Science of the Total Environment 613–614 (2018) 1250–1262

Contents lists available at ScienceDirect

Science of the Total Environment

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

Tracking pesticide fate in conventional banana cultivation in Costa Rica: A

disconnect between protecting ecosystems and consumer health
Annelle Mendez a, Luisa E. Castillo b, Clemens Ruepert b, Konrad Hungerbuehler a, Carla A. Ng c,⁎
a
Safety and Environmental Technology Group, ETH Zurich, Vladimir-Prelog-Weg 1, CH-8093 Zurich, Switzerland
b
Central American Institute for Studies on Toxic Substances, Universidad Nacional, Heredia, Costa Rica
c
Department of Civil and Environmental Engineering, University of Pittsburgh, 3700 O'Hara St, Pittsburgh, PA 15261, USA

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

• New multimedia model describes pesti-

cide dynamics from banana production.
• This model contains a dynamic vegeta-
tion compartment integrated with
environment.
• Pesticide concentrations in water regu-
larly exceed thresholds for ecosystem
health.
• Pesticide residues in bananas are below
maximum residue limits.
• Consumers are safe, but the environ-
ment is not likewise protected.

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

Article history: Conventional banana cultivation in Costa Rica relies on heavy pesticide use. While pesticide residues in exported
Received 27 July 2017 bananas do not generally represent a safety concern for consumers abroad, ecosystem and human health in pro-
Received in revised form 17 September 2017 ducing regions are not likewise protected. In Costa Rica, most studies on pesticide residues in the environment
Accepted 17 September 2017
are snapshots, limiting our ability to identify temporal dynamics that can inform risk mitigation strategies. To
Available online 25 September 2017
help bridge this gap, we created a dynamic multimedia model for the Caño Azul River drainage area, which is
Editor: D. Barcelo heavily influenced by banana and pineapple plantations. This model estimates chemical concentrations in
water, air, soil, sediments, and banana plants through time, based on pesticide properties and emission patterns
Keywords: and on variable environmental conditions. Case studies for three representative chemicals—the herbicide diuron,
Pesticides the nematicide ethoprofos, and the fungicide epoxiconazole—show that concentrations in fruit remain below EU
MRL and US maximum residue limits set to ensure consumer health, while those in the environment are highly var-
Ecosystem health iable, reaching peak concentrations in water that can exceed thresholds for ecosystem health. Critical research
Environmental justice needs, including incorporating sediment dynamics and the effects of adjuvants on the properties and transport
Export crops
of active ingredients into multimedia models, were identified.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction scale monocultures, where intensive agrochemical use is the main

Costa Rica is the 3rd largest exporter of bananas in the world
(FAO, 2014). Most export bananas in Costa Rica are grown in large-
⁎ Corresponding author.
E-mail address: carla.ng@pitt.edu (C.A. Ng).
method of pest control (Bellamy, 2013). Approximately 76 kg of active
ingredient (a.i.) are applied per hectare each year (Echeverría-Sáenz
et al., 2016), including a variety of pre- and post-harvest fungicides,
nematicides, herbicides, and insecticides (Castillo et al., 2006, 1997;
Polidoro et al., 2008). Banana cultivation in Costa Rica is concentrated
on the Caribbean Coast, where soils and weather conditions are optimal,

