Science of the Total Environment 625 (2018) 168–177

Contents lists available at ScienceDirect

Science of the Total Environment

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

Pesticide load dynamics during stormwater flow events in

Mediterranean coastal streams: Alexander stream case study
Tom Topaz a, Roey Egozi b,⁎, Gil Eshel b, Benny Chefetz a
a
Dept. of Soil and Water Sciences, Faculty of Agriculture, Food and Environment, The Hebrew University of Jerusalem, Rehovot 76100, Israel
b
The Soil Erosion Research Station, Soil Conservation and Drainage Division, Ministry of Agriculture and Rural Development, Bet Dagan 50250, Israel

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Pesticide-load dynamic is examined for

both dissolved and adsorbed phases.
• Most pesticides are transported in the
dissolved phase.
• Flood event magnitude and pesticide
application method affect its concentra-
tion.
• Mix of pesticides is dominated by diu-
ron (herbicide) and tebuconazole
(fungicide).

a r t i c l e i n f o a b s t r a c t

Article history: Cultivated land is a major source of pesticides, which are transported with the runoff water and eroded soil dur-
Received 6 September 2017 ing rainfall events and pollute riverine and estuarine environments. Common ecotoxicological assessments of
Received in revised form 20 November 2017 riverine systems are mainly based on water sampling and analysis of only the dissolved phase, and address a sin-
Accepted 19 December 2017
gle pesticide's toxicological impact under laboratory conditions. A clear overview of mixtures of pesticides in the
Available online 27 December 2017
adsorbed and dissolved phases is missing, and therefore the full ecotoxicological impact is not fully addressed.
Editor: Ouyang Wei The aim of this study was to characterize and quantify pesticide concentrations in both suspended sediment
and dissolved phases, to provide a better understanding of pesticide-load dynamics during storm events in coast-
Keywords: al streams in a Mediterranean climate. High-resolution sampling campaigns of seven flood events were conduct-
Soil erosion ed during two rainy seasons in Alexander stream, Israel. Samples of suspended sediments were separated from
Suspended sediment the solution and both media were analyzed separately for 250 pesticides. A total of 63 pesticides were detected;
River water quality 18 and 16 pesticides were found solely in the suspended sediments and solution, respectively. Significant differ-
Flow regime ences were observed among the pesticide groups: only 7% of herbicide, 20% of fungicide and 42% of insecticide
Non-point contamination source
load was transported with the suspended sediments. However, in both dissolved and adsorbed phases, a mix
Ecotoxicology
of pesticides was found which were graded from “mobile” to “non-mobile” with varied distribution coefficients.
Diuron, and tebuconazole were frequently found in large quantities in both phases. Whereas insecticide and fun-
gicide transport is likely governed by application time and method, the governing factor for herbicide load was
the magnitude of the stream discharge. The results show a complex dynamic of pesticide load affected by exces-
sive use of pesticides, which should be taken into consideration when designing projects to monitor riverine and
estuarine water quality.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction

Agricultural lands in Mediterranean climates are characterized by

⁎ Corresponding author. high soil-erosion rates, due to the coupling of high-intensity rainfall
E-mail address: regozi@moag.gov.il (R. Egozi). events (erosivity index) and bare soils (Garcia-Ruiz et al., 2015;

https://doi.org/10.1016/j.scitotenv.2017.12.213
0048-9697/© 2017 Elsevier B.V. All rights reserved.
 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177 169

