Full text of DOI:10.1002/etc.5620210928

Environmental Toxicology and Chemistry, Vol. 21, No. 9, pp. 1977–1992, 2002
2002 SETAC
Printed in the USA
0730-7268/02 $9.00 .00
᭧
ϩ
EFFECTS ANALYSIS OF TIME-VARYING OR REPEATED EXPOSURES IN AQUATIC
ECOLOGICAL RISK ASSESSMENT OF AGROCHEMICALS
K
EVIN H. REINERT,*† JEFFREY M. GIDDINGS,‡ and LAURA JUDD§
†Rohm and Haas Company, Toxicology Department, 727 Norristown Road, Spring House, Pennsylvania 19477, USA
‡The Cadmus Group, 177 Spring Street, Marion, Massachusetts 02738, USA
§ERM, Environmental Resources Management, 855 Springdale Drive, Exton, Pennsylvania 19341, USA
(
Received 19 March 2001; Accepted 21 January 2002)
Abstract—Exposure to agrochemicals in the aquatic environment often occurs as time-varying or repeated pulses. Time-varying
exposures may occur due to runoff events and spray drift associated with precipitation and application events. Hydrologic dilution,
dispersion, and degradation also produce pulsed exposures. Standard laboratory toxicity tests using constant exposure concentrations
typically do not investigate the toxicity of time-varying or repeated exposures. Detoxification, elimination, and recovery may occur
within organisms or populations during the periods between exposures. The difficulty of estimating effects of realistic time-varying
exposures from measurements made under constant exposure conditions is often an important source of uncertainty in ecological
risk assessment of pesticides. This article discusses the criteria and tools for deciding whether time-varying exposures are relevant
in a particular risk assessment, approaches for laboratory toxicity testing with time-varying exposure, modeling approaches for
addressing effects of time-varying exposure, deterministic and probabilistic ecological risk characterization of time-varyingexposures
and toxicity, and uncertainty analysis.
Keywords—Aquatic ecological risk assessment
Agrochemicals
Time-varying exposure
Effects analysis
Repeated
exposures
INTRODUCTION
iments (Fig. 1). Such exposures can be pulsed and nonre-
peating or repeating. Effects from the exposure(s) may be cu-
mulative or reversible.
Input to aquatic environments by agrochemicals often oc-
curs in pulses rather than by continuous exposure [1]. For
example, substances released through spills, application of pes-
ticides or agrochemicals, surface runoff, precipitation of air-
borne pollutants, or discharge of industrial effluents result in
maximum concentrations just after reaching a water body.
These concentrations then change rapidly as a result of varying
rates of input and dilution, changes in form, and solubility and
degradation. Standard laboratory toxicity tests utilizing con-
tinuous exposure scenarios typically do not investigate the
toxicity of short-term pulsed or intermittent exposures of sub-
stances to aquatic organisms [2–4]. Some level of detoxifi-
cation and/or elimination (depuration) of the substance during
the substance-free period may reduce the toxic effects of the
earlier exposures and is dependent on the length of time be-
tween pulses (recovery interval, postevent interval).
The value of investigating the relationship between expo-
sure duration and toxicity of compounds is relevant when eval-
uating the potential field effects of agrochemicals. Input to
surface waters from agrochemicals typically occurs in pulses
or time-varying and/or repeated exposures (application inter-
val) due to agricultural runoff [5] and spray drift; continuous
exposure may not provide an appropriate estimate of a toxic
effect, especially for agrochemcials with short half-lives. Ap-
plication often results in a pulse concentration in the environ-
ment, followed by an interval during which concentrations of
the agrochemical may fall to below detectable levels (recovery
interval) prior to the subsequent application due to hydrologic
dilution, runoff events, or degradation in soil, water, and sed-
POTENTIAL CONSEQUENCES OF A TIME-VARYING OR
PULSED EXPOSURE
Time-varying or repeated exposures can have a variety of
consequences. The first pulse may select more hardy, more
robust individuals (individual selection for tolerance), causing
an apparent lessening in toxic response. Induced individual
tolerance from previous pulses may strengthen survivors (mak-
ing them more tolerant [6]) through acclimation, induction of
detoxification or biotransformation enzymes, and so on (e.g.,
induction of cytochrome P450–dependent mono-oxygenases
[7] and mixed-function oxygenases in fish [8,9]). Cumulative
individual effects occur when the first pulse weakens the or-
ganism (making it less tolerant), dissipating energy and low-
ering organism fitness [6], or the substance irreversibly inter-
acts with the receptor [10,11]. Postexposure, latent, or delayed
effects occur after a typical study is terminated. This may be
an important factor in the interpretation of toxicity studies and
in risk assessment for agrochemicals [12]. Ecological recovery
may occur after an exposure that affects a significant propor-
tion of the population. The population (and higher organiza-
tion) may recover adequately during the period of negligible
exposure (recovery interval). Exposure–response reciprocity
may occur; for example, a 2-d exposure at 2
␮
g/L may cause
the same effects as a 1-d exposure at 4
L [13].
␮
g/L or 4 d at 1 g/
␮
Independence or dependence of effects from repeated
exposures
The effects of a first exposure pulse may influence the
biological response to a second pulse. Using time to event
* To whom correspondence may be addressed
(kreinert@rohmhaas.com).
1977

1978
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
but more frequent exposure pulses. Generally for these ag-
rochemicals, the longer the pulse (shorter the recovery), the
closer the toxicity value was to the 96-h lethal concentration
(LC50) from continuous exposure.
In a study by Williams and Holdway [25], the Australian
crimson spotted rainbow fish (Melanotaenia fluviatilis) was
exposed to either 2 h or continuous levels of Cd or Zn. Larval
fish (9–10 d) demonstrated 96-h LC50 values of 0.56 and 0.01
mg/L Cd and 1.57 and 0.27 mg/L of Zn for 2 h and continuous
exposures, respectively. Substantially higher levels were re-
quired to produce an LC50 for the 2 h compared to the 96-h
exposure.
Heming et al. [26] exposed early life stages of rainbow
trout to pulsed and continuous doses of methoxychlor. The
pulsed exposure scenario mimicked environmentally realistic
exposures from the use of the agrochemical to control biting
fly larvae in western Canadian rivers. Survival, growth, and
development of early life stages were assessed for 68 d after
Fig. 1. Schematic showing pulse, time-varying, or repeated-exposure
relationships and terminology.
a pulsed exposure of 580
water after 2 h) and a spiked concentration of 30
design in which the substance degraded over 68 d). The me-
thoxychlor concentration decreased due to fate processes. No
biologically significant effects were observed on any growth
stage for either exposure scenario, although the pulsed con-
␮
g/L (biota returned to clean test
g/L (static
approaches [14], one can determine whether enhanced, re-
duced, and independent effects are produced with pulsed ex-
posures. The value of the effect parameter (e.g., mortality) at
a particular time (e.g., 96 h) after several exposures can be
compared with the value after a single exposure. Again, the
effect may be similar, enhanced, or reduced with respect to a
single exposure. The response may also depict a split probit
[15,16], which is often attributed to different mode of actions
occurring at different times during an exposure (e.g., pulsed
exposure events) or distinct subsets of individuals with dif-
ferent resistance to a stressor (or pulsed exposures in this case).
Note that use of the typical model (lognormal or probit) could
also produce a split probit because of poor data fit [16]; al-
ternative toxicity models should be investigated in such in-
stances. Such a sharp break in the plot may also indicate the
presence of two different substances [15].
␮
centration was
Ͼ10 times the spiked level.
Stuijfzand et al. [27] reported that increasing the exposure
duration to diazinon from 2 to 4 d decreased the survival of
midges (Chironomus riparius) and caddisflies (Hydropshyche
angustipennis) by a factor of 1.4 to 1.6 and 2.2 to 8.4, re-
spectively. In a recent study [28], the effects of fenoxycarb on
growth and reproduction of D. magna were examined using
a single pulse dose. This realistic exposure regime was based
on laboratory fate data and field observations and mimicked
the reduction in fenoxycarb following field application to nat-
ural waters. For both the pulsed and the continuous exposure
studies, reproduction was the most sensitive endpoint. A sub-
stantial reduction in toxicity (maximum acceptable toxicant
Acute versus chronic exposure
concentrations of 26 and 0.0016 ␮g/L from the pulsed and the
continuous exposure studies, respectively) was observed from
exposure to environmentally realistic levels of fenoxycarb.
Intermittent or pulsed exposures most often are associated
with acute responses [17–19]; however, chronic effects may
also be important when considering postexposure toxicity (de-
layed effects) such as effects on growth or when sublethal
responses occur [1,19–21]. However, growth rates may not be
a reliable indicator of delayed effects because reproductive
failure may occur even when poor growth is not observed.
Also, the relationship between cumulative mortality and ex-
posure duration may not be proportional in intermittent ex-
posure events when the peak concentrations are constant [1].
Naddy et al. [22] supports a curvilinear response for pulsed
studies with chlorpyrifos using Daphnia magna. The rela-
tionship between concentration and duration of exposure was
not proportional at low concentrations and longer durations;
once higher concentrations and shorter exposure durations
were reached, the relationship became proportional (i.e., if the
concentration is halved, the exposure duration must double to
produce the same effect).
Effect related to peak concentration, not duration
Curtis et al. (in [3]) reported that intermittent exposure to
fenvalerate was more toxic to fish than continuous exposure
when daily mean concentrations were equal. However, the re-
sults are difficult to interpret because the concentration ad-
ministered intermittently was higher than that administered
continuously to achieve equal daily mean concentrations and
may have resulted in increased toxicity, which would not be
representative of field situations.
In a more environmentally realistic study, Schulz and Liess
[29] exposed caddisfly larvae (Limnephilus lunatus) to three
different equivalent fenvalerate concentrations for 1 or 10 h.
After transfer to a stream microcosm with pesticide-free water,
chronic effects (emergence, dry wt) after 240 d were more
pronounced in equivalent exposures (
to 10 h.
␮g/h) of 1 h compared
Shorter exposure causes less effect
Jarvinen et al. [23] studied the acute toxicity of chlorpyr-
ifos, endrin, and fenvalerate to fathead minnows in both single
pulsed and continuous exposures. Varying combinations of
pulse (1–96 h) and recovery (0–95 h) periods were used, to-
taling 96 h. Also, Naddy and Klaine [24] demonstrated that
chlorpyrifos toxicity to D. magna was reduced using shorter
Latency (delayed effects)
Van der Hoeven and Gerritsen [10] studied the effect of
chlorpyrifos on Daphnia pulex, reporting that the agrochem-
ical immobilized daphnids several days before death. Even
when exposure was discontinued, immobilized D. pulex died,