https://doi.org/10.1016/j.scitotenv.2017.09.172
0048-9697/© 2017 Elsevier B.V. All rights reserved.
 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
particularly the frequent rainfall, which precludes the need for irriga-
tion. This area also harbors highly biodiverse wetland areas and coastal
lagoons. Most of Costa Rican banana exports go to Europe (55%) and the
United States (36%) (CORBANA, 2016). Both regions monitor pesticide
residues in high consumption crops, including bananas, to ensure that
maximum residue limits (MRL) set to protect consumers are not
exceeded. The latest results of the European and US monitoring pro-
grams have shown that the vast majority of pesticide levels in bananas
are below these MRLs and hence rarely pose a risk to consumers (EFSA,
2015; EFSA, 2014; EFSA, 2013; EFSA, 2011; FDA, 2017; USDA, 2016a;
USDA, 2016b; USDA, 2014a; USDA, 2014b; USDA, 2008; USDA, 2007).
MRLs are intended to reflect both toxicologically acceptable intakes
for consumers and good agricultural practices (European Union,
2005). Yet evidence suggests that populations and ecosystems in pro-
ducing regions, unlike consumers in importing countries, are not ade-
quately protected by these practices. Although the rise in pesticide
imports and use in Costa Rica is well documented (de la Cruz et al.,
2014; Ramírez et al., 2009) there is little environmental regulatory in-
frastructure, and risk assessment is only required for pesticide registra-
tion purposes (Polidoro and Morra, 2016; Rämö et al., 2016). Moreover,
water quality criteria are not enforced and are too general, only covering
the sum of organochlorines and of organophosphates (de la Cruz and
Castillo, 2002). Until recently, no coordinated national monitoring pro-
gram for surface water quality has been implemented, so monitoring of
pesticide residues in the environment is mainly carried out as indepen-
dent research projects at national universities (de la Cruz et al., 2014; de
la Cruz and Castillo, 2002). During the last two decades, these studies
have regularly detected residues of pesticides used for banana cultiva-
tion in drainage canals, packaging plant effluents, streams, rivers and
coastal lagoons in the Caribbean Coast, in association with negative ef-
fects on biota, such as fish mortality incidents and changes in macroin-
vertebrate community structure and species richness (Castillo et al.,
2006; Diepens et al., 2014; Echeverría-Sáenz et al., 2016). Pesticide res-
idues have also been found in blood of spectacled caimans, which due to
their high trophic level are good indicators of ecosystem health (Grant
et al., 2013). Although banana plantations are often close to villages
whose residents use rivers and groundwater for consumption and fish-
ing (Polidoro et al., 2008; van Wendel de Joode et al., 2014) only few
studies have investigated non-occupational human exposure to the
pesticides used in these plantations. Recent biomonitoring studies
found neurotoxic and thyroid-impairing metabolites of the fungicide
mancozeb and the insecticide chlorpyrifos in the urine of pregnant
women and children, respectively, living near banana plantations
(van Wendel de Joode et al., 2014, 2012). Pesticide concentrations in
72% of the women and 82% of the children resulted in a daily intake
above USEPA reference doses for chronic exposure.
While studies documenting the presence and effects of agricultural
pesticides in non-target sites are key to inform risk assessment, most
of them represent snapshots, resulting in scattered insights on pesticide
fate and transport. The few studies that do take into account the tempo-
ral aspect use sampling schemes that may not capture application
events and climate conditions that could result in peak concentrations
(Castillo et al., 2000; Rämö et al., 2016), or only track the short-term
evolution of pesticide concentrations for a few days (Castillo et al.,
2006; Mortensen et al., 1998). Most studies have detected high variabil-
ity in pesticide concentrations in water, and associated toxicity risks,
within and among sampling sites (Rämö et al., 2016). In addition, not
only the amounts but also the types and number of chemicals detected
vary among studies. It is unclear whether this variability is due to chem-
ical properties, environmental factors (i.e., Diepens et al., 2014; Polidoro
and Morra, 2016), changes in the use pattern, or targeting different
chemicals.
The spatial and temporal distribution of pesticide concentrations in
the environment is a result of the interplay of many elements, including
the timing and amount of pesticide applications, climate, and chemical
properties. Thus, current approaches based on pesticide imports, use
1251
intensity, or chemical properties alone are not sufficient to identify pri-
ority chemicals for risk reduction in Costa Rica (Castillo et al., 2006; de la
Cruz and Castillo, 2002; Diepens et al., 2014; Mena et al., 2014;
Mortensen et al., 1998). In response to this, we created a dynamic mul-
timedia model that integrates the aforementioned elements to describe
pesticide fate and transport in the drainage area of the Caño Azul River, a
tributary of the Madre de Dios River (MDR) influenced by three banana
and one pineapple plantation. Measurement campaigns in the MDR wa-
tershed from 2011 to 2012 and results of ecological assessment models
showed that among all study sites, pesticide concentrations measured
in the Caño Azul River from 2011 to 2012 posed the highest risks to
aquatic biota and contributed significantly to the pesticide load of
downstream surface waters in the MDR watershed (Echeverría-Sáenz
et al., 2016; Rämö et al., 2016).
The Caño Azul model estimates chemical concentrations in water,
air, soil, sediments and banana plants through time. To represent the
different types of pesticides and application modes used in banana cul-
tivation, the model tracks the fluxes of ethoprofos, diuron and
epoxiconazole, which have been the most commonly detected nemati-
cide, herbicide, and fungicide, respectively, in the MDR watershed and
throughout the Madre de Dios coastal lagoon (Diepens et al., 2014;
Rämö et al., 2016). These chemicals have been interpreted as the main
contributors to adverse changes in the macroinvertebrate community
composition (Echeverría-Sáenz et al., 2016), and to risk for primary pro-
ducers and arthropods (Rämö et al., 2016) in the MDR watershed.
Drawing on earlier models of chemical uptake and transport in vegeta-
tion (Cousins and Mackay, 2001; Fantke et al., 2011; Trapp, 2007;
Undeman et al., 2009; Wania and McLachlan, 2001) we created a vege-
tation compartment that captures the growth dynamics of banana
plants and their interactions with the environment at the plantation
scale. Existing dynamic models have largely focused on single plants
and on estimating human exposure via food consumption (Fantke et
al., 2011; Trapp and Eggen, 2013; Undeman et al., 2009). The goals of
our newly developed banana-environment model are to a) assess
when and whether pesticide residues in bananas and the environment
exceed benchmarks for both human and ecosystem health under cur-
rent application practices, and b) identify the most important factors
that determine pesticide fate in the Caño Azul drainage area.
2. Methods
2.1. Study site description
The Caño Azul River is located in the canton of Siquirres, province of
Limón, in the Caribbean Lowlands of Costa Rica (see Appendix A,
Fig. A.1). Caño Azul is a 12.3 km long shallow river with water levels
ranging from 0.2 to 0.7 m, depending on rainfall, which flows into the
MDR. The MDR ultimately drains into a coastal lagoon with an outlet
to the Caribbean Sea. Banana, mostly conventionally grown, is the
main crop in the MDR watershed, covering nearly one third of its area
(Rämö et al., 2016). Three banana plantations, with a total area of
1187 ha, as well as one pineapple plantation, with an area of 386 ha,
are located within the drainage area of the Caño Azul River (Appendix
A, Fig. A.1).
2.2. Multimedia model
2.2.1. General description
A dynamic multimedia mass balance model for the Caño Azul River
basin was developed using the fugacity approach, briefly described in
Appendix A.2 (see Mackay, 2001 for further details). The system is rep-
resented by 10 compartments: air, water, sediment, soil cultivated with
pineapple, 3 soil compartments cultivated with bananas (one for each
plantation), and 3 banana vegetation compartments (Appendix A, Fig.
A.2). Each vegetation compartment consists of 4 subcompartments:
leaf surface deposit, leaf, fruit, and pseudostem (see Appendix A.3.2
1252 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
for details). Chemical fluxes to and from each compartment are
expressed as D-values that describe pesticide transport across environ-
mental media using mass transfer coefficients and chemical loss pro-
cesses from each medium through first-order degradation rate
constants and advective flow rates. While it is assumed that each com-
partment is spatially homogenous (perfectly mixed), chemical inputs
into the system through pulse emissions, parameters describing inter-
compartmental transport and loss, and the dimensions of the water
and vegetation compartments change through time. To take this vari-
ability into account, dynamic mass balances for each compartment
were solved numerically, yielding daily chemical concentrations and
mass fluxes. Details on the derivation of D-values and mass balance cal-
culations are presented in Appendix A.2 and A.6.
Whenever available, environmental properties specific to the Caño
Azul River basin, the MDR watershed, or the Atlantic Coast of Costa
Rica were used. Parameters specific to the Caño Azul watershed includ-
ed suspended solid concentrations, water flow and depth, and the or-
ganic matter content, density and porosity of the banana soil. Water
runoff from surface soils and leaching were based on water balance cal-
culations for well-drained soils under natural conditions or with drains
at 50 or 100 m spacing for the Atlantic Zone of Costa Rica (Sevenhuysen
and Maebe, 1995). Soil solids runoff was based on the annual average
soil erosion rates for perennial crops in Costa Rica estimated through
the Universal Soil Loss Equation (USLE) (Rubin and Hyman, 2000).
The depth of the soil was considered to be that containing 80–90% of
the banana roots, measured by Araya (2005) in a Costa Rican banana
plantation. The height of the air compartment was calculated as a
function of the total area modeled, as recommended in the CALTOX
model (McKone et al., 1997), which agrees with minimum boundary
layer height measurements in the central valley of Costa Rica
(Esquivel-Hernández et al., 2015; Xi et al., 2016). Changes in air
boundary layer height were incorporated in the uncertainty analysis
(Table A.9). Daily rain, temperature, relative humidity, and wind for
the last decade (2004–2014) were obtained from the nearest meteoro-
logical stations in Limón Province (Table A.3).
A comprehensive list of all model parameters and associated as-
sumptions are provided in Appendix A.3 and Tables A.1-A.3. The
model was run for 5 years, as plantation lifetimes of 5–10 years have
been reported (Damour et al., 2015).
2.2.2. Vegetation compartment
The model describes pesticide uptake and transport in banana plants
through their life-cycle, drawing on earlier efforts to model pesticide
dynamics in plants, particularly the apple tree model of Trapp (2007),
the DynamiCROP model (Fantke et al., 2011), the vegetation compart-
ments of Cousins and Mackay (2001), and the plant model of
Undeman et al. (2009).
The banana plant was represented as 4 compartments: fruit,
pseudostem (which consists of tight overlapping layers of leaf sheaths),
leaf, and leaf surface deposit (Fig. A.3). The latter consists of the pesti-
cide residue deposited on the leaf surface, and was thus used only for
chemicals that are applied aerially. Most aerially sprayed fungicides in
Costa Rica are prepared in mineral oil emulsions or suspensions, and ap-
plied on a weekly basis (CORBANA, 2011). Given the high application
frequency, the low volatility of mineral oils, and their adherence to
plant surfaces (Zabkiewicz, 2002), it was assumed that the volume of
the leaf surface deposit remains constant. For the other plant compart-
ments, logistic growth was simulated based on dry matter accumulation
curves for Grande Naine banana plants (Martínez Acosta and Cayón
Salinas, 2011), one of the main varieties cultivated in Costa Rica
(CORBANA, 2011) (see Appendix A.3.2 for details).
Leaf and plant emergence and fall dynamics are also included in the
model. Banana plants usually have 9–15 leaves present (CORBANA,
2011; Turner et al., 2007). In the humid tropics, a new leaf emerges
every 7–12 days (Gómez Ossa et al., 2011; Pérez Valvidia, 2012;
Robinson and Galán, 2010; van Velzen, 1994) and lives approximately
50 days (Paull and Duarte, 2011; Robinson and Galán, 2010). The last
leaves emerge at flowering, approximately 200 days after planting,
and live until harvest (Robinson and Galán, 2010; van Velzen, 1994). Ba-
nana fruits start growing after flowering and are harvested approxi-
mately 10 months after planting (Robinson and Galán, 2010). The
banana plant is then cut down. In the model, when harvest and leaf-
shedding occur, the chemical mass stored in the falling pseudostem
and/or leaves is incorporated into the soil, given that plant material
is left to degrade in the field (CORBANA, 2011; Hernandez and Witter,
1996). Bananas are perennial plants: several suckers grow around
the banana plant, one of which is chosen to succeed the adult
plant (Robinson and Galán, 2010). This succession of plant cohorts is
represented in the model, with a daughter banana plant growing
approximately 2 months before the adult banana tree flowers and
eventually replacing the adult banana tree. Harvest-to-harvest
intervals were assumed to be 5 months, as values between 3.5 and
8 months have been reported in Central America (Stover and
Simmonds, 1987; Svanes and Aronsson, 2013). Synchronous growth
throughout the plantation is assumed (Fernando Ramirez, IRET, person-
al communication).
Soil-to-plant and within-plant chemical transfer occurs through ad-
vection via the xylem channels, which transport water and nutrients
from the roots to the rest of the plant, and the phloem channels,
which transport the products of photosynthesis to all growing tissues
and storage organs (Doucette et al., 2010; Trapp and Legind, 2011).
Chemical uptake from the soil into the pseudostem through the roots
is described by the transpiration stream concentration factor (TSCF),
which is the concentration ratio of the chemical in the xylem sap to
that in the root-zone solution, following Dettenmaier et al. (2009) and
Trapp and Legind (2011). Chemical transport from the pseudostem to
the leaves and fruit through xylem flow and from the leaves to the
pseudostem and fruit through phloem flow was simulated as in the
apple tree model of Trapp (2007), assuming unidirectional transloca-
tion (see Appendix A.3.2).
Banana plants also receive chemical inputs from the air. Diffusive ex-
changes between air and leaf, air and the leaf sheaths on the outer layer
of the pseudostem, and air and fruit were adopted from the resistance
approach of Trapp (2007), with some adjustments for air-fruit diffusion.
Similar to Undeman et al. (2009), we also include the air boundary layer
resistance when calculating air-fruit diffusion, and diffusion to the fruit
via stomata is neglected due to the low stomatal density on the banana
peel, which is approximately 30 times lower than on the leaves (Brat et
al., 2016; Carr, 2012).
Furthermore, chemical dissolved in the air and sorbed to aerosols
