Environ. Sci. Technol. 2005, 39, 5219-5226
Solar Radiation, Relative Humidity, and Soil Water Effects on Metolachlor Volatilization J O H N H . P R U E G E R , * ,† T I M O T H Y J . G I S H , ‡ LAURA L. MCCONNELL,§ LYNN G. MCKEE,‡ JERRY L. HATFIELD,† AND WILLIAM P. KUSTAS‡ USDA-ARS National Soil Tilth Laboratory, Ames, Iowa 50010, USDA-ARS Hydrology and Remote Sensing Laboratory, Beltsville, Maryland 20705, and USDA-ARS Environmental Quality Laboratory, Beltsville, Maryland 20705
Pesticide volatilization is a significant loss pathway that may have unintended consequences in nontarget environments. Field-scale pesticide volatilization involves the interaction of a number of complex variables. There is a need to acquire pesticide volatilization fluxes from a location where several of these variables can be held constant. Accordingly, soil properties, tillage practices, surface residue management, and pesticide formulations were held constant while fundamental information regarding metolachlor volatilization (a pre-emergent pesticide) was monitored over a five-year period as influenced by meteorological variables and soil water content. Metolachlor vapor concentrations were measured continuously for 120 h after each application using polyurethane foam plugs in a logarithmic profile above the soil surface. A flux gradient technique was used to compute volatilization fluxes from metolachlor concentration profiles and turbulent fluxes of heat and water vapor (as determined from eddy covariance measurements). Differences in meteorological conditions and surface soil water contents resulted in variability of the volatilization losses over the years studied. The peak volatilization losses for each year occurred during the first 24 h after application with a maximum flux rate in 2001 (1500 ng m-2 s-1) associated with wet surface soil conditions combined with warm temperatures. The cumulative volatilization losses for the 120-hour period following metolachlor application varied over the years from 5 to 25% of the applied active ingredient, with approximately 87% of the losses occurring during the first 72 h. In all of the years studied, volatilization occurred diurnally and accounted for between 43 and 86% during the day and 14 and 57% during the night of the total measured loss. The results suggest that metolachlor volatilization is influenced by multiple factors involving meteorological, surface soil, and chemical factors.
* Corresponding author phone: (515) 294-7694; fax: (515) 2948125; email:
[email protected]. † National Soil Tilth Laboratory. ‡ Hydrology and Remote Sensing Laboratory. § Environmental Quality Laboratory. 10.1021/es048341q Not subject to U.S. Copyright. Publ. 2005 Am. Chem. Soc. Published on Web 06/21/2005
Introduction Pesticides continue to be an essential component of modern agricultural production systems. Weed control agents, or herbicides, are routinely applied to major crops such as wheat, corn, and soybeans, resulting in millions of kilograms used each year. A large body of research exists in the area of surface runoff and subsurface leaching of herbicides and other pesticides from production fields; however, measurements of volatile losses are rarely included in studies of the environmental fate of herbicides. The atmospheric transport and deposition of pesticides can be limited to short distance spray drift of pesticide formulation droplets (1-4) or can extend to wind erosion of pesticide adsorbed soil particulates and volatilization from soil and plant surfaces into the boundary layer of the atmosphere. Once in the boundary layer, pesticide residues may be transported and deposited to nontarget regions such as lakes and rivers (5-9) where they may create toxic conditions for sensitive species of plants or animals (10-13). During the 1970s, a number of groundbreaking studies revealed that volatilization was a major dissipation route for chlorinated insecticides such as dieldrin, heptachlor, lindane, and chlordane (14-19). Cumulative losses from volatilization for highly volatile compounds such as trifluralin and lindane approached 90% within a week after application (20). Glotfelty et al. (21) showed results from field experiments that pesticides applied to the surface of fallow soil initially volatilize at rates proportional to the vapor density of the pure chemical. Herbicides are generally much less volatile and more water soluble relative to the early chlorinated insecticides and are also more susceptible to microbial degradation in soil. Atrazine was reported to volatilize just 2% of the applied amount after a period of over 3 weeks (20). In another study, atrazine and metolachlor losses from bare soil were just 3.6 and 6.5%, respectively, after 21 days (22). However, Prueger et al. (23) reported variation in the cumulative metolachlor losses as a function of the herbicide application method. The cumulative metolachlor losses from the banded and broadcast applications were found to be 9 and 21%, respectively, of the applied mass over approximately 7 days after application. Unlike insecticides, which are generally used under conditions of