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1979
further supporting the concept of irreversible effects. However,
in a study by Naddy and Klaine [24], no latent effects were
observed for daphnids that survived initial chlorpyrifos ex-
posure, provided adequate recovery time between exposures
was allowed. Without an adequate recovery period, effects
were observed. In Naddy and Klaine [24], the authors sus-
pected that in the van der Hoeven and Gerritsen [10] study, a
combination of higher exposure concentrations, coupled with
longer exposure periods (less recovery time), allowed the
daphnids to accumulate chlorpyrifos, exceeding their critical
toxicity threshold.
when at least 2 to 6 h of clean water was provided between
doses [35]. Mancini et al. (in [4]) and Clark et al. [36] have
also conducted tests that suggest that multiple-pulse exposures
of substances are less toxic to aquatic organisms than contin-
uous exposures of equal total duration. This suggests that some
level of detoxification and/or elimination of the substance dur-
ing the substance-free period can reduce the toxic effects of
the earlier exposures [3] and is dependent on the length of
time between pulses (Wang and Hanson, as cited in [3]).
Using algae (Pseudokirchneriella subcapitata, formerly
Selenastrum capricornutum) and pulsed exposure to atrazine
from the field, Klaine et al. [37] demonstrated that recovery
Unless study results indicate otherwise, it may be appro-
priate to assume latency. Although latency is a function of the
toxic substance, it is the degree of this latency that is important
in the design and interpretation of the pulsed toxicity study
(K.R. Solomon, 1998, ‘‘Latency of Responses,’’ ECOFRAM,
unpublished data). For example, if latency does not occur,
short-term exposures should be evaluated against toxicity tests
with similar exposure times. In order to demonstrate latency
(or lack of) in acute studies, observation intervals must con-
tinue after the exposure is completed and the organism has
been removed from the stressor. For example, in Hurd et al.
[12], diflubenzuron effects on aquatic macroinvertebrates were
not observed until molting began, some two to four weeks
after a single exposure. Lack of latency in chronic studies can
be determined from observation during exposure, just as in
acute studies. However, since endpoints such as growth are
typically measured only at study termination, one could argue
that such observations are the result of events that occurred
early in the exposure period.
from exposure up to 50
␮g/L of atrazine was nearly instan-
taneous once the herbicide was removed from the overlying
water. Klaine et al. [37] also showed that D. magna survival
after exposure to chlorpyrifos was age related, with older
daphnids being more sensitive. Daphnids exposed to two puls-
es had a higher mortality if the second pulse occurred later in
their life cycle. Naddy and Klaine [24] demonstrated that pro-
vided sufficient time between exposures occurred, no effects
on D. magna were observed. Naddy and Klaine [24] also stated
that
Ն72 h were needed for recovery of D. magna from ex-
posure to chlorpyrifos. In some cases, recovery did not occur,
and the authors hypothesized that the critical exposure thresh-
old for the compound had been exceeded (see later section).
Based on mobility data, it appeared that daphnids could recover
from environmentally realistic chlorpyrifos exposures below
a critical threshold.
Irreversible, additive, or cumulative effects
Often, knowledge about the mode of action of a substance
may be all that is needed to determine whether a substance
has latent effects (e.g., diflubenzuron). Close analogues that
demonstrate the presence or absence of latency can be used
to deduce the likelihood of latency occurring. Most organic
substances display a baseline or narcotic mode of action [30],
which is believed to be a general disruption of membrane
integrity [31]. Narcosis occurs when the compound reaches a
critical threshold or critical body residue (CBR) [32]; such
effects are not latent and are reversible unless death has oc-
curred. This concept can also be applied to agrochemicals
where the mode of action may not be well understood in non-
target organisms. At an ecosystem level, complexity and non-
linear biological dynamics may create a latency period between
the exposure event and effects [33,34]. Where it is possible
to redesign bioassay protocols to determine whether latency
exists, the protocols should be modified as long as the results
are compatible with historical data from similar protocols.
For some compounds, no difference in the toxicity has been
observed between time-varying or repeated and continuous
exposures when these exposures are additive for compounds
with slow depuration times or with only slowly reversible or
irreversible effects (organism memory; Breck, as cited in [3];
van der Hoeven and Gerritsen [10]). In a study by Kallander
et al. [35], no difference in the toxicity of organophosphorus
compounds to midges was observed between pulsed and con-
tinuous exposures, whereas chronic pulsed exposure over 28
d to bromoxynil was reported to be more toxic to D. magna
when compared to 28-d continuous exposure [38]. Reproduc-
tive parameters and growth (weight) were adversely affected
at pulsed exposure levels of 20
uous levels of 40 (sublethal effects) to 80 (survival)
␮
g/L as compared to contin-
g/L.
␮
In an investigation of five insecticides (technical permeth-
rin, microencapsulated permethrin, fenitrothion, carbaryl, and
carbofuran), mosquito larvae were exposed to the same in-
secticide concentrations in the pulsed (two 1-h exposures, sep-
arated from between 6 and 24 h) and continuous (2-h exposure)
tests [3]. Results of this study demonstrated that double-pulse
exposures were equally toxic as or more toxic than a single
continuous exposure of equal total duration.
Microencapsulated permethrin was the only insecticide of
the five tested that exhibited greater toxicity in the double-
pulse exposure test. The investigators speculated that some of
the larvae immobilized after the first 1-h exposure to micro-
encapsulated permethrin recovered between exposures, which
allowed them to ingest more capsules than those exposed con-
tinuously. With respect to the other insecticides tested in dou-
ble-pulse exposures, recovery from immobilization during
clean water periods between exposures did not result in lower
toxicity of the compound. Therefore, either insufficient recov-
ery intervals (duration between events) occurred (6 h for all
Reversible effects with recovery between pulses
If the mode or mechanism of action for the agrochemical
has noncumulative, reversible effects (e.g., triazines) or a pe-
riod of recovery returns the organism to pre-exposure state
(e.g., organophosphates), time-varying exposure testing may
be useful to provide more realistic estimates of the effects
associated with exposure. Also, for rapidly acting compounds
like pyrethroids, maximum or peak concentrations are more
important than the area under the curve (AUC) or cumulative
dose.
Toxicity tests with the midge (Chironomus riparius) to
organophosphorus and carbamate compounds showed that two
1-h pulses caused significantly fewer symptoms of intoxication
than 2 h of continuous exposure to carbamate compounds,

1980
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
except carbaryl, which was tested with a 24-h period) or other
irreversible toxic effects were occurring [3].
tem. The pesticide root zone model (PRZM) [40] coupled with
the exposure analysis modeling system (EXAMS) [41] are
higher-tier (beyond screening-level tools and approaches)
aquatic exposure models. When associated with the multiple
scenario risk assessment tool (MUSCRAT) (G. Mangels and
J.M. Cheplick, 1997, MUSCRAT, Ver 1.0, Beta Ver, unpub-
lished data), numerous agrochemical-specific exposure sce-
narios can be run. Only those scenarios indicating substantial
risk (as defined by a particular regulatory paradigm) will be
subject to additional scrutiny using time-varying or pulsed
exposure testing. A risk assessment tool to evaluate duration
and recovery (RADAR) (M. Williams, 1998, RADAR, Ver
1.10, Waterborne Environmental, Reston, VA, USA, unpub-
lished data), a postprocessor for PRZM/EXAMS, can be used
to address time-varying exposures, providing output from
PRZM/EXAMS corresponding to a particular threshold con-
centration (the concentration where a peak event begins and
ends). The RADAR requires a complete time-series record (a
consistent sampling frequency). The analysis is based on the
units of the data file. If the time series is hourly, durations and
recoveries are expressed in hours. If the time series is daily
(as is the case here), all output is expressed in days. Note that
the user needs to fill in missing data if the time-series data are
not continuous (if sampling frequency is not consistent).
This threshold concentration should be a function of the
toxicity of the compound. For example, the acute threshold
could be 10% of the LC50, LC10, or other acute endpoint and
the no-observed-effect concentration (NOEC) or other chronic
endpoint (e.g., effect concentration [EC10]) for chronic effects.
The slope of the exposure response could also be used to assist
in determining an acceptable threshold. For each 36-year run,
the generated information may appear as event, date, peak
concentration, average (arithmetic mean) concentration, geo-
metric mean concentration, duration (in days), and recovery
period (in days). Recovery refers to the time between events.
For each event, the following information may be generated:
number of peaks, peak height/average concentration, peak
height/geometric mean concentration, and concentration–time
above threshold (in AUC units of concentration–day).
Panter et al. [39] studied plasma vitellogenin levels in fat-
head minnows (Pimephales promelas) in response to pulsed
and continuous exposures of estradiol for 21 or 42 d. Inter-
mittent exposures produced approximately equal vitellogenin
concentrations to those in response to the same continuous
exposures but were significantly higher than continuous ex-
posure to the equivalent time-weighted average concentration.
The intermittent exposure response was greater than expected
from a simple integration of exposure and duration using the
AUC.
Population recovery
In a simulation of single- and multiple-pulsed exposure on
the population dynamics of Gammarus pulex, Kedwards and
Wood (T. Kedwards and S. Wood, 1998, ‘‘The Influence of
Simulated Perturbations on the Population Dynamics of Gam-
marus pulex [L.],’’ SETAC Europe Annual Meeting, Bor-
deaux, France, April 1998, unpublished data) observed pop-
ulation reductions proportional to the lack of immigration of
individuals and to the magnitude of the exposure. The mag-
nitude of effects was inversely related to the age class of the
amphipod. Recovery interval of the population to original lev-
els was also proportional to the immigration numbers and in-
versely to the level of the exposure.
DECISION CRITERIA FOR CONSIDERING PULSED
EXPOSURE ASSESSMENT
The decision to consider pulsed exposure assessment is best
addressed on a case-by-case basis with several aspects that
may move one to consider pulsed testing.
Application interval and half-life
Agrochemicals with short half-lives and those applied with
intervals longer than two to three half lives can easily exhibit
pulsed behavior. A pulsed toxicity study, based on simple cal-
culations of peak height and pulse duration, can be designed
using conservative estimates of the expected environmental
concentration in ponds and other aquatic exposure scenarios.
Numerous questions should be addressed when choosing this
option, such as which chemical half-life (water or soil), drift
versus runoff input (or both), and which specific habitat
(stream, river, pond, lake) should be considered. Lower-order
streams are more susceptible to time-varying or pulsed ex-
posures than higher-order streams and rivers, lakes, or reser-
voirs because their hydrologic dampening is less and these
systems are more directly connected to drift and runoff events
in the watershed. Ponds adjacent to agricultural areas are also
susceptible to pulsed exposures. Each of these parameters may
influence the pulse exposure design.
Hypothetically, if the peak height ratio (either arithmetic
or geometric mean) is similar for all events, then one can
assume that the peak shapes are similar. The concentration
above the threshold is calculated using the AUC (see Fig. 1)
method. These data can easily be processed into cumulative
distributions, subjected to trend analysis, sorted by duration,
recovery times, and so on, using RADAR to provide outputs
of particular pulses (e.g., 90th-percentile pulse length, 10th-
percentile recovery length, or the most likely pulse and re-
covery length). If an unmanageable number of values is gen-
erated (e.g., too many pulses are identified), additional noneco-
toxicologically based thresholds (filters or event trigger values)
may need to be added to create an interpretable data set (e.g.,
modify the ecotoxicologically based threshold by 10 to arti-
ficially reduce the number of events).
Exposure modeling
GENEEC. As currently available, output from GENEEC
(generic environmental exposure concentration model; U.S.
Environmental Protection Agency Office of Pesticide Pro-
grams Environmental Fate and Effects Division) can be used
to indicate potential time-varying behavior. Peak and time-
weighted average concentrations as well as the aquatic half-
life are generated by this model. This information could be
used to develop pulsed testing protocols for particular agro-
chemicals.
Figure 2 shows a modeled agrochemical concentration in
a low-order headwater stream obtained using RADAR (figure
text added). Most events have a duration of 1 d, and all events
have a duration of 2 d or less before returning to below 2 ␮g/
L. This is due to an extremely rapid dissipation rate due to
hydrologic dilution in a fast-moving stream. Note that except
for two events, when the precipitation event lasted for 2 d, the
peak concentration equals the averages (single-day events).
Other years have a similar multiple-pulse pattern. The number
Pesticide root zone model/exposure analysis modeling sys-