reaches the leaf through wet and dry deposition. Following the ap-
proach of Cousins and Mackay (2001), Wegmann et al. (2004), Wania
and McLachlan (2001), and Glüge et al. (2016), wet gaseous and parti-
cle-bound deposition to the leaf and the vegetation-covered soil is esti-
mated through a rain scavenging factor and an interception fraction.
The latter is based on field measurements for banana plants and de-
scribes the amount of rain captured by the leaves (Carr, 2012; Cattan
et al., 2007; Macinnis-Ng et al., 2012), while the rest falls through the
canopy and reaches the soil. Dry, particle-bound deposition from the
air to the leaf surface was described using the dry particle deposition ve-
locity of PCBs to deciduous trees in the summer from Glüge et al. (2016)
as a proxy for broadleaf trees in the Tropics. This velocity changes in par-
allel to changes in canopy volume, as in Wania and McLachlan (2001)
and Wegmann et al. (2004).
Finally, chemicals on the leaf surface deposit can volatilize, degrade,
be transported into the leaf, or be washed off by rain and reach the soil.
Volatilization of the leaf surface deposit was not taken into account
given that minimal volatilization of epoxiconazole from the surface,
ranging from 3 to 5% within 1 day after application (EFSA, 2008;
Lichiheb et al., 2015), and reduced volatilization of pesticides in oil
(Houbraken et al., 2015; Zabkiewicz, 2002) have been reported. Pene-
tration from leaf surface deposit into the leaf and pesticide removal
 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
from the leaf surface through rainfall were based on experiments from
Lichiheb et al. (2015).
2.2.3. Emissions scenarios
To represent the different types of pesticides used and capitalize on
existing data on pesticide application and measured concentrations in
air and water for model validation, the herbicide diuron, the fungicide
epoxiconazole, and the nematicide ethoprofos were chosen for this
study. These chemicals were measured in bulk surface water at 3 sam-
pling sites within the Caño Azul River from 2009 to 2012 (n = 52 for
epoxiconazole and ethoprofos, n = 45 for diuron) as part of a larger pro-
ject by the Central American Institute for Studies on Toxic Substances
(IRET) at the Universidad Nacional of Costa Rica, which monitored pes-
ticide residues in surface waters throughout the MDR Watershed
(Arias-Andrés et al., 2016; Rämö et al., 2016). Ethoprofos and
epoxiconazole were also detected in the majority of air samples taken
100 m from banana plantations in the neighboring canton of Matina
from June 2010 to December 2011 (Córdoba, 2015).
Pesticide application amounts were based on emissions inventories
for Limón Province and application frequencies were based on assump-
tions detailed in Appendix A.4. Emissions are highly variable over time
and can include substantial uncertainties, changing according to specific
practices at individual plantations. Two surveys were available for
the chemicals studied, one conducted during the sampling period
(Ramirez et al., 2011) and one from 2013 (Bravo et al., 2015). Results
from the most recent survey indicate that application patterns may
have changed substantially, particularly for diuron and epoxiconazole.
However, insufficient data were available to resolve these changes
with time. Because of this, we decided to base our model parameteriza-
tion and analysis on the earlier survey, but we include a scenario analy-
sis, as detailed in Appendix A.4, to evaluate the implications of using the
more recent emissions survey on the model results.
It was assumed that yearly application amounts were equally distrib-
uted among the number of applications per year. In banana plantations,
ethoprofos was emitted twice per year, on a non-rainy day in March
and September, the driest months. Ethoprofos was applied at a rate of
1.38 kg a.i. ha−1 year−1 (Bravo Durán et al., 2013; LAREP-IRET, 2017).
Two diuron applications per year were simulated: one in May and one
8 weeks later (CORBANA, 2011; Córdoba, 2015; Vargas, 2013), with
0.04 kg a.i. emitted per year (Bravo Durán et al., 2013; LAREP-IRET,
2017). In pineapple plantations, ethoprofos was applied 6 times per
year on dry days, at a rate of 9 kg a.i. ha−1 year−1 (LAREP-IRET, 2017). Di-
uron was applied once every 2 months, at 0.5 kg a.i. ha−1 year−1(LAREP-
IRET, 2017). Given that ethoprofos is applied directly to the soil (Castillo
et al., 2006; CORBANA, 2011; Córdoba, 2015) while diuron is sprayed on
the soil and weeds (Bellamy, 2013; Córdoba, 2015), it was assumed for
both banana- and pineapple-covered soils that 99.5–100% of ethoprofos
is emitted to the soil and the rest to the air, while 80–90% of diuron emis-
sions are to the soil and the rest are to the air (Camenzuli et al., 2012;
Linders et al., 2002).
Epoxiconazole, an aerially sprayed fungicide, was emitted every 40–
50 days on a non-rainy day. For each application, 80 g a.i. ha−1 are used,
prepared in a water and mineral oil emulsion (BASF, 2016; CORBANA,
2011). After the emulsion is sprayed, a fraction reaches the banana
leaves (0–40%, proportional to the leaf surface area), some the soil
(12–52%), and the rest remains in the air (48%). These fractions were
based on a study of fungicide-oil mixtures sprayed aerially on a banana
plantation in São Paulo, Brazil, which calculated deposition volumes per
hectare above and beneath the canopy during application (Corrêa et al.,
2004). Fruits are covered with insecticide-treated bags since the begin-
ning of growth (Paull and Duarte, 2011), so it was assumed that fungi-
cides are not deposited on the fruit surface during spraying. However,
as these bags are porous (CORBANA, 2011; Vargas-Calvo and
Valle-Ruiz, 2011) and some chemical might enter the bags, we include
diffusive exchange between the fruit and the air (see Appendix A.3.2).
In addition, as aircrafts spray continuously over the entire plantation
1253
area, we assumed that the epoxiconazole emitted over primary canals
(see Fig. A.1) is directly transported to the river (Fernando Ramirez,
2017; IRET, personal communication).
2.2.4. Chemical properties
Degradation rate constants in the water, soil, and sediments were
obtained from measured values in literature (Table A.4). For ethoprofos,
it was assumed that degradation in the sediments was similar to that
measured in anaerobic soils. Measured degradation rate constants in
the air were only available for diuron. Therefore, for ethoprofos and
epoxiconazole we assumed photochemical degradation was the domi-
nant pathway, and estimated degradation in air based on hydroxyl rad-
ical rate constants from the AOPWin Program in Epi Suite v. 4.11
(USEPA, 2016) and average hydroxyl radical concentrations from
Bahm and Khalil (2004) for 5–15°N latitude. Degradation on the leaf
surface deposit, only relevant for epoxiconazole, was assumed to also
occur mainly through indirect photolysis (Birkved and Hauschild,
2006; Leistra and Wolters, 2004), and was thus assigned the same rate
constant value as degradation in the air. The same degradation half-
life was used for the pseudostem, leaf, and fruit. This half-life was esti-
mated from plant dissipation half-lives, as only these are available for
our study chemicals and for most chemicals in general (Fantke et al.,
2014; Jacobsen et al., 2015). As a conservative assumption, it was as-
sumed that degradation accounts for 30% of the reported dissipation,
based on Jacobsen et al. (2015). Dissipation rate constants for diuron
and epoxiconazole in vegetation were obtained from studies summa-
rized in the EFSA Draft Assessment Report (2003) and in Fantke and
Juraske (2013), respectively. The degradation half-life for ethoprofos
in vegetation was estimated from soil degradation half-lives, following
Thomas et al. (2011).
Experimental values for octanol-water partition coefficients (KOW)
(USEPA, 2016) and air-water partition coefficients (KAW) (Ma et al.,
2006; USEPA, 2016) were used for all chemicals. Plant-water partition
coefficients were calculated according to Trapp et al. (1994), as a func-
tion of water and lipid contents of plant tissue, KOW, and a correction
factor accounting for the differences between octanol and plant lipids
(see Appendix A.5). We assumed that the leaf surface deposit has the
same chemical properties as octanol.
2.2.5. Uncertainty and sensitivity analysis
Variance in model input parameters associated with true uncertain-
ty, spatial and temporal variability, and model assumptions (e.g. param-
eter values chosen for the washoff fraction and air boundary layer
height), was propagated to model outputs through Monte Carlo uncer-
tainty analysis using 250 Latin hypercube samples. Input parameters
were assigned distributions based on data availability. Input parameters
were assigned uniform, triangular, normal or log-normal distributions,
depending on data availability (see Appendix A.7 and Table A.9). Log-
normal distributions were assigned when only one value was available
and their standard deviations were estimated based on confidence fac-
tors (CF), following MacLeod et al. (2002). The contribution of each pa-
rameter to the variance in the model outputs was calculated as its
Spearman's rank correlation coefficient (r) divided by the sum of r2 for
all parameters (MacLeod et al., 2002). Vegetation growth curves and
emissions amounts and frequencies were considered as scenario pa-
rameters and not included in the sensitivity and uncertainty analysis.
3. Results
3.1. Plant compartments
3.1.1. Model results
A similar pattern in chemical concentrations and masses in all com-
partments occurs each year; hence, the middle year of the model run is
used as a reference for discussing the model results and their fit to mea-
surements. For all chemicals, concentrations in plant compartments
1254 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
peak when emissions occur and then decrease (Fig. 1). These peaks are
also observed, particularly in the banana leaf, when emissions occur to
pineapple soil because chemicals are rapidly transferred to the common
air compartment that covers both bananas and pineapples, in which in-
stantaneous mixing is assumed. Discontinuities in concentration pro-
files in the plant compartments represent cutting down of the banana
plant at harvest. For all chemicals, concentrations in the fruit increase
rapidly as the fruit bunch begins to grow and receives inputs through
xylem and phloem flow. The concentration and mass in the leaves is
1–2 orders of magnitude higher than in the other plant compartments
(Figs. 1, A.4, A.7, A.10) because leaves have a much higher surface area
for diffusion and also receive additional inputs through wet and dry par-
ticle deposition. The leaf also contains a higher proportion of organic
phases, with high capacity to accumulate all three chemicals, than the
fruit and the pseudostem.
3.1.1.1. Epoxiconazole. Among the studied chemicals, the highest mass
and concentrations in the leaf are for epoxiconazole, as expected from
direct inputs during aerial applications (Fig. 1). The leaves contain 93–
99% of the epoxiconazole mass in the banana plants, while the
pseudostem contains 0.7–6%, and the fruit contain up to 2%, with mass
in the latter two increasing as the plants grow (Fig. A.4). Epoxiconazole
penetrates quickly into the leaf interior and within 2 weeks the leaf sur-
face deposit is essentially depleted. Afterwards, xylem flow via the
pseudostem becomes the main input into the leaf. Diffusive inputs
and deposition from the air are important sources of epoxiconazole
for the leaf and the fruit only on the day of emissions. Xylem flow
from the pseudostem is the main input into the fruit on all other days.
A higher variability is observed in epoxiconazole concentrations in the
leaf and fruit than in the pseudostem because the latter has a stable
source of epoxiconazole in the soil. Degradation is the main loss process
of epoxiconazole from the leaf and the fruit, while chemical transport
into the leaf via xylem flow is the main loss process from the
pseudostem, followed by degradation.
3.1.1.2. Ethoprofos. Of the total ethoprofos mass in the vegetation, 46–
99% is in the leaves, 0.3–50% in the pseudostem, and up to 5% is in the
fruit (Fig. A.7). Concentrations in the pseudostem and in the fruit are
more variable for ethoprofos than for epoxiconazole (Fig. 1), reflecting
the variability of ethoprofos concentrations in the soil (Section 3.3.2),
which is the main chemical source to the pseudostem. The main input
into the leaves is xylem flow for up to three weeks after pesticide
Fig. 1. Concentrations of epoxiconazole (a), ethoprofos (b), and diuron (c) in the banana vegetation compartments during the 3rd year of the model run. Concentrations in the leaf and
pseudostem are given in terms of the overall biomass in the plantation (sum of adults and suckers). The solid lines represent the 50th percentile of 250 Latin hypercube samples,
while the shaded area represents the 95% confidence interval. The dashed line is the maximum residue limit for the European Union.
applications, while diffusion from the air is the main source the rest of
the time. Similarly, when fruit growth overlaps with emissions, most
ethoprofos inputs to the fruit are through xylem and phloem flow
from the pseudostem for up to 60 days; otherwise, the main input is dif-
fusion from the air. Most ethoprofos losses in the fruit and leaves occur
through degradation, while xylem flow to the leaves and degradation
are equally important loss processes for ethoprofos in the pseudostem.
3.1.1.3. Diuron. The leaves contain most of the diuron mass in the banana
plants (80–99%), while the pseudostem and fruit contain 1.5–16% and
0.05–8%, respectively, depending on the growth stage (Fig. A.10). Even
though diuron has a similar degradation half-life in the soil and vegeta-
tion as epoxiconazole, diuron concentrations in the plant compartments
span a larger range than those of epoxiconazole (Fig. 1), as it is applied
less frequently to banana plantations. Similar to epoxiconazole, the
main input into the leaves and fruit is xylem flow from the pseudostem,
while diffusion and deposition from the air are only important on the
days that emissions occur. The latter have a smaller influence on diuron
concentrations in the leaf than for epoxiconazole, evidenced by the less
pronounced peaks, given that 90% of diuron is applied to the soil. Hence
concentrations in the leaf follow the accumulation pattern in the
pseudostem. Degradation accounts for most diuron losses from the
leaf and fruit, while xylem flow (to the leaves and fruit) is the main
loss process from the pseudostem.
3.1.2. Comparison to measurements
Diuron, ethoprofos, and epoxiconazole concentrations modeled in
the fruit at harvest are all below the EU and US MRLs (Fig. 1), indicating
that bananas are safe for human consumption. This agrees with results
of the 2012–2014 monitoring program of the United States Department
of Agriculture (USDA) for diuron and ethoprofos concentrations in
peeled bananas (USDA, 2016a; USDA, 2014a; USDA, 2014b) and of the
2009–2014 monitoring program of the US Food and Drug Administra-
tion (FDA) for diuron, ethoprofos and epoxiconazole concentrations in
whole, unwashed bananas at the point of import (FDA, 2017).
Ethoprofos and epoxiconazole residues in whole bananas also did not
exceed MRLs in the EU-coordinated monitoring program and the EU na-
tional control programs from 2009 to 2013 (EFSA, 2014; EFSA, 2013;
EFSA, 2013; EFSA, 2011), as detailed in Appendix A.8. If increased emis-
sions from the most recent survey are considered, the upper 50th per-
centile of modeled diuron concentrations in fruit occasionally exceed
the MRL (Fig. A.19).
 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262 1255