Montgomery, 2007; Morin and Benyamini, 1977; Panagos et al., 2015a, Resh, 1999). Once the pesticides that are bound to soil colloids are re-
2015b). Whereas the former is related to the natural rainfall suspended in the aquatic environment, they may undergo geochemical
regime—rainfall amount, distribution, intensity, and raindrop kinetic changes and become available to biota (Chapman et al., 2013; Turner,
energy, the latter is an outcome of agricultural practices—for example, 2003). Therefore, the aim of this study was to characterize and quantify
intensive weed-control practices aimed at eliminating competition on pesticide concentrations in both the suspended sediments and dis-
water and nutrient resources with the desired crop (e.g., plantations, or- solved phases, while following the flood events in Alexander stream.
chards, vineyards and other field crops) (Oerke, 2006). Soil erosion and We hypothesized that flood events are the dominant contributor of pes-
transport by water initiates with particle detachment, due to the kinetic ticides to this stream, and that significant amounts of pesticides are
energy of the raindrop's impact on bare ground, and continues with transported adsorbed to suspended sediments as well as in the dis-
inter-rill, rill and gully erosion. When soils are water-saturated, erosion solved phase.
intensifies, and full hydrological connectivity is reached between the
fields on the hillslopes and the riverine system's drainage network 2. Material and methods
(Covino, 2017; Sharma et al., 1991; Stieglitz et al., 2003; Valentin et
al., 2005). In the riverine system, the eroded soil particles can potential- 2.1. Study site
ly travel substantial distances in suspension, contributing significantly
to global sediment fluxes to seas and oceans (Syvitski et al., 2005). Alexander stream lies within a typical coastal watershed located at
Surface-runoff water containing eroded soil particles from cultivated the eastern part of the Mediterranean Sea. The watershed spans a
land often includes nutrients and applied pesticides released from the total area of 553 km2 with mean annual rainfall of 570 mm. The main
topsoil. This may result in direct and indirect physical, chemical and bi- stream channel runs over 45 km in a general east–west direction, with
ological effects on water quality, thereby impacting the aquatic ecosys- three main tributaries (Bahan, Shchem and Teenim) that all start from
tems (e.g., toxic algal blooms, poisoning and disturbances to food web the Samaria mountains, and continue through the lowlands of Hefer
structures), and posing a major global threat to human water security valley, where they converge at different locations (Fig. 1). Alexander
and river biodiversity (Beketov et al., 2013; Blann et al., 2009; stream flows into a well-defined micro-estuary before entering the
Schwarzenbach et al., 2006; Vörösmarty et al., 2010). Pesticide fluxes Mediterranean Sea. The watershed drainage area can be roughly divid-
are more difficult to quantify and trace (e.g., Bundschuh et al., 2014; ed into four parts: (i) the upper part consists of mountain slopes, com-
Kronvang et al., 2003) than excessive fertilizer fluxes originating from posed mainly of limestone and dolomite rocks and characterized by
cultivated land (e.g., Neal et al., 2012; Segal-Rozenhaimer et al., 2004). Cambisols (Terra Rossa and Rendzina soils; ~ 40% clay, 0.3–2% soil or-
This is because fertilizer dynamics is easier to study by monitoring ganic carbon (SOC)); (ii) the middle part consists of rolling hills, made
only a few key tracers in the surface water, such as phosphorus and ni- of limestone and chalk lithology covered with Chromic Luvisols
trogen (e.g. Bennett et al., 2001; Minami and Fukushi, 1984). In addi- (Brown Hamra; ~7% clay, 0.3–1% SOC); (iii) the lowland part is a typical
tion, pesticide-load dynamics is affected by the partitioning of the mixture of Chromic Vertisols (Grumusol; ~65% clay, 0.3–2% SOC) and
chemical between dissolved and adsorbed phases. Some pesticides sandy Luvisols (Hamra; 4% clay, 0.3–0.6% SOC); (iv) a narrow section
tend to bind with suspended sediments and are transported as particu- consisting of sand dunes along the estuary. The western part of the wa-
late-bound pollutants, whereas others are more stable in their dissolved tershed is a highly fertile valley, characterized by intensive agricultural
form in the surface runoff (Birch et al., 2015; Carpenter et al., 2016; activity covering 45% of its watershed area (Fig. 1). The eastern water-
Chapman et al., 2013; Ng et al., 2012). This dynamic can be highly influ- shed is characterized by steep terrain, with agricultural land use domi-
enced by field conditions (e.g., time of pesticide application and loca- nated mainly by citrus and olives trees, as well as vineyards. A total of
tion; Pérez et al., 2017) and soil characteristics (e.g., clay and organic 66% of the entire watershed is used for agriculture, mainly orchards,
content; Larson et al., 1995). plantations and field crops (Tal et al., 2007).
While awareness of sediment-bound pesticide toxicity has risen in
the last few decades (Bundschuh et al., 2014; Kronvang et al., 2003), ef- 2.2. Data acquisition and field sampling
forts to establish benchmarks have only recently been initiated (e.g.,
Nowell et al., 2016; Rodrigues et al., 2017), along with studies that con- Discharge records were retrieved from the Eliashiv hydrometric sta-
tinually monitor pesticide concentrations (e.g., Pérez et al., 2017). tion which has been operated by the Israel Hydrological Services since
Significant knowledge gaps still exist regarding pesticide-load dynamics 1967. Flood events were defined as times with stream flow having a
during flood events, in particular with respect to differentiating the peak discharge above 2 m3 s−1, providing a clear distinction between
pesticide distribution between (i) baseflows and flood events, and (ii) floods and baseflows. This flow limit was defined to differentiate occa-
dissolved and particulate-bound phases. This basic information may sional releases of treated wastewater and fishponds into the stream,
contribute to a more realistic environmental risk assessment of non- reaching up to ~1.5 m3 s−1, from baseflow discharge, which fluctuates
point contamination sources, and knowledge of the mitigation require- around a daily mean value of 0.14 m3 s− 1. Data on land use and soil
ments (e.g., Bereswill et al., 2013) required to reduce the potential type were obtained from the Israel National Geographic Information
threat to biological functioning of different aquatic species, ranging Database and analyzed using ArcGIS (v10.1 by ESRI).
from diatoms and bacteria to fish, vertebrates, and birds (e.g., Ayas et Data pertaining to potential pollutant sources, agricultural activities
al., 1997; Beketov et al., 2013; DeLorenzo et al., 2001; Fasola et al., and potential pesticide usage were obtained via interviews conducted
1998; Köhler and Triebskorn, 2013; Liess and Von Der Ohe, 2005; with local farmers and professionals representing different stake-
Mora, 1997). holders, such as Yad-Hanna wastewater-treatment plant, Ministry of
This potential risk is a profound threat, especially in Mediterranean Environmental Protection, Ministry of Agriculture and Rural Develop-
streams, which are ecosystems that are under continuous water stress. ment, Nature & Parks Authority, local municipal authority, Sharon
These streams are characterized by seasonal rain periods, with dry sum- Drainage Basin Authority. This source of information will be referred
mers and wet winters, during which the dominant riverine flows are to herein as ‘personal communication’.
stormwater flow events (hereafter flood events), while low baseflows Runoff water samples were collected during flood events automati-
or dry riverbeds are often maintained between storm events (Gasith cally using an automated sampler (Sigma 900, Hach Company, Love-
and Resh, 1999; Hershkovitz and Gasith, 2013). Deposition of land, CO) equipped with twenty-four 350-mL glass bottles, and
suspended sediments with bound (or attached) pesticides may occur manually as grab samples using 2- and 4-L glass bottles. Baseflow
in streambeds and estuary banks and floodplains, where they may re- water was sampled four times during the 2014 dry season and analyzed
main during the dry season or until the next flood event (Gasith and for dissolved pesticides to determine typical baseflow pesticide load.
170 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177

Fig. 1. Alexander watershed geographical location and boundaries, main tributaries, and spatial distribution of land use within the western part. Note the location of Eliashiv hydrometric
station where the water for this study was sampled.

The small number of samples collected was sufficient because drip irri- in the centrifuge (Sorvall SLA-1500) for one run of 40 min at
gation is performed in the watershed, and therefore no significant pulse 13,000 rpm (25,500g). The solution was then extracted and passed
leaching is expected. All containers used for sample collection were pre- through a 0.45-μm filter prior to analysis of pesticide residues. The
cleaned according to the US Environmental Protection Agency (EPA) tubes with the solids and remaining extracted solution were weighed
protocol (EPA, 2007a). After acquisition, samples were kept in coolers and freeze-dried. The dry solids were removed and weighed separately
and frozen (− 20 °C) within a maximum of 24 h until processing and before analysis of adsorbed pesticide residues. Suspended sediment
analysis. concentration was calculated from the solid-to-solution weight ratio
We used an event-based approach for data acquisition and sampling, (Fig. 2b). Pesticides were analyzed by both GC/MS and LC/MS/MS to
i.e., each flood event was treated separately. The flood hydrograph was
divided into segments of water volume or discharge that were associat-
ed with matching suspended sediment concentrations, and dissolved
and adsorbed pesticide concentrations (Fig. 2a). For example, the
suspended sediment load was calculated as an average, based on six
samples with different suspended sediment concentrations collected
during six matching water-discharge measurements, during a 1-h
flow segment. The total flow volume and total sediment load for each
flood event were defined as the sum value of all segments. In total,
110 samples were collected during seven flood events, from which 65
samples were selected, prepared and screened for residues of 250 differ-
ent pesticides (43 were floodwater samples; 22 were suspended sedi-
ment samples). These samples were selected as best representing the
temporal changes in discharge, as sufficient solid material had been col-
lected to enable pesticide analysis (5 g was required by the laboratory).
Therefore, some of the samples were grouped before analysis, resulting
in higher detection resolution but lower temporal resolution, e.g., at the
end of the falling limb of the hydrograph (S5 in Fig. 2a).