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1981
Fig. 2. Example of event threshold of 2 ␮g/L and influence on number
of peaks, event duration, recovery duration, and shape of peak for a
low-order stream in 1948 (e.g., edge of field for a headwater stream)
using pesticide root zone model/exposure analysis modeling system
(PRZM/EXAMS) with risk assessment tool to evaluate duration and
recovery (RADAR) postprocessing.
of peaks decreases as the threshold increases (Fig. 3a). Because
each event contains only one peak, the maximum number of
peaks occurs at the lowest threshold concentration.
All the events are at least 1 d in length with only 20% at
2 d and longer (Fig. 3b). Peak concentrations with
p values
of 0.2 to 1.0 range from 7.5 to 3 g/L (Fig. 3c). Most events
␮
have recovery intervals of 1 to 7 d (Fig. 3d). From these data,
a laboratory study representing these exposure parameters
could be designed using a 1-d pulse at 5
␮g/L with a recovery
interval of 4 d. Other scenarios are also possible if alternative
p
values are chosen from the probability distributions.
Similar information for a higher-order (lower-reach) stream
(river) and a pond are shown in Figures 4 and 5. The same
input to PRZM/EXAMS is used for these scenarios as used
for the lower-order stream with the exception that the char-
acteristics of the receiving water body are changed. Different
event threshold levels (1 and 2 ␮g/L) and the effect on number
and shape of the peaks and associated parameters are described
for the higher-order stream (Fig. 4). The spikiness in the graph-
ical outputs is an artifact of model operation and output.
As in the headwater stream, the number of events versus
threshold decreases for both the higher-order stream and the
pond as the threshold increases. Fewer, broader peaks are ob-
served when compared to the headwater stream in Figures 4
and 5. Hydrologic dilution produces peaks of lower magnitude
in the higher-order stream as compared to the headwater stream
and similar concentrations to the pond. Pulse durations in the
pond are longer than both stream scenarios because of the lack
of flow in the pond system. Recovery intervals are similar
between the stream scenarios (subject to similar runoff re-
gimes); however, the pond recovery intervals can be quite long
due to the lack of flow in and out of the pond. One can conclude
that pulsed behavior is particularly relevant in the headwater
stream and less important in the higher-order stream and pond.
Postprocessor (RADAR) enhanced PRZM/EXAMS outputs al-
low one to observe whether the scenario produces potential
time-varying responses. Using the tools described previously,
decisions can be made whether additional efforts to address
time-varying exposure and effects are warranted. The follow-
Fig. 3. (a) Influence of event threshold on event parameter distribution
for a low-order stream (results from Fig. 2) using risk assessment tool
to evaluate duration and recovery (RADAR) threshold values ranging
from 0.1 to 10 ␮g/L. (b) Probability of duration for the same scenario.
(c) Probability of peak maximum for the same scenario. (d) Probability
of recovery over 36 years for a low-order stream with a threshold at
2
␮
g/L using pesticide root zone model/exposure analysis modeling
system (PRZM/EXAMS) with RADAR postprocessing.

1982
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
Fig. 6. Decision matrix for time-varying exposures. RADAR
assessment tool to evaluate duration and recovery; TWA
weighted average; AUC area under the curve; NOEC
served-effect concentration; EC effective concentration.
ϭ risk
ϭ
time-
Fig. 4. Examples of event threshold values (1 [top] and 2 ␮g/L [bot-
ϭ
ϭ
no-ob-
tom]) from a single year and influence on number of peaks, event
duration, recovery duration, and shape of peak for a large stream
(lower reach) using pesticide root zone model/exposure analysis mod-
eling system (PRZM/EXAMS) with risk assessment tool to evaluate
duration and recovery (RADAR) postprocessing.
ϭ
single-pulse or multiple-pulse laboratory study could be run;
if the pulses cannot be adequately or appropriately represented
in the laboratory, pulse toxicity modeling could be conducted.
Acute and/or chronic studies can be designed, producing mea-
surement endpoints appropriate to the risk assessment end-
points. If expected, studies should be extended to include de-
layed or latent effects. Chronic studies could address either
individual [28] or population endpoints (T. Kedwards and S.
Wood, 1998, ‘‘The Influence of Simulated Perturbations on
the Population Dynamics of Gammarus pulex [L.],’’ SETAC
Europe Annual Meeting, Bordeaux, France, unpublished data).
Note that acute studies under a multiple-pulse regime are not
particularly relevant because short multiple-pulse exposures
approach continuous exposures unless a rapid loss of the sub-
stance occurs in the environment and test system between
pulses.
ing section describes the next steps to move to a more detailed,
higher tier or level of risk assessment.
SELECTING TOOLS FOR INVESTIGATING TIME-
VARYING AND REPEATED EXPOSURE
Figure 6 provides guidance on which direction to take when
assessing the effects of time-varying exposures on ecological
receptors. Depending on the time-varying exposure pattern, a
For the single-pulse case, a joint probability distribution
can be used to characterize the risks (Fig. 6). In the multiple-
exposure case, the pulses are assessed for their independence
in order to determine whether the responses are enhanced,
reduced, or independent for multiple pulses. An understanding
of the independence (or lack of) of the exposures is helpful
in characterizing the risks using population analysis or joint
probability distribution analysis.
Fig. 5. Number of pulses, event duration, and recovery duration for
a pond with a 2-␮g/L threshold (events in 1956) using pesticide root
Laboratory approaches
zone model/exposure analysis modeling system (PRZM/EXAMS)
with risk assessment tool to evaluate duration and recovery (RADAR)
postprocessing.
Pulsed exposure toxicity studies can easily be conducted
using current study designs provided the exposure profile is

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1983
known and the endpoint is defined to provide toxicologically
relevant estimates [1]. Interpretation of the exposure concen-
trations may be problematic, as one could use the AUC, de-
termine arithmetic or geometric averages, and so on (e.g.,
where on the concentration curve should the concentration be
set—arithmetic mean, geometric mean, 90th percentile, and so
on?). Consultation with the appropriate regulatory authority is
recommended when developing pulsed toxicity test designs
because these studies are not yet standardized. Also, study
designs need to be determined to test the appropriate hypoth-
eses.
Single pulses based on maximum concentration or pulse
duration (width of pulse) can easily be designed in the labo-
ratory. Designs based on degradation or hydraulic parameters
are more difficult to test in the laboratory. Flow rates could
be altered in a flow-through study, semistatic- or static-renewal
rates could be controlled to mimic these pulses, or static de-
signs with periodic additions of agrochemical could be used
if the compound degraded rapidly. An understanding of the
behavior of the substance in the test system and under envi-
ronmentally realistic conditions prior to study conduct is im-
portant.
Static exposure. Some of the processes that can affect the
concentration of a agrochemical in the water column, espe-
cially volatilization, hydrolysis, photolysis, and biodegrada-
tion, may also take place in the exposure medium of a toxicity
test. The exposure concentration of such a compound will
decline during the course of a static toxicity test. Moreover,
the test organisms will be exposed to the agrochemical deg-
radation products. If the rate of decline in the concentration
is environmentally realistic, the results of a static test may be
more relevant to the risk assessment than other designs; how-
ever, if rates in the exposure solution are very different from
those in the environment, the results may be meaningless.
Static-renewal exposure. Virtually any temporal exposure
pattern may be created in a static-renewal toxicity test, pro-
vided the renewals are frequent enough. In standard practice,
test solutions are renewed daily, but exposures shorter than 24
h are certainly feasible. Static-renewal exposures would be
especially suitable for generating square pulses (exposure to
a constant concentration for a set duration, followed by transfer
to clean water) with relatively stable test substances—a situ-
ation most relevant when real-world exposure is expected to
be dominated by water flow, such as in small streams. Static-
renewal designs may also be used by laboratories not equipped
to run flow-through studies.
Flow-through exposure. Flow-through exposure systems
(e.g., diluters) are typically used to create constant exposure
concentrations in toxicity tests (especially for substances that
have short half-lives), but they can be adapted to simulate
pulsed exposures as well. Changes in exposure concentration
can be brought about by changing the concentration of the
stock solution, by changing the stock solution pumping rate,
or both. The resulting changes in the exposure solution may
be gradual (depending on the water replacement rate in the
exposure container), not immediate, as in a static-renewal test.
Microcosms (and mesocosms). Microcosms simulating nat-
ural dissipation processes can be used to generate exposure
solutions for static, static-renewal, or flow-through toxicity
tests. For example, the agrochemical can be applied to an
outdoor tank containing natural sediment and natural water in
which it will be subjected to the influences of volatilization,
photolysis, hydrolysis, biodegradation, and sorption, all under
conditions more realistic than laboratory single-species stud-
ies. Water can be taken from the microcosm, periodically or
continuously, for use in toxicity tests. Alternatively, organisms
can be exposed in cages or other containers placed directly in
the microcosm.
The design of the pulsed study (pulse magnitude, duration,
frequency) could be based on the application interval and half-
life, the GENEEC output, or PRZM/EXAMS–RADAR out-
puts. These studies can be conducted when pulses in exposure
are either expected or observed. The results would be used to
characterize the exposure–response relationship in terms of
pulse height (maximum concentration), pulse width (duration),
frequency, or other exposure metrics.
Time-to-event analysis
If the maximum toxic response occurs primarily in a period
shorter than the duration of the typical toxicity study (96 h
for fish, 48 h for daphnids), time-to-event (TTE) analysis may
be useful for designing and interpreting pulsed toxicity studies.
Exposure duration, a critical determinant of exposure conse-
quences, is explicitly included in TTE, affording possibly more
accurate prediction of effects for different exposure durations.
Using conventional testing methods, TTE methods draw from
an experimental design in which groups of individual organ-
isms are monitored through time (M.C. Newman, 1997, ‘‘Not-
ing Time-to-Event Enhances Prediction from Toxicity Test
Data,’’ ECOFRAM, unpublished data). The times (specific
time or interval) until some event occurs are recorded for each
individual. Time-to-event analysis may provide more ecolog-
ically meaningful estimates of lethal (and sublethal) effects
than exposure–response methods alone [42]. Such TTE anal-
yses include both exposure intensity and duration as well as
other covariates [14], allowing for design of a study to spe-
cifically test pulse durations and intensities that may alter the
response of an organism to a more realistic exposure regime.
Results from a TTE analysis could also be used to estimate
time-varying exposure model parameters discussed later in this
article or even to project possible consequences in the field.
Time-to-event analysis may assist in determining whether the
events are independent, antagonistic, additive, or synergistic.
If independent, time-varying exposure testing would not be
needed, as each pulse event occurs as a singular event, without
organism memory or effect overlap. Slopes can be compared
or the effect endpoint at time t can be compared using a chi-
square analysis. Time-to-event analysis can also be used to
model the proportion of the population affected by an expo-
sure, based on
log TTE
ϭ
ϭ
a
ϩ
b
(log concn.) ϩ ␴
(
W
)
or
log TTE
Ϫ
a
Ϫ
␴
b(log concn.)
W
where
reflecting the proportion affected (e.g.,
[0.5] would be 0 for the lognormal model).
W
is a characteristic of the specific underlying model
W
for 50% affected
As a closely related alternative, Mancini [43] proposed a
method to use data from a classical (constant exposure con-
centration) toxicity study to calculate the effects from time-
variable exposures. This approach assumes reversibility of ef-
fects and can be used for both acute and chronic toxicity stud-
ies. The method is able to calculate the probable mortality and/
or the percentage of mortality expected due to any temporally