3.2. Environmental compartments
Chemical concentrations in environmental compartments, like in
the banana plant, peak after the emissions pulse, and then decrease at
rates dictated by different loss processes, which are a function of chem-
ical or environmental properties, or both. Concentrations in the water
also peak after rain events. Day-to-day variability of chemical concen-
trations is higher in the air and the water than in the soil, as these are
influenced by parameters that change on a daily basis (particularly
rain, wind velocity, and water flow).
3.2.1. Epoxiconazole
The highest amounts of epoxiconazole are found in the banana soil
(70–94% of the total mass), followed by the banana plants (5–27%,
Figs. A.4, A.5). Water, sediments, and air together contain b0.5% of the
total mass in the system.
This distribution is a result of a degradation half-life that is one order
of magnitude longer in soil (geometric mean = 226 days) than in the
water and sediments, fast advective removal processes from the air
and water, and constant inputs into the soil during application and via
litterfall (leaf shedding and incorporation of plant residues into the
soil after harvest). During the plantation lifetime, litterfall accounts for
22% of all inputs into the soil and for 61% of indirect inputs. Chemical
runoff from the leaf surface deposit is the most important soil input
on rainy days within 1 week after spraying, amounting to 34% of all in-
direct soil inputs throughout the plantation's life. On days when litterfall
or pesticide application does not occur, most epoxiconazole inputs to
the soil (70–99%) originate from wet and dry deposition. Epoxiconazole
concentrations in the soil decrease slowly after emissions, driven mainly
by degradation, which accounts for 60% of all losses from the soil. On dry
days, most of the remaining outputs from the soil occur through uptake
by the pseudostem, while on rainy days, they occur via leaching and
runoff.
In the river, most epoxiconazole losses (N98%) occur via water flow
out of the system. As a result, epoxiconazole concentrations in the sed-
iment are higher than in the water (Fig. A.6), despite similar degrada-
tion half-lives in the water and sediments and sediment inputs only
through deposition of suspended solids and diffusion from water. Of
the total epoxiconazole emissions to the Caño Azul agricultural area,
4% (20 kg) is exported from the river through advection. The single
biggest source of epoxiconazole to water is runoff from soil. Air-water
exchange processes (deposition and diffusion) and direct inputs via ca-
nals are the dominant sources of epoxiconazole to the water only on the
days of spraying.
The overall variability in the median predicted concentrations
(0.4–3.2 μg/L) is similar to that observed in the measurements
(0.08–0.7 μg/L), both spanning one order of magnitude. The
epoxiconazole grab samples had concentrations within the lower 50th
percentile of model results (Fig. 2a). If increased emissions from the
most recent survey are considered, the majority of the measured
data fall within the lower 25th percentile. The overestimation of
epoxiconazole concentrations in the water could be a result from our
worst-case emissions scenario, in which this fungicide was applied
the maximum allowable number of times per year. The absence of sea-
sonal trends in the model and measurements supports the assumption
of constant aerial applications throughout the year. Aquatic quality
criteria for specific substances have not been developed for Costa Rica,
and studies comparing toxic effects of pesticides on tropical and tem-
perate species have not shown consistent differences (Arias-Andrés et
al., 2016; Diepens et al., 2014). Therefore, we compared the model re-
sults and measurements to environmental quality standards (EQS)
from Europe. Most measured epoxiconazole concentrations are be-
tween the annual average EQS (AA-EQS) for Germany (BMUB, 2016),
Switzerland (Oekotoxzentrum, 2016) and the Netherlands (CML and
WVL, 2013), and the maximum allowable concentration (MAC-EQS)
for the Netherlands (CML and WVL, 2013), which protect aquatic
ecosystems from chronic and acute exposure to micropollutants,
respectively. Peak median modeled concentrations exceed the AA-EQS
(0.19 μg/L) on multiple occasions.
In air, epoxiconazole concentrations peak on the day of emissions
and then sharply decrease (Fig. A.6). On days without emissions,
diffusion from the leaf interior is the single main source of
epoxiconazole. When it rains, wet deposition to the soil removes 76–
90% of epoxiconazole mass in the air compartment. On dry days, wind
advection removes 57–77% of all inputs; the residence time of the air
compartment is only 0.4–0.8 h. Gaseous diffusion and dry particle depo-
sition to the leaf account for most of the remaining losses. Approximate-
ly 18% of the epoxiconazole mass in the air is associated with aerosols,
which are quickly transferred to the leaf due to the high dry deposition
velocity (570 m/h). Over the plantation lifetime, constant advection,