increased pest pressure, herbicides are used at planting for seasonal weed control. Planting of major crops such as corn and soybeans in agricultural regions typically occurs at approximately the same period (May-June) each year, resulting in a large number of applications occurring within a short period of time. This mass application period may contribute to a very large nonpoint pulse of herbicides entering the atmosphere, creating conditions for larger-thanexpected wet or dry deposition events in the surrounding region. This springtime regional signal of herbicide volatilization and wet deposition has been observed in recent studies by Kuang et al. (6) on the Delmarva Peninsula of the Chesapeake Bay, by Goolsby et al. (24) and Majewski et al. (25) in the Midwestern U.S., and by Nations and Hallberg (26) and Hatfield et al. (27) in Northeastern Iowa. Although the fundamental physical properties of a pesticide can be used to estimate the volatilization potential of a particular chemical (28, 29), most field-scale studies are limited to a single season under a particular set of meteorological conditions. More information is needed to understand and predict the volatilization behavior of these more polar widely used pesticides under varying meteorological conditions. VOL. 39, NO. 14, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5219
TABLE 1. Physical and Chemical Properties of Metolachlor propertiesa
structure chemical name molecular weight physical state boiling point solubility vapor pressure Henry’s law constant
2-chloro-6′-methyl-N-(1-methoxy-2-methoxyethyl)acetanilide 283.8 g/mol liquid 100 °C 530 mg/L at 20 °C 1.7 mPa at 20 °C 3.21 × 10-6 dimensionless at 20 °Cb
a All of the property values are from Montgomery (30), except Henry’s law constant cited from Rice et al. (31). b The dimensionless value can be converted to units of Pa m3/mol by multiplying by RT, where R is the universal gas constant, 8.3143 Pa m3/mol K, and T ) 298 K.
New management strategies and pesticide application guidelines are needed to limit the volatile losses of herbicides to the atmosphere. This type of tool development requires adequate pesticide flux measurement data acquired in a variety of environmental conditions followed by careful examination of the factors or combination of factors that significantly enhance volatile losses. This project satisfies the initial stage of this work, in which metolachlor has been used as a model herbicide for a multiyear study of volatile pesticide emissions from a continuous corn production site. Metolachlor, a pre-emergence herbicide, is an important weed control agent for corn and soybean production and has been used extensively by producers in the Midwest in recent years (Table 1). This paper presents metolachlor flux measurements from the same location over five years from 1998 to 2002 using high-resolution air concentration measurements from five heights above the field combined with corresponding data from standard meteorological and eddy covariance measurements. This data is used to evaluate ambient meteorological factors and initial soil conditions that may influence volatilization.
Experimental Section Site Description and Pesticide Application. The research site is a 21-ha agricultural production field located at the USDA Henry A. Wallace Beltsville Agricultural Research Center, in Beltsville, Maryland (39° 01′ 00′′ N, 76° 52′ 00′′ W). The soils are variable, with the majority being typic hapludults, coarse-loamy, siliceous, and mesic. The soil series with the proportional extent of the respective mapping units are: Downer-Muirkirk-Matawan sandy loam, 49%; Bourne fine sandy loam, 23%; Matawan-Hammonton loamy sand, 23%; and Downer-Ingleside loamy sand, 5%. The organic matter content in these soils is less than 3%. The method of pesticide application was spray broadcast onto bare soil surface shortly after the corn was planted. In 1998, the application rate was 2.24 kg/ha, and in the remaining years, the application rate was 1.51 kg/ha. This change was due to the introduction by Syngenta Crop Protection (Guelph, Ontario, Canada) of the more herbicidally active S-Metolachlor enriched formulations, Dual II Magnum and Bicep II Magnum, which reduced the manufacturer’s label application rate by approximately 30% (32). Micrometeorological Instrumentation. Micrometeorological and eddy covariance instrumentation were mounted on a 10-m tower near the center of application area to measure the surface energy balance components of net radiation (Rn), soil heat flux (G), sensible heat flux (H), and latent heat flux (λE) densities (W m-2). The net radiation and soil heat flux were measured with a Kipp & Zonen, Inc. (U.S.A.) CNR-1 net radiometer and 3 Radiation Energy Balance Systems, Inc. (U.S.A.) HFT-1 soil heat flux plates. The CNR-1 was placed 4 m above the surface, and the soil heat flux 5220
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 14, 2005
