Comparison of annual dry and wet deposition fluxes of selected ...
Environmental Pollution 157 (2009) 303–312
Contents lists available at ScienceDirect
Environmental Pollution journal homepage: www.elsevier.com/locate/envpol
Comparison of annual dry and wet deposition fluxes of selected pesticides in Strasbourg, France Nathalie Sauret a,1, Henri Wortham a, *, Rafal Strekowski a, Pierre Hercke`s b, Laura Ines Nieto a a b
Marseilles University, Laboratoire Chimie Provence – UMR 6264, Campus Saint Charles, Case 29, 3 Place Victor Hugo, 13331 Marseilles Cedex 03, France Arizona State University, Department of Chemistry and Biochemistry, Tempe, AZ 85287-1604, USA
A modified one-dimensional cloud water deposition model is used to estimate the deposition fluxes of pesticides in the particle phase and compare the relative importance of dry and wet depositions.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 April 2008 Received in revised form 19 June 2008 Accepted 22 June 2008
This work summarizes the results of a study of atmospheric wet and dry deposition fluxes of Deisopropyl-atrazine (DEA), Desethyl-atrazine (DET), Atrazine, Terbuthylazine, Alachlor, Metolachlor, Diflufenican, Fenoxaprop-p-ethyl, Iprodione, Isoproturon and Cymoxanil pesticides conducted in Strasbourg, France, from August 2000 through August 2001. The primary objective of this work was to calculate the total atmospheric pesticide deposition fluxes induced by atmospheric particles. To do this, a modified one-dimensional cloud water deposition model was used. All precipitation and deposition samples were collected at an urban forested park environment setting away from any direct point pesticide sources. The obtained deposition fluxes induced by atmospheric particles over a forested area showed that the dry deposition flux strongly contributes to the total deposition flux. The dry particle deposition fluxes are shown to contribute from 4% (DET) to 60% (cymoxanil) to the total deposition flux (wet þ dry). Ó 2008 Elsevier Ltd. All rights reserved.
Keywords: Pesticides Wet deposition Dry deposition Model development for dry deposition
1. Introduction Ubiquitous in today’s farming culture, pesticides are used to protect the crops against weeds (herbicides), fungal parasites (fungicides), insects (insecticides) and slugs and snails (molluscicides). Not surprisingly, the current agricultural practice is considered to be the main source of atmospheric pesticide pollution (Samsonov et al., 1998; Scholtz et al., 2002). Other important sources of pesticide pollution include production and industrial and urban applications. Globally, about 2.5 million tons of pesticides are sprayed onto cultivated land per year (Bergstrom and Stenstrom, 1998; Centner, 1998). In the late 1990s, about 500 106 kg of pesticides were sprayed annually in Europe (Candela, 2003). This corresponds to an average pesticide dose of 4.4 kg ha1 (Candela, 2003). In France, about 19 million hectares of crops are sprayed annually with pesticides, i.e. 35% of the total surface area of France. Once sprayed over the desired area, pesticides may then enter the atmosphere via direct volatilization at the moment of their application or volatilize later from the ground surface or vegetation. The
* Corresponding author. Tel.: þ33 4 91 10 62 44; fax: þ33 4 91 10 63 77. E-mail address: [email protected] (H. Wortham). 1 Present address: University of Nice Sophia-Antipolis, Laboratoire de Radiochimie, Sciences Analytiques et Environnement (LRSAE), Parc Valrose, 06108 Nice, France. 0269-7491/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.envpol.2008.06.034
volatilized pesticides may then undergo further volatilization/ sublimation or be adsorbed on particle surfaces (Glotfelty et al., 1989). This ‘post-application’ volatilization represents a secondary source of pesticide pollution that can last for a relatively long time. For example, about 80–90% of certain pesticides can be lost to volatilization over several days (Glotfelty et al., 1984; Majewski et al., 1993; Taylor et al., 1976). Once airborne, pesticides are mixed within the boundary-layer where their atmospheric fate is mostly determined by given meteorological conditions (Van Pul et al., 1999). However, once the pesticides enter the free troposphere their atmospheric residence time is considerably prolonged and then may be transported over great distances on the global scale including the polar regions (Buser, 1990; Rawn et al., 1999a,b; Waite et al., 1995; Wania and Mackay, 1996). Once in the atmosphere, pesticides are partitioned among the gas, solid (particles) and liquid phases depending on their physical and chemical properties and the environmental conditions (Tsal and Cohen, 1991). Primary physical/chemical properties of the selected pesticides are shown in Table 1 (Sauret, 2002). The choice of pesticides and metabolites used in this work and shown in Table 1 is based on their different physico-chemical properties to better identify key factors that characterize the pesticide’s atmospheric behavior and fate. The understanding of the partition among the three phases is essential to model and predict the atmospheric transport and the environmental fate of pesticides.
304
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
Table 1 Primary physical properties of the selected pesticides Pesticide
Molecular Vapor pressure, weight, (g mol1) pvap (Pa)
DEAa DETa Atrazine Terbuthylazine Alachlor Metolachlor Diflufenican Fenoxaprop-p-ethyl Iprodione Isoproturon Cymoxanil
137.7 187.6 215.7 229.7 269.8 283.8 394.3 361.8 330.2 206.3 198.2
a
2.4& 104 1.5& 104 2.6& 104 4& 103 2.1& 103 2.4& 103 3.3& 102 2.7& 104 7& 106 9.7& 106 2.7& 105
were prepared in ethyl acetate solvent and stored at 4 C. Deionized water was used with resistivity above 18 MU cm at 25 C (corresponding roughly to less than 1 ppb total ionic contamination in water) and was prepared by passing tap water through a reverse osmosis demineralization filter followed by a commercial deionizer (Millipore, Milli-Q). Sodium chloride (Prolabo, France) had the stated minimum purity of >99.5%.
– 6.9 4.7 13.5 2.8 2.3 40.2 3.8 1.5 2.3 21.3
Triazines metabolite products.
Atmospheric sinks of pesticides include: (1) wet deposition (Kumar, 2001; Majewski et al., 1998), (2) dry deposition (Kumar, 2001; Majewski et al., 1998), (3) direct (Le Person et al., 2007) and indirect photochemical degradation (Feigenbrugel et al., 2006; Grover et al., 1997), and (4) microbial biotic (Alexander, 1994; Moorman, 1993; Skipper and Turco, 1995) and (5) microbial abiotic degradation (Wolfe et al., 1990). Other pesticide degradation pathways include hydrolysis in soil and redox reactions, which occur in aerobic soil (oxidation reactions) and anaerobic soil (reduction reactions). The indirect photochemical lifetime of pesticides in the three phases is mostly governed by the reactions with the OH radicals (Martin et al., 2002), NO3 radicals and O3 molecules among others. Although extensive studies have been devoted to the deposition of trace gases and particles to forest canopies, less work has been reported in pesticide deposition models for other vegetated surfaces (Kumar, 2001; Le Person et al., 2007; Majewski et al., 1998). The latter has been a challenge to the scientific community of environment science. As a result, to better understand the atmospheric fate of pesticides, this work focuses on the impact of pesticide removal by wet and dry deposition by vegetative surfaces. To date, the atmospheric deposition of pesticides is often estimated using wet deposition only. However, particle phase pesticides may also be removed from the atmosphere by dry deposition. In turn, pesticide dry deposition rates are evaluated using aerosol deposition. This is problematic since based on the results obtained in this work, the selected pesticides presented here are mostly found in the particulate phase. Since the selected pesticides in this work are of different physical/chemical properties and multiple phases in environmental media, their dry deposition processes cannot be simply calculated using wet deposition. In effect, if the atmospheric particulate phase pesticide concentration is known, no information about particle deposition rates is available. As a result, in this work, the atmospheric pesticide deposition fluxes induced by atmospheric particles are evaluated by modifying the original one-dimensional cloud water (droplets) deposition model to the new sampling and experimental objectives of this work to measure wet and dry particle deposition fluxes. 2. Experimental