2.3. Pesticide analysis and extraction

Fig. 2. Event-based approach for data acquisition and analysis of runoff water samples (a).
Schematic diagram of the extraction process for pesticides, and differentiation between
Each sample was homogenized and then transferred into 250-mL dissolved and bound particulate matter (b). Note that S1 to S5 indicates suspended
carbon centrifuge tubes. The tubes were weighed, balanced and placed sediment samples merged for pesticide analysis.
 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177 171

achieve higher limits of detection (LOD). This procedure was followed relationship had a regression coefficient (R2) of 0.75, suggesting that
for both dissolved and suspended particulate phases. All of the detected overall water discharge explains 75% of the variation in suspended sed-
pesticides were quantified based on specific standards. The GC/MS anal- iment concentration. However, high variation remained for a given dis-
ysis consisted of the following steps: pesticide extraction was per- charge value, for example, a water discharge of 8 m3 s−1 could transport
formed three consecutive times using liquid–liquid extraction with suspended sediments at concentrations ranging between 110 and
dichloromethane; 100 mL dichloromethane was added to 500 mL of 10,100 mg L−1. This variation likely results from the temporal and spa-
sample, and the mixture was shaken for 1 min and allowed to settle tial distribution of the rainfall events over the entire watershed and the
until separate layers were visible. The lower level was drained into a different sediment sources. Differences between the flood events (peak
500-mL glass vessel through a filter containing sodium thiosulfate to en- discharge N 2 m3 s−1) and baseflow are also shown (Fig. 3b). The yearly
sure water removal, and the procedure was repeated twice. The vessel average baseflow discharge (excluding the specific days with
containing the extraction was placed in a fume-chamber until the di- stormwater flows) was 0.15 m3 s−1, and the total volume of 4.5 m3
chloromethane was completely vaporized. The residue in the vessel × 106 with an average suspended sediment concentration of
was redissolved using 10% acetone in isooctane solution and analyzed 50 mg L−1 gave a yearly sediment load of 200 tons. This estimate is sig-
in a GC/MS system. The LC/MS/MS analysis was conducted according nificantly lower than the measured amounts of 4250 and 7180 tons of
to the following steps: samples were extracted using a solid-phase ex- sediment discharged during flood events F4 and F7 with peak dis-
traction C18 cartridge. After passing the water samples through the car- charges of 21.5 m3 s− 1 (recurrence interval of 1.3 years) and
tridge, the residues were washed into glass tubes using methanol. The 64 m3 s−1 (recurrence interval of 3.5 years), respectively (Table 1, Fig.
solution was then concentrated and injected directly into the LC/MS/ 3). This indicates that even one small flood event in the rainy season
MS system. The suspended particulate phase was extracted according contributes 10 times more suspended load than the entire annual
to the QuEChERS method for LC/MS/MS (Restek, 2012) and an ultrason- baseflow, which emphasizes the significant role of floods in
ic method for GC/MS (EPA, 2007b). Pesticides detected below the limit transporting suspended sediments as compared to baseflows, and fur-
of quantification (LOQ) were conservatively given their LOD value, ther highlights the importance of water-quality monitoring during
which, in most compounds, was 0.03 μg L−1 and 7 μg kg−1 for dissolved flood events.
and suspended sediments, respectively. Given the positive relationship between water discharge and
We used different basic statistical methods to describe correlations suspended sediment concentration, we expected to find similarities in
and trends in pesticide dynamics, such as linear and non-linear regres- the number, composition and concentration of pesticides transported
sion models. We also applied the kernel distribution method, which is a by these two phases, i.e., dissolved in the water column and adsorbed
non-parametric method of estimating a pesticide concentration's prob- to the suspended sediments. Based on the positive relationship between
ability density function (for more details on the method see Parzen, water discharge and suspended sediment concentration, and the fact
1962 and Silverman, 1986). We used a MATLAB script (‘ksdensity’) to that two-thirds of the runoff-contributing area is cultivated, higher con-
run this analysis. The script was based on the kernel estimator equation: centrations of pesticides were expected in the larger flood events. How-
ever, Fig. 4 indicates that this was not necessarily the case. The number
 
^f ðxÞ ¼ 1 ∑n K x−xi ; ð1Þ
of pesticides detected or quantified in each phase was not the same
h
nh i¼1
h within a particular flood event. These differences consistently appeared
within the types of pesticides—herbicides, fungicides, and insecticides.
where x1, x2, …, xn are random samples from an unknown distribution, n Moreover, high variation was observed in the detected and quantified
is the sample size, K(·) is the kernel smoothing function, and h is the compounds between flood events. For example, F4 and F5 were equal,
bandwidth at which the density is estimated. In our case these are the with the highest number of different pesticides (10) above the LOQ in
density of pesticide concentrations (x1, x2, …, xn) over the default the dissolved phase. However, F2, which had a relatively low peak dis-
value of 100 points (h) (MathWorks, 2017). charge, was identified with the highest number of pesticide types (29)
To estimate the partitioning of pesticides between the adsorbed and in the adsorbed phase. This suggests a strong effect of pesticide applica-
dissolved phases, we followed Lewis et al. (2017) and used the mobility tion methodology and timing, which can be affected by specific ante-
scale. This scale is based on Koc and Kfoc (with omission of the 1/n pa- cedent soil moisture conditions in the field. This was especially salient
rameter) values, representing distribution coefficients for sorption–de- in the case of F2, in which we observed a large difference in the number
sorption processes. The method converts the distribution coefficients to of pesticides between the two phases. An exceptionally high number of
a scale that ranges from “very mobile” for compounds that are almost detected pesticides (39) produced an exceptionally large load with high
entirely present in the dissolved phase, to “non-mobile” for compounds concentrations, bound to suspended sediments (Fig. 4). This finding has
that are mainly adsorbed to sediments. Between these two end mem- two possible explanations: (i) the F2 flood event occurred in March, late
bers, the grades are “mobile”, “moderately mobile” and “slightly in the rainy season, and the preceding event occurred in December, pro-
mobile”. viding a long period for pesticide application and accumulation; (ii) the
limited amount of suspended sediments obtained for analysis allowed a
3. Results and discussion single batch measurement for the entire event, creating a bias in
adsorbed pesticide load. Nevertheless, the dissolved pesticide concen-
Seven flood events (identified as F1 to F7) were sampled over tration during the same flood event (F2) was notable in terms of load,
2 years during the rainy seasons of 2013/14 (F1 and F2) and 2014/15 producing 3 kg, composed almost entirely of tebuconazole, despite the
(F3–F7). The flood event hydrographs varied in magnitude and dura- fact that this compound is defined as ‘slightly mobile’ (Lewis et al.,
tion, ranging between 2.6 and 89.7 m3 s− 1 and between 0.9 and 2017, see Section 2.3). The exceptional presence of such a compound
5.2 days, respectively (Table 1). The flood event magnitudes appeared in the dissolved phase is the result of excessive usage of a strongly per-
representative within the distributions of the flood peak discharge se- sistent compound that accumulated in the cultivated soil, and during
ries for Alexander stream (Fig. 3a). Four events' peak discharges had re- the storm event the compound distributed to runoff water. Thus, the
currence intervals of 1–1.5 years, and three events had recurrence F2 scenario, a time gap between storm events which increases the like-
intervals of 2–8 years (Fig. 3a). These data suggested that our samples, lihood of additional use of fungicides such as the persistent
collected during the storm events, reflect the typical flow regime of Al- tebuconazole, might reflect a possible outcome of increased environ-
exander stream. mental risk to the riverine system, given climate change forecasts for
The flow regime (water discharge) exhibited a positive correlation the Mediterranean region of longer dry periods between rainfall storm
with suspended sediment concentration (Fig. 3b). The derived power events (Ziv et al., 2013).
172 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177