plates were buried 0.08 m below the soil surface. Above each soil heat flux plate at 0.02 and 0.06 m depth were two Type-T (copper-constantan) soil thermocouples. The soil temperature data were used to compute the storage component of the soil heat energy above the flux plates. A Campbell Scientific, Inc. (U.S.A.) 3D sonic anemometer and a LICOR, Inc. (U.S.A.) LI7500 infrared hygrometer were used to measure H and λE flux densities as the covariance of the vertical wind velocity (w) with air temperature (Ta) and water vapor density (q). Friction velocity (u*) and the standard deviation of vertical velocity (σw) were computed from the sonic anemometer. Standard local surface meteorological instrumentation were also mounted on the tower to measure mean wind speed, direction, Ta, relative humidity (RH), solar radiation (Rs), precipitation, and surface radiometric temperatures (surface temperature). The sampling frequency was 20 Hz for the eddy covariance and 10 s for the energy balance and meteorological instrumentation. All of the data were stored as 30-min averages on Campbell 23x and 21 data loggers. Soil Water and Soil Pesticide Measurements. Because soil water content in the surface layer is difficult to quantify (because no instrument exists which can monitor soil water content in the top few millimeters), shallow 2-cm samples were collected during the early morning when this thin layer of surface soil should be in near-equilibrium with the shallow subsurface (33). The surface soil water content samples were acquired each sampling day before sunrise. During the course of this study, we evaluated over 600 surface soil measurements. The surface soil samples of 38.5 cm2 were collected and sealed in clean aluminum cans from at least 20 randomly selected locations extending a maximum of 100 m radially around the air sampling tower. The soils were weighed before and after drying at 100 °C to determine the percent water content by mass. Because direct surface soil moisture data is not available for 1998, the values were estimated from a capacitance probe (EnviroSCAN, Sentek Pty Ltd., South Australia) with a sensor at a depth of 10 cm that was located within 10 m of the air sampling tower. On the day of application for each year, separate surface soil samples were collected for determining the metolachlor concentration of the soil surface (20 samples, in the same random radial pattern as the soil water) within 30 min of the time of application. The metolachlor recovery from these samples was greater than 97%. Air Sampling. The pesticide sampling masts consisted of a 2-m length by 0.0254-m i.d. galvanized steel pipe erected near the micrometeorological instruments. The sampling canisters were made of glass tubes 0.0254-m i.d. by 0.15-m length. The glass tubes were tapered at one end to a stem of 0.0085-m diameter and were connected with Tygon tubing to a Staplex, Inc. (U.S.A.) high volume air sampler Model TFIA that employs a universal rotary motor with a variable orifice flow meter calibrated to a flow rate of approximately
TABLE 2. Pesticide Formulations Used and Sample Measurement Period for 1998-2002 1998
1999
2000
2001
2002
pesticide formulation
Dual II 8E
metolachlor application rate (kg/ha) application date and start time
2.24 6:15 a.m. June 10
Dual II Magnum Bicep II Magnum 1.51 5:30 a.m. June 1
Dual II Magnum Bicep II Magnum 1.51 5:30 a.m. June 13
Dual II Magnum Bicep II Magnum 1.51 5:20 a.m. June 20
Dual II Magnum Bicep II Magnum 1.51 6:00 a.m. April 24
50 L min-1 through each sampling canister. The individual canisters were wrapped with aluminum reflective tape to prevent photodegradation of the PUF plugs. Five canisters were mounted on to the sampling mast in a logarithmic configuration 0.15, 0.30, 0.60, 1.2, and 1.95 m above the soil surface in the same location for each of the five years. Pesticide Sampling Procedure. The timing of planting and herbicide application varied over the years as a function of local precipitation patterns and technical/logistical problems typically encountered with any planting operation (Table 2). The actual pesticide sampling for each year began approximately 1 h after the pesticide was applied. Although some minor differences in sample collection frequencies between the years exist, in general, the air in the dynamic surface layer above the treated field was sampled continuously every 1-2 h during the first 56 h after application and then extended to 4-h sampling intervals until the end of the study. Sampling between midnight and 6 am was generally extended to 6-h sampling intervals. Each sampling canister contained two polyurethane foam (PUF) plugs with dimensions of 0.0254 m in diameter by 0.075 m in length. The first PUF plug served as the primary pesticide vapor trap, and the second plug served as the breakthrough trap. The airflow rates through the PUF canisters at each height were measured and recorded at the beginning and end of each sampling interval. After each sampling period, the PUF plugs were placed in glass containers, secured with Teflon-lined lids, and stored in a freezer at a temperature of -20 °C. Polyurethane Foam Plug Extraction and Analysis. Prior to sampling, PUF plugs were precleaned using separate methanol and ethyl acetate washes and allowed to air dry in a fume hood. After precleaning, 25 PUF