Fig. 2. Comparison of results for the 3rd year of the model run for epoxiconazole (a), ethoprofos (b), and diuron (c) with concentrations measured in bulk water grab samples from the
Caño Azul river from 2009 to 2012 (LAREP-IRET, 2017) and toxicological thresholds (black lines). For epoxiconazole the number of measurements taken (n) was 52, of which 1 contained
traces (T) and 28 were below the LOQ; for ethoprofos, n = 52, T = 7, 15 b LOQ; for diuron n = 45, T = 2, 4 b LOQ. The solid blue line represents the 50th percentile of 250 Latin hypercube
samples, while the shaded area represents the 95% confidence intervals. AA-EQS and MAC-EQS are the annual average and maximum allowable concentration environmental quality
standards of European countries, respectively (see Section 3.3 for details). HC5 is the hazard concentration affecting 95% of the species calculated by Rämö et al. (2016). For ethoprofos
(b), the solid black line refers to the maximum permissible concentration (MPC) for the Netherlands (CML and WVL, 2013; Crommentuijn et al., 2000).
1256 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
diffusion to the leaf, and dry aerosol deposition to the leaf account for
70%, 15%, and 12% of the losses from the air, respectively, while inter-
mittent wet deposition accounts for only 1%. Of all emissions into the
system during the model run (475 kg), 33% (159 kg) are lost via wind
advection.
On the day of emissions, peak air concentrations predicted by the
model reach maximum air concentrations measured by passive air sam-
plers at sites within 100 m of banana plantations in the neighboring
canton of Matina (Fig. A.6). The rest of the time, model results for the
air are 3–4 orders of magnitude lower than the measurements. Passive
air samplers were deployed for an average period of 47 days to record
air concentrations in the gaseous phase (Córdoba, 2015). As information
on the exact dates of deployment could not be retrieved and only overall
maximum and minimum air concentrations are available, the 47-day
moving average of modeled air concentrations was selected as a more
suitable comparison to the measurements (Fig. 3a). Using these aver-
ages, the upper 50th percentile of the model results encompasses the
minimum measured concentrations and is above the limit of detection
(LOD) (note that the dip in Fig. 3a corresponds to a 47-day moving
average that does not capture the day of emissions). Median model re-
sults are overall 1.5–2 times lower than the minimum measured con-
centrations. It is important to underscore that epoxiconazole was only
detected in 13/52 passive air samples in sites within 100 m from banana
plantations and was not detected in background sites (LOD = 1 ng/m3).
Epoxiconazole was also found in 1/13 filter samples in sites within
100 m from banana plantations but not in active samplers deployed
for 24 h. Considering the more recent higher emissions, the model
would be in better agreement with the air compartment (Fig. A.15)
but would overestimate concentrations in water (Fig. A.14).
3.2.2. Ethoprofos
As in the case of epoxiconazole, soils contain most of the ethoprofos
mass in the system (97–99% in pineapple and banana soils and b1% in
the plant, water, sediments, and air; see Figs. A.7, A.8). This accumula-
tion pattern is driven by direct applications to the soil. As a result of a
shorter soil degradation half-life (15 days) and less frequent applica-
tions to the banana soil than epoxiconazole (2 vs. 8 times per year),
ethoprofos concentrations in soils decrease rapidly after emissions
(Fig. A.9). The pineapple soil contains more ethoprofos than the banana
soil, as the amount applied per year is 6.5 times higher and applications
Fig. 3. Comparison of model results for simulation year 3 for epoxiconazole (a), and ethoprofos (b) with maximum and minimum values measured by passive air samplers at sites located
within 100 m of banana plantations in Matina, province of Limon (Córdoba, 2015). The solid red line is the 47-day moving average of the 50th percentile of the model results, while the
shaded area represents the 95% confidence intervals.
are more frequent (6 per year). In both banana and pineapple soils, deg-
radation is responsible for 96–99% of all daily ethoprofos losses. Gaseous
wet deposition contributes to 70% and 95% of the indirect inputs into ba-
nana soil and pineapple soil, respectively, throughout the plantation's
lifetime. During this time, litter incorporation accounts for 27% of all
the indirect inputs into the banana soil.
Ethoprofos concentrations in the sediments exceed those in water
and in banana soils approximately 50 days after the emission peaks
(Fig. A.9), as expected from its longer degradation half-life in sediments
(geometric mean = 92 days) than in all other media. Similar to
epoxiconazole, water advection removes most of the ethoprofos from
the Caño Azul River. Throughout the plantations' lifetime, 0.72%
(183 kg) of the total emissions into the system are exported out of the
Caño Azul agricultural region via water flow, which is approximately
equivalent to the percent of the emissions to the soil lost through runoff.
Ethoprofos concentrations in the water are more variable than those of
epoxiconazole, as inputs are less constant and input fluxes from runoff
quickly decrease as the chemical degrades in the soil. After the first
yearly application into the soil, runoff accounts for nearly 100% of the in-
puts into the water compartment.
Median model results for the water and measurements have a sim-
ilar, wide range of variability spanning 2 orders of magnitude (0.07–5
μg/L for model and 0.04–2.7 μg/L for measurements). Peak measured
concentrations agree with peak median concentrations, while the low-
est measured concentrations agree with the lower 5th percentile of
model results (Fig. 2b). Most measured concentrations (83%) and medi-
an modeled concentrations are above the maximum permissible con-
centration in the Netherlands of 0.063 μg/L, which is based on the no
observable adverse effect level for 95% of species considered (CML and
WVL, 2013; Crommentuijn et al., 2000). Peak median and measured
concentrations reach the hazard concentrations for 5% of species consid-
ered (HC5), calculated by Rämö et al. (2016) from species sensitivities
distributions based on toxicity values in the U.S. EPA Ecotox database
and E-toxBase.
Ethoprofos concentrations in the air above the Caño Azul drainage
basin are mainly driven by background concentrations in the incoming
air. Diffusion from the leaf to the air becomes an important source of
ethoprofos to the air once the adult banana plantation is established, ac-
counting for approximately 5–30% of the daily inputs. Most losses from
the air compartment occur through wind advection (40–80% of the
 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
daily losses), followed by diffusion into the banana leaves (10–29% of
the daily losses). Wind advection removes 2% (423 kg) of the total emis-
sions, much less than for epoxiconazole, as most of the ethoprofos is ap-
plied directly to the soil, and only mobilized into the air via diffusion
from the soil and the vegetation. Dry aerosol deposition to the leaves
is negligible compared to epoxiconazole, as only 0.075% of ethoprofos
is associated with particles.
Median ethoprofos concentrations in the air predicted by the model
are within the maximum and minimum concentrations measured by
passive air samplers (Figs. 3b, A.9). Ethoprofos was detected in 77% of
all samples, with no statistically significant differences in the frequency
of detection or concentrations between sites located N 1.65 km away
from banana plantations and those adjacent to banana plantations
(Córdoba, 2015). These observations agree with model outputs indicat-
ing that incoming background air concentrations are the main source of
ethoprofos in the air above the Caño Azul River Basin. In the case of
ethoprofos, there is little difference in emissions found in the two avail-
able surveys, indicating a stable emissions profile and greater confi-
dence in our model parameterization and results.
3.2.3. Diuron
Diuron accumulates mainly in pineapple soil (52–96%) and banana
soil (3–51%) (Figs. A.10, A.11). Most of the remaining mass is stored in
the plants (up to 5%), while water, sediment, and air together contain
4% or less of the chemical mass in the system. When emissions occur, di-
uron concentrations in the banana soil are similar to those in the water
and sediments (Fig. A.12). Eventually, the sediments and water have
higher diuron concentrations than the banana soil, as the water com-
partment receives runoff from pineapple soil as well, which itself
receives 10 times more emissions than the banana soil. Diuron concen-
trations in the sediment and water are less variable than those of
ethoprofos, even though the degradation half-life of ethoprofos in
water is higher. Chemical inputs into the water, and hence into the sed-
iments, are more constant for diuron than for ethoprofos because these
originate mainly from the soil, in which diuron has a long degradation
half-life (221 days).
In pineapple soils, leaching and runoff constitute the main outputs
on rainy days, accounting for 45 and 23% of all losses, respectively,
while degradation is the single main output on dry days. In banana
soils, most losses (34% of total) occur via leaching, while runoff, degra-
dation, and pseudostem advection have a similar contribution (approx-
imately 20%) to the daily losses from the soil, which varies through time.
Advection from the soil to the pseudostem becomes the main output
when fruit growth begins. Overall, 21% of the diuron mass that reaches
the soil during application is lost through runoff. Litterfall accounts for
37% of all non-emission inputs into the banana soil. Wet deposition is
an important source of diuron in the banana and pineapple soils, ac-
counting for 63% and 99% of all indirect inputs, respectively.
In the water compartment, runoff from pineapple and banana soils
constitutes 85–97% of all inputs on rainy days, while diffusion from
the sediment is the main input on dry days. Advection is the single
major loss process. The amount of diuron exported from the Caño
Azul River throughout the plantation lifetime is 249.9 kg, equivalent to
20.7% of all emissions. The 95% confidence interval of the model results
encompasses 92% of the measurements in the water (Fig. 2c). The good
agreement between measured and modeled water concentrations sug-
gests that the baseline emissions scenario for diuron is appropriate for
the sampling period. All median modeled concentrations and the ma-
jority of measured concentrations are above the European Union AA-
EQS of 1.8 μg/L. Peak modeled and measured concentrations are above
the European Union MAC-EQS of 0.2 μg/L and the HC5 of 2.62 μg/L
(Rämö et al., 2016).
Following emissions pulses, the main source of diuron to air is diffu-
sion from the leaves. Wet deposition to banana soil and to pineapple soil
is the main loss process from the air on rainy days, responsible for 65%
and 20% of daily outputs, respectively. On dry days, wind advection
1257
removes 70–78% of the diuron mass in the air. Overall, 45% of all losses
from the air occur through advection, 34% through wet deposition, and
11% through diffusion to the leaf. Diuron losses from the air via wet de-
position are much higher than those of epoxiconazole even though they
have a similar KAW because, in the model, we assume applications fol-
low recommended guidelines: diuron can be applied on rainy days,
and is thus subject to scavenging by raindrops, while epoxiconazole is
only applied on dry days, and hence most of the chemical is advected
out of the system before it can be removed from the air via rainfall.
Throughout the plantation's lifetime, about 57 kg of diuron leave the
Caño Azul agricultural area via air, equivalent to 4.7% of total emissions.
Measurements for diuron concentrations in air were not available for
comparison to model results.
3.3. Uncertainty and sensitivity analysis
The magnitude of the contribution to variance (CV) of a parameter
on model predictions depends both on its uncertainty and on the sensi-
tivity of the model to the parameter. Fig. 4 illustrates the average CV of
the three most influential parameters to chemical concentrations in all
model compartments throughout the third generation of banana plants
in the simulation. Averages are presented, as only minor differences oc-
curred in the absolute values corresponding to different stages in the
growth cycle (i.e.: pre- and post-fruit emergence) and the ranking of
the most influential parameters remained the same. Given that mea-
surements are available for the water, air, and fruit (which are also the
main vectors for exporting chemicals out of the Caño Azul system),
and soil runoff is the main source of chemicals to the water, the follow-
ing discussion will focus on these four compartments.
3.3.1. Water
Rain is the most influential parameter for concentrations of all
chemicals in the water, as it drives runoff in the model (Appendix A,
Eq. (27)) and is also highly variable. Rain accounts for 27%, 43%, and
25% of the variance of epoxiconazole, diuron, and ethoprofos concentra-
tions in water, respectively. The higher contributions for diuron and
epoxiconazole are a result of their higher solubilities in water and a
higher uncertainty in the Henry's law constant for diuron. Degradation
half-lives in the soil are also among the most influential parameters
for ethoprofos (CV = 14%) and epoxiconazole (CV = 10%) concentra-
tions in the water, while the water velocity is important for diuron
(CV = 15%) and epoxiconazole (CV = 11%). This is expected, as soil is
the main source of chemical to water while advection, determined by
the water velocity, is the main removal process from water.
3.3.2. Air
The rain rate is the most influential parameter for all chemical con-
centrations in air, accounting for approximately 20% of the variance.
The air-water partition coefficient is also an important parameter, con-
tributing to 12% of the variance for ethorpofos and epoxiconazole, and
20% of the variance for diuron concentrations in the air. Epoxiconazole
and diuron have a very low KAW (−7.79 and −7.28, respectively) and
thus a high fugacity capacity in the water and higher dissolution in
rain droplets. In the model, the rain parameter determines the days in
which ethoprofos and epoxiconazole emissions occur (on days with
no rain), while diuron can be applied on rainy days and thus be removed
from the air through rain scavenging. Leaf-water partitioning is also in-
fluential for epoxiconazole concentrations in the air. As a result of direct
applications to their surface, the leaves have a higher epoxiconazole
load that can diffuse to the air compared to the other chemicals.
3.3.3. Soil
The fraction of organic carbon is the main contributor to variance of
diuron concentrations in the banana (39%) and pineapple (45%) soils.
The degradation half-life in the soil is the single most influential param-
eter for ethoprofos concentrations in the banana and pineapple soils,
1258 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262