Table 1 tebuconazole (96%; fungicide) were the most frequent compounds de-
Characteristics of seven flood events, F1–F7, and their dates (dd/mm/yyyy). The tected in runoff samples during the rainy season. Two other compounds,
hydrograph describes the change in water discharge (Q in m3 s−1 on the y axis) over time
(T in days on the x axis); QP is peak water discharge; duration refers to the total duration of
bromacil (67%; herbicide) and carbendazim (54%; fungicide), were also
the flood event; nw and ns refer to the total number of water and sediment samples collect- frequently detected. Out of these 4, diuron, tebuconazole, and bromacil,
ed, respectively. with the additional herbicide fluometuron, accounted for 95% of the
total amount of pesticides found dissolved in the floodwater samples
# Hydrograph Date Qp Duration nw ns
(m3s−1) (days) (Fig. 5a). Tebuconazole exhibited the highest measured concentration
(23 μg L−1).
F1 11/12/2013 89.7 5.2 6 4
A similar trend was observed in the adsorbed phase: 5 compounds
were detected in N50% of the sediment samples. Diuron was the most
frequent pesticide, found in all sediment samples analyzed (100%).
Two other herbicides—oxyfluorfen (73%) and bromacil (68%)—and
F2 13/03/2014 2.6 3.1 4 1 two fungicides—tebuconazole (50%) and boscalid (73%)—were fre-
quently identified. Similar to the dissolved phase, tebuconazole exhibit-
ed the highest measured concentration (5.5 μg kg− 1). However, the
cumulative contribution of the five compounds to the total amount of
pesticides in the adsorbed phase was less significant than their contri-
F3 16/11/2014 5.8 1.4 5 0
bution to the dissolved phase. A mix of 11 herbicides, fungicides and in-
secticides made up 95% of the total amount of pesticides in the adsorbed
phase (Fig. 5b). Of these, 8 were characterized as “slightly” to “non-mo-
bile” on the mobility scale explaining the larger number of pesticides in
F4 26/11/2014 21.5 2.3 7 4 the adsorbed-phase load. The other 3 compounds: bromacil,
fluometuron and carbendazim, were graded as “moderately mobile”
to “mobile” (Lewis et al., 2017). Fluometuron and bromacil, along with
diuron, are the most commonly used herbicides in the western part of
the Alexander watershed (personal communication). Although their
F5 07/01/2015 31.5 3.3 7 4
distribution coefficients favor the dissolved phase and they are scaled
as “mobile”, excessive use of this compounds in the watershed increases
their adsorbed fraction and they were abundantly detected and contrib-
ute significantly to the adsorbed phases. This trend is similar, but oppo-
F6 12/02/2015 10.2 0.9 7 4 site, to that observed for tebuconazole. The general result was a mix of
pesticides graded from “mobile” to “non-mobile” with varied distribu-
tion coefficients in both the dissolved and adsorbed phases. According
to Arora et al. (2010), the partitioning between adsorbed and dissolved
F7 13/02/2015 64.5 2.5 10 5 phases is direct linked to the identification of soil erosion as the main
loss pathway if a specific threshold per pesticide is crossed for a
known specific sediment source area. However, often, on the watershed
scale as with Alexander stream which drains a total area of 540 km2, this
is difficult to obtain for several reasons: (i) the intensity of pesticide
usage, which is often unknown, is expected to affect the pesticides'
amounts at any given time, including in the less dominant phase, ac-
These findings (Fig. 4) reveal a more complex dynamic than expect- cording to their distribution coefficients; (ii) the pesticides' physical
ed. We were interested to see how this complex dynamic is expressed in properties, combined with their application methods and field condi-
terms of pesticide composition, i.e., which pesticides are more dominant tions (e.g., water content), govern sorption processes and partitioning,
than others. Fig. 5 lists most of the pesticides (40) detected in each i.e., the affinity of the compound to the soil organic matter and clay frac-
phase. In the dissolved phase, only 4 compounds were found in N50% tion in the cultivated fields. However, these steady-state conditions may
of the water samples. Among them, diuron (89%; herbicide) and change during the storm event, e.g., additional excessive rainfall may

Fig. 3. The Alexander stream hydrological regime (1967–2015). Peak discharge values vs. number of flood events (a) and peak discharge vs. recurrence interval (RI), inset. Note that F1–F7
flood events are distributed according to their peak discharge. (b) Suspended sediment concentration (SSC) relative to discharge measured in flood events (F1–F7) in Alexander stream
during the hydrological years 2013/14 and 2014/15.
Baseflow data retrieved from Ruppin Estuarine and Coastal Observatory, http://reco.ruppin.ac.il.
 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177 173