plugs were selected randomly and analyzed as blanks. No interfering peaks were observed above our detection limits. The PUF plug samples were individually extracted with ethyl acetate for 4 h using a Soxhlet technique. The blank and spike recovery controls were included in the sample extraction batches to determine the extraction efficiency (93 ( 11, n ) 23) and to detect contamination from laboratory procedures (all blanks were free of interfering peaks). The extracts were then reduced to 5-10 mL using a rotary evaporator. The extracts were further reduced to a volume of 1 mL using a gentle stream of nitrogen gas with a 5-µg spike of trifluralin added to each sample as an internal standard. The samples were analyzed on a Hewlett-Packard 5890 Series II GC equipped with a nitrogen phosphorus detector. The method quantification limits for metolachlor were 10 ng, well above the baseline noise level of 2 ng m-3 for ambient air concentrations. Pesticide Flux Calculations. The transport of mass, energy, and scalars from a surface to the boundary layer of the atmosphere can be estimated by using a first-order closure (K-theory) approach. This approach has been used routinely to estimate the fluxes of heat, water, and momentum over land and vegetated surfaces. Assuming similarity of transport mechanisms, this approach has been extended to trace gases such as carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), and pesticides (34, 35).
In this study, the flux-gradient theory was employed using the aerodynamic profile. The flux-gradient theory is based on the assumption that the turbulent transfer of mass or scalars is analogous to molecular diffusion and can thus be determined as the product of a mean vertical mixing ratio gradient and a turbulent-transfer coefficient (36). Expressed in general form, the aerodynamic profile for any atmospheric entity can be expressed as
F ) Kz
∂χ ∂z
(1)
where F is the vertical flux of the entity in question, Kz the eddy diffusivity coefficient (m2 s-1), χ the mixing ratio and z the height (m) above a surface. Specifically, the flux-gradient theory has been developed to represent the transport of momentum, heat, and water vapor in response to a gradient and eddy diffusivity in the following forms
τ ) FaKm
∂u ∂z
H ) FaCpKh E ) FaKq
(2)
∂u ∂z
(3)
∂q ∂z
(4)
where τ is the surface shear stress (kg m-1 s-2), H is the sensible heat flux (W m-2), E is the water vapor flux (kg m-2 s-1), Fa is the air density (kg m-3), Cp is the specific heat of air (J kg-1 K-1), and Km, Kh, and Kq are eddy diffusivities for momentum, heat and water vapor, respectively (m2 s-1). The mass and scalar gradients for momentum, heat, and specific humidity are expressed as ∂u/∂z, ∂θ/∂z, and ∂q/∂z, respectively. The Ki (where i ) m, h, or q) term is known by various names: eddy viscosity, eddy diffusivity, eddy-transfer coefficient, turbulenttransfer coefficient, and gradient-transfer coefficient (37). The eddy diffusivity is not a fluid property as is molecular diffusivity but rather a flow property that is related to the state of turbulence. Collectively, eqs 1-4 represent first-order closure or “K-theory” as the product of a scalar concentration gradient and a turbulent diffusivity (K). All of the complexities and uncertainties of turbulent transport for momentum, heat, and water vapor are embedded in the K term, thus greatly simplifying a complicated and highly nonlinear process. Using the flux gradient approach for pesticide flux estimates is based on extending the assumption that transport similarity exists for pesticide vapor as it does for scalar and mass properties of sensible heat and water vapor. This is reasonable because only the vapor phase of the chemical is of interest. On this assumption, the general flux-gradient expression (eq 1) can be rewritten easily to express a pesticide flux as
∂c P ) Kc ∂z
(5)
where P is the pesticide flux (ng m-2 s-1), Kc (m2 s-1) is now VOL. 39, NO. 14, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5221
TABLE 3. Meteorological Conditions, Surface Soil Water Content, Maximum Metolachlor Fluxes and Cumulative Total Losses over the 5 days after Each Pesticide Application from 1998 to 2002
1998 1999 2000 2001 2002
air T, daya (°C)
air T, nightb (°C)
total solar radiation (MJ m-2)
soil water contentc (% wt wt-1)
wind speedd (m s-1)
max fluxe (ng m-2 s-1)
cumulative lossf (ng m-2)
19 (12-24) 27 (21-32) 25 (17-33) 24 (20-31) 13 (2-18)
13 (6-20) 21 (10-29) 20 (17-27) 24 (20-29) 13 (4-18)
44 120 63 110 94
3.8 4.0 27 32 22
1.2 1.9 1.6 1.4 2.0
146 240 280 1500 68
10 000 12 000 19 000 32 000 14 000
a Represents average and range of temperatures (T ) from 800 to 2000 hours. b Represents the average and range of temperatures (T ) from a a 2000 to 800 h. c Units are in percent water mass of total soil mass. Average of n ) 100 (20 per day) surface soil samples collected each morning before sunrise. d Average measured wind speed at 3-m height. e The 1998 value has been normalized for differences in application rate. f Total measured flux over 5 days after application. 1998 values have been normalized for differences in application rate.