1984
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
variable substance level. See Mancini [43] for details on the
approach.
Time-varying toxicokinetic models
Pulsed exposure toxicity tests can be coupled with toxi-
cokinetic modeling as a solution to the problems associated
with predicting substance effects in non-steady-state field ex-
posures. Such models may be useful for long recovery inter-
vals, long pulses, or numerous repeated exposures (exposure
scenarios not easily duplicated in the lab). Toxicokinetic mod-
els may be compartment, physiological, or bioenergetic based.
These models can be useful tools in estimating changes in
tissue concentrations resulting from absorption, distribution,
metabolism, and elimination of a substance from an organism.
Compartment-based models describe substance movement
among compartments. For example, a simple two-compart-
ment model may contain water (source) and organism (sink)
compartments. Physiological-based models describe the ac-
cumulation and internal distribution of substances among mul-
tiple tissues. They can account for different rates of elimination
from various tissues but can require a great deal of detailed
physiological data, some of which may not be available. Bio-
energetic-based models describe substance accumulation and
loss in terms of the organism’s energy requirements where the
organism is treated as a single compartment and substance
uptake is a function of the flux of water across the gills or
food/sediment through the gut. These models become partic-
ularly important in estimating the effect of a substance when
evaluating non-steady-state, nonequilibrium exposures that
may vary temporally or spatially. Understanding substance
accumulation and distribution in an organism ultimately con-
tributes to predicting its toxic effect [44].
Fig. 7. Output from the food and gill exchange of toxic substances
model (FGETS)—aqueous and fish concentrations.
(bioaccumulation from water and food) (Fig. 7). An important
aspect of FGETS is that it is a chemical–biological mechanistic
model in which uptake is not considered an arbitrary action.
The model can calculate the time to reach lethality in fish
assuming that the substance has a narcotic mode of action.
The FGETS model can be run in three modes: laboratory, food
chain (simulates one or two fish) and food web (simulates
more realistic predator–prey interactions). Inputs for the three
modes and various model assumptions are found in the Ap-
pendix. The FGETS model incorporates data on the compo-
sition, structure, and morphology of the organism in calcu-
lating uptakes. An important aspect to this is tracking indi-
vidual organism weight as a dynamic variable. Fundamental
to the formulation of a dynamic energetic-based model is the
relationship between the type of chemical, its mode of action
or partitioning capability, and the physiological and genetic
characteristics of the individual [46].
To be able to define toxic effects based on body residue
levels would also eliminate uncertainties associated with the
bioavailability of the substance in question. While a number
of poorly defined species-specific and site-specific variables
still affect the refinement of toxicokinetic models, they are
gaining more widespread use in ecotoxicological evaluations
[44].
In this article, three relatively simple toxicokinetic models
are described: the food and gill exchange of toxic substances
model (FGETS), PULSETOX (as defined in the following),
and the dynamic energy budget toxicology model (DEBtox).
Their use in pulsed exposure testing and assessment is ad-
dressed. The decision as to which of these kinetic models is
most appropriate will depend on data availability; however,
use of the simplest model that will adequately address the study
question will minimize potential errors [44].
Pulsed exposure toxicity tests can be coupled with toxi-
cokinetic modeling as a solution to the problems associated
with predicting substance effects in non-steady-state field ex-
posures. Such models may be useful for long recovery inter-
vals, long pulses, or numerous repeated exposures (difficult to
test the exposure scenario in the lab due to cost and logistics).
The results of these models will be useful in a weight-of-
evidence approach to help understand the compound dynamics
in time-varying exposures and the potential for effects.
FGETS. This model is an example of a bioenergetic-based
toxicokinetic model that addresses the chemical uptake in fish
from both food and environmental pathways [45] using ther-
modynamic potential. It models the chemical exchange be-
tween fish and the aqueous environment that occurs across gill
membranes and across gut walls from ingestion of substances
A similar model, modified from FGETS, was developed for
Daphnia spp. [47]. Acute effects (mortality) of a lipophilic
narcotic on a dynamic daphnid population are modeled, using
uptake from both water and food. The model is based on the
Lassiter and Hallam [48] survival of the fattest static theory
in which the effect of a toxic exposure is analyzed by relating
the Kow to the partition coefficient of the fat (lipid) and aqueous
phases in the aquatic organism. For exposure of equal chemical
activity, increasing lipid content and hydrophobicity increase
the exposure duration tolerance without an effect [48]. A dy-
namic approach was developed in order to couple the dynamic
behavior of individuals with chronic or multiple acute sub-
stance (e.g., pulsed) exposures. The authors note that ecolog-
ical risk assessment needs to be based not only on the attributes
of the substance but also on the biology of the exposed or-
ganism and its population dynamics.
PULSETOX. Another model, PULSETOX, is a residue-
based, pulse exposure toxicokinetic model for aquatic toxi-
cology based on a simple one-compartment first-order kinetics
(1CFOK) equation [2,44]. This model tracks the whole-body
accumulation of a chemical in fish and predicts acute toxicity
using previously established relationships among aqueous con-
centrations, whole-body residues, and lethality using the non-
polar narcosis CBR theory (Fig. 8). This simple toxicokinetic

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1985
proach for predicting effects of pulsed exposures and provided
a clear link between pharmacokinetics of the chemical and the
time course of toxicity both within standard continuous tox-
icity tests and in independent pulse toxicity tests. Such models
may not be useful for chemicals that are highly degradable,
are reactive, or do not readily bioconcentrate. Although the
model is thought to be less effective for substances that cause
cumulative damage where toxicity could increase through time
even though peak fish residues do not change with successive
pulses, Meyer et al. (in [2]) have modified the toxicokinetic
model by the addition of a power term that was effective in
describing the intermittent exposure toxicity with this type of
substance.
DEBtox. Another model that may be used for pulse toxicity
results and body residue analysis is the dynamic energy budget
toxicology model (DEBtox), an energetics-based model [51].
The DEBtox (Ver 1.0) software package is designed to analyze
results from standard aquatic toxicity tests. Results from tests
on acute and chronic survival studies, fish early life stage
studies, D. magna 21-d reproduction studies, and algal growth
inhibition can be input. Outputs from the program are estimates
of study parameters with standard deviations and correlation
coefficients using different models for differing modes of ac-
tion; goodness of fit for the selected model with graphical
presentation; graphic presentation of the likelihood function
of the no-effect concentration with confidence intervals; cal-
culation of time-, concentration-, and/or response-dependent
EC values; statistical analysis of any single or combination of
parameters based on the likelihood ratio test; and an analysis
of residuals, including plots as functions of concentration, ex-
posure time, or response.
The DEBtox model may prove useful for analysis of data
generated from time-varying or repeated-exposure studies. The
program estimates primarily parameter values by maximizing
a nonlinear likelihood function for a survival experiment com-
bined with a weighted least-squares method. For growth and
reproduction, a set of differential equations at initial conditions
is solved using a fourth-order Runge Kutta numerical method.
An elimination rate is essential to this analysis and dictates
how fast a response occurs during exposure. In addition to
various tabular outputs, graphs of time profiles, concentration
profiles, the likelihood function for the no-effect concentration,
and various response surfaces (e.g., effect surface) are pro-
duced.
Fig. 8. Pulsed exposure toxicity test (PULSETOX) output—aqueous
and fish concentrations.
model accounts for the effect of non-steady-state, nonequilib-
rium accumulation from temporally varying exposures on tox-
icity using a fugacity approach. The model can be run in two
modes: repeating exposure and cyclic or random exposure (see
Appendix). The only loss mechanism from the system is via
tank outflow; volatilization, degradation, and so on are not
considered.
A pulse LC50 defines the duration and magnitude of the
exposure required for an organism to accumulate a dose equiv-
alent to the lethal or critical body residue. The CBR is the
minimum tissue concentration associated with an adverse ef-
fect. The concentration can be based on the whole organism
or on a particular target organ and can provide a more direct
measure of a predicted adverse effect than can external ex-
posure concentrations [49]. Such an approach may be an in-
appropriate indicator if the whole-body residue is not repre-
sentative of accumulation in specific target tissues [1]. Hickie
et al. [2] demonstrated that the toxicity resulting from pulsed
exposures is controlled primarily by the rate of substance ac-
cumulation and depuration rate in the exposed fish and that
the level of biological response is associated more with the
accumulated dose than exposure concentration. Because the
model is mechanistically based, it is useful in illustrating the
effects on toxicity resulting from interactions among the num-
ber, duration, and frequency of pulses. Although inadequate
information exists relating biological responses with CBRs,
the model assumes that the CBR is an acceptable, but not
ideal, surrogate to the lethal dose and that residues predicted
by PULSETOX can be interpreted in relation to lethal effects.
The CBR-based techniques are useful to see if a particular
exposure or series of exposures approach the residue level
where effects may be realized without having to test an or-
ganism in difficult laboratory study designs or prior to begin-
ning laboratory testing programs.
Use of time-varying models
Models such as PULSETOX can assist in the interpretation
of time-varying or repeated-exposure studies or to estimate
effects from long pulses, long recovery intervals, or numerous
repeated exposures that would be difficult to test in the lab-
oratory. The results from this modeling are not used directly
in risk characterization but are useful in understanding the
dynamics of the compound in environmental receptors. For
agrochemicals with log Kow at or above 3, results from a fish
bioconcentration study will usually be available. A biocon-
centration factor, often at 1 or 2 concentrations, a Ku and Kd
(uptake and depuration rate constants, respectively) will be
available from this study. Using these inputs, as well as others
for PULSETOX, the time to reach a lethal body residue can
be estimated using default CBRs [11,44]. Estimates of body
residues and/or time to lethality can be estimated for other
exposure concentrations or pulse durations. If a compound has
a log Kow less than 3, FGETS could be used to determine the
Klee [50] presented a successful validation trial of the basic
modeling algorithm used in PULSETOX, the 1CFOK model
[44]; the uptake, depuration, and the bioconcentration factor
(BCF) of chlorinated benzenes and phenols in rainbow trout
were adequately described by the model when compared to
laboratory measurements using predicted versus measured
concentrations.
The PULSETOX model provided a useful deterministic ap-