2.2. Sampling All gas, particle and rain samples were collected in an urban environment, a noncultivated park of the Strasbourg observatory. Strasbourg is a major metropolitan area in the East of France with a total urban and suburban population of 500,000. Industrial activities, which comprise petrochemical industries, incinerators and breweries, are located outside the city limits. An important agricultural area is located about 5 km away from the city center (thus the sampling site) with the agricultural activities focusing on growing wheat, corn, beets, hop, and grapes (vines). Because of the distance between the agricultural areas and the sampling site, the pesticide concentrations measured on this site were assumed to be representative of background rural atmospheric conditions away from direct pesticide sources. Similar to the study by Sauret and coworkers (Sauret et al., 2000), gas phase and particulate matter samples were collected using 20 g of the XAD-2 resin and a 30 cm diameter glass fiber filter, respectively. Briefly, the atmospheric sample was first allowed to pass through the fiber glass filter and then through the XAD-2 resin connected is series. All sampling (particulate matter and gas samples) was stopped once a volume of 400 m3 of gas has been collected (typically 3–4 days). To prevent filter clogging and resin saturation, both, the filters as well as the XAD-2 traps were changed every 24 h. Further, before use all filters and resin traps were Soxhlet washed for 24 h with 85/15 hexane/dichloromethane solution. To limit contamination, the filters were wrapped in Joseph paper and stored in thermally sealed plastic bags. The XAD-2 resins were stored in polyethylene containers. Following sampling, all traps were immediately placed in a freezer and stored at 30 C until analysis. Meteorological measurements including temperature, pressure, relative humidity, wind speed and direction plus the total suspended matter (TSP) were assured by the local Atmospheric Pollution Monitoring Association (ASPA). The TSP was collected for particles with an aerodynamic diameter (Dp) ranging from 0.1 mm to 10 mm. As a result, particles with an aerodynamic diameter greater than 10 mm were not sampled. Moreover, due to experimental limitations, particles with an aerodynamic diameter less than 0.1 mm, i.e. fine particles, were not trapped. Atmospheric temperature, pressure, relative humidity and TSP for particles with an aerodynamic diameter in the range from 0.1 mm to 10 mm were measured and recorded at the DRIRE site located in the city center. However, to better characterize dominant wind conditions, the wind speed and direction were measured by the national meteorological agency, Meteo France, at the city of Reichstett located in the greater Strasbourg area. Prior to sampling, the glass fiber filters and the XAD-2 resins were Soxhlet preextracted with n-hexane/CH2Cl2 (85/15) for 24 h (Albanis et al., 1986), dried at 60 C and then stored in thermally sealed plastic bags. Blanks were used to verify the efficiency of the cleaning and conservation procedures. Rain samples were collected on a weekly basis. An automatic wet-only rainwater collector was used to collect the rain while an open collector was used to directly measure the precipitation levels (Sanusi et al., 2000). After each sampling collection, the funnel and the collection bottle were cleaned and rinsed thoroughly with Milli-Q water. Samples were stored in the dark in pre-cleaned glass bottles at 4 C prior to analysis (Sanusi et al., 2000; Sauret-Szczepanski et al., 2006). 2.3. Sample extraction Similar to the situation listed above, the loaded XAD-2 resin and the glass fiber filters were separately Soxhlet-extracted for 12 h using 250 mL of the n-hexane/ CH2Cl2 (85/15) mix (Atlas and Schauffler, 1990; Sauret, 2002) All samples were extracted within 24 h following sampling to reduce any pesticide degradation, evaporation, contamination and reaction among others. The resulting extraction solutions were then concentrated to a volume of approximately 1 mL in a rotary evaporator at 30 C at a reduced pressure of 280 mbar. Reduced system pressure was used to limit pesticide degradation (Sauret, 2002). Rain samples were extracted using the Solid Phase Micro Extraction method. The sampling and extraction details for rain samples have recently been published and will not be discussed further (Sauret-Szczepanski et al., 2006). Only rain results pertaining to this work are discussed below.
2.1. Chemicals 2.4. Sample analysis The solvents used in this work with stated minimum purities were n-hexane (HPLC Grade, Carlo Erba) and methylene dichloride (HPLC Grade, SDS, Peypin, France). The AmberliteÔ XAD-2 resin (Rohm and Hass) and glass fiber filters (Whatman, GF/A) were obtained from Prolabo and used without any further modification. All reagents used in this work were obtained from the U.I.P.P. (Union des Industries de la Protection des Plantes, France) member companies and were 98– 99% pure: Atrazine, Terbuthylazine and Metolachlor (Syngenta, France); Alachlor (Monsanto, France); Diflufenican, Isoproturon, Iprodione and Fenoxaprop-p-ethyl (Aventis CropScience, France); DIA, DEA and DET (Promochem). Stock solutions
The analysis approach used in this work is similar to the one used in previous studies for the determination of atmospheric pesticides (Sauret-Szczepanski et al., 2006; Sauret et al., 2000). Only details relevant to this study are presented here. All extracted samples were analyzed using the Varian Star 3400 CX gas chromatograph (GC) equipped with a split/splitless injector and coupled to a Saturn 4D Varian ion trap mass spectrometer. Both, the GC and the ion trap were controlled using the Saturn IV acquisition software. The ion trap mass spectrometer was operated in the MS/MS mode. The gas chromatograph was fitted with a 30 m 0.32 mm i.d. J&W
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312 Scientific fused-silica DB-5-MS capillary column with a film thickness of 0.25 mm. The carrier gas was helium and the inlet pressure was 12 psi that corresponded to a flow rate of about 1 mL min1. Initially, the GC oven temperature was held at 60 C for 2 min. After the 2 min delay, the oven temperature was increased from 60 C to 145 C at 15 C min1 and then maintained at that temperature (T ¼ 145 C) for 3 min. After 3 min, the oven temperature was increased to 151 C at 1 C min1. Then, once the oven temperature reached 151 C, the oven temperature was immediately increased to 250 C at 11 C min1 and maintained at this temperature for 10 min. The injector, transfer line and the manifold temperatures were kept at 290, 280 and 200 C, respectively. Linearity and sensitivity of the ion trap detector were optimized by adjusting various parameters such as filament emission current, voltage of the electron multiplier and the total number of ions in the trap (target). These adjustments and the improvement induced by the MS/MS mode compared to the simple-MS mode were discussed in detail previously (Sauret et al., 2000). Even though the electron impact (EI) ionization induced a much higher molecular fragmentation (relative intensity of the parent ion very small or nonexistent) than chemical ionization (CI), the CI signal resulted in a much higher background noise than EI. As a result, the mass spectrometer was operated in the EI ionization mode where the ionization energy was 70 eV. Then, in the MS/MS mode, the daughter ion spectrum was obtained by isolating the most intense fragment ion (created in the simple-MS mode) followed by acceleration and collision with the helium carrier gas molecules (collision induced dissociation, CID). The daughter ions generated by CID were then used for identification and quantification. The details of the MS/MS identification/quantification procedure are described elsewhere (Sauret et al., 2000). The MS/MS parameters used for the pesticides analysis under study are summarized in Table 2 (Sauret-Szczepanski et al., 2006; Sauret, 2002; Sauret et al., 2000). All standard pesticide solutions were prepared in hexane. The quantification limits (QL) for the selected pesticides were obtained assuming that the sampling volume was 700 cm3 and that, after extraction, samples were concentrated to 1 mL. The quantification limits were (in ng m3): 0.08 for Cymoxanil and Isoproturon, 0.04 for DEA, Iprodione and Fenoxaprop-p-ethyl, and 0.016 for DET, Atrazine, Terbuthylazine, Alachlor, Metolachlor and Diflufenican. 2.5. Model description To better evaluate the atmospheric pesticide deposition, a modified version of the original one-dimensional cloud water deposition model first developed by Lovett (Lovett, 1984; Lovett et al., 1982) and then described in detail by Pahl et al. (1994) was used. Briefly, the original cloud deposition model is a one-dimensional resistance model that divides the vegetation into 1 m layers (Lovett et al., 1982). The flux of cloud water is calculated for each layer using an analogue of the Ohm’s law. F ¼
DC R
¼ DC vd
(1)
In Eq. (1) above F is the cloud water flux, DC is the cloud droplet concentration gradient, R is the resistance acting against the droplet deposition in each layer and vd is the deposition flux. The resistance, R, is the sum of the aerodynamic resistance and boundary-layer resistance of individual canopy components, such as needles, branches and twigs. The resistance values are calculated for each layer using characteristic vegetation profiles. By definition, R is a resistance to the droplet transfer; therefore, R is inversely proportional to the deposition flux and vice versa. Since its initial application, the one-dimensional cloud water deposition model has been modified several times to better adapt new sampling parameters and experimental objectives (Hercke`s et al., 2002; Pahl et al., 1994; Wobrock et al., 1994). One of the major assumptions of the modified model used in this work is that the attachment (deposition) is definitive and irreversible with no evaporation. Moreover, while the original cloud deposition model measures wet deposition of liquid droplets with a 2–50 mm aerodynamic diameter (Pahl et al.,1994), the modified version used in Table 2 The MS/MS parameters used in this work (Sauret-Szczepanski et al., 2006; Sauret, 2002) Pesticide