example of this screening is the herbicides, which in many cases are

quite mobile, yet they are still dominant in the adsorbed phase, proba-
bly due to their excessive use. Indeed, according to the Israel Central
Bureau of Statistics (2016), herbicides are the most commonly used
pesticide.
The analysis of pesticide concentrations confirmed this information.
The data set that included various pesticides was grouped according to
their functionality—herbicides, fungicides and insecticides—to examine
whether a generic trend could be more easily detected for each phase.
Fig. 6a, b shows the kernel probability density function (KPDF) fit to
concentrations for the dissolved phase. In general, herbicide concentra-
tions were frequently observed in water samples; the maximum insec-
ticide concentration was two orders of magnitude lower than the
maximum herbicide and fungicide concentrations, though the estimat-
ed KPDF for the three groups demonstrated unimodal distribution
skewed to the left, toward lower concentrations. However, additional
peaks, i.e., fungicide and insecticide kernels, are clearly represented.
This may result from water samples having exceptionally high pesticide
concentrations that affect KPDF values, specifically, water samples mea-
Fig. 4. Number of different pesticides (herbicides, fungicides, and insecticides) (a) sured in flood events F6 and F7 for herbicides, F2 for fungicides, and F4
identified NLOD and bLOQ, and (b) quantified NLOQ, dissolved in the water column or and F5 for insecticides. Surprisingly, these samples originated from five
adsorbed to suspended sediment in different flood events (F1–F7). different flood events, suggesting the absence of a direct relationship to
the physical control of water discharge and flow regime.
lead to additional sorption and desorption processes, such as the release The KPDF fitting for pesticide concentrations in the adsorbed phase
of bound compounds to flowing water; (iii) monitoring water quality at (Fig. 6c, d) demonstrated a difference of one order of magnitude be-
one location at the outlet of the watershed integrates different hydro- tween fungicides and insecticides, and between herbicides and insecti-
logical responses and sediment sources from different sub-watersheds cides. For example, the highest concentration of herbicides and
and therefore, the effect of soil type and field conditions that determine fungicides was 1000 μg kg−1, whereas the highest insecticide concen-
the steady states at the source areas, weakens along the flow path due to tration was 100 μg kg−1. Exceptional extreme values, e.g., 600 and
the mixing effect and additional sorption–desorption processes. A good 4000 μg kg−1 for insecticides and herbicides, respectively (attributes

Fig. 5. Active ingredient relative frequencies (bars) and cumulative amounts (red dots) identified in the dissolved phase in water samples (a) and in suspended sediment samples (b). Note
that the black and red dashed lines refer to the 50th and 95th percentiles, respectively. Different bar patterns distinguish between insecticides (i), fungicides (f) and herbicides (h). Note
that the lists of pesticides in (a) and (b) are not identical and only 40/45 and 40/47 compounds are shown in the dissolved phase and suspended sediments, respectively.
174 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177

Fig. 6. KPDF for fungicides, herbicides, and insecticides in the dissolved (a, b) and adsorbed (c, d) phases. Note that insecticides in both phases are plotted separately due to differences in
scale for their PDF values. This is because the normal kernel function integrates over 100 equally spaced points that cover the range of the input univariate data.

of F2, with a single sample for the entire event) affected the KPDF values regression models applied to the physical properties (floodwater peak
(Fig. 6). Additional high values were also measured in F4, F6 and F7 for discharge and suspended sediments) that control the amount of each
all three groups in samples that were collected along the falling limb of pesticide group transported in Alexander stream. Fig. 7 shows the statis-
the hydrograph. The contribution per flood event of each pesticide tically significant relationship between floodwater peak discharge and
group of agrochemicals is shown in Table 2. Of the total pesticide load, herbicide amounts, as well as between suspended sediments and herbi-
10% was contributed by suspended sediments, with the variability be- cide amounts, for the hydrological year 2014/15. We analyzed this year
tween flood events ranging from 4% to 33% in F7 and F2, respectively. independently because the hydrological year 2013/14 was a drought
Furthermore, significant differences were observed between the pesti- year with only two flood events, of which F2 exhibited a strong anomaly
cide groups. Only 7% of the herbicide load was transported with (Figs. 4 and 6). The hydrological year 2014/15 provided a sufficient
suspended sediments, whereas 20% and 42% of fungicides and insecti- number of flood events to enable best-fit power-trend lines to describe
cides, respectively, were bound to suspended sediments. these relationships. Interestingly, the relationships showed a strong cor-
Based on these findings, it may be that no positive correlation exists relation with flood event magnitudes, with respect to both dissolved
between suspended sediments and pesticide concentrations and there- and adsorbed phases. This means that in general, large flood events
fore, we are unable to make any generalizations with regard to transported high amounts of herbicides and the chronological sequence
streamwater quality. We examined this issue through a series of of the event along the hydrological year was less significant. For

Table 2
Amount of pesticides (g) in seven flood events (F1–F7) sampled during the hydrological years 2013/14 and 2014/15. Numbers in parentheses represent the number of pesticides com-
posing 95% of the total amount in each flow event.

Flow event F1 F2 F3 F4 F5 F6 F7
Peak discharge (m3 s−1) 89.7 2.6 5.8 21.5 31.5 10.2 64.5
Fungicides Dissolved 1249 (1) 3011 (1) 720 (1) 854 (4) 955 (4) 289 (2) 471 (3)
Adsorbed 183 (3) 919 (5) NA 553 (5) 57 (2) 21 (6) 47 (3)
Herbicides Dissolved 218 (3) 17 (3) 303 (1) 2391 (1) 7688 (3) 1531 (1) 24,694 (3)
Adsorbed 270 (3) 471 (6) NA 385 (4) 576 (4) 141 (5) 1002 (4)
Insecticides Dissolved 10 (1) 0 (0) 15 (3) 196 (6) 573 (6) 17 (1) 29 (1)
Adsorbed 187 (4) 70 (9) NA 179 (3) 10 (1) 4 (1) 8 (1)
 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177 175

Fig. 7. Herbicide amounts vs. peak discharge (a) and sediment load (b) during the hydrological year 2014/2015 (F3–F7). Note flood event F2 is only fitted for the dissolved phase.