Fig. 4. Contributions to variance of the 3 most influential parameters for epoxiconazole (a), ethoprofos (b), and diuron (c) concentrations in the Caño Azul agricultural area.

accounting for 68% of the variance. As presented in Section 3.3.2, degra-
dation is responsible for most of the ethoprofos losses from the soil. The
degradation half-life in the soil contributes to 33% of the variance of
epoxiconazole concentrations in the soil, while the KOW and the organic
carbon content contribute to 19% and 8% of the variance, respectively.
Degradation is the main constant loss process from the soil, and is also
highly uncertain (CF = 4.5) for epoxiconazole. The KOW and the soil or-
ganic carbon content are used in the model to describe partitioning to
the soil, so concentrations in the soil are sensitive to these parameters.
Also, high uncertainty is associated with the KOW (CF = 3.5), while
the organic carbon content varies an order of magnitude from 0.8%–
5.4%.
3.3.4. Fruit
The concentration of each pesticide in the fruit is influenced by
different parameters. Degradation in vegetation has the highest
 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
contribution to the variance of diuron concentrations in fruit (17%),
followed by xylem flow (13%) and the TSCF (10%), which describes
chemical uptake from the soil via the roots. The degradation half-life
in vegetation is the single main contributor (33%) to the variance of
ethoprofos concentrations in the banana fruit. This parameter is highly
uncertain (CF = 4.2), as it is extrapolated from soil degradation half-
lives, while measured values for dissipation in plants were used for
epoxiconazole and diuron. Parameters associated with chemical trans-
fer from the soil to the banana plant are the most influential for
epoxiconazole concentrations in the fruit, with degradation in the soil,
the TSCF and the soil organic carbon content each contributing to 14%,
13% and 9% of the variance, respectively.
4. Discussion
4.1. Water
Model results are in overall good agreement with concentrations
measured in the Caño Azul River, with a similar range of variability.
However, additional information on the timing of emissions would be
required to accurately capture peak concentrations. The quantification
of modeled chemical fluxes reveals that rainfall-driven runoff is the
dominant source of pesticides to the water, as has been proposed by
several field studies (Castillo et al., 2006, 2000; Polidoro and Morra,
2016). Among the studied chemicals, diuron has the highest transfer
(20% of the yearly emissions) to downstream sites via water advection.
This agrees with the widespread presence of diuron in surface waters
throughout the RMD watershed, at higher concentrations and detection
frequency than epoxiconazole and ethoprofos (Rämö et al., 2016). The
transport of diuron throughout tropical catchments and into coastal la-
goons, and its accumulation in sediments, has also been widely docu-
mented in Queensland, Australia, where it is mainly used in sugar
cane cultivation (Haynes et al., 2000; Mitchell et al., 2005). For example,
using a multimedia model, Camenzuli et al. (2012) estimated that 25%
of diuron applied in the Tully River catchment reaches the Great Barrier
Reef lagoon, mainly via water runoff.
4.2. Air
Model results show daily changes in air concentrations and higher
average concentrations in seasons during which chemicals are applied,
congruent with observations from passive and active air sampling of
current-use agricultural pesticides in the central valley of Costa Rica
(Gouin et al., 2008). While the model results encompass measured
ethoprofos concentrations in the air, it is unclear whether they repre-
sent or overestimate epoxiconazole measurements, most of which
were below the LOD. Additional air measurements from active and pas-
sive air samplers that capture emission peaks and seasonal trends, re-
spectively, would be useful for model validation. As wind advection
accounts for most chemical losses from the air, local wind measure-
ments would also be very valuable, as those used in the model were
from a meteorological station approximately 13 km north of our study
site.
Furthermore, model estimation of epoxiconazole concentrations in
the air could be improved by refining influential parameters and as-
sumptions. Multimedia models usually assume that the pesticides emit-
ted to the air behave as the pure active ingredient, as was done in the
present version of the Caño Azul model. However, this assumption
should be revisited as adjuvants such as mineral oil can alter the trans-
port and physicochemical properties of chemicals (Ellis et al., 2009). For
example, oil suspensions most likely remain air-borne for longer pe-
riods than water (Cotteux et al., 2013). In addition, constant dry particle
deposition and gaseous diffusion, as commonly implemented in multi-
media models, could overestimate chemical removal from the air. Fur-
ther research on dry particle deposition velocities to vegetation for
different chemical families and of chemical permeation through cuticles
1259
representing key plant species would be valuable to better describe
chemical transfer from the air to the vegetation.
4.3. Soils and sediment
Our model predicts that epoxiconazole, diuron and ethoprofos con-
centrations in soils and sediments are as high or higher than peak con-
centrations in the water. However, no studies documenting pesticide
residues in banana plantation soils were found in the scientific litera-
ture. Analyses of sediments from rivers and streams draining banana
plantations have detected none (Polidoro et al., 2009) or only a few
(Castillo et al., 2006, 2000, 1997; de la Cruz and Castillo, 2002) pesti-
cides used in banana cultivation. Studies in rice and cotton-producing
regions in Central America (Carvalho et al., 2002; Rasmussen et al.,
2016) and agricultural areas in the Mediterranean and Australia
(Moreno-González and León, 2017) have shown that stream waters
and suspended sediments transport pesticides to coastal lagoons and
wetlands, which accumulate pesticides in sediments. Studies in the
Mediterranean strongly suggest that contaminated riverbed sediments
are removed in high-flow conditions and exported to coastal lagoons
downstream (David et al., 2012). Thus, pesticide dynamics in sediments
merit further attention both in field studies and in the model presented
here. Seasonal data on suspended solid concentrations would be partic-
ularly useful, as this is the most influential parameter for chemical con-
centrations in sediments and has a wide uncertainty range as a result of
temporal variability. Addition of a bed load transport process to account
for horizontal advection of sediments as in Blaser et al. (2008) would
also allow better characterization of pesticide export from the Caño
Azul system.
4.4. Vegetation
Plants act as both a sink and a source of chemicals for other environ-
mental media. Model results show that chemicals taken up by the veg-
etation are mainly lost via degradation, which is a highly uncertain
parameter as usually only dissipation half-lives in vegetation are avail-
able (Jacobsen et al., 2015). Like in the air compartment, taking into ac-
count the role of adjuvants would allow a more accurate description of
chemical transport in banana plantations. In the model, runoff from the
leaf surface deposit, an important source of epoxiconazole to the soil,
and the leaf penetration rate constant were based on measurements
for epoxiconazole emulsifiable concentrate (EC) prepared in a water
solution (Lichiheb et al., 2015). However, the epoxiconazole EC applied
to bananas is prepared in oil, which most likely increases rainfastness
(Rae, 2002) and affects the rate of chemical penetration into the leaf
(Forster et al., 2006; Lichiheb et al., 2016). Therefore, we echo the re-
search call by Lichiheb et al. (2016) for additional penetration measure-
ments for different types of formulations of systemic pesticides (such as
epoxiconazole) on different types of leaves. With these data, the effect
of adjuvants could be incorporated into multimedia models by adjusting
first order rate constants for pure substances with empirical factors,
following Lichiheb et al. (2016).
Finally, model results and measurements also show the strong influ-
ence of nearby pineapple plantations on pesticide concentrations in the
water, given that diuron and ethoprofos are applied more frequently
and at higher amounts than in banana plantations. Thus, adding a vege-
tation compartment to the pineapple soil, like on the banana soil, would
result in a better assessment of pesticide fate and transport in the Caño
Azul drainage area.
5. Conclusions
In countries with low environmental regulatory quality (as defined
by Stehle and Schulz (2015)), both due to the lack of regulations and en-
forcement, such as Costa Rica (Polidoro and Morra, 2016; Rämö et al.,
2016), MRLs derived from information on Good Agricultural Practice
1260 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
(GAP) are in a way the only official control of pesticide residues associ-
ated with crop production. According to the FAO (2003) definition of
GAP, these practices should “address the environmental sustainability
for on-farm processes” and ultimately produce safe food items. Howev-
er, as presented in this study, ensuring that pesticide residues in banan-
as are below the MRL does not necessarily protect ecosystem quality.
The model developed here shows promise as a decision-support tool
that complements ongoing efforts to assess the ecological effects of
exposure to chemical mixtures and transient pulses (Echeverría-Sáenz
et al., 2016; Rämö et al., 2016), both of which are enhanced in tropical
climates (Rasmussen et al., 2016). For example, the model could assist
with modifying application patterns to minimize pesticide pulses, re-
main within desired safe levels in water, and prevent the onset of con-
ditions that may lead to fish mortalities. The model could also help to
elucidate the fate and transport of pesticides that are widely used for ba-
nana cultivation, yet haven't been measured, such as mancozeb (de la
Cruz and Castillo, 2002; Rämö et al., 2016), and to estimate the overall
pesticide burden in surface waters. The flexible model structure we em-
ploy can incorporate new knowledge as it becomes available, for exam-
ple in response to our call for further insights into sediment dynamics
and the effects of adjuvants on the physico-chemical properties and
transport of active ingredients.
Acknowledgments
We thank Geannina Moraga, Seiling Vargas, and Fernando Ramirez
from the Central American Institute for Studies on Toxic Substances of
the Universidad Nacional de Costa Rica (UNA) for their timely, valuable
inputs for model parameterization, and Dr. Juliane Glüge from ETH
Zürich for her technical assistance with model development. We also
thank Dr. Christian Bogdal, Dr. Kimberly Hageman, and Dr. Cleo Davie-
Martin for their insights.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.scitotenv.2017.09.172.
References
Araya, M., 2005. Stratification and spatial distribution of the banana (Musa AAA, Caven-
dish subgroup, cvs “Valery” and “Grande naine”) root system. Banana Root System:
Towards a Better Understanding for Its Productive Management: Proceedings of an
International Symposium. International Network for the Improvement of Banana
and Plantain, Montpellier, France.
Arias-Andrés, M., Rämö, R., Mena Torres, F., Ugalde, R., Grandas, L., Ruepert, C., Castillo,
L.E., Van den Brink, P.J., Gunnarsson, J.S., 2016. Lower tier toxicity risk assessment
of agriculture pesticides detected on the Río Madre de Dios watershed, Costa Rica. En-