DEA DET Atrazine Terbuthylazine Alachlor Isoproturon Metolachlor Diflufenican Iprodione Fenoxaprop-p-ethyl Cymoxanil
Base peak ion used as parent ion (m/z)
CID voltage (V)
172 186 200 214 188 191 162 266 314 361 177
49 49 55 56 45 40 45 81 69 75 42
Daughter ions used for quantification (m/z) 79, 104 104, 145 94, 104, 174 104, 173 132, 160 146 117, 120, 134 218, 238, 246 245, 271 244, 261, 288 119, 135, 149
305
this work measures deposition of particles (solid) with the aerodynamic diameter in the range from 0.1 to 10 mm with a 0.1 mm resolution, i.e. 100 classes of particles. As a result, the sedimentation rate had to be modified to account for the particle aerodynamic diameter difference. Other parameters that were modified include distribution of particles using proper aerosol distribution.(Hercke`s, 1999; Sauret, 2002) The original model version developed by Pahl et al. (1994) used hourly meteorological data provided by stations at different altitudes. Here, in a similar situation to the work of Hercke`s (Hercke`s, 1999; Hercke`s et al., 2002), the meteorological data was measured continuously and directly incorporated into the model. The most important model modification carried out on Pahl’s model and adapted it to the particle deposition model used in this work is that the Total Suspended Particles (TSP, g cm3) was substituted for the Liquid Water Content (LWC, g cm3). This data was measured on an hourly basis and the average value was incorporated into the model. 2.6. Wind speed The logarithmic wind speed profile above the canopy can be expressed using Eq. (2) (Thom, 1975). The wind speed profile (Eq. (2)) applies only in the neutral surface boundary-layer where the atmospheric stability is ignored. uðzÞ ¼
u* z d ln z0 K
ðz > forest height; HÞ
(2)
where u(z) is the wind speed at the measurement height z ¼ 10 m (this work), u* is the friction velocity, K is the von Karman constant (K ¼ 0.41), d is the zero plane displacement height in meters and z0 is the roughness height. Using the wind speed calculated using Eq. (2) above, the wind speed u1 over the sampling site is determined (Eq. (3)). u1 ðzÞ ¼
u*1 z ln z0G K
(3)
In Eq. (3) above, z0G ¼ 0.02 m (this work) and is the lawn grass roughness height and u*1 is the friction velocity. For the roughness lengths of less than 0.02 m, d is negligible and is ignored (Pahl, 1996). The friction velocity, u*1, is calculated for the measurement height zM ¼ 10 m used in this work using following Eq. (4). K M lnzz0G
u*1 ¼ u1 ðzM Þ
(4)
To transpose the wind profile u1 to the wind profile over the forest (u2), the profile u1 undergoes a vertical translation equivalent to 2/3 of the forest height corresponding to the difference between the zero plane displacement height (d) at the wind speed measurement site and in the forest. The friction velocity u2* is then directly proportional to u*1 (Eq. (5)). u1 ð2HÞ ¼ u2 ð2H þ dÞ
(5)
In Eq. (5) above, u(H) is the wind speed at the tree top height H. The wind speed at the tree top level can be calculated using the following Eq. (6), uðHÞ ¼
u*2 H d ln zoF K
(6)
where zoF is the forest roughness height (in Europe, typically zoF ¼ 1–3 m) and is equivalent to 2/3 of the forest height (Brutsaert, 1975; Flemming, 1982; Pahl, 1996). If z < H, then the wind profile can be expressed using the following relation (Eq. (7)). uðzÞ ¼ uðHÞexp½bu CSAIðzÞ
(7)
In Eq. (7) above bu is the experimentally determined wind reduction coefficient and CSAI(z) is the Cumulative Surface Area Index. Based on the experimental work of Lovett and coworkers bu ¼ 0.27 (Lovett et al., 1982). The CSAI depends on the tree height and is calculated as a function of leaf/needle surface area and the planar canopy surface area (horizontal surface projection). The expression for the vertical turbulent coefficient profile that directly determines the aerodynamic resistance within the canopy is written in the similar form as u(z) Kd ðzÞ ¼ Kd ðHÞexp½bk CSAIðzÞ
ðz < HÞ
(8)
where Kd(z) (m2 s1) is the turbulent diffusion coefficient at the height z, and bk is the turbulent reduction coefficient. According to Lovett and Pahl, bk ¼ 0.14 (Lovett et al., 1982; Pahl, 1996). The Kd(H) is the forest limit coefficient and is calculated using the following Eq. (9): Kd ðHÞ ¼ Ku* ðH dÞ
(9)
As a result, the forest aerodynamic resistance Ra (s m1) at level i becomes
Ra ðiÞ ¼ zðiÞ
1 1 þ Kd ðiÞ Kd ði 1Þ
(10)
where z(i) (m) is the forest level height and Kd is the turbulent diffusion coefficient at the level i.
306
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
2.7. Vegetation The Strasbourg sampling site forest is a closed spruce stand (Picea abies). The forest vegetation is classified in terms of its Surface Area Index (SAI) which corresponds to the interface between the ‘free’ atmosphere and trees. The SAI represents the total impact surface for aerosols (droplets or particles) and is the algebraic sum of the total tree surface, i.e. needles, branches, twigs among others, normalized to the ground surface area. The sampling site forest SAI (m2 m2) is calculated using the following Eq. (11) (Hercke`s et al., 2002; Pahl, 1996): (11)
SAI ¼ 91:8 þ 14:3 d1:3
where d1.3 (cm) is the tree trunk diameter 1.3 m above the ground. Sampling site forest characteristics are shown in Table 3 (Asae¨l, 1990; Sauret, 2002). 2.8. Granulometric particle distribution As stated above, the cloud deposition model was originally used for droplets. In order to modify the model from droplet deposition to particle deposition several parameters had to be changed. For one, it is important to estimate the total mass contribution as a function of the particle size to the total suspended particles (TSP). The sedimentation and impact processes will be directly depended on the particle aerodynamic diameter. As a result, in order to better classify the total atmospheric distribution spectrum (Seinfeld and Pandis, 1998) particle class with an aerodynamic diameter in the range from 0.1 to 10 mm with a 0.1 mm resolution, i.e. 100 classes of particles, was used. Further, it was not possible to experimentally evaluate the granulometric size particle distribution in this work. Given the geographical location and size of the sampling site (Strasbourg observatory park) standard rural particle size distribution was used (Seinfeld and Pandis, 1998). The atmospheric aerosol size distribution can be described using the following Eq. (12) (Seinfeld and Pandis, 1998): n X N pffiffiffiffiffiffi i exp n0N log Dp ¼ i ¼ 1 2plog si
2 ! log Dp log Dpi 2log2 si
(12)
where n0N ðlog Dp Þ (cm3) is the number of particles as a function of logarithmic diameter Dp (mm), Ni (cm3) is the average number of particles, Dpi is the average particle diameter and si is the standard deviation. Then, surface volume distributions can be calculated using Eqs. (13) and (14), respectively. n0S log Dp ¼ pD2p n0N log Dp mm2 cm3
(13)
p n0V log Dp ¼ D3p n0N log Dp mm3 cm3 6
(14)
Fig. 1. Rural particle size distribution in volume.
2.9. Sedimentation velocity Once the particle distribution is known (see above), the sedimentation (gravitational) velocity is calculated using the following Eq. (17):
vs ¼
2 1 Dp rp gCc m 18
Dp in m
(17)
In the above Eq. (17) g is the gravitational constant (9.806 m s2), rp is the volume mass of a single particle (1.7 106 g m3), m is the air viscosity (1.72& 102 g m1 s1 at T ¼ 20 C) and Cc is the Cunningham factor (sliding factor) and is dependent on the particle diameter (Seinfeld and Pandis, 1998). For large particles the sliding factor can be considered negligible. However, for small particles the sedimentation velocity can be multiplied by values ranging from Cc ¼ 2.85 for Dp ¼ 0.1 mm to Cc ¼ 216 for Dp ¼ 0.001 mm (Seinfeld and Pandis, 1998). Correction sliding factors for spherical particles in air at T ¼ 298 K and p ¼ 1 atm are given in Table 4. 2.10. Resistance model
Modification of the particle distribution did not show to influence the total particle quantities deposited. Actually, rural particle deposition has been shown to be slightly greater than urban particle deposition. Typically, ‘bigger’ particles are more abundant in rural regions (Dp > 1 mm) compared to urban regions. In urban regions smaller diameter particles are more abundant (Dp < 1 mm). Using the rural particle size distribution slightly overestimates the total particle deposition (higher deposition flux), a result that is in line with the research objectives of this study to estimate the maximum particle deposition. Rural volume aerosol distribution used in this work is shown in Fig. 1. Once the particle size distribution is known, the total standard particle volume, VT, can be calculated (Eq. (15)). VT ¼
p 6
Z
þN N
D3p n0N log Dp dlog Dp
(15)
Typical Total Suspended Particles, TSP , are then calculated by multiplying VT by aerosol density (rp) (g mm3) (Eq. (16)).