example, F7 was the last flood event during 2014/15, but it still mobi- suspended sediment-bound pesticides is available. To evaluate the actu-
lized the largest quantity of herbicides in both phases. This implies al toxicity potential of the total pesticide load, this adsorbed fraction
that in Alexander stream, and probably in other similar coastal Mediter- must be taken into consideration.
ranean streams in Israel, herbicides are abundant in the watershed, and Insecticide toxicity was found to be relatively high considering its
sediment supply is not a limiting factor. However, we observed differ- smaller contribution to the total pesticide load (Table 2). Herbicides,
ences in the range of herbicides transported in the dissolved vs. on the other hand, are used in the watershed intensively and applied
adsorbed phases. For example, F7 transported almost 24 kg of herbi- over large areas, posing a potentially higher risk to environmental
cides in the dissolved phase at its peak discharge, but only 1 kg of herbi- stream health. Fungicides seem to play a minor role in the overall aquat-
cides in the adsorbed phase, representing an order of magnitude ic toxicity; nevertheless, the concentrations are well above the chronic
difference. No similar significant relationships were found for fungicides threat benchmark for fish, suggesting a specific stress targeting fish pop-
or insecticides (Table 2). Fungicide and insecticide dynamics were char- ulations. This additional information highlights the importance, both in
acterized by pulses rather than a positive correlation with flood event magnitude and in the diversity of toxicity stress, of chemical transport
magnitude. This outcome can be partly explained by the fact that herbi- during flood events, whereas no such potential was observed during
cides are sprayed over vast areas of bare soil, mostly twice a year, at the baseflow sampling. This pulse-like stress may also affect microorgan-
end of the winter, whereas fungicides and insecticides are applied di- isms since they are not exposed to the pesticides constantly at low
rectly to the plant or to its root system year-round. chronic levels (DeLorenzo et al., 2001).
The coupling between flood magnitude and pesticide application
methodology and timing directly affects potential aquatic toxicity of 4. Conclusions
transported pesticides into the riverine environment. Initially, the indi-
vidual pesticides' toxicity potential was estimated using two major da- This study provides significant basic information on pesticide-load
tabases (USEPA, 2012; Lewis et al., 2017). These benchmarks are dynamics in Alexander stream—a Mediterranean watershed with inten-
based on freshwater experimental toxicity studies and reflect the com- sive agriculture over half of its contributing area. It is evident that pesti-
pounds' chemical potential to impact biota. Of the 63 pesticides detect- cides enter the riverine system during flood events at concentrations
ed in this study, hexachlorobenzene, metolachlor and DDT are known to that greatly exceed those entering the stream via the baseflow through-
be carcinogenic agents and 29 are marked as “suspected carcinogenic out the year. Pesticide loads within the riverine system were seen to ex-
agents”. The ecotoxicology benchmarks for aquatic environments in- hibit complex dynamics controlled by flood event magnitude and
clude acute and chronic values for fish, invertebrates, crustaceans, sed- sediment supply in the case of herbicides, and pesticide-application
iment-dwelling organisms, algae and vascular and non-vascular methods and timing in the case of insecticides and fungicides. Because
plants. A total of 9 pesticides (2 fungicides: carbendazim and herbicide concentrations are high and they are frequently found dis-
tebuconazole; 4 herbicides: atrazine, bromacil, diuron, linuron; 3 insec- solved in the water or adsorbed to suspended sediments, development
ticides: chlorpyrifos, diazinon, permethrin) were detected in our sam- of pesticide pollution-monitoring programs should be based on the re-
ples at least once, with higher concentrations than the defined toxicity lations found with flood peak discharge and with suspended sediment
benchmark (Table 3). However, this information does not provide a concentration.
complete ecotoxicology assessment because no such database for We suggest that the coupling between flood magnitude and pesti-
cide application methodology and timing directly affects potential
aquatic toxicity of pesticides transported into the riverine environment.
Table 3
Pesticide (f = fungicide; h = herbicide; i = insecticide) peak concentrations (μg L−1) de- The significant toxicity contribution from dissolved-phase transport rel-
tected in 7 flood events (F1–F7) and the aquatic taxonomic groups that are potentially ative to the smaller contribution of suspended sediment-bound
threatened (Lewis et al., 2017; USEPA, 2017): 1 – fish, 2 – invertebrates, 3 – crustaceans, chemicals, however, emphasizes the importance of the former fraction
4 – algae, 5 – non-vascular plants, 6 – vascular plants. to ecotoxicology assessment in floodwater. This pulse-like stress may
Agricultural chemical Peak dissolved Taxonomic Taxonomic groups affect microorganisms and may lead to a chain reaction in the aquatic
generic name and concentration groups acutely chronically threatened environment. However, application of this information is still limited
functional group (μg L−1) threatened since this study does not consider the effects of geochemical conditions,
Carbendazimf 1 0 2 which are dynamic as well and may change, e.g., due to sediment depo-
Tebuconazolef 23 0 1, 2 sition or resuspension, water salinity, or dissolved oxygen concentra-
Atrazineh 0.6 5 5
tions, among others.
Bromacilh 8 5 5
Diuronh 16 4, 5, 6 4, 5, 6 Among the detected pesticides, only a limited number of herbicides
Linuronh 0.2 0 2 entered Alexander stream, and were detected in all flood events and all
Chlorpyrifosi 0.05 2, 3 2, 3 samples, both in the dissolved phase and bound to suspended sedi-
Diazinoni 0.3 2 2 ments. Diuron's potential toxicity to aquatic organisms, frequent abun-
Permethrini 0.03 2, 3 2, 3
dance in the Alexander watershed and in other agricultural watersheds,
176 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177