1986
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
depuration rate from the uptake rate (slope of the uptake curve)
and the BCF (calculated from Kow or water solubility). Ac-
Physiologically based pharmacokinetic or toxicokinetic
models
cording to Spacie et al. [52], the BCF
ϭ Ku/Kd.
Physiologically based pharmacokinetic or toxicokinetic
Besides measuring the BCF for compounds, numerous
equations are available that can be used [52–54] to calculate
the BCF. Such quantitative structure–activity relationship
(QSAR) equations can be based on many or specific classes
of chemicals and are based on one or more measured param-
models would be employed only at higher levels of risk as-
sessment refinement (higher tiers) due to costs and complexity
and typically only for specialty products or special cases of
need. Considering all the information presented, significant
factors need to be considered that greatly affect the interpre-
tation of substance residues in wildlife tissues and highlight
the important role of physiologically based toxicokinetic mod-
els to elucidate the intraspecies and interspecies differences
affecting the development of CBRs. The coupling of these
more complex kinetic models (more compartments, more link-
ages, real model of an organism) with CBRs allows for the
estimation of when a chronic or acute toxic response is ex-
pected to occur under various exposure scenarios. This type
of modeling also predicts the time course for the toxic effect.
Such methods should be able to elucidate the intraspecies and
interspecies differences affecting the development of CBRs,
making them useful for extrapolating responses to other spe-
cies. Besides being an effective tool for predicting acute tox-
icity from pulse exposures, the concept of CBRs and toxicok-
inetics could be applied to other cases where exposure–re-
sponse relationships are complex, for instance, exposure via
sediments, diet, and for chemical mixtures [44,55].
As with any model, the user needs to have a good under-
standing of the assumptions of the particular toxicokinetic
model being used. For instance, many models will assume
steady-state conditions whereby all organisms are assumed to
internally act on a substance in the same way when exposed
to the same source of chemical activity. When these assump-
tions are clearly not appropriate, more complex toxicokinetic
modeling should be investigated [56]. Pharmacokinetic models
in general could be improved by a better understanding of
CBRs, true dose–response distributions, and the link between
CBRs and modifying factors such as fat content, which may
alter the toxicokinetics and toxicodynamics of a substance [2].
Pharmacokinetic modeling is the process of developing
mathematical descriptions of absorption, distribution, metab-
olism, and excretion of chemicals in biota [57]. Such models
can be used for interpolation but should not be used for ex-
trapolation outside the dose ranges, routes of exposure, and
species used to develop the model without an explicit under-
standing of the model constraints. Models have been developed
to describe the disposition of more than 100 chemicals in
mammalian species, and although more limited in scope than
mammals, models for nonmammalian vertebrate species such
as fish have also been developed [55,58–63; K.J. Clark, W.L.
Hayton, W.H. Gingerich, and G.R. Stehly, ‘‘Pharmacokinetics,
Metabolism and a PBPK Model of Benzocaine in Channel
Catfish,’’ SETAC Annual Meeting, November 1997, San Fran-
cisco, CA, USA, unpublished data] and aquatic invertebrates
[47].
eters (e.g.,
Kow, water solubility). Several equations for cal-
culating fish BCF values are listed by the European Center for
Ecotoxicology and Toxicology of Chemicals (ECETOC) [53]
based on general chemicals (see ECETOC document for spe-
cific references) using either log Kow or water solubility (e.g.,
BCF ϭ Ϫ1.32
ϩ log Kow and BCF ϭ 2.02 Ϫ 0.47 log WS
[water solubility in mg/L], respectively).
The correlation coefficients for the equations based on wa-
ter solubility (0.7–0.87) are lower than those based on log Kow
(0.93–0.97). Additional equations, some specific for a partic-
ular class of compounds and others that use parameters such
as molecular connectivity indices, may also be found in the
previously cited references.
According to ECETOC [53], uptake from food is important
only if the log
K
ow Ͼ4.5 and the concentration in food is about
ow Ͻ4,
105 greater than water. Also, for compounds with log
K
the typical time to equilibrium is less than 28 d. Therefore,
studies in which the uptake phase is conducted for 28 d (e.g.,
Federal Insecticide Fungicide and Rodenticide Act [FIFRA]—
Office of Pollution Prevention and Toxic Substances [OPPTS]
850.1730, Organization for Economic Cooperation and De-
velopment 305) can produce meaningful steady-state BCF val-
ues for compounds between log Kow of 3 and 4. Above log
Kow of 4, BCF values may be underestimated because equi-
librium may not have been achieved. Additionally, uptake of
compounds with molecular weights greater than 700 g/mole
may not be predicted by log Kow because of steric hindrances
[53].
As an alternative to calculating the CBR, the whole-body
concentration may be measured during a typical toxicity study
consistent with the time of death (or other effect). This is
obviously a more costly method than using empirically derived
ranges, calculating the value using TTE and FGETS, or other
methods mentioned previously. Although the true CBR for a
particular effect would be measured, the cost and potential
time lag between death and analysis of the body residue may
be problematic.
Additional considerations exist when working with CBRs.
The CBR may be based on a threshold (instantaneous) or cu-
mulative body residue (refer to [11]). A threshold is probably
the typical response since many compounds are considered
nonpolar narcotics. In these cases, the arithmetic or geometric
mean could be used to represent both the exposure concen-
tration and the body residue. For cumulative CBRs, derived
from reactive, irreversible, or slowly reversible target inter-
actions, the body residue should be determined using AUC or
critical area under the curve. Verhaar et al. [11] have found
this parameter to be constant and independent of exposure time
for particular single species–compound interactions. For re-
ceptor-mediated or reactive toxicity, LC50 versus time values
decrease after achieving steady-state bioconcentration, and the
incipient LC50 will be substantially lower than the 96-h LC50
for many compounds.
These higher-level, more detailed models require more spe-
cies-specific or habitat (environment)-specific information and
assumptions; because they are data intensive, the amount of
parameterization needed may be prohibitively costly for most
applications, and the cumulative uncertainty in these models
can be large if parameters are poorly specified or understood
[64]. These models must be interpretable in terms of physi-
cochemical, biochemical, and physiological properties of the
organism. They still represent significant simplifications of the

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1987
true complexities of biological systems. Such detailed models
are often simplified to one- or two-compartment models as
described previously in order to reduce data and parameteri-
zation needs and possibly increase their utility [58].
narcotics is about 10% of that required to elicit an acute re-
sponse [44].
This approach also appears to be useful with sediment-
bound narcotic chemicals. Driscoll and Landrum [70] exposed
Hyalella azteca and Diporeia spp. to varying levels of fluor-
anthene in sediments. Rough estimates of the CBR for H.
The principal application of PBPK models is the prediction
of a dose to a target tissue, a body residue for a parent chemical
or reactive metabolite, or chemical concentration time course
for specific tissues and organs [57,58]. The PBPK models can
help reduce uncertainty associated with conventional dose ex-
trapolation, dose–effect relationships, and CBR estimation
methods. However, in many cases, it might be more feasible
and cost effective to conduct a microcosm or mesocosm ex-
periment designed with specific effects endpoints or to monitor
associated effects in the aquatic environment.
azteca after only 10-d exposures were 3.6 to 5.6 and 1 ␮mol/
g after 30 d. Because H. azteca is known to metabolize fluor-
anthene to more polar and possibly more toxic metabolites,
fluoranthene was not acting solely as a narcotic compound in
this study. For Diporeia, however, body residues of 2.7 to 6.5
␮
mol/g after 30-d exposure resulted in very little toxicity.
Diporeia is known not to be able to readily metabolize fluor-
anthene and has a higher lipid content (which facilitates se-
questration of nonpolar narcotics) than H. azteca, resulting in
higher body residues and negligible toxicity in Diporeia
.
Interpreting tissue concentrations of substances in aquatic
organisms
One method for determining CBRs for substances in aquatic
organisms assumes that the LC50 multiplied by the biocon-
centration factor represents the lethal or effect concentration
in tissue (L. S. McCarty, as cited in [53,71]). Actual body
residues are expected to be lower due to either non-steady-
state conditions or incomplete transfer of the substance from
the water column into the organism. Also, according to McKim
and Schmeider [72], BCF values are considered at steady state,
a condition that may not be reached in toxicity tests of short
A number of difficulties exist in the interpretation of sub-
stance residues in wildlife tissues. Toxicity due to the body
residue of a contaminant will depend on the age, sex, fat con-
tent, and other variables of an individual within a species.
Interspecies differences are also expected to be large. For in-
stance, many toxicokinetic models assume steady-state con-
ditions whereby all organisms are assumed to internally act
on a substance in the same way when exposed to the same
source of chemical activity [56].
duration with chemicals having log Kow Ͼ 3. Note that a recent
publication [73] calculates the CBR for marine organisms us-
ing the NOEC from an acute study and not the LC50.
Van Loon et al. [65] addresses the problem associated with
threshold levels for individual substances through the identi-
fication of CBRs for a chemical class of compounds. It is
generally thought that narcosis-type substances are completely
concentration additive and are intrinsically all equally toxic
[65]. Therefore, body residue levels that would cause a certain
effect would be the same for all substances within this toxi-
cological class (toxic equivalence factor of 1.0). The differ-
ences in aqueous effect concentrations within this class are
related to differences in bioconcentration factors. As reported
in Van Loon et al. [65], no-effect CBRs (mmol/kg lipid) for
baseline toxicity are mortality (fish), 25; sublethal effects
(fish), 5.0; and, estimated ecosystem no-effect level, 0.25.
Total body residues may be quantified by measuring the
total molar concentrations of chemical mixtures in one mea-
surement. An advantage of working with CBRs is that for
chemicals and chemical mixtures with only baseline toxicity,
the CBR is relatively constant for a certain endpoint [65].
Landrum et al. [44] also investigated CBRs for the narcosis-
type class of compounds (also cited in [66]). Their findings
indicated that residues that yield 50% mortality for acute ex-
posures range from 2 to 6 mmol/kg for small fish and inver-
tebrates to a wide range of neutral narcotics. If a tissue con-
centration required to produce 50% acute mortality is below
0.5 mmol/kg, the substance acts by a specific mode of action
that is indeterminate between 0.5 and 2 mmol/kg; however,
recent data with halobenzenes and fish indicate that CBRs are
not constant and decrease with greater exposure time [67,68].
Likewise, Deneer et al. [69] found that the CBRs in the guppy
for 13 organophosphorus insecticides differed widely, partially
due to hydrophobicity. These results make the sole use of
CBRs more complicated for compounds with specific modes
of action and cannot replace the LC50 in such instances. Res-
idue concentrations needed to elicit chronic effects are much
lower than those needed for acute mortality. For 50% mortality,
the residue concentration for chronic exposure to nonpolar
McCarty et al. [74] investigated the estimation of CBR
values in fathead minnows for the polar narcotic compounds,
substituted phenols. The ability to estimate CBRs from the
experimental data was affected by pH-dependent ionization.
When ionization-corrected toxicity data were assessed, several
categories of CBRs were observed that were apparently related
to different modes of action and generally fit the CBR clas-
sification schemes presented earlier in this article. In addition,
Vaes et al. [75] reports that CBRs and LC50s for nonpolar
(narcosis I) and polar narcotic (narcosis II) chemicals are dif-
ferent with the higher toxicity of the polar narcotic chemicals
often related to the polar moeity on the molecule. However,
these researchers have found that membrane–water partition
coefficients (based on dimyristoyl phosphatidyl-choline) are
higher for polar narcotics and generally are better able to pre-
dict the LC50 or CBR for these chemicals than using the log
K
ow. Differences in toxicity between the compounds can be
explained by different membrane lipid partitioning.
Although additional analytical measurement is required, es-
timated CBRs can be compared with tissue concentrations
from laboratory studies to verify these data. Critical body res-
idues may also be compared to tissue levels from field-col-
lected specimens in monitoring programs or incidents. Critical
body residues can be useful components in ecological risk
assessment [49] when used in a weight-of-evidence approach.
Population analysis
Life history or age-stage models can be used to design
pulsed toxicity tests a priori. In screening or initial levels of
ecological risk assessment, generic life history tables for sev-
eral life history strategies in aquatic invertebrates and fish
could be developed. For each of these strategies, a particular
period of growth suppression and/or lethality and its corre-
sponding impact on population dynamics can be described. A
pulsed toxicity study could be conducted for a particular ex-