viron. Sci. Pollut. Res. 1–10. https://doi.org/10.1007/s11356-016-7875-7.
Bahm, K., Khalil, M.A.K., 2004. A new model of tropospheric hydroxyl radical concentra-
tions. Chemosphere 54:143–166. https://doi.org/10.1016/j.chemosphere.2003.08.006.
BASF, 2016. Opus 12,5 SC. URL. http://www.agrodac.com/assets/opus-12%2c5-sc-or.-esp-
panfleto.pdf, Accessed date: 9 August 2016 (WWW Document).
Bellamy, A.S., 2013. Banana production systems: identification of alternative systems for
more sustainable production. Ambio 42:334–343. https://doi.org/10.1007/s13280-
012-0341-y.
Birkved, M., Hauschild, M.Z., 2006. PestLCI—a model for estimating field emissions of pes-
ticides in agricultural LCA. Ecol. Model. 198:433–451. https://doi.org/10.1016/
j.ecolmodel.2006.05.035.
Blaser, S.A., Scheringer, M., MacLeod, M., Hungerbühler, K., 2008. Estimation of cumula-
tive aquatic exposure and risk due to silver: contribution of nano-functionalized
plastics and textiles. Sci. Total Environ. 390:396–409. https://doi.org/10.1016/
j.scitotenv.2007.10.010.
Brat, P., Lechaudel, M., Segret, L., Morillon, R., Hubert, O., Gros, O., Lambert, F., Benoit, S.,
Bugaud, C., Salmon, F., 2016. Post-harvest banana peel splitting as a function of rela-
tive humidity storage conditions. Acta Physiol. Plant. 38:234. https://doi.org/10.1007/
s11738-016-2253-0.
Bravo Durán, V., de la Cruz Malavassi, E., Herrera Ledezma, G., Ramírez Muñoz, F., 2013.
Pesticide Use in Agriculural Produce as a Tool for Monitoring Health Hazards. p. 27
(2013).
Bravo, V., Herrera, G., Ramirez, F., 2015. Uso de plaguicidas en cultivos de arroz, banano,
palma, pastos y piña en las cuencas Pacuare, Madre de Dios y Matina, 2013. (Internal
Report TROPICA Proyect). IRET, Universidad Nacional, Heredia, Costa Rica.
Camenzuli, L., Scheringer, M., Gaus, C., Ng, C.A., Hungerbühler, K., 2012. Describing the en-
vironmental fate of diuron in a tropical river catchment. Sci. Total Environ. 440:
178–185. https://doi.org/10.1016/j.scitotenv.2012.07.037.
Carr, M.K.V., 2012. Advances in Irrigation Agronomy: Plantation Crops. Cambridge Uni-
versity Press, Cambridge https://doi.org/10.1017/CBO9780511998263.
Carvalho, F.P., Villeneuve, J.-P., Cattini, C., Tolosa, I., Montenegro-Guillen, S., Lacayo, M.,
Cruz, A., 2002. Ecological risk assessment of pesticide residues in coastal lagoons of
Nicaragua. J. Environ. Monit. 4:778–787. https://doi.org/10.1039/B203728A.
Castillo, L.E., de la Cruz, E., Ruepert, C., 1997. Ecotoxicology and pesticides in tropical
aquatic ecosystems of Central America. Environ. Toxicol. Chem. 16:41–51. https://
doi.org/10.1002/etc.5620160104.
Castillo, L.E., Ruepert, C., Solis, E., 2000. Pesticide residues in the aquatic environment of
banana plantation areas in the North Atlantic Zone of Costa Rica. Environ. Toxicol.
Chem. 19:1942–1950. https://doi.org/10.1002/etc.5620190802.
Castillo, L.E., Martínez, E., Ruepert, C., Savage, C., Gilek, M., Pinnock, M., Solis, E., 2006.
Water quality and macroinvertebrate community response following pesticide appli-
cations in a banana plantation, Limon, Costa Rica. Sci. Total Environ. 367:418–432.
https://doi.org/10.1016/j.scitotenv.2006.02.052.
Cattan, P., Bussière, F., Nouvellon, A., 2007. Evidence of large rainfall partitioning patterns
by banana and impact on surface runoff generation. Hydrol. Process. 21:2196–2205.
https://doi.org/10.1002/hyp.6588.
CML, WVL, 2013. Atlas Bestrijdingsmiddelen in Oppervlaktewater. Universiteit Leiden and
Rijkswaterstaat Water, Verkeer en Leefomgeving. URL. www.bestrijdingsmiddelatlas.nl
(WWW Document).
CORBANA, 2011. Good Agricultural Practices Manual for Banana Cultivation (San José,
Costa Rica).
CORBANA, 2016. [National Banana Corporation]. Estadísticas Bananeras [Banana Statis-
tics]. URL. https://www.corbana.co.cr/categories/categoria_1348198131, Accessed
date: 26 July 2017 (WWW Document).
Córdoba, L., 2015. Evaluation of Environmental Contamination in Air and Dust by Pesti-
cides in 12 Schools in the Matina Canton, Limon. University of Costa Rica (Masters
Thesis).
Corrêa, H.G., Benez, S.H., Berton, R.S., Sáes, L.A., 2004. Spray deposits obtained after aerial
application of pesticides in banana crops. Bragantia 63, 121–128.
Cotteux, E., Rombaut, M., Douzals, J.P., 2013. Comparison of vertical and horizontal
collecting methods for spray deposits in crop canopy and airborn spray drift assess-
ment. 12th Suprofruit. Editorial Universitat Polytecnica de Valencia, Valencia, Spain,
pp. 5–7.
Cousins, I.T., Mackay, D., 2001. Strategies for including vegetation compartments in
multimedia models. Chemosphere 44:643–654. https://doi.org/10.1016/S0045-
6535(00)00514-2.
Crommentuijn, T., Sijm, D., de Bruijn, J., van Leeuwen, K., van de Plassche, E., 2000. Max-
imum permissible and negligible concentrations for some organic substances and
pesticides. J. Environ. Manag. 58:297–312. https://doi.org/10.1006/jema.2000.0334.
Damour, G., Garnier, E., Navas, M.L., Dorel, M., Risède, J.-M., 2015. Chapter three - using
functional traits to assess the services provided by cover plants: a review of potenti-
alities in banana cropping systems. In: Sparks, Donald L. (Ed.), Advances in Agrono-
my. Academic Press:pp. 81–133. https://doi.org/10.1016/bs.agron.2015.06.004.
David, A., Bancon-Montigny, C., Salles, C., Rodier, C., Tournoud, M.-G., 2012. Contamina-
tion of riverbed sediments by hazardous substances in the Mediterranean context:
influence of hydrological conditions. J. Hydrol. 468–469:76–84. https://doi.org/
10.1016/j.jhydrol.2012.08.015.
de la Cruz, E.M., Castillo, L.E., 2002. The use of pesticides in Costa Rica and their impact on
coastal ecosystems. Pesticide Residues in Coastal Tropical Ecosystems. CRC Press.
de la Cruz, E., Bravo-Duran, V., Ramirez, F., Castillo, L.E., 2014. Environmental hazards as-
sociated with pesticide import into Costa Rica, 1977-2009. J. Environ. Biol. 35, 43–55.
Dettenmaier, E.M., Doucette, W.J., Bugbee, B., 2009. Chemical hydrophobicity and uptake
by plant roots. Environ. Sci. Technol. 43:324–329. https://doi.org/10.1021/es801751x.
Diepens, N.J., Pfennig, S., Van den Brink, P.J., Gunnarsson, J.S., Ruepert, C., Castillo, L.E.,
2014. Effect of pesticides used in banana and pineapple plantations on aquatic eco-
systems in Costa Rica. J. Environ. Biol. 35, 73–84.
Doucette, W., Dettenmaier, E., Bugbee, B., Mackay, D., 2010. Mass transport from soil to
plants. Handbook of Chemical Mass Transport in the Environment. CRC Press:
pp. 389–411. https://doi.org/10.1201/b10262-15.
Echeverría-Sáenz, S., Mena, F., Arias-Andrés, M., Vargas, S., Ruepert, C., Van den Brink, P.J.,
Castillo, L.E., Gunnarsson, J.S., 2016. In situ toxicity and ecological risk assessment of
agro-pesticide runoff in the Madre de Dios River in Costa Rica. Environ. Sci. Pollut.
Res. 1–13. https://doi.org/10.1007/s11356-016-7817-4.
EFSA, 2003. Diuron Draft Assessment Report. Volume 3 Annex B.7 Residue Data. URL.
http://dar.efsa.europa.eu/dar-web/provision (WWW Document).
EFSA, 2008. Conclusion regarding the peer review of the pesticide risk assessment of the ac-
tive substance epoxiconazole. EFSA J. 6:138r. https://doi.org/10.2903/j.efsa.2008.138r.
EFSA, 2011. The 2009 European Union report on pesticide residues in food. EFSA J. 9.
https://doi.org/10.2903/j.efsa.2011.2430.
EFSA, 2013. The 2010 European Union report on pesticide residues in food. EFSA J. 11.
https://doi.org/10.2903/j.efsa.2013.3130.
EFSA, 2014. The 2012 European Union report on pesticide residues in food. EFSA J. 12.
https://doi.org/10.2903/j.efsa.2014.3942.
EFSA, 2015. The 2013 European Union report on pesticide residues in food. EFSA J. 13.
https://doi.org/10.2903/j.efsa.2015.4038.
Ellis, M.C., Lane, A.G., O'Sullivan, C.M., Miller, P.C., 2009. Investigations into the source of
two fungicides measured in the air for 24 hours following application to a cereal
crop. Commun. Agric. Appl. Biol. Sci. 74, 37–46.
Esquivel-Hernández, G., Villalobos-Forbes, M., Sánchez-Murillo, R., Birkel, C., Valdés-
González, J., Boll, J., 2015. Near surface carbon dioxide and methane in urban areas
of Costa Rica. Open J. Air Pollut. 4:208–223. https://doi.org/10.4236/ojap.2015.44018.
 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
European Union, 2005. Regulation (EC) No 396/2005 of the European Parliament and of
the Council on maximum residue levels of pesticides in or on food and feed of
plant and animal origin. Off. J. Eur. Union 48, 1–16.
Fantke, P., Juraske, R., 2013. Variability of pesticide dissipation half-lives in plants. Envi-
ron. Sci. Technol. 47, 3548–3562.
Fantke, P., Charles, R., de Alencastro, L.F., Friedrich, R., Jolliet, O., 2011. Plant uptake of pesti-
cides and human health: dynamic modeling of residues in wheat and ingestion intake.
Chemosphere 85:1639–1647. https://doi.org/10.1016/j.chemosphere.2011.08.030.
Fantke, P., Gillespie, B.W., Juraske, R., Jolliet, O., 2014. Estimating half-lives for pesticide
dissipation from plants. Environ. Sci. Technol. 48:8588–8602. https://doi.org/
10.1021/es500434p.
FAO, 2003. Development of a Framework for Good Agricultural Practices (No. Seven-
teenth Session) (Rome).
FAO, 2014. Banana Market Review and Banana Statistics 2012–2013. Food and Agricul-
ture Organization of the United Nations, Rome.
FDA, 2017. Pesticide Residue Monitoring Program. URL. U.S. Food and Drug
Administrationhttps://www.fda.gov/Food/FoodborneIllnessContaminants/Pesti-
cides/ucm2006797.htm, Accessed date: 15 January 2017 (WWW Document).
Forster, W.A., Zabkiewicz, J.A., Liu, Z., 2006. Cuticular uptake of xenobiotics into living
plants. Part 2: influence of the xenobiotic dose on the uptake of bentazone,
epoxiconazole and pyraclostrobin, applied in the presence of various surfactants,
into Chenopodium album, Sinapis alba and Triticum aestivum leaves. Pest Manag. Sci.
62:664–672. https://doi.org/10.1002/ps.1227.
German Federal Ministry for the Environment, Nature Conservation, Building and
Nuclear Safety (BMUB), 2016. Verordnung zum Schutz der Oberflächengewässer
(Oberflächengewässerverordnung - OGewV).
Glüge, J., Bogdal, C., Scheringer, M., Hungerbühler, K., 2016. What determines PCB concen-
trations in soils in rural and urban areas? Insights from a multi-media fate model for
Switzerland as a case study. Sci. Total Environ. 550:1152–1162. https://doi.org/
10.1016/j.scitotenv.2016.01.097.
Gómez Ossa, D.M., Sánchez, J.D., Velásquez, J.E., Gamboa, J.A., Bedoya, C.D., 2011. Influence
of the micronutrient (B-Zn) balance on banana yield and the severity of
Mycosphaerella fijinesis. Ing. Amazon. 4, 88–102.
Gouin, T., Wania, F., Ruepert, C., Castillo, L.E., 2008. Field testing passive air samplers for
current use pesticides in a tropical environment. Environ. Sci. Technol. 42:
6625–6630. https://doi.org/10.1021/es8008425.
Grant, P.B.C., Woudneh, M.B., Ross, P.S., 2013. Pesticides in blood from spectacled caiman
(Caiman crocodilus) downstream of banana plantations in Costa Rica. Environ.
Toxicol. Chem. 32:2576–2583. https://doi.org/10.1002/etc.2358.
Haynes, D., Müller, J., Carter, S., 2000. Pesticide and herbicide residues in sediments and
seagrasses from the Great Barrier Reef world heritage area and Queensland coast.
Mar. Pollut. Bull. 41:279–287. https://doi.org/10.1016/S0025-326X(00)00097-7.
Hernandez, C.E., Witter, S.G., 1996. Evaluating and managing the environmental impact of
banana production in Costa Rica: a systems approach. Ambio 25, 171–178.
Houbraken, M., Senaeve, D., Fevery, D., Spanoghe, P., 2015. Influence of adjuvants on
the dissipation of fenpropimorph, pyrimethanil, chlorpyrifos and lindane on the
solid/gas interface. Chemosphere 138:357–363. https://doi.org/10.1016/
j.chemosphere.2015.06.040.
Jacobsen, R.E., Fantke, P., Trapp, S., 2015. Analysing half-lives for pesticide dissipation in
plants. SAR QSAR Environ. Res. 26:325–342. https://doi.org/10.1080/