and relatively high persistence in the environment—found both dis- Garcia-Ruiz, J.M., Begueria, S., Nadal-Romero, E., Gonzalez-Hidalgo, J.C., Lana-Renault, N.,
Sanjuan, Y., 2015. A meta-analysis of soil erosion rates across the world. Geomorphol-
solved in water and bound to sediments (Lewis et al., 2017)—make it ogy 239:160–173. https://doi.org/10.1016/j.geomorph.2015.03.008.
a potentially reliable marker for herbicides in water quality-monitoring Gasith, A., Resh, V.H., 1999. Streams in Mediterranean climate regions: abiotic influences
projects. and biotic responses to predictable seasonal events. Annu. Rev. Ecol. Syst. 30:51–81.
https://doi.org/10.1146/annurev.ecolsys.30.1.51.
An effective soil-conservation measure, such as protection of the soil Hershkovitz, Y., Gasith, A., 2013. Resistance, resilience, and community dynamics in med-
surface with sufficient mulch (live or dead), may reduce the runoff iterranean-climate streams. Hydrobiologia 719:59–75. https://doi.org/10.1007/
amount by 50% and peak discharge by up to 80% (Eshel et al., 2015). s10750-012-1387-3.
Israel Central Bureau of Statistics, 2016. Survey of pesticides in agriculture and sanitation
This could substantially decrease agrochemical transport and the 2011–2013. http://www.cbs.gov.il/www/hodaot2016n/07_16_294b.doc.
resulting stress on aquatic environments, particularly that imposed by Köhler, H.-R., Triebskorn, R., 2013. Wildlife ecotoxicology of pesticides: can we track ef-
herbicides, which showed high concentrations in the surface flow de- fects to the population level and beyond? Science 341:759–765. https://doi.org/
10.1126/science.1237591.
rived from sheet erosion, and made up about 80% of the total pesticide
Kronvang, B., Laubel, A., Larsen, S.E., Friberg, N., 2003. Pesticides and heavy metals in Dan-
load. Implementing soil-conservation measures will not only limit fur- ish streambed sediment. Hydrobiologia 494:93–101. https://doi.org/10.1023/A:
ther contamination of surface water but will also encourage farmers to 1025441610434.
reduce quantities of applied chemicals, improve application methods, Larson, S.J., Capel, P.D., Goolsby, D.A., Zaugg, S.D., Sandstrom, M.W., 1995. Relations be-
tween pesticide use and riverine flux in the Mississippi River basin. Chemosphere
and promote soil conservation, resulting in improved efficiency, lower 31:3305–3321. https://doi.org/10.1016/0045-6535(95)00176-9.
costs and protection of stream quality. Lewis, K.A., Tzilivakis, J., Warner, D., Green, A., 2017. The PPDB: Pesticide Properties Data-
Base. University of Hertfordshirehttp://sitem.herts.ac.uk/aeru/ppdb/en/index.htm,
Accessed date: 17 January 2017.
Acknowledgment Liess, M., Von Der Ohe, P.C., 2005. Analyzing effects of pesticides on invertebrate commu-
nities in streams. Environ. Toxicol. Chem. 24, 954–965.
MathWorks, 2017. Kernel distibution. https://www.mathworks.com/help/stats/kernel-
This manuscript was improved by comments made by two anonyms distribution.html.
reviewers. Support provided to the first author by the Mediterranean Minami, K., Fukushi, S., 1984. Methods for measuring N2O flux from water surface and
Sea Research Center of Israel helped to complete this research project. N2O dissolved in water from agricultural land. Soil Sci. Plant Nutr. 30 (4):495–502.
https://doi.org/10.1080/00380768.1984.10434716.
Fieldwork assistance provided by the team at the Ruppin Estuarine Montgomery, D.R., 2007. Soil erosion and agricultural sustainability. Proc. Natl. Acad. Sci.
and Coastal Observatory (RECO) of the School of Marine Science, Ruppin U. S. A. 104:13268–13272. https://doi.org/10.1073/pnas.0611508104.
Academic Center, is highly appreciated. Mora, M.A., 1997. Transboundary pollution: persistent organochlorine pesticides in mi-
grant birds of the southwestern United States and Mexico. Environ. Toxicol. Chem.
16:3–11. https://doi.org/10.1897/1551-5028(1997)016b0003:TPPOPIN2.3.CO;2.
References Morin, J., Benyamini, Y., 1977. Rainfall infiltration into bare soils. Water Resour. Res. 13,
813–817.
Arora, K., Mickelson, S.K., Helmers, M.J., Baker, J.L., 2010. Review of pesticide retention Neal, C., Bowes, M., Jarvie, H.P., Scholefield, P., Leeks, G., Neal, M., Rowland, P., Wickham,
processes occurring in buffer strips receiving agricultural runoff1. J. Am. Water H., Harman, S., Armstrong, L., Sleep, D., Lawlor, A., Davies, C.E., 2012. Lowland river
Resour. Assoc. 46:618–647. https://doi.org/10.1111/j.1752-1688.2010.00438.x. water quality: a new UK data resource for process and environmental management
Ayas, Z., Barlas, N.E., Kolankaya, D., 1997. Determination of organochlorine pesticide res- analysis. Hydrol. Process. 26:949–960. https://doi.org/10.1002/hyp.8344.
idues in various environments and organisms in Göksu Delta, Turkey. Aquat. Toxicol. Ng, C.M., Weston, D.P., Lydy, M.J., 2012. Pyrethroid insecticide transport into Monterey
39:171–181. https://doi.org/10.1016/S0166-445X(96)00849-1. Bay through riverine suspended solids. Arch. Environ. Contam. Toxicol. 63:461–470.
Beketov, M.A., Kefford, B.J., Schäfer, R.B., Liess, M., 2013. Pesticides reduce regional biodi- https://doi.org/10.1007/s00244-012-9796-x.
versity of stream invertebrates. Proc. Natl. Acad. Sci. U. S. A. 110:11039–11043. Nowell, L.H., Norman, J.E., Ingersoll, C.G., Moran, P.W., 2016. Development and application
https://doi.org/10.1073/pnas.1305618110. of freshwater sediment-toxicity benchmarks for currently used pesticides. Sci. Total
Bennett, E.M., Carpenter, S.R., Caraco, N.F., 2001. Human impact on erodable phosphorus Environ. 550:835–850. https://doi.org/10.1016/j.scitotenv.2016.01.081.
and eutrophication: a global perspective: increasing accumulation of phosphorus in Oerke, E.C., 2006. Crop losses to pests. J. Agric. Sci. 144:31–43. https://doi.org/10.1017/
soil threatens rivers, lakes, and coastal oceans with eutrophication. AIBS Bull. 51 S0021859605005708.
(3), 227–234. Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., Tadić, M.P.,
Bereswill, R., Streloke, M., Schulz, R., 2013. Current-use pesticides in stream water and Michaelides, S., Hrabalíková, M., Olsen, P., Aalto, J., Lakatos, M., Rymszewicz, A.,
suspended particles following runoff: exposure, effects, and mitigation requirements. Dumitrescu, A., Beguería, S., Alewell, C., 2015a. Rainfall erosivity in Europe. Sci.
Environ. Toxicol. Chem. 32 (6), 1254–1263. Total Environ. 511:801–814. https://doi.org/10.1016/j.scitotenv.2015.01.008.
Birch, G.F., Drage, D.S., Thompson, K., Eaglesham, G., Mueller, J.F., 2015. Emerging contam- Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K., Montanarella, L.,
inants (pharmaceuticals, personal care products, a food additive and pesticides) in Alewell, C., 2015b. The new assessment of soil loss by water erosion in Europe. Envi-
waters of Sydney estuary, Australia. Mar. Pollut. Bull. 97:56–66. https://doi.org/ ron. Sci. Pol. 54:438–447. https://doi.org/10.1016/j.envsci.2015.08.012.
10.1016/j.marpolbul.2015.06.038. Parzen, E., 1962. On estimation of a probability density function and mode. Ann. Math.
Blann, K.L., Anderson, J.L., Sands, G.R., Vondracek, B., 2009. Effects of agricultural drainage Stat. 33 (3), 1065–1076.
on aquatic ecosystems: a review. Crit. Rev. Environ. Sci. Technol. 39:909–1001. Pérez, D.J., Okada, E., De Gerónimo, E., Menone, M.L., Aparicio, V.C., Costa, J.L., 2017. Spatial and
https://doi.org/10.1080/10643380801977966. temporal trends and flow dynamics of glyphosate and other pesticides within an agricul-
Bundschuh, M., Goedkoop, W., Kreuger, J., 2014. Evaluation of pesticide monitoring strat- tural watershed in Argentina. Environ. Toxicol. Chem. https://doi.org/10.1002/etc.3897.
egies in agricultural streams based on the toxic-unit concept–experiences from long- Restek, 2012. Fast, simple sample prep for multiresidue pesticide analysis. http://
term measurements. Sci. Total Environ. 484:84–91. https://doi.org/10.1016/ www.restek.com/pdfs/FFBR1183D-UNV.pdf.
j.scitotenv.2014.03.015. Rodrigues, E.T., Pardal, M.Â., Gante, C., Loureiro, J., Lopes, I., 2017. Determination and val-
Carpenter, K.D., Kuivila, K.M., Hladik, M.L., Haluska, T., Cole, M.B., 2016. Storm-event- idation of an aquatic Maximum Acceptable Concentration-Environmental Quality
transport of urban-use pesticides to streams likely impairs invertebrate assemblages. Standard (MAC-EQS) value for the agricultural fungicide azoxystrobin. Environ.
Environ. Monit. Assess. 188:1–18. https://doi.org/10.1007/s10661-016-5215-5. Pollut. 221:150–158. https://doi.org/10.1016/j.envpol.2016.11.058.
Chapman, P.M., Wang, F., Caeiro, S.S., 2013. Assessing and managing sediment contamina- Schwarzenbach, R.P., Escher, B.I., Fenner, K., Hofstetter, T.B., Johnson, C.A., von Gunten, U.,
tion in transitional waters. Environ. Int. 55:71–91. https://doi.org/10.1016/ Wehrli, B., 2006. The challenge of micropollutants in aquatic systems. Science 313:
j.envint.2013.02.009. 1072–1077. https://doi.org/10.1126/science.1127291.
Covino, T., 2017. Hydrologic connectivity as a framwork for understanding biogeochemi- Segal-Rozenhaimer, M., Shavit, U., Vengosh, A., Gavrieli, I., Farber, E., Holtzman, R., Mayer,
cal flux through watersheds and along fluvial networks. Geomorphology 277, B., Shaviv, A., 2004. Sources and transformations of nitrogen compounds along the
133–144 (do). Lower Jordan River. J. Environ. Qual. 33:1440–1451. https://doi.org/10.2134/
DeLorenzo, M.E., Scott, G.I., Ross, P.E., 2001. Toxicity of pesticides to aquatic microorganisms: jeq2004.1440.
a review. Environ. Toxicol. Chem. 20:84–98. https://doi.org/10.1002/etc.5620200108. Sharma, P.P., Gupta, S.C., Rawls, W.J., 1991. Soil detachment by single raindrops of varying
EPA, 2007a. Test Methods for Evaluating Solid Waste, Physical/Chemical Methods, EPA kinetic energy. Soil Sci. Soc. Am. J. 55:301–307. https://doi.org/10.2136/
publication SW-846, Third Edition, Final Updates I (1993), II (1995), IIA (1994), IIB sssaj1991.03615995005500020001x.
(1995), III (1997), IIIA (1999), IIIB (2005), IV (2008), and V (2015). Silverman, B.W., 1986. Density Estimation for Statistics and Data Analysis. vol. 26. CRC
EPA, 2007b. Ultrasonic Extraction. Cambridge University Press, Cambridge https:// Press.
doi.org/10.1017/CBO9781107415324.004. Stieglitz, M., Shaman, J., McNamara, J.P., Engel, V., Shanley, J., Kling, G.W., 2003. An ap-
Eshel, G., Egozi, R., Goldwasser, Y., Kashti, Y., Fine, P., Hayut, E., Kazukro, H., Rubin, B., Dar, proach to understanding hydrologic connectivity on the hillslope and the implica-
Z., Keisar, O., DiSegni, D.M., 2015. Benefits of growing potatoes under cover crops in a tions for nutrient transport. Glob. Biogeochem. Cycles 17:1–15. https://doi.org/
Mediterranean climate. Agric. Ecosyst. Environ. 211, 1–9. 10.1029/2003GB002041.
Fasola, M., Movalli, P.A., Gandini, C., 1998. Heavy metal, organochlorine pesticide, and PCB Syvitski, J.P.M., Vörösmarty, C.J., Kettner, A.J., Green, P., 2005. Impact of humans on the
residues in eggs and feathers of herons breeding in Northern Italy. Arch. Environ. flux of terrestrial sediment to the global coastal ocean. Science 308:376–380.
Contam. Toxicol. 34:87–93. https://doi.org/10.1007/s002449900289. https://doi.org/10.1126/science.1109454.
 T. Topaz et al. / Science of the Total Environment 625 (2018) 168–177 177