1988
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
posure period or pulse that causes an ecologically significant
effect on a population. Age-specific mortality or reproduction
information can be used to determine the time required for
recovery of a particular population assuming that delayed ef-
fects do not occur and the effects are reversible and not cu-
mulative. These analyses can address questions such as, Are
single- or multiple-pulsed studies relevant, and are recovery
intervals sufficient to allow biological recovery? For example,
if the recovery period is not sufficient, a time-varying study
may be warranted. Such a study could be conducted regardless
of the application interval and half-life of a compound or the
output from specific exposure models.
These models may be used a posteriori to address whether
the results from a pulsed toxicity study could be ecologically
significant, that is, cause ecologically meaningful changes in
population numbers. Information on the recovery times could
be integrated with population analysis results, indicating re-
quired recovery intervals for particular populations and per-
centage reductions in some ecological parameter. They could
also be used to project and interpret responses or results for
other populations.
USE OF TIME-VARYING RESULTS IN RISK
CHARACTERIZATION
Figure 6 describes the risk characterization step for both
single- and multiple-pulsed exposures. This step is consistent
with current U.S. Environmental Protection Agency guidance,
such as the guidelines for ecological risk assessment [76]. The
probability of an exposure, derived from RADAR, for the time-
varying exposure is then compared to the probability of an
effect magnitude from a pulse exposure toxicity study or to
the result from PULSETOX if additional effects testing was
not conducted. In the former comparison, a probability of an
effect curve is generated. Concentration units for both expo-
sure and effects must be the same (e.g., peak, time-weighted
average, AUC). In the latter comparison, a weight-of-evidence
approach is used to decide whether the probability of exposures
will cause a CBR to be reached, producing an effect.
Fig. 9. Joint probability distribution: use of time-varying results in
risk characterization (peak concentrations in a lower-reach stream
using a threshold of 2
␮g/L [bottom]). Exposure data are from risk
assessment tool to evaluate duration and recovery (RADAR; see Fig.
4); effects data are hypothetical Daphnia magna survival results (30-
d study) (top).
effect endpoint at time
analysis.
t can be compared using a chi-square
Population analysis approaches can be used a posteriori to
address whether the results from a pulsed toxicity study could
be ecologically significant, that is, cause ecologically mean-
ingful changes in population numbers. Age-structured models
can be used to extrapolate acute or chronic measurement end-
points to population effects. Risk characterization relates the
exposure distribution to the population responses to determine
a time to recovery, probability of a percentage of population
decline, probability that the population will decline below a
quasi-extinction threshold, and so on.
Laboratory-based approach
Results from laboratory-based pulsed exposure studies can
be used in this risk assessment scheme in the same manner as
typical toxicity studies. Effects based on peak concentration
or AUC integration can be used and compared to exposure
concentrations derived in the same manner (units) (Fig. 9).
In Figure 9, exposure, characterized by a distribution of the
probability of a particular concentration (peak concentration
[shown] using a 2-␮g/L threshold; AUC, concentration–pulse
Toxicokinetic modeling-based approach
duration; geometric mean; or arithmetic mean) occurring, cou-
pled with the effect (e.g., percentage mortality) versus con-
centration (peak concentration [shown], AUC, geometric
mean, or arithmetic mean), generates a probability curve of
As mentioned earlier, time-varying models, such as
PULSETOX, may be used to assist in the interpretation of
effects from long pulses, long recovery intervals, or numerous
repeated exposures that would be difficult to test in the lab-
oratory. The time to reach a lethal body residue can be esti-
mated using default CBRs. Estimates of body residues and/or
time to lethality can be estimated for other exposure concen-
trations or pulse durations. The results from this modeling are
not used directly in the risk characterization but are useful in
understanding the dynamics of the compound in environmental
receptors. The proximity of these concentrations to the CBR
would be used in a weight-of-evidence approach in deciding
whether effects are expected based on the particular exposure
scenario.
the
p
of a particular effect occurring. In this example, there is
0.8 that a 20% effect will occur in D. magna (acute
toxicity in this case) in this exposure scenario and a 50% (
0.5) probability that 90% of the daphnids will be affected.
Consistent with Figure 9, time-to-event analysis can be used
a
p ϭ
p
ϭ
to characterize the risks from time-varying exposures. Time-
to-event analysis assists in determining whether the events are
independent, antagonistic, additive, or synergistic. If indepen-
dent, time-varying exposure testing would not be needed, as
each pulse event occurs as a singular event, without organism
memory or effect overlap. Slopes can be compared, or the

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1989
Risk characterization in higher tiers
RECOMMENDATIONS
Recommendations supported in this document are that (1)
time-varying exposure testing and modeling should be con-
sidered if exposure profiles and chemical behavior suggest
pulsed scenarios (more realistic exposures); (2) time-varying
exposures should be considered if the application interval,
compound half-life, and/or exposure modeling indicate that
such exposures may occur; (3) pulse testing methods or guide-
lines (both acute and chronic) should be developed and vali-
dated for general use; (4) where possible to redesign bioassay
protocols to determine whether latency exists, the protocols
should be modified as long as the results are compatible with
historical data from the same protocol; (5) time-to-event anal-
yses of time-varying responses are useful tools for interpreting
population effects; (6) PULSETOX is useful when laboratory
testing of time-varying exposure is not practical (e.g., long
pulses, long recovery intervals, or numerous repeated expo-
sures); (7) detailed PBPK models may be useful at higher tiers
of ecological risk assessment, but monitoring is often more
cost effective and less uncertain; (8) PBPK models are useful
for extrapolation to other species, especially larger species
where contaminant dynamics are not considered instantaneous;
and (9) consideration of time-varying exposure results will
reduce uncertainty and help refine the determinations of the
risks associated with the use of many agrochemicals.
Risk characterization in higher tiers (or levels of refine-
ment) would be subject to professional judgment and not a
direct comparison of exposure and effects distributions as in
lower tiers. Particular aspects of the pulsed exposure, either
via monitoring (exposure and/or effects) or higher-level mod-
eling (e.g., PBPK), would be refined and used in a weight-of-
evidence approach to assessing the ecological risks.
UNCERTAINTY ANALYSIS
Uncertainties are grouped into laboratory and modeling cat-
egories.
Laboratory studies
Several issues and uncertainties are associated with pulse
exposure toxicity testing. The exposure regime used in the
laboratory should represent that encountered in the field. Pulse
exposure toxicity testing may not be able to test pulses of long
duration and/or long recovery intervals. The relationship be-
tween exposure time and acute toxicity will differ among sub-
stances. The postexposure observation period may reveal that
some test organisms recovered from their toxic effects [3].
Toxicokinetic models
Uncertainties associated with and barriers to the use of
toxicokinetic models in predicting toxic effects in aquatic or-
ganisms from whole-body residues are that (1) the toxicodyn-
amics of the compound must be considered and are frequently
poorly defined; (2) the distribution of the compound between
tissue compartments must be at a steady state; (3) if equilib-
rium is not met, then the residue approach may not work for
very short-term exposures; (4) defining toxic effects based on
body residue levels reduces uncertainties associated with the
bioavailability of the substance in question, but a number of
poorly defined species-specific and site-specific variables exist
that affect the refinement of toxicokinetic models; (5) consid-
erable research is needed to generate data in support of the
establishment of CBRs for the most common environmental
contaminants and for a wide range of receptors, including sen-
sitive species [44]; (6) when measuring CBRs, an operational
lag time exists between when the effect occurs and the or-
Acknowledgement—The authors acknowledge technical comments
from M. McKee, B. Sebastien, K. Solomon, and S. Klaine as well as
the technical support from L. Burg, E. Kaminski, and V. Watts of the
Rohm and Haas Company. We are also grateful for the detailed com-
ments provided by two anonymous reviewers. This effort was based
on discussions conducted during the Ecological Committee for FIFRA
risk assessment methods (ECOFRAM) process from 1997 to 1999
(ECOFRAM Aquatic and Terrestrial Final Draft Reports, http://
www.epa.gov/oppefed1/ecorisk/index.htm, June 1, 1999).
REFERENCES
1. Handy RD. 1994. Intermittent exposure to aquatic pollutants: As-
sessment, toxicity and sublethal responses in fish and inverte-
brates. Comp Biochem Physiol C 107:171–184.
2. Hickie BE, McCarty LS, Dixon DG. 1995. A residue-based tox-
icokinetic model for pulse-exposure toxicity in aquatic systems.
Environ Toxicol Chem 14:2187–2197.
3. Parsons JT, Surgeoner GA. 1991. Acute toxicities of permethrin,
fenitrothion, carbaryl and carbofuran to mosquito larvae during
single- or multiple-pulse exposures. Environ Toxicol Chem 10:
1229–1233.
4. Parsons JT, Surgeoner GA. 1991. Effect of exposure time on the
acute toxicities of permethrin, fenitrothion, carbaryl and carbof-
uran to mosquito larvae. Environ Toxicol Chem 10:1219–1227.
5. Solomon KR, Baker DB, Richards RP, Dixon KR, Klaine SJ, La
Point TW, Kendall RJ, Weisskopf CP, Giddings JM, Hall LW Jr,
Williams WM. 1996. Ecological risk assessment of atrazine in
North American surface waters. Environ Toxicol Chem 15:31–
76.
6. Dauterman WC. 1994. Adaptation to toxicants. In Hodson E, Levi
PE, eds, Introduction to Biochemical Toxicology, 2nd ed. Ap-
pleton and Lange, Norwalk, CT, USA, pp 569–581.
7. Sipes IG, Gandolfi AJ. 1986. Biotransformation of toxicants. In
Klaasen CD, Amdur MO, Doull J, eds, Casarett and Doull’s
Toxicology: The Basic Science of Poisons, 3rd ed. Macmillan,
New York, NY, USA, pp 64–98.
8. Burnison BK, Hodson PV, Nuttley DJ, Efler S. 1996. A bleached-
kraft mill effluent fraction causing induction of a fish mixed-
function oxygenase enzyme. Environ Toxicol Chem 15:1524–
1531.
ganism is retrieved and analyzed [44]; (7) high Kow compounds
will be more difficult to predict due to bioavailability limi-
tations; (8) the development of CBRs for compounds with
specific mechanisms of action will need to include exposures
to a range of taxa to account for differences in sensitive species
[44]; (9) bioaccumulation is a complex process involving bi-
ological and chemical factors, including kinetics, equilibrium
levels, tissue translocation, sequestration, excretion, and dep-
uration (body residues depend on age, health, sex, reproductive
state, lipid content, trophic level, feeding habits, and ambient
tissue concentrations); (10) analytical factors such as matrix
interferences, extraction efficiency, methodological sensitivity,
instrumentation, and variation in methodology will effect mea-
sured body residue levels [77]; and (11) a threshold level in
tissue will generally refer to a single chemical. However, toxic
potentials are associated with residues resulting from mixtures
of compounds that will not be adequately described by de-
veloping threshold levels for individual compounds. For in-
stance, DDT may stimulate the metabolism of dieldrin in an-
imals, and different organochlorines may interact to effect the
accumulation of residues in fish [65].
9. Parrott JL, Hodson PV, Servos MR, Huestis SL, Dixon DG. 1995.
Relative potency of polychlorinated dibenzo-p-dioxins and di-
benzofurans for inducing mixed-function oxygenase activity in
rainbow trout. Environ Toxicol Chem 14:1041–1050.