TSP ¼ rp VT g cm
3
(16)
Table 3 The Strasbourg forest sampling site characteristics Parameter
Average characteristic
Tree type Age Height Density d1,3a SAIb
Picea abies 80–100 yr 25.8 m 525 branches ha1 35.8 cm 22 m2 m2
a b
Rij ¼
1 CEij uðiÞ
(18)
Then, the surface resistance at level i (Ri) can be written as the inverse of the sum of both the vertical surface repartition of different vegetation components normalized to the sum of resistances Rij in parallel (Eq. (19)). Ri ¼ P 7
1
(19)
SAIij j ¼ 1 Rij
0
0
To calculate surface resistance, Rij (s m1), i.e. transport resistance of vegetation particle component j at the tree (forest) level i, the wind speed (u(i)) and capture coefficient of vegetation particle component j (CEij) at the height i has to be known (Eq. (18)).
Tree trunk diameter 1.3 m above the ground. Calculated using Eq. (11).
The independent collection efficiency of particle deposition on vegetation can be written as the sum of the three different contributions (Slinn, 1982), CE ¼ RðEB þ EIN þ EIM Þ
(20)
where R is the particle rebound factor, EB is the collection efficiency resulting from Brownian diffusion, EIN is the collection efficiency resulting from interception by vegetation and EIM is the collection efficiency resulting from direct impact of particles on vegetation. Since the Brownian diffusion collection efficiency is valid only for particles with the aerodynamic diameter less than 0.1 mm (Dp < 0.1 mm) it was ignored. Moreover, to better determine the maximum particle deposition flux
Table 4 Correction sliding factors for spherical particles in air at T ¼ 298 K and p ¼ 1 atm Particle diameter, Dp (mm)
Cc
0.1 0.2–0.5 0.6–1 1.1–10
2.85 1.5955 1.245 1.05
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312 R is set equal to 1 (no particle rebound). Thus, Eq. (20) can be rewritten to give Eq. (21). CE ¼ EIN þ EIM
(21)
The EIN is defined as follows (Slinn, 1982): EIN ¼
Dp Dp cv þ ð1 FÞ F c d LS þ D p LL þ Dp
(22)
where cv =cd is the ratio of the viscous drag to the total drag (usually cv =cd ¼ 1=3), LS is the characteristic small vegetation collection diameter (LS < 1 mm), LL is the characteristic large vegetation collection diameter (LL > 1 mm) and F is the interception fraction by small collectors (usually about 1%) (Ruijgrok et al., 1997; Slinn, 1982). In this work, small collectors with diameter less than 1 mm were ignored; therefore, F ¼ 0. Then, Eq. (22) is reduced to give Eq. (23). EIN ¼
Dp cv cd LL þ Dp
(23)
To evaluate particle deposition on smooth surfaces, the following semi-empirical relation can be used (Chamberlain, 1967): EIM ¼ 103=St
(24)
where St is the Stokes number (Eq. (25)). St ¼
vs uðzÞ gð2L
(25)
In Eq. (25) above, vs is the particle sedimentation velocity, u(z) is the wind velocity at the height z and L is the impact surface diameter. A different evaluation of EIM was proposed by Slinn (1982) where the strong dependence on the Stokes number increases the impact rate (Eq. (26)). EIM ¼
ðStÞ2 1 þ ðStÞ2
(26)
If the deposition fluxes are calculated using Eqs. (24) and (26) as a function of the measured wind speed u(zm) and fixing the TSP to the average value of 0.03 g m3, completely different particle deposition profiles are obtained (Fig. 2). As shown in Fig. 2, deposition flux calculations using the Slinn relation demonstrate that the deposition by impact becomes more important than sedimentation at wind speeds 2 m s1 and higher. On the other hand, deposition flux calculations using the Chamberlain relation show that impact deposition becomes predominant only at wind speeds 8 m s1 and higher. Wind speeds of 8 m s1 were rarely encountered under the experimental conditions of this work. Typical wind speeds ranged from 1 to 4 m s1. As a result, similar to the situation listed above (rebound factor), the Slinn relation (Eq. (26)) was used to obtain the maximum deposition flux for pesticides under study. To calculate CEij for each type of vegetation j, the following relation was used (Pahl et al., 1994): CEij ¼
CE Cj
307
months for triazines, chloroacetanilides and fenoxaprop-p-ethyl. Observed concentration peaks during the spring months corresponded to the spraying periods of these pre-emergent herbicides on maize and soybeans. We observed the same seasonal pattern for triazines metabolites (DEA and DET) which appeared a few days after their parent compounds. Nevertheless, the reactions producing the above degradation products could occur both in surface water and in the atmosphere and with the current data we have no way to estimate the relative importance of these two modes of formation. Moreover, because the same metabolites are produced in the two media, we cannot use the present results to demonstrate the efficiency of the atmospheric reactions on the degradation of atmospheric pesticides. Nevertheless, because the metabolite concentrations in precipitation were in the same order of magnitude as the parent compounds, these molecules have to be considered in the determination of the pesticide contamination assessment. Also, high concentrations of diflufenican and isoproturon in precipitation were observed during winter months according to their spraying periods. Once the pesticide concentration in precipitation is known, the wet deposition flux is obtained using the following Eq. (28):
Di ¼ H C i i
(28) 2
where D (mg m ) is a deposition of compound i, H is the precipitation height in mm (1 mm ¼ 1 L m2) and Ci (mg L1) is the pesticide concentration in precipitation. The precipitation was collected using the permanently open rainwater collector. The precipitation height was obtained from Meteo France, a national meteorological agency. Obtained wet deposition fluxes are presented in Table 6.
(27)
where Cj is the experimentally determined correction factor (typically Cj values varied from 3 to 8).
3. Results and discussions 3.1. Pesticide wet deposition Pesticide concentrations in precipitation for the 11 molecules under study are shown in Fig. 3. As shown in Fig. 3, important seasonal pesticide concentration variation is observed with the maximum values occurring during spring when the agricultural activities are greatest. The concentrations of the 11 compounds under study were measured in the rain events occurring in Strasbourg between August 3, 2000 and August 6, 2001. Table 5 presents the concentration characteristics of the target molecules in rain samples. For technical reasons, cymoxanil was measured from January 2001 through August 2001, i.e. 7 months. The detection percentage calculation was based upon the 29 samples collected during the sampling campaign. Rain concentrations showed patterns of high concentrations of the selected pesticides in precipitation during spring months (Fig. 3). These maximal concentrations of the selected pesticides in precipitation decreased rapidly down to a level of less than 0.1 mg L1 in the following
Fig. 2. Sensitivity of the deposition flux as a function of the wind speed (TSP of 0.03 g m3).
308
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
Fig. 3. Concentrations of the target compounds (mg L1) in rain samples during the field campaign.
3.2. Model sensitivity The particle deposition model properties are affected by many physical parameters including (as discussed above) wind speed and TSP. The sensitivity of the model’s behavior to changes in parameters is known as parametric sensitivity. When a system operates in a parametrically sensitive region, its
performance becomes unreliable and changes sharply with small variations in parameters. Thus, it would be of great interest to predict sensitivity behavior in the one-dimensional particle deposition model used in this work. As a result, the influence of the (1) TSP, (2) ground surface roughness, (3) forest roughness, and (4) measured wind speed on the deposition flux is evaluated.