Tal, A., Al Khateeb, N., Asaf, L., Assi, A., Nassar, A., Muath Sadah, A., Gasith, A., Laronne, J., Valentin, C., Poesen, J., Li, Y., 2005. Gully erosion: impacts, factors and control. Catena 63:
Ronen, Z., Hirshkowitz, Y., Halawani, D., Nagouker, N., Angel, R., Akerman, H., 132–153. https://doi.org/10.1016/j.catena.2005.06.001.
Diabat, M., Abramson, A., 2007. Watershed Modeling: BioMonitoring and Economic Vörösmarty, C.J., McIntyre, P.B., Gessner, M.O., Dudgeon, D., Prusevich, A., Green, P.,
Analysis to Determine Optimal Restoration Strategies for Transboundary Streams. Glidden, S., Bunn, S.E., Sullivan, C.A., Reidy Liermann, C., Davies, P.M., 2010. Global
The Arava Institute for Environmental Studies, Israel. threats to human water security and river biodiversity. Nature 467:555–561.
Turner, A., 2003. Salting out of chemicals in estuaries: implications for contaminant https://doi.org/10.1038/nature09549.
partitioning and modelling. Sci. Total Environ. 314–316:599–612. https://doi.org/ Ziv, B., Saaroni, H., Pargament, R., Harpaz, T., Alpert, P., 2013. Trends in rainfall regime over
10.1016/S0048-9697(03)00076-7. Israel, 1975–2010, and their relationship to large-scale variability. Reg. Environ.
USEPA, 2017. Aquatic Life Benchmarks and Ecological Risk Assessments for Registered Chang. 14:1751–1764. https://doi.org/10.1007/s10113-013-0414-x.
Pesticides. https://www.epa.gov/pesticide-science-and-assessing-pesticide-risks/
aquatic-life-benchmarks-and-ecological-risk. Accessed 17 January 2017.