1990
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
10. Van der Hoeven N, Gerritsen AAM. 1997. Effects of chlorpyrifos
on individuals and populations of Daphnia pulex in the laboratory
and field. Environ Toxicol Chem 16:2438–2447.
11. Verhaar HJM, de Wolf W, Dyer S, Legierse KCHM, Seinen W,
Hermens JLM. 1999. An LC50 vs. time model for the aquatic
toxicity of reactive and receptor-mediated compounds: Conse-
quences for bioconcentration kinetics and risk assessment. En-
viron Sci Technol 33:758–763.
12. Hurd MK, Perry SA, Perry WB. 1996. Nontarget effects of a test
application of diflubenzuron to the forest canopy on stream ma-
croinvertebrates. Environ Toxicol Chem 15:1344–1351.
13. Giesy JP, Solomon KR, Coates JR, Dixon KR, Giddings JM,
Kenaga EE. 1999. Chlorpyrifos ecological risk assessment in
North American environments. Rev Environ Contam Toxicol 160:
1–129.
correlation for narcosis in aquatic organisms: The key role of
partitioning. Environ Toxicol Chem 7:469–481.
32. McCarty LS, Mackay D, Smith AD, Ozburn GW, Dixon DG.
1992. Residue-based interpretation of toxicity and bioconcentra-
tion QSARs from aquatic bioassays: Neutral narcotic organics.
Environ Toxicol Chem 11:917–930.
33. Landis WG, Matthews RA, Matthews GB. 1996. The layered and
historical nature of ecological systems and the risk assessment of
pesticides. Environ Toxicol Chem 15:432–440.
34. Matthews RA, Landis WG, Matthews GB. 1996. The community
conditioning hypothesis and its application to environmental tox-
icology. Environ Toxicol Chem 15:597–603.
35. Kallander DB, Fisher SW, Lydy MJ. 1997. Recovery following
pulsed exposure to organophosphorus and carbamate insecticides
in the midge, Chironomus riparius
. Arch Environ Contam Tox-
icol 33:29–33.
14. Newman MC, McCloskey JT. 1996. Time-to-event analyses of
36. Clark JR, Goodman LR, Borthwick PW, Patrick JM Jr, Moore
JC, Lores EM. 1986. Field and laboratory toxicity tests with
shrimp, mysids and sheepshead minnows exposed to fenthion. In
Poston TM, Purdy R, eds, Aquatic Toxicology and Environmental
Fate, Vol 9. STP 921. American Society for Testing and Materials,
Philadelphia, PA, pp 161–176.
37. Klaine SJ, Richards P, Baker D, Naddy R, Brown T, Joab B,
Casey R, Fernandez D, Overmeyer J, Benjamin R. 1997. Agro-
chemical fate and effects in terrestrial, aquatic and estuarine eco-
systems. In Environmental Behavior of Crop Protection Chem-
icals. IAEA-SM-343/13. International Atomic Energy Agency,
Vienna, Austria, pp 247–263.
38. Buhl KJ, Hamilton SJ, Schmulbach JC. 1993. Chronic toxicity
of the bromoxynil formulation Buctril to Daphnia magna exposed
continuously and intermittently. Arch Environ Contam Toxicol
25:152–159.
39. Panter GH, Thompson RS, Sumpter JP. 2000. Intermittent ex-
posure of fish to estradiol. Environ Sci Technol 34:2756–2760.
40. Carsel RF, Imhoff JC, Hummel PR, Cheplick JM, Donigan AS
Jr. 1997. PRZM-3, A Model for Predicting Pesticide and Nitrogen
Fate in the Crop Root and Unsaturated Soil Zones: User’s Man-
ual for Release 3.0. Environmental Research Laboratory, U.S.
Environmental Protection Agency, Athens, GA.
41. Burns LA, Cline DM, Lassiter RR. 1982. Exposure Analysis Mod-
eling System (EXAMS): User Manual and System Documenta-
tion. EPA-600/3-82/023. Environmental Research Laboratory,
U.S. Environmental Protection Agency, Athens, GA.
42. Newman MC, Dixon PM. 1996. Ecologically meaningful esti-
mates of lethal effect in individuals. In Newman MC, Jagoe CH,
eds, Ecotoxicology: A Hierarchical Treatment. Lewis, Boca Ra-
ton, FL, USA, pp 225–253.
43. Mancini JL. 1983. A method for calculating effects, on aquatic
organisms, of time varying concentrations. Water Res 17:1355–
1362.
44. Landrum PF, Lee H II, Lydy MJ. 1992. Toxicokinetics in aquatic
systems: Model comparisons and use in hazard assessment. En-
viron Toxicol Chem 11:1709–1725.
45. Barber MC, Suarez LA, Lassiter RR. 1988. Bioconcentration of
nonpolar organic pollutants by fish. Environ Toxicol Chem 7:
545–558.
46. Hallam TG, Lassiter RR. 1994. Individual-based mathematical
modeling approaches in ecotoxicology: A promising direction for
aquatic population and community ecological risk assessment. In
Kendall RJ, Lacher TE Jr, eds, Wildlife Toxicology and Popu-
lation Modeling: Integrated Studies of Agroecosystems. SETAC,
Pensacola, FL, USA, pp 531–542.
47. Hallam TG, Lassiter RR, Li J, McKinney W. 1990. Toxicant-
induced mortality in models of Daphnia populations. Environ
Toxicol Chem 9:597–621.
48. Lassiter RR, Hallam TG. 1990. Survival of the fattest: Implica-
tions for acute effects of lipophilic chemicals on aquatic popu-
lations. Environ Toxicol Chem 9:585–595.
49. Environment Canada. 1997. Environmental assessments of pri-
ority substances under the Canadian Environmental Protection
Act. Guidance Manual, Ver 1.0. EPS/2/CC/3E. Ottawa, ON.
50. Klee U. 1998. Development and testing of a body-residue based
one-compartment first-order kinetic model for estimating the tox-
icity of mixtures of chlorinated organic contaminants to rainbow
trout (Oncorhynchus mykiss). PhD thesis. University of Waterloo,
Waterloo, ON, Canada.
ecotoxicity data. Ecotoxicology 5:187–196.
15. Sprague JB. 1969. Measurement of pollutant toxicity to fish: I.
Bioassay methods for acute toxicity. Water Res 3:793–821.
16. Newman MC. 1995. Quantitative Methods in Aquatic Ecotoxi-
cology. Lewis, Boca Raton, FL, USA.
17. Holdway DA, Dixon DG. 1985. Acute toxicity of pulse-dosed
methoxychlor to juvenile American flagfish (Jordanella floridae
Goode and Bean) as modified by age and food availability. Aquat
Toxicol 6:243–250.
18. Holdway DA, Dixon DG, Kaiser KLE. 1991. The acute toxicity
of pulse-dosed, para-substituted phenols to larval American flag-
fish (Jordanella floridae): A comparison with toxicity to pho-
toluminescent bacteria and predicted toxicity using log Kow
Total Environ 104:229–237.
19. Fischer DJ, Burton DT, Yonkos LT, Turley SD, Turley BS, Ziegler
GP, Zillioux EJ. 1994. Acute and short-term chronic effects of
continuous and intermittent chlorination on Mysidopsis bahia and
. Sci
Menidia beryllina. Environ Toxicol Chem 13:1525–1534.
20. Holdway DA, Dixon DG. 1986. Impact of pulse exposure to
methoxychlor on flagfish (Jordanella floridae) over one repro-
ductive cycle. Can J Fish Aquat Sci 43:1410–1415.
21. Barry MJ, O’Halloran K, Logan DC, Ahokas JT, Holdway DA.
1995. Sublethal effects of esfenvalerate pulse-exposure on spawn-
ing and non-spawning Australian crimson-spotted rainbow fish
(
Melanotaenia fluviatilis). Arch Environ Contam Toxicol 28:
459–463.
22. Naddy RB, Johnson KA, Klaine SJ. 2000. Response of Daphnia
magna to pulsed exposures of chlorpyrifos. Environ Toxicol
Chem 19:423–431.
23. Jarvinen AW, Tanner DK, Kline ER. 1988. Toxicity of chlorpyr-
ifos, endrin, or fenvalerate to fathead minnows following episodic
or continuous exposure. Ecotoxicol Environ Saf 15:78–95.
24. Naddy RB, Klaine SJ. 2001. Effect of pulse frequency and interval
on the toxicity of chlorpyrifos to Daphnia magna. Chemosphere
45:497–506.
25. Williams ND, Holdway DA. 2000. The effects of pulse-exposed
cadmium and zinc on embryo hatchability, larval development
and survival of Australian crimson spotted rainbow fish (Melan-
otaenia fluviatilis). Environ Toxicol 15:165–173.
26. Heming TA, McGuinness EJ, George LM, Blumhagen KA. 1988.
Effects of pulsed- and spiked-exposure to methoxychlor on early
life stages of rainbow trout. Bull Environ Contam Toxicol 40:
764–770.
27. Stuijfzand SC, Poort L, Greve GD, van der Geest HG, Krook
MHS. 2000. Variables determining the impact of diazinon on
aquatic insects: Taxon, developmental stage and exposure. En-
viron Toxicol Chem 19:582–587.
28. Hosmer AJ, Warren LW, Ward TJ. 1998. Chronic toxicity of pulse-
dosed fenoxycarb to Daphnia magna exposed to environmentally
realistic concentrations. Environ Toxicol Chem 17:1860–1866.
29. Schulz R, Liess M. 2000. Toxicity of fenvalerate to caddisfly
larvae: Chronic effects of 1- vs 10-h pulse exposure with constant
doses. Chemosphere 41:1511–1517.
30. Lipnick RL. 1993. Baseline toxicity QSAR models: A means to
assess mechanism of toxicity for aquatic organisms and mammals.
In Gorsuch J, Dwyer F, Ingersoll C, La Point TW, eds, Environ-
mental Toxicology and Risk Assessment, Vol 2. STP 1216. Amer-
ican Society for Testing and Materials, Philadelphia, PA, pp 610–
619.
31. Abernethy S, Mackay D, McCarty L. 1988. ‘‘Volume fraction’’
51. Kooijman SALM, Bedaux JJM. 1996. The Analysis of Aquatic