1062936X.2015.1034772.
LAREP-IRET, 2017. Base de datos del proyecto TROPICA. Laboratorio de Análisis de
Residuos de Plaguicidas (LAREP), Central American Institute for Studies on Toxic Sub-
stances (IRET), Universidad Nacional, Heredia, Costa Rica.
Leistra, M., Wolters, A., 2004. Computations on the volatilisation of the fungicide
Fenpropimorph from plants in a wind tunnel. Water Air Soil Pollut. 157:133–148.
https://doi.org/10.1023/B:WATE.0000038883.86688.83.
Lichiheb, N., Bedos, C., Personne, E., Benoit, P., Bergheaud, V., Fanucci, O., Bouhlel, J.,
Barriuso, E., 2015. Measuring leaf penetration and volatilization of chlorothalonil
and epoxiconazole applied on wheat leaves in a laboratory-scale experiment.
J. Environ. Qual. 44:1782–1790. https://doi.org/10.2134/jeq2015.03.0165.
Lichiheb, N., Personne, E., Bedos, C., Van den Berg, F., Barriuso, E., 2016. Implementation of
the effects of physicochemical properties on the foliar penetration of pesticides and
its potential for estimating pesticide volatilization from plants. Sci. Total Environ.
550:1022–1031. https://doi.org/10.1016/j.scitotenv.2016.01.058.
Linders, J., Mensink, H., Stephenson, G., Wauchope, D., Racke, K., 2002. Foliar interception
and retention values after pesticide application: a proposal for standardised values
for environmental risk assessment. Pest Manag. Sci. 58:315. https://doi.org/
10.1002/ps.448.
Ma, Kuo-Ching, Mackay, Donald, Lee, Sum Chi, Shiu, Wan Ying, 2006. Herbicides. Handbook
of Physical-Chemical Properties and Environmental Fate for Organic Chemicals, second
edition CRC Press:pp. 3457–3710. https://doi.org/10.1201/9781420044393.ch17.
Macinnis-Ng, C.M.O., Flores, E.E., Müller, H., Schwendenmann, L., 2012. Rainfall
partitioning into throughfall and stemflow and associated nutrient fluxes: land use
impacts in a lower montane tropical region of Panama. Biogeochemistry 111:
661–676. https://doi.org/10.1007/s10533-012-9709-0.
Mackay, D., 2001. Multimedia Environmental Models. The Fugacity Approach. 2nd ed.
CRC Press.
MacLeod, M., Fraser, A.J., Mackay, D., 2002. Evaluating and expressing the propagation of
uncertainty in chemical fate and bioaccumulation models. Environ. Toxicol. Chem.
21:700–709. https://doi.org/10.1002/etc.5620210403.
Martínez Acosta, A.M., Cayón Salinas, D.G., 2011. Growth and development dynamics of
banana (Musa AAA Simmonds cvs. Grande Nain and Valery). Rev. Fac. Nac. Agron. 64.
McKone, T.E., Lawrence, E.O., Hall, D., Kastenberg, W.E., 1997. CalTOX Version 2.3 Descrip-
tion of Modifications and Revisions. California Environmental Protection Agency,
California.
1261
Mena, F., Fernandez San Juan, M., Campos, B., Sanchez-Avila, J., Faria, M., Pinnock, M., de la
Cruz, E., Lacorte, S., Soares, A.M., Barata, C., 2014. Pesticide residue analyses and bio-
marker responses of native Costa Rican fish of the Poeciliidae and Cichlidae families
to assess environmental impacts of pesticides in Palo Verde National Park.
J. Environ. Biol. 35, 19–27.
Mitchell, C., Brodie, J., White, I., 2005. Sediments, nutrients and pesticide residues in event
flow conditions in streams of the Mackay Whitsunday Region, Australia. Catchment
Reef Water Qual. Issues Gt. Barrier Reef Reg. 51:23–36. https://doi.org/10.1016/
j.marpolbul.2004.10.036.
Moreno-González, R., León, V.M., 2017. Presence and distribution of current-use pesti-
cides in surface marine sediments from a Mediterranean coastal lagoon (SE Spain).
Environ. Sci. Pollut. Res. 24:8033–8048. https://doi.org/10.1007/s11356-017-8456-0.
Mortensen, S.R., Johnson, K.A., Weisskopf, C.P., Hooper, M.J., Lacher, T.E., Kendall, R.J.,
1998. Avian exposure to pesticides in Costa Rican banana plantations. Bull. Environ.
Contam. Toxicol. 60:562–568. https://doi.org/10.1007/s001289900662.
Oekotoxzentrum, 2016. Proposals for Acute and Chronic Quality Standards.
Paull, R.E., Duarte, O., 2011. Tropical Fruits, Crop Production Science in Horticulture Series.
CABI, Wallingford.
Pérez Valvidia, E., 2012. Response of nine Banana Cultivars in the Vegetative State to four
Levels of Shade. Centro Agronómico Tropical de Investigacion y Enseñanza (CATIE),
Costa Rica (MSC Tropical Agroforestry).
Polidoro, B.A., Morra, M.J., 2016. An ecological risk assessment of pesticides and fish kills
in the Sixaola watershed, Costa Rica. Environ. Sci. Pollut. Res. 23:5983–5991. https://
doi.org/10.1007/s11356-016-6144-0.
Polidoro, B.A., Dahlquist, R.M., Castillo, L.E., Morra, M.J., Somarriba, E., Bosque-Pérez, N.A.,
2008. Pesticide application practices, pest knowledge, and cost-benefits of plantain
production in the Bribri-Cabécar Indigenous Territories, Costa Rica. Environ. Res.
108:98–106. https://doi.org/10.1016/j.envres.2008.04.003.
Polidoro, B.A., Morra, M.J., Ruepert, C., Castillo, L.E., 2009. Pesticide sequestration in pas-
sive samplers (SPMDs): considerations for deployment time, biofouling, and stream
flow in a tropical watershed. J. Environ. Monit. 11:1866–1874. https://doi.org/
10.1039/B904329B.
Rae, D.J., 2002. Use of spray oils with synthetic insecticides, acaricides and fungicides.
Spray Oils Beyond 2000: Sustainable Pest and Disease Management. University of
Western Sydney, Sydney, Australia, pp. 248–284.
Ramírez, F., Chaverri, F., de la Cruz, E., Wesseling, C., Castillo, L., Bravo, V., 2009.
Importación de Plaguicidas en Costa Rica Periodo 1977–2006 [Pesticide Import into
Costa Rica from 1997–2006] (No. Informes Tecnicos IRET no. 6). Universidad
Nacional, Heredia, Costa Rica.
Ramirez, F., Moraga, G., Berrocal, S., Bravo, V., 2011. Uso del suelo, cultivos y plaguicidas.
Area de influencia Laguna Madre de Dios. (Internal Report TROPICA Proyect). IRET,
Universidad Nacional, Heredia, Costa Rica.
Rämö, R.A., van den Brink, P.J., Ruepert, C., Castillo, L.E., Gunnarsson, J.S., 2016. Environ-
mental risk assessment of pesticides in the River Madre de Dios, Costa Rica using
PERPEST, SSD, and msPAF models. Environ. Sci. Pollut. Res. 1–16. https://doi.org/
10.1007/s11356-016-7375-9.
Rasmussen, J.J., Reiler, E.M., Carazo, E., Matarrita, J., Muñoz, A., Cedergreen, N., 2016. Influ-
ence of rice field agrochemicals on the ecological status of a tropical stream. Sci. Total
Environ. 542 (Part A):12–21. https://doi.org/10.1016/j.scitotenv.2015.10.062.
Robinson, J.C., Galán, V., 2010. Bananas and Plantains. CABI, Wallingford https://doi.org/
10.1079/9781845936587.0000.
Rubin, B.D., Hyman, G.G., 2000. Chapter 15 - the extent and economic impacts of soil ero-
sion in Costa Rica. In: Leclerc, C.A.S.H.L.P. (Ed.), Quantifying Sustainable Development.
Academic Press, San Diego, pp. 449–471.
Sevenhuysen, R., Maebe, P., 1995. Hydrology and drainage options atlantic zone of Costa
Rica. Serie Técnica. Centro Agronómico Tropical de Investigacion y Enseñanza
(CATIE), Turrialba, Costa Rica.
Stehle, S., Schulz, R., 2015. Agricultural insecticides threaten surface waters at the global
scale. Proc. Natl. Acad. Sci. 112, 5750–5755.
Stover, R., Simmonds, N., 1987. Bananas. Tropical Agriculture Series, 3rd ed Longman Sci-
entific & Technical, Harlow, Essex, England.
Svanes, E., Aronsson, A.K.S., 2013. Carbon footprint of a Cavendish banana supply chain.
Int. J. Life Cycle Assess. 18:1450–1464. https://doi.org/10.1007/s11367-013-0602-4.
Thomas, P.J., Mineau, P., Juraske, R., 2011. Determining pesticide foliar half-lives from soil
half-life value: not so “cut-and-dry”. Chemosphere 84:1531–1533. https://doi.org/
10.1016/j.chemosphere.2011.05.030.
Trapp, S., 2007. Fruit tree model for uptake of organic compounds from soil and air. SAR
QSAR Environ. Res. 18:367–387. https://doi.org/10.1080/10629360701303693.
Trapp, S., Eggen, T., 2013. Simulation of the plant uptake of organophosphates and other
emerging pollutants for greenhouse experiments and field conditions. Environ. Sci.
Pollut. Res. 20:4018–4029. https://doi.org/10.1007/s11356-012-1337-7.
Trapp, S., Legind, C.N., 2011. Uptake of organic contaminants from soil into vegetables and
fruits. In: Swartjes, F.A. (Ed.), Dealing With Contaminated Sites: From Theory To-
wards Practical Application. Springer Netherlands, Dordrecht:pp. 369–408. https://
doi.org/10.1007/978-90-481-9757-6_9.
Trapp, S., Matthies, M., McFarlane, C., 1994. Model for uptake of xenobiotics into plants:
validation with bromacil experiments. Environ. Toxicol. Chem. 13:413–422. https://
doi.org/10.1002/etc.5620130308.
Turner, D.W., Fortescue, J.A., Thomas, D.S., 2007. Environmental physiology of the bananas
(Musa spp.). Braz. J. Plant Physiol. 19, 463–484.
Undeman, E., Czub, G., McLachlan, M.S., 2009. Addressing temporal variability when
modeling bioaccumulation in plants. Environ. Sci. Technol. 43:3751–3756. https://
doi.org/10.1021/es900265j.
USDA, 2007. Pesticide Data Program Annual Summary, Calendar Year 2006.
USDA, 2008. Pesticide Data Program Annual Summary, Calendar Year 2007.
USDA, 2014a. Pesticide Data Program Annual Summary, Calendar Year 2012.
1262 A. Mendez et al. / Science of the Total Environment 613–614 (2018) 1250–1262
USDA, 2014b. Pesticide Data Program Annual Summary, Calendar Year 2013.
USDA, 2016a. Pesticide Data Program Annual Summary, Calendar Year 2014.
USDA, 2016b. Pesticide Data Program Annual Summary, Calendar Year 2015.
USEPA, 2016. Estimation Programs Interface Suite™ for Microsoft® Windows, v 4.11.
United States Environmental Protection Agency, Washington, DC, USA.
van Velzen, J., 1994. Estimation of the Potential and Actual Yield of Banana in the Atlantic
Zone of Costa Rica, Atlantic Zone Programme. Centro Agronómico Tropical de
Investigacion y Enseñanza (CATIE), Ministerio de Agricultura y Ganadería, Agricultur-
al University Wageningen, Wageningen, Cost Rica.
van Wendel de Joode, B., Barraza, D., Ruepert, C., Mora, A.M., Córdoba, L., Öberg, M.,
Wesseling, C., Mergler, D., Lindh, C.H., 2012. Indigenous children living nearby plan-
tations with chlorpyrifos-treated bags have elevated 3,5,6-trichloro-2-pyridinol
(TCPy) urinary concentrations. Environ. Res. 117:17–26. https://doi.org/10.1016/
j.envres.2012.04.006.
van Wendel de Joode, B., Mora, A.M., Córdoba, L., Cano, J.C., Quesada, R., Faniband, M.,
Wesseling, C., Ruepert, C., Öberg, M., Eskenazi, B., Mergler, D., Lindh, C.H., 2014. Aerial
application of mancozeb and urinary ethylene thiourea (ETU) concentrations among
pregnant women in Costa Rica: the Infants' Environmental Health Study (ISA). Envi-
ron. Health Perspect. 122:1321–1328. https://doi.org/10.1289/ehp.1307679.
Vargas, S., 2013. Effect of Runoff on the Variation of Pesticide Concentrations in Surface
Waters of the Caño Azul River, Limón, Costa Rica. University of Costa Rica, Heredia,
Costa Rica (Masters Thesis).
Vargas-Calvo, A., Valle-Ruiz, H., 2011. Effect of two types of bags on banana (Musa AAA)
fruits. Agron. Mesoam. 22. https://doi.org/10.15517/am.v22i1.8670.
Wania, F., McLachlan, M.S., 2001. Estimating the influence of forests on the overall fate of
semivolatile organic compounds using a multimedia fate model. Environ. Sci.
Technol. 35:582–590. https://doi.org/10.1021/es0011919.
Wegmann, F., Scheringer, M., Möller, M., Hungerbühler, K., 2004. Influence of vegetation
on the environmental partitioning of DDT in two global multimedia models. Environ.
Sci. Technol. 38:1505–1512. https://doi.org/10.1021/es034262n.
Xi, X., Johnson, M.S., Jeong, S., Fladeland, M., Pieri, D., Diaz, J.A., Bland, G.L., 2016.
Constraining the sulfur dioxide degassing flux from Turrialba volcano, Costa Rica
using unmanned aerial system measurements. J. Volcanol. Geotherm. Res. 325:
110–118. https://doi.org/10.1016/j.jvolgeores.2016.06.023.
Zabkiewicz, J.A., 2002. Enhancement of pesticide activity by oil adjuvants. Spray Oils Be-
yond 2000: Sustainable Pest and Disease Management. University of Western Syd-
ney, Sydney, Australia, pp. 52–54.