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
309
Table 5 Concentration characteristics of the target molecules in rain samples from 08/03/2000 to 08/06/2001 Compound
Wet deposition flux (g ha1 yr1)
Minimum (mg L1)
Maximum (mg L1)
Observed maximum pesticide concentration date (dd.mm.yy–dd.mm.yy)
Pesticide application period
DEA DET Atrazine Terbuthylazine Alachlor Metolachlor Diflufenicana Fenoxaprop-p-ethyl Iprodione Isoproturona Cymoxanilb
0.44 0.56 1.54 1.11 1.9 1.39 1.87 0.39 0.11 0.22 14.5
Contents lists available at ScienceDirect
Environmental Pollution journal homepage: www.elsevier.com/locate/envpol
Comparison of annual dry and wet deposition fluxes of selected pesticides in Strasbourg, France Nathalie Sauret a,1, Henri Wortham a, *, Rafal Strekowski a, Pierre Hercke`s b, Laura Ines Nieto a a b
Marseilles University, Laboratoire Chimie Provence – UMR 6264, Campus Saint Charles, Case 29, 3 Place Victor Hugo, 13331 Marseilles Cedex 03, France Arizona State University, Department of Chemistry and Biochemistry, Tempe, AZ 85287-1604, USA
A modified one-dimensional cloud water deposition model is used to estimate the deposition fluxes of pesticides in the particle phase and compare the relative importance of dry and wet depositions.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 April 2008 Received in revised form 19 June 2008 Accepted 22 June 2008
This work summarizes the results of a study of atmospheric wet and dry deposition fluxes of Deisopropyl-atrazine (DEA), Desethyl-atrazine (DET), Atrazine, Terbuthylazine, Alachlor, Metolachlor, Diflufenican, Fenoxaprop-p-ethyl, Iprodione, Isoproturon and Cymoxanil pesticides conducted in Strasbourg, France, from August 2000 through August 2001. The primary objective of this work was to calculate the total atmospheric pesticide deposition fluxes induced by atmospheric particles. To do this, a modified one-dimensional cloud water deposition model was used. All precipitation and deposition samples were collected at an urban forested park environment setting away from any direct point pesticide sources. The obtained deposition fluxes induced by atmospheric particles over a forested area showed that the dry deposition flux strongly contributes to the total deposition flux. The dry particle deposition fluxes are shown to contribute from 4% (DET) to 60% (cymoxanil) to the total deposition flux (wet þ dry). Ó 2008 Elsevier Ltd. All rights reserved.
Keywords: Pesticides Wet deposition Dry deposition Model development for dry deposition
1. Introduction Ubiquitous in today’s farming culture, pesticides are used to protect the crops against weeds (herbicides), fungal parasites (fungicides), insects (insecticides) and slugs and snails (molluscicides). Not surprisingly, the current agricultural practice is considered to be the main source of atmospheric pesticide pollution (Samsonov et al., 1998; Scholtz et al., 2002). Other important sources of pesticide pollution include production and industrial and urban applications. Globally, about 2.5 million tons of pesticides are sprayed onto cultivated land per year (Bergstrom and Stenstrom, 1998; Centner, 1998). In the late 1990s, about 500 106 kg of pesticides were sprayed annually in Europe (Candela, 2003). This corresponds to an average pesticide dose of 4.4 kg ha1 (Candela, 2003). In France, about 19 million hectares of crops are sprayed annually with pesticides, i.e. 35% of the total surface area of France. Once sprayed over the desired area, pesticides may then enter the atmosphere via direct volatilization at the moment of their application or volatilize later from the ground surface or vegetation. The
* Corresponding author. Tel.: þ33 4 91 10 62 44; fax: þ33 4 91 10 63 77. E-mail address: [email protected] (H. Wortham). 1 Present address: University of Nice Sophia-Antipolis, Laboratoire de Radiochimie, Sciences Analytiques et Environnement (LRSAE), Parc Valrose, 06108 Nice, France. 0269-7491/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.envpol.2008.06.034
volatilized pesticides may then undergo further volatilization/ sublimation or be adsorbed on particle surfaces (Glotfelty et al., 1989). This ‘post-application’ volatilization represents a secondary source of pesticide pollution that can last for a relatively long time. For example, about 80–90% of certain pesticides can be lost to volatilization over several days (Glotfelty et al., 1984; Majewski et al., 1993; Taylor et al., 1976). Once airborne, pesticides are mixed within the boundary-layer where their atmospheric fate is mostly determined by given meteorological conditions (Van Pul et al., 1999). However, once the pesticides enter the free troposphere their atmospheric residence time is considerably prolonged and then may be transported over great distances on the global scale including the polar regions (Buser, 1990; Rawn et al., 1999a,b; Waite et al., 1995; Wania and Mackay, 1996). Once in the atmosphere, pesticides are partitioned among the gas, solid (particles) and liquid phases depending on their physical and chemical properties and the environmental conditions (Tsal and Cohen, 1991). Primary physical/chemical properties of the selected pesticides are shown in Table 1 (Sauret, 2002). The choice of pesticides and metabolites used in this work and shown in Table 1 is based on their different physico-chemical properties to better identify key factors that characterize the pesticide’s atmospheric behavior and fate. The understanding of the partition among the three phases is essential to model and predict the atmospheric transport and the environmental fate of pesticides.
304
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
Table 1 Primary physical properties of the selected pesticides Pesticide
Molecular Vapor pressure, weight, (g mol1) pvap (Pa)
DEAa DETa Atrazine Terbuthylazine Alachlor Metolachlor Diflufenican Fenoxaprop-p-ethyl Iprodione Isoproturon Cymoxanil
137.7 187.6 215.7 229.7 269.8 283.8 394.3 361.8 330.2 206.3 198.2
a
2.4& 104 1.5& 104 2.6& 104 4& 103 2.1& 103 2.4& 103 3.3& 102 2.7& 104 7& 106 9.7& 106 2.7& 105
were prepared in ethyl acetate solvent and stored at 4 C. Deionized water was used with resistivity above 18 MU cm at 25 C (corresponding roughly to less than 1 ppb total ionic contamination in water) and was prepared by passing tap water through a reverse osmosis demineralization filter followed by a commercial deionizer (Millipore, Milli-Q). Sodium chloride (Prolabo, France) had the stated minimum purity of >99.5%.
– 6.9 4.7 13.5 2.8 2.3 40.2 3.8 1.5 2.3 21.3
Triazines metabolite products.
Atmospheric sinks of pesticides include: (1) wet deposition (Kumar, 2001; Majewski et al., 1998), (2) dry deposition (Kumar, 2001; Majewski et al., 1998), (3) direct (Le Person et al., 2007) and indirect photochemical degradation (Feigenbrugel et al., 2006; Grover et al., 1997), and (4) microbial biotic (Alexander, 1994; Moorman, 1993; Skipper and Turco, 1995) and (5) microbial abiotic degradation (Wolfe et al., 1990). Other pesticide degradation pathways include hydrolysis in soil and redox reactions, which occur in aerobic soil (oxidation reactions) and anaerobic soil (reduction reactions). The indirect photochemical lifetime of pesticides in the three phases is mostly governed by the reactions with the OH radicals (Martin et al., 2002), NO3 radicals and O3 molecules among others. Although extensive studies have been devoted to the deposition of trace gases and particles to forest canopies, less work has been reported in pesticide deposition models for other vegetated surfaces (Kumar, 2001; Le Person et al., 2007; Majewski et al., 1998). The latter has been a challenge to the scientific community of environment science. As a result, to better understand the atmospheric fate of pesticides, this work focuses on the impact of pesticide removal by wet and dry deposition by vegetative surfaces. To date, the atmospheric deposition of pesticides is often estimated using wet deposition only. However, particle phase pesticides may also be removed from the atmosphere by dry deposition. In turn, pesticide dry deposition rates are evaluated using aerosol deposition. This is problematic since based on the results obtained in this work, the selected pesticides presented here are mostly found in the particulate phase. Since the selected pesticides in this work are of different physical/chemical properties and multiple phases in environmental media, their dry deposition processes cannot be simply calculated using wet deposition. In effect, if the atmospheric particulate phase pesticide concentration is known, no information about particle deposition rates is available. As a result, in this work, the atmospheric pesticide deposition fluxes induced by atmospheric particles are evaluated by modifying the original one-dimensional cloud water (droplets) deposition model to the new sampling and experimental objectives of this work to measure wet and dry particle deposition fluxes. 2. Experimental