Effects analysis of time-varying aquatic exposures
Environ. Toxicol. Chem. 21, 2002
1991
Toxicity Data (includes DEBtox, Ver 1.0). VU University Press,
Amsterdam, The Netherlands.
64. Suter GW. 1993. Ecological Risk Assessment. Lewis, Boca Raton,
FL, USA.
52. Spacie A, McCarty LS, Rand GM. 1995. Bioaccumulation and
bioavailability in multiphase systems. In Rand GM, ed, Funda-
mentals of Aquatic Toxicology. Taylor & Francis, Washington
DC, USA, pp 493–521.
53. European Center for Ecotoxicology and Toxicology of Chemicals.
1995. The role of bioaccumulation in environmental risk assess-
ment: The aquatic environment and related food webs. ECETOC
Technical Report 67. Brussels, Belgium.
54. Bysshe SE. 1990. Bioconcentration factor in aquatic organisms.
In Lyman WJ, Reehl WF, Rosenblat DH, eds, Handbook of Chem-
ical Property Estimation Methods. American Chemical Society,
Washington DC, USA, pp 5-1–5-30.
55. Lien GJ, Nichols JW, McKim JM, Gallinat CA. 1994. Modeling
the accumulation of three waterborne chlorinated ethanes in fat-
head minnows (Pimephales promelas): A physiologically based
approach. Environ Toxicol Chem 13:1195–1205.
65. Van Loon WMGM, Verwoerd ME, Wijnker FG, van Leeuwen
CJ, van Duyn P, van deGuchte C, Hermens JLM. 1997. Estimating
total body residues and baseline toxicity of complex organic mix-
tures in effluents and surface waters. Environ Toxicol Chem 16:
1358–1365.
66. McCarty LS, Mackay D. 1993. Enhancing ecotoxicological mod-
eling and assessment. Environ Sci Technol 27:1719–1728.
67. Chaisuksant Y, Yu Q, Connell D. 1997. Internal lethal concen-
trations of halobenzenes with fish (Gambusia affinis). Ecotoxicol
Environ Saf 37:66–75.
68. Yu Q, Chaisuksant Y, Connell D. 1999. A model for non-specific
toxicity with aquatic organisms over relatively long periods of
exposure time. Chemosphere 38:909–919.
69. Deneer JW, Budde BJ, Weijers A. 1999. Variations in the lethal
body burdens of organophosphorus compounds in the guppy.
Chemosphere 38:1671–1683.
56. Lassiter RR. 1986. Design criteria for a predictive ecological
effects modeling system. In Poston TM, Purdy R, eds, Aquatic
Toxicology and Environmental Fate, Vol 9. STP 921. American
Society for Testing and Materials, Philadelphia, PA, pp 42–54.
57. Krishnan K, Andersen ME. 1994. Physiologically based phar-
macokinetic modeling in toxicology. In Hayes AW, ed, Principles
and Methods of Toxicology, 3rd ed. Raven, New York, NY, USA,
pp 149–188.
58. Nichols J, Rheingans P, Lothenbach D, McGreachie R, Skow L,
McKim J. 1994. Three-dimensional visualization of physiologi-
cally based kinetic model outputs. Environ Health Perspect 102:
952–956.
59. Bradbury SP, Coats JR, McKim JM. 1986. Toxicokinetics of fen-
valerate in rainbow trout (Salmo gairdneri). Environ Toxicol
Chem 5:567–576.
60. Erickson RJ, McKim JM. 1990. A model for exchange of organic
chemicals at fish gills: Flow and diffusion limitations. Aquat Tox-
icol 18:175–198.
61. Hayton WL, Barron MG. 1990. Rate-limiting barriers to xeno-
biotic uptake by the gill. Environ Toxicol Chem 9:151–157.
62. Nichols JW, McKim JM, Lien GJ, Hoffman AD, Bertelsen SL,
Elonen CM. 1996. A physiologically based toxicokinetic model
for dermal absorption of organic chemicals by fish. Fundam Appl
Toxicol 31:229–242.
63. Tarr BD, Barron MG, Hayton WL. 1990. Effect of body size on
the uptake and bioconcentration of di-2-ethylhexyl phthalate in
rainbow trout. Environ Toxicol Chem 9:989–995.
70. Driscoll SK, Landrum PF. 1997. A comparison of equilibrium
partitioning and critical body residue approaches for predicting
toxicity of sediment-associated fluoranthene to freshwater am-
phipods. Environ Toxicol Chem 16:2179–2186.
71. Beyer WN, Heinz GH, Redmon-Norwood AW, eds. 1996. En-
vironmental Contaminants in Wildlife: Interpreting Tissue Con-
centrations. Lewis, Boca Raton, FL, USA.
72. McKim JM, Schmeider PK. 1991. Bioaccumulation: Does it re-
flect toxicity? In Nagel R, Loskill R, eds, Bioaccumulation in
Aquatic Systems. VCH, Weinheim, Germany, pp 161–188.
73. Nendza M, Herbst T, Kussatz C, Gies A. 1997. Potential for
secondary poisoning and biomagnification in marine organisms.
Chemosphere 35:1875–1885.
74. McCarty LS, Mackay D, Smith AD, Ozburn GW, Dixon DG.
1993. Residue-based interpretation of toxicity and bioconcentra-
tion QSARs from aquatic bioassays: Polar narcotic organics. Eco-
toxicol Environ Saf 25:253–270.
75. Vaes WHJ, Ramos EU, Verhaar HJM, Hermens JLM. 1998. Acute
toxicity of nonpolar versus polar narcosis: Is there a difference?
Environ Toxicol Chem 17:1380–1384.
76. U.S. Environmental Protection Agency. 1998. Guidelines for eco-
logical risk assessment. Fed Reg 63(93):26846–26924.
77. Maciorowski AF, McCulloch WL, Fava JA. 1985. An approach
to sewage sludge bioaccumulation potential tests. In Cardwell
RD, Purdy R, Bahner RC, eds, Aquatic Toxicology and Hazard
Assessment, 7th Symposium. STP 854. American Society for
Testing and Materials, Philadelphia, PA, pp 253–268.

1992
Environ. Toxicol. Chem. 21, 2002
K.H. Reinert et al.
APPENDIX
APPENDIX Continued
Run modes, input parameters, and assumptions for food and gill
exchange of toxic substances model (FGETS)
Common inputs to the laboratory, food chain, and food web modes
include the following:
●
Assume: unlimited prey resources and the physiologically active
fraction of the gill is constant across all year classes for a par-
ticular species
●
●
●
●
●
●
●
Molecular weight, volume
Melting point
Log Kow
Fish species
Fish weight
Water temperature
Assume: food is in thermodynamic equilibrium with the water
Pulsed exposure toxicity tests (PULSETOX) modes and
input parameters
For both cyclic and random exposures:
●
●
●
●
●
●
●
●
●
●
Henry’s law constant
Volume of organism
Volume fraction of lipid
Kow
Molecular weight
Bioconcentration factor
Uptake and clearance rate constants
Test chamber volume
Laboratory is defined as an aquarium with constant in- and outflows
of water.
●
●
●
●
●
Tank flow rate, volume
Number of fish
Concentration of chemical in food
Length of study
Assume: aquarium with constant in- and outflows of water and
volume of the tank and the number of fish from a single species
remains constant
Water flow rate
Substance stream flow rate and concentration (from GENEEC or
pesticide root zone model/exposure analysis modeling system
[PRZM/EXAMS])
Food chain simulates one or two fish interaction. With one fish, the
fish is a predator that feeds on either plankton, benthic organisms, or
generic fish. With two fish, one fish is the prey of the other. The prey
species feeds on either plankton, benthic organisms, or generic fish,
and the predator feeds on the prey species according to a specific
length–length relationship.
●
●
Number, duration, and interval of pulses
Assume: the test chamber acts as a well-mixed vessel; the rate
of change and absolute water concentration (Cw) at any point in
time is a function of the test chamber volume, the flow rates of
water, and the substance stream flow rate and concentration; the
only clearance of chemical from the system is via the tank out-
flow; the organism is a single compartment (works well for small-
er organisms [50]); toxicokinetic rate uptake and depuration con-
●
●
●
●
Plankton or benthic chemical concentration (constant)
Prey identification (only in two-fish system)
Predator identification
Assume: concentration of chemical in water, plankton, benthos,
and generic fish as well as water is constant during the length of
simulation
stants, Ku and Kd, are independent of exposure concentration and
do not change when the organism is intoxicated; the substance
concentration in the organism reaches steady state if the external
substance concentration is constant; the population of organisms
responds in classic dose–response manner; individual response
near instantaneous when the organism achieves its critical whole-
body dose; and the whole-body substance concentration is an
adequate surrogate for the dose at the site(s) of action.
Food web is designed to describe more realistic predator–prey interac-
tions. Fish may feed on each other, plankton, or benthos, according to a
user-specified diet. The exposure conditions (water temperature and con-
centrations of chemicals in plankton, benthos, and water) are arbitrary.
For the random exposure mode, the exposure conditions must be
described for each exposure phase.
●
●
Percentage diet makeup of prey (time or weight dependent)
Percentage diet makeup of predator (time or weight dependent)