dave
in Orange County, September 1999 (Study 183). Memorandum from Johana Walters, Kim, and Kean S. Goh to John S. Sanders, July 17, 2001. Sacramento, CA: California
Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, January 31, 2000. Sacra- Department of Pesticide Regulation; 2001h [9 pp.].
mento, CA: California Department of Pesticide Regulation; 2000k [8 pp.]. Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test-
Walters J, Kim D, Sava R, Goh KS. Results of pesticide analysis and acute toxicity testing of ing of monthly surface water monitoring for the red imported fire ant project in
rain runoff monitoring for the red imported fire ant project in Orange County, January Orange County, November 2000 (Study 183). Memorandum from Johana Walters,
2000 (Study 183). Memorandum from Johana Walters, Dave Kim, Roger Sava, and Dave Kim, and Kean S. Goh to John S. Sanders, February 6, 2001. Sacramento, CA: California
Kean S. Goh to John S. Sanders, May 30, 2000. Sacramento, CA: California Department Department of Pesticide Regulation; 2001i [10 pp.].
of Pesticide Regulation; 2000l [10 pp.]. Walters J, Kim D, Goh KS. Revised preliminary results of pesticide analysis and acute tox-
Walters J, Kim D, Sava R, Goh KS. Revised-preliminary results of pesticide analysis and icity testing of monthly surface water monitoring for the red imported fire ant project
acute toxicity testing of monthly surface water monitoring for the red imported in Orange County, April 2001 (Study 183). Memorandum from Johana Walters, Dave
fire ant project in Orange County, May 2000 (Study 183). Memorandum from Kim, and Kean S. Goh to John S. Sanders, September 7, 2001. Sacramento, CA: California
Johana Walters, Dave Kim, Roger Sava, and Kean S. Goh to John S. Sanders, Department of Pesticide Regulation; 2001j [8 pp.].
September 29, 2000. Sacramento, CA: California Department of Pesticide Regulation; Warne M. A review of the ecotoxicity of mixtures, approaches to, and recommendations
2000m [9 pp.]. for, their management. In: Langgley A, Gilbey B, Kennedy B, editors. Proceedings of
Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test- the Fifth National Workshop on the Assessment of Site Contamination. Adelaide,
ing of monthly surface water monitoring for the red imported fire ant project in Australia: National Environment Protection Council Service Corp.; 2003.
Orange County, August 2001 (Study 183). Memorandum from Johana Walters, Warne M, Hawker D. The number of components in a mixture determines whether syn-
Dave Kim, and Kean S. Goh to John S. Sanders, November 8, 2001. Sacramento, CA: ergistic and antagonistic or additive toxicity predominate: the funnel hypothesis.
California Department of Pesticide Regulation; 2001a [11 pp.]. Ecotoxicol Environ Saf 1995;31:23–8.
Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test- Whiteside M, Mineau P, Morrison C, Knopper LD. Comparison of a score-based approach
ing of monthly surface water monitoring for the red imported fire ant project in with risk-based ranking of in-use agricultural pesticides in Canada to aquatic recep-
Orange County, December 2000 (Study 183). Memorandum from Johana Walters, tors. Integr Environ Assess Manag 2008;4:215–36.
Dave Kim, and Kean S. Goh to John S. Sanders, April 26, 2001. Sacramento, CA: California Yuan LL, Hawkins CP, Van Sickle J. Effects of regionalization decisions on an O/E index for
Department of Pesticide Regulation; 2001b [10 pp.]. the US national assessment. J N Am Benthol Soc 2008;27:892–905.
Walters J, Kim D, Goh KS. Preliminary results of pesticide analysis and acute toxicity test- Yuan LL, Pollard AI, Carlisle DM. Using propensity scores to estimate the effects of insec-
ing of monthly surface water monitoring for the red imported fire ant project in ticides on stream invertebrates from observational data. Environ Toxicol Chem
Orange County, February 2001 (Study 183). Memorandum from Johana Walters, 2009;28:1518–27.

Science of The Total Environment

Uploaded by

Copyright:

Available Formats

Science of The Total Environment

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Science of The Total Environment

Uploaded by

Copyright:

Available Formats

Science of the Total Environment 476–477 (2014) 144–157

Contents lists available at ScienceDirect

Science of the Total Environment

Pesticide Toxicity Index—A tool for assessing potential toxicity of

• Pesticide Toxicity Index (PTI) is a tool to assess pesticide mixtures.

0048-9697/$ – see front matter. Published by Elsevier B.V.

Bioassay data by taxonomic group Pesticides Pesticide degradates

Fish Cladocerans Benthic invertebrates Fish Cladocerans Benthic invertebrates

No. of total compounds 435 423 205 48 47 15

CA DPR Sacramento River watershed (Wadsworth Canal)

intermediate (CV = 500%) and benthic invertebrates, which is the (A)

Ceriodaphnia dubia survival (percent of control)

resent the relative toxicity of sites, samples, or individual pesticides—

test conditions, from skewing the results. Median-PTI

well the PTI approach predicted the incidence of observed toxicity,

analytical reporting levels, are not considered to represent a robust test

Samples classified correctly by threshold, in percent

may have contributed to the observed toxicity in these samples.

Median-PTI values N1 is more difﬁcult to explain, but this occurred in 20 Median-PTI

You might also like