2.2. Sampling All gas, particle and rain samples were collected in an urban environment, a noncultivated park of the Strasbourg observatory. Strasbourg is a major metropolitan area in the East of France with a total urban and suburban population of 500,000. Industrial activities, which comprise petrochemical industries, incinerators and breweries, are located outside the city limits. An important agricultural area is located about 5 km away from the city center (thus the sampling site) with the agricultural activities focusing on growing wheat, corn, beets, hop, and grapes (vines). Because of the distance between the agricultural areas and the sampling site, the pesticide concentrations measured on this site were assumed to be representative of background rural atmospheric conditions away from direct pesticide sources. Similar to the study by Sauret and coworkers (Sauret et al., 2000), gas phase and particulate matter samples were collected using 20 g of the XAD-2 resin and a 30 cm diameter glass fiber filter, respectively. Briefly, the atmospheric sample was first allowed to pass through the fiber glass filter and then through the XAD-2 resin connected is series. All sampling (particulate matter and gas samples) was stopped once a volume of 400 m3 of gas has been collected (typically 3–4 days). To prevent filter clogging and resin saturation, both, the filters as well as the XAD-2 traps were changed every 24 h. Further, before use all filters and resin traps were Soxhlet washed for 24 h with 85/15 hexane/dichloromethane solution. To limit contamination, the filters were wrapped in Joseph paper and stored in thermally sealed plastic bags. The XAD-2 resins were stored in polyethylene containers. Following sampling, all traps were immediately placed in a freezer and stored at 30 C until analysis. Meteorological measurements including temperature, pressure, relative humidity, wind speed and direction plus the total suspended matter (TSP) were assured by the local Atmospheric Pollution Monitoring Association (ASPA). The TSP was collected for particles with an aerodynamic diameter (Dp) ranging from 0.1 mm to 10 mm. As a result, particles with an aerodynamic diameter greater than 10 mm were not sampled. Moreover, due to experimental limitations, particles with an aerodynamic diameter less than 0.1 mm, i.e. fine particles, were not trapped. Atmospheric temperature, pressure, relative humidity and TSP for particles with an aerodynamic diameter in the range from 0.1 mm to 10 mm were measured and recorded at the DRIRE site located in the city center. However, to better characterize dominant wind conditions, the wind speed and direction were measured by the national meteorological agency, Meteo France, at the city of Reichstett located in the greater Strasbourg area. Prior to sampling, the glass fiber filters and the XAD-2 resins were Soxhlet preextracted with n-hexane/CH2Cl2 (85/15) for 24 h (Albanis et al., 1986), dried at 60 C and then stored in thermally sealed plastic bags. Blanks were used to verify the efficiency of the cleaning and conservation procedures. Rain samples were collected on a weekly basis. An automatic wet-only rainwater collector was used to collect the rain while an open collector was used to directly measure the precipitation levels (Sanusi et al., 2000). After each sampling collection, the funnel and the collection bottle were cleaned and rinsed thoroughly with Milli-Q water. Samples were stored in the dark in pre-cleaned glass bottles at 4 C prior to analysis (Sanusi et al., 2000; Sauret-Szczepanski et al., 2006). 2.3. Sample extraction Similar to the situation listed above, the loaded XAD-2 resin and the glass fiber filters were separately Soxhlet-extracted for 12 h using 250 mL of the n-hexane/ CH2Cl2 (85/15) mix (Atlas and Schauffler, 1990; Sauret, 2002) All samples were extracted within 24 h following sampling to reduce any pesticide degradation, evaporation, contamination and reaction among others. The resulting extraction solutions were then concentrated to a volume of approximately 1 mL in a rotary evaporator at 30 C at a reduced pressure of 280 mbar. Reduced system pressure was used to limit pesticide degradation (Sauret, 2002). Rain samples were extracted using the Solid Phase Micro Extraction method. The sampling and extraction details for rain samples have recently been published and will not be discussed further (Sauret-Szczepanski et al., 2006). Only rain results pertaining to this work are discussed below.
2.1. Chemicals 2.4. Sample analysis The solvents used in this work with stated minimum purities were n-hexane (HPLC Grade, Carlo Erba) and methylene dichloride (HPLC Grade, SDS, Peypin, France). The AmberliteÔ XAD-2 resin (Rohm and Hass) and glass fiber filters (Whatman, GF/A) were obtained from Prolabo and used without any further modification. All reagents used in this work were obtained from the U.I.P.P. (Union des Industries de la Protection des Plantes, France) member companies and were 98– 99% pure: Atrazine, Terbuthylazine and Metolachlor (Syngenta, France); Alachlor (Monsanto, France); Diflufenican, Isoproturon, Iprodione and Fenoxaprop-p-ethyl (Aventis CropScience, France); DIA, DEA and DET (Promochem). Stock solutions
The analysis approach used in this work is similar to the one used in previous studies for the determination of atmospheric pesticides (Sauret-Szczepanski et al., 2006; Sauret et al., 2000). Only details relevant to this study are presented here. All extracted samples were analyzed using the Varian Star 3400 CX gas chromatograph (GC) equipped with a split/splitless injector and coupled to a Saturn 4D Varian ion trap mass spectrometer. Both, the GC and the ion trap were controlled using the Saturn IV acquisition software. The ion trap mass spectrometer was operated in the MS/MS mode. The gas chromatograph was fitted with a 30 m 0.32 mm i.d. J&W
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312 Scientific fused-silica DB-5-MS capillary column with a film thickness of 0.25 mm. The carrier gas was helium and the inlet pressure was 12 psi that corresponded to a flow rate of about 1 mL min1. Initially, the GC oven temperature was held at 60 C for 2 min. After the 2 min delay, the oven temperature was increased from 60 C to 145 C at 15 C min1 and then maintained at that temperature (T ¼ 145 C) for 3 min. After 3 min, the oven temperature was increased to 151 C at 1 C min1. Then, once the oven temperature reached 151 C, the oven temperature was immediately increased to 250 C at 11 C min1 and maintained at this temperature for 10 min. The injector, transfer line and the manifold temperatures were kept at 290, 280 and 200 C, respectively. Linearity and sensitivity of the ion trap detector were optimized by adjusting various parameters such as filament emission current, voltage of the electron multiplier and the total number of ions in the trap (target). These adjustments and the improvement induced by the MS/MS mode compared to the simple-MS mode were discussed in detail previously (Sauret et al., 2000). Even though the electron impact (EI) ionization induced a much higher molecular fragmentation (relative intensity of the parent ion very small or nonexistent) than chemical ionization (CI), the CI signal resulted in a much higher background noise than EI. As a result, the mass spectrometer was operated in the EI ionization mode where the ionization energy was 70 eV. Then, in the MS/MS mode, the daughter ion spectrum was obtained by isolating the most intense fragment ion (created in the simple-MS mode) followed by acceleration and collision with the helium carrier gas molecules (collision induced dissociation, CID). The daughter ions generated by CID were then used for identification and quantification. The details of the MS/MS identification/quantification procedure are described elsewhere (Sauret et al., 2000). The MS/MS parameters used for the pesticides analysis under study are summarized in Table 2 (Sauret-Szczepanski et al., 2006; Sauret, 2002; Sauret et al., 2000). All standard pesticide solutions were prepared in hexane. The quantification limits (QL) for the selected pesticides were obtained assuming that the sampling volume was 700 cm3 and that, after extraction, samples were concentrated to 1 mL. The quantification limits were (in ng m3): 0.08 for Cymoxanil and Isoproturon, 0.04 for DEA, Iprodione and Fenoxaprop-p-ethyl, and 0.016 for DET, Atrazine, Terbuthylazine, Alachlor, Metolachlor and Diflufenican. 2.5. Model description To better evaluate the atmospheric pesticide deposition, a modified version of the original one-dimensional cloud water deposition model first developed by Lovett (Lovett, 1984; Lovett et al., 1982) and then described in detail by Pahl et al. (1994) was used. Briefly, the original cloud deposition model is a one-dimensional resistance model that divides the vegetation into 1 m layers (Lovett et al., 1982). The flux of cloud water is calculated for each layer using an analogue of the Ohm’s law. F ¼
DC R
¼ DC vd
(1)
In Eq. (1) above F is the cloud water flux, DC is the cloud droplet concentration gradient, R is the resistance acting against the droplet deposition in each layer and vd is the deposition flux. The resistance, R, is the sum of the aerodynamic resistance and boundary-layer resistance of individual canopy components, such as needles, branches and twigs. The resistance values are calculated for each layer using characteristic vegetation profiles. By definition, R is a resistance to the droplet transfer; therefore, R is inversely proportional to the deposition flux and vice versa. Since its initial application, the one-dimensional cloud water deposition model has been modified several times to better adapt new sampling parameters and experimental objectives (Hercke`s et al., 2002; Pahl et al., 1994; Wobrock et al., 1994). One of the major assumptions of the modified model used in this work is that the attachment (deposition) is definitive and irreversible with no evaporation. Moreover, while the original cloud deposition model measures wet deposition of liquid droplets with a 2–50 mm aerodynamic diameter (Pahl et al.,1994), the modified version used in Table 2 The MS/MS parameters used in this work (Sauret-Szczepanski et al., 2006; Sauret, 2002) Pesticide
DEA DET Atrazine Terbuthylazine Alachlor Isoproturon Metolachlor Diflufenican Iprodione Fenoxaprop-p-ethyl Cymoxanil
Base peak ion used as parent ion (m/z)
CID voltage (V)
172 186 200 214 188 191 162 266 314 361 177
49 49 55 56 45 40 45 81 69 75 42
Daughter ions used for quantification (m/z) 79, 104 104, 145 94, 104, 174 104, 173 132, 160 146 117, 120, 134 218, 238, 246 245, 271 244, 261, 288 119, 135, 149
305
this work measures deposition of particles (solid) with the aerodynamic diameter in the range from 0.1 to 10 mm with a 0.1 mm resolution, i.e. 100 classes of particles. As a result, the sedimentation rate had to be modified to account for the particle aerodynamic diameter difference. Other parameters that were modified include distribution of particles using proper aerosol distribution.(Hercke`s, 1999; Sauret, 2002) The original model version developed by Pahl et al. (1994) used hourly meteorological data provided by stations at different altitudes. Here, in a similar situation to the work of Hercke`s (Hercke`s, 1999; Hercke`s et al., 2002), the meteorological data was measured continuously and directly incorporated into the model. The most important model modification carried out on Pahl’s model and adapted it to the particle deposition model used in this work is that the Total Suspended Particles (TSP, g cm3) was substituted for the Liquid Water Content (LWC, g cm3). This data was measured on an hourly basis and the average value was incorporated into the model. 2.6. Wind speed The logarithmic wind speed profile above the canopy can be expressed using Eq. (2) (Thom, 1975). The wind speed profile (Eq. (2)) applies only in the neutral surface boundary-layer where the atmospheric stability is ignored. uðzÞ ¼
u* z d ln z0 K
ðz > forest height; HÞ
(2)
where u(z) is the wind speed at the measurement height z ¼ 10 m (this work), u* is the friction velocity, K is the von Karman constant (K ¼ 0.41), d is the zero plane displacement height in meters and z0 is the roughness height. Using the wind speed calculated using Eq. (2) above, the wind speed u1 over the sampling site is determined (Eq. (3)). u1 ðzÞ ¼
u*1 z ln z0G K
(3)
In Eq. (3) above, z0G ¼ 0.02 m (this work) and is the lawn grass roughness height and u*1 is the friction velocity. For the roughness lengths of less than 0.02 m, d is negligible and is ignored (Pahl, 1996). The friction velocity, u*1, is calculated for the measurement height zM ¼ 10 m used in this work using following Eq. (4). K M lnzz0G
u*1 ¼ u1 ðzM Þ
(4)
To transpose the wind profile u1 to the wind profile over the forest (u2), the profile u1 undergoes a vertical translation equivalent to 2/3 of the forest height corresponding to the difference between the zero plane displacement height (d) at the wind speed measurement site and in the forest. The friction velocity u2* is then directly proportional to u*1 (Eq. (5)). u1 ð2HÞ ¼ u2 ð2H þ dÞ
(5)
In Eq. (5) above, u(H) is the wind speed at the tree top height H. The wind speed at the tree top level can be calculated using the following Eq. (6), uðHÞ ¼
u*2 H d ln zoF K
(6)
where zoF is the forest roughness height (in Europe, typically zoF ¼ 1–3 m) and is equivalent to 2/3 of the forest height (Brutsaert, 1975; Flemming, 1982; Pahl, 1996). If z < H, then the wind profile can be expressed using the following relation (Eq. (7)). uðzÞ ¼ uðHÞexp½bu CSAIðzÞ
(7)
In Eq. (7) above bu is the experimentally determined wind reduction coefficient and CSAI(z) is the Cumulative Surface Area Index. Based on the experimental work of Lovett and coworkers bu ¼ 0.27 (Lovett et al., 1982). The CSAI depends on the tree height and is calculated as a function of leaf/needle surface area and the planar canopy surface area (horizontal surface projection). The expression for the vertical turbulent coefficient profile that directly determines the aerodynamic resistance within the canopy is written in the similar form as u(z) Kd ðzÞ ¼ Kd ðHÞexp½bk CSAIðzÞ
ðz < HÞ
(8)
where Kd(z) (m2 s1) is the turbulent diffusion coefficient at the height z, and bk is the turbulent reduction coefficient. According to Lovett and Pahl, bk ¼ 0.14 (Lovett et al., 1982; Pahl, 1996). The Kd(H) is the forest limit coefficient and is calculated using the following Eq. (9): Kd ðHÞ ¼ Ku* ðH dÞ
(9)
As a result, the forest aerodynamic resistance Ra (s m1) at level i becomes
Ra ðiÞ ¼ zðiÞ
1 1 þ Kd ðiÞ Kd ði 1Þ
(10)
where z(i) (m) is the forest level height and Kd is the turbulent diffusion coefficient at the level i.
306
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
2.7. Vegetation The Strasbourg sampling site forest is a closed spruce stand (Picea abies). The forest vegetation is classified in terms of its Surface Area Index (SAI) which corresponds to the interface between the ‘free’ atmosphere and trees. The SAI represents the total impact surface for aerosols (droplets or particles) and is the algebraic sum of the total tree surface, i.e. needles, branches, twigs among others, normalized to the ground surface area. The sampling site forest SAI (m2 m2) is calculated using the following Eq. (11) (Hercke`s et al., 2002; Pahl, 1996): (11)
SAI ¼ 91:8 þ 14:3 d1:3
where d1.3 (cm) is the tree trunk diameter 1.3 m above the ground. Sampling site forest characteristics are shown in Table 3 (Asae¨l, 1990; Sauret, 2002). 2.8. Granulometric particle distribution As stated above, the cloud deposition model was originally used for droplets. In order to modify the model from droplet deposition to particle deposition several parameters had to be changed. For one, it is important to estimate the total mass contribution as a function of the particle size to the total suspended particles (TSP). The sedimentation and impact processes will be directly depended on the particle aerodynamic diameter. As a result, in order to better classify the total atmospheric distribution spectrum (Seinfeld and Pandis, 1998) particle class with an aerodynamic diameter in the range from 0.1 to 10 mm with a 0.1 mm resolution, i.e. 100 classes of particles, was used. Further, it was not possible to experimentally evaluate the granulometric size particle distribution in this work. Given the geographical location and size of the sampling site (Strasbourg observatory park) standard rural particle size distribution was used (Seinfeld and Pandis, 1998). The atmospheric aerosol size distribution can be described using the following Eq. (12) (Seinfeld and Pandis, 1998): n X N pffiffiffiffiffiffi i exp n0N log Dp ¼ i ¼ 1 2plog si
2 ! log Dp log Dpi 2log2 si
(12)
where n0N ðlog Dp Þ (cm3) is the number of particles as a function of logarithmic diameter Dp (mm), Ni (cm3) is the average number of particles, Dpi is the average particle diameter and si is the standard deviation. Then, surface volume distributions can be calculated using Eqs. (13) and (14), respectively. n0S log Dp ¼ pD2p n0N log Dp mm2 cm3
(13)
p n0V log Dp ¼ D3p n0N log Dp mm3 cm3 6
(14)
Fig. 1. Rural particle size distribution in volume.
2.9. Sedimentation velocity Once the particle distribution is known (see above), the sedimentation (gravitational) velocity is calculated using the following Eq. (17):
vs ¼
2 1 Dp rp gCc m 18
Dp in m
(17)
In the above Eq. (17) g is the gravitational constant (9.806 m s2), rp is the volume mass of a single particle (1.7 106 g m3), m is the air viscosity (1.72& 102 g m1 s1 at T ¼ 20 C) and Cc is the Cunningham factor (sliding factor) and is dependent on the particle diameter (Seinfeld and Pandis, 1998). For large particles the sliding factor can be considered negligible. However, for small particles the sedimentation velocity can be multiplied by values ranging from Cc ¼ 2.85 for Dp ¼ 0.1 mm to Cc ¼ 216 for Dp ¼ 0.001 mm (Seinfeld and Pandis, 1998). Correction sliding factors for spherical particles in air at T ¼ 298 K and p ¼ 1 atm are given in Table 4. 2.10. Resistance model
Modification of the particle distribution did not show to influence the total particle quantities deposited. Actually, rural particle deposition has been shown to be slightly greater than urban particle deposition. Typically, ‘bigger’ particles are more abundant in rural regions (Dp > 1 mm) compared to urban regions. In urban regions smaller diameter particles are more abundant (Dp < 1 mm). Using the rural particle size distribution slightly overestimates the total particle deposition (higher deposition flux), a result that is in line with the research objectives of this study to estimate the maximum particle deposition. Rural volume aerosol distribution used in this work is shown in Fig. 1. Once the particle size distribution is known, the total standard particle volume, VT, can be calculated (Eq. (15)). VT ¼
p 6
Z
þN N
D3p n0N log Dp dlog Dp
(15)
Typical Total Suspended Particles, TSP , are then calculated by multiplying VT by aerosol density (rp) (g mm3) (Eq. (16)).
TSP ¼ rp VT g cm
3
(16)
Table 3 The Strasbourg forest sampling site characteristics Parameter
Average characteristic
Tree type Age Height Density d1,3a SAIb
Picea abies 80–100 yr 25.8 m 525 branches ha1 35.8 cm 22 m2 m2
a b
Rij ¼
1 CEij uðiÞ
(18)
Then, the surface resistance at level i (Ri) can be written as the inverse of the sum of both the vertical surface repartition of different vegetation components normalized to the sum of resistances Rij in parallel (Eq. (19)). Ri ¼ P 7
1
(19)
SAIij j ¼ 1 Rij
0
0
To calculate surface resistance, Rij (s m1), i.e. transport resistance of vegetation particle component j at the tree (forest) level i, the wind speed (u(i)) and capture coefficient of vegetation particle component j (CEij) at the height i has to be known (Eq. (18)).
Tree trunk diameter 1.3 m above the ground. Calculated using Eq. (11).
The independent collection efficiency of particle deposition on vegetation can be written as the sum of the three different contributions (Slinn, 1982), CE ¼ RðEB þ EIN þ EIM Þ
(20)
where R is the particle rebound factor, EB is the collection efficiency resulting from Brownian diffusion, EIN is the collection efficiency resulting from interception by vegetation and EIM is the collection efficiency resulting from direct impact of particles on vegetation. Since the Brownian diffusion collection efficiency is valid only for particles with the aerodynamic diameter less than 0.1 mm (Dp < 0.1 mm) it was ignored. Moreover, to better determine the maximum particle deposition flux
Table 4 Correction sliding factors for spherical particles in air at T ¼ 298 K and p ¼ 1 atm Particle diameter, Dp (mm)
Cc
0.1 0.2–0.5 0.6–1 1.1–10
2.85 1.5955 1.245 1.05
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312 R is set equal to 1 (no particle rebound). Thus, Eq. (20) can be rewritten to give Eq. (21). CE ¼ EIN þ EIM
(21)
The EIN is defined as follows (Slinn, 1982): EIN ¼
Dp Dp cv þ ð1 FÞ F c d LS þ D p LL þ Dp
(22)
where cv =cd is the ratio of the viscous drag to the total drag (usually cv =cd ¼ 1=3), LS is the characteristic small vegetation collection diameter (LS < 1 mm), LL is the characteristic large vegetation collection diameter (LL > 1 mm) and F is the interception fraction by small collectors (usually about 1%) (Ruijgrok et al., 1997; Slinn, 1982). In this work, small collectors with diameter less than 1 mm were ignored; therefore, F ¼ 0. Then, Eq. (22) is reduced to give Eq. (23). EIN ¼
Dp cv cd LL þ Dp
(23)
To evaluate particle deposition on smooth surfaces, the following semi-empirical relation can be used (Chamberlain, 1967): EIM ¼ 103=St
(24)
where St is the Stokes number (Eq. (25)). St ¼
vs uðzÞ gð2L
(25)
In Eq. (25) above, vs is the particle sedimentation velocity, u(z) is the wind velocity at the height z and L is the impact surface diameter. A different evaluation of EIM was proposed by Slinn (1982) where the strong dependence on the Stokes number increases the impact rate (Eq. (26)). EIM ¼
ðStÞ2 1 þ ðStÞ2
(26)
If the deposition fluxes are calculated using Eqs. (24) and (26) as a function of the measured wind speed u(zm) and fixing the TSP to the average value of 0.03 g m3, completely different particle deposition profiles are obtained (Fig. 2). As shown in Fig. 2, deposition flux calculations using the Slinn relation demonstrate that the deposition by impact becomes more important than sedimentation at wind speeds 2 m s1 and higher. On the other hand, deposition flux calculations using the Chamberlain relation show that impact deposition becomes predominant only at wind speeds 8 m s1 and higher. Wind speeds of 8 m s1 were rarely encountered under the experimental conditions of this work. Typical wind speeds ranged from 1 to 4 m s1. As a result, similar to the situation listed above (rebound factor), the Slinn relation (Eq. (26)) was used to obtain the maximum deposition flux for pesticides under study. To calculate CEij for each type of vegetation j, the following relation was used (Pahl et al., 1994): CEij ¼
CE Cj
307
months for triazines, chloroacetanilides and fenoxaprop-p-ethyl. Observed concentration peaks during the spring months corresponded to the spraying periods of these pre-emergent herbicides on maize and soybeans. We observed the same seasonal pattern for triazines metabolites (DEA and DET) which appeared a few days after their parent compounds. Nevertheless, the reactions producing the above degradation products could occur both in surface water and in the atmosphere and with the current data we have no way to estimate the relative importance of these two modes of formation. Moreover, because the same metabolites are produced in the two media, we cannot use the present results to demonstrate the efficiency of the atmospheric reactions on the degradation of atmospheric pesticides. Nevertheless, because the metabolite concentrations in precipitation were in the same order of magnitude as the parent compounds, these molecules have to be considered in the determination of the pesticide contamination assessment. Also, high concentrations of diflufenican and isoproturon in precipitation were observed during winter months according to their spraying periods. Once the pesticide concentration in precipitation is known, the wet deposition flux is obtained using the following Eq. (28):
Di ¼ H C i i
(28) 2
where D (mg m ) is a deposition of compound i, H is the precipitation height in mm (1 mm ¼ 1 L m2) and Ci (mg L1) is the pesticide concentration in precipitation. The precipitation was collected using the permanently open rainwater collector. The precipitation height was obtained from Meteo France, a national meteorological agency. Obtained wet deposition fluxes are presented in Table 6.
(27)
where Cj is the experimentally determined correction factor (typically Cj values varied from 3 to 8).
3. Results and discussions 3.1. Pesticide wet deposition Pesticide concentrations in precipitation for the 11 molecules under study are shown in Fig. 3. As shown in Fig. 3, important seasonal pesticide concentration variation is observed with the maximum values occurring during spring when the agricultural activities are greatest. The concentrations of the 11 compounds under study were measured in the rain events occurring in Strasbourg between August 3, 2000 and August 6, 2001. Table 5 presents the concentration characteristics of the target molecules in rain samples. For technical reasons, cymoxanil was measured from January 2001 through August 2001, i.e. 7 months. The detection percentage calculation was based upon the 29 samples collected during the sampling campaign. Rain concentrations showed patterns of high concentrations of the selected pesticides in precipitation during spring months (Fig. 3). These maximal concentrations of the selected pesticides in precipitation decreased rapidly down to a level of less than 0.1 mg L1 in the following
Fig. 2. Sensitivity of the deposition flux as a function of the wind speed (TSP of 0.03 g m3).
308
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
Fig. 3. Concentrations of the target compounds (mg L1) in rain samples during the field campaign.
3.2. Model sensitivity The particle deposition model properties are affected by many physical parameters including (as discussed above) wind speed and TSP. The sensitivity of the model’s behavior to changes in parameters is known as parametric sensitivity. When a system operates in a parametrically sensitive region, its
performance becomes unreliable and changes sharply with small variations in parameters. Thus, it would be of great interest to predict sensitivity behavior in the one-dimensional particle deposition model used in this work. As a result, the influence of the (1) TSP, (2) ground surface roughness, (3) forest roughness, and (4) measured wind speed on the deposition flux is evaluated.
N. Sauret et al. / Environmental Pollution 157 (2009) 303–312
309
Table 5 Concentration characteristics of the target molecules in rain samples from 08/03/2000 to 08/06/2001 Compound
Wet deposition flux (g ha1 yr1)
Minimum (mg L1)
Maximum (mg L1)
Observed maximum pesticide concentration date (dd.mm.yy–dd.mm.yy)
Pesticide application period
DEA DET Atrazine Terbuthylazine Alachlor Metolachlor Diflufenicana Fenoxaprop-p-ethyl Iprodione Isoproturona Cymoxanilb
0.44 0.56 1.54 1.11 1.9 1.39 1.87 0.39 0.11 0.22 14.5