the eddy diffusivity for a pesticide of interest, and ∂c/∂z is the height dependent pesticide concentration gradient (ng m-3). As expressed in eq 5, the pesticide flux formulation can appear to be deceptively simple. The concentration gradient term in eq 5 is in fact straightforward providing that the sampling rate and trapping media are appropriate for the chemical of interest. Deriving the eddy diffusivity (Kc) for a pesticide compound is accomplished by analogy to diffusivity for momentum, heat, or water vapor. In the atmospheric boundary layer, the eddy diffusivity for momentum (Km) is well understood and can be expressed as
Km )
ku*zm φm
(6)
where k is the von Karman constant (0.4), u* is the friction velocity (m s-1), zm is the mean geometric measurement height (m), and φm is a nondimensional correction to account for the enhanced upward vertical displacement of a parcel of air due to buoyancy effects from surface heating (unstable) or a downward suppressing displacement due to surface cooling (stable) in the surface atmospheric layer. When there is negligible enhancement of surface layer vertical motions (typically associated with dawn and dusk periods), conditions are said to be neutral and φm ) 1. There have been numerous contributions in the literature over the last 40 years that describe functional forms of the φ functions for momentum, heat, and water vapor. For a review, see Dyer and Hicks (38), Dyer (39), and Hogstrom (40). In this study, we have used relationships suggested by Hogstrom (40) to express φm for the unstable case (z/L < 0) as
z φm ) 1 - 19 L
(
-0.25
)
(7)
and for the stable case (z/L > 0) as
z φm ) 1 + 5.3 L
(
)
(8)
where z is measurement height (m) above the surface and L is a scaling length (the Obukov length) expressed as
L)
-u*3Fa H kg + 0.61E T aC p
[( )
]
(9)
where u*, Fa, k, H, and E have been previously defined, g is the acceleration due to gravity (9.81 m s-2), and Ta is air temperature (°C). Having derived the eddy diffusivity for momentum (Km), we can proceed to compute an estimate for the eddy diffusivity for pesticides (Kc) following Flesch et al. (41). This was expressed in the form of a turbulent Schmidt number 5222
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 14, 2005
from a similar metolachlor volatilization study in central Iowa as
Sc )
Km (ku*z/φm) ) Kc 2(σ 4 + u 4)/C w * 0
(10)
where the numerator is defined in eq 6, σw (m s-1) is the standard deviation of the vertical wind velocity, C0 is a “universal” turbulent constant with reported values from 3 to 9, and , the turbulent energy dissipation rate, is often presumed to be proportional to u*3. This approach was selected so that the estimates of the turbulent Schmidt number could be compared with those from Flesch et al. (41) because both studies were conducted using similar techniques.
Results and Discussion Ambient Conditions. The meteorological conditions, surface soil water content, maximum metolachlor flux rates, and total cumulative losses are presented as averages (Ta, soil water content, wind speed, maximum volatilization flux) and total solar radiation, and cumulative loss for each five day sampling period of each year (Table 3). The mean Ta values (day and night) were similar for 1999-2001 with the coolest Ta in 2002. The average daily wind speeds ranged from 1.2 (1998) to 2.0 m s-1 (2002) and were considered similar among the years. The total Rs value was variable and ranged from 44 (1998) to 120 MJ m-2 (1999). The gravimetric surface soil water contents (2 cm depth) were very low in 1998 and 1999 followed by considerably larger soil water contents in each of the remaining years. The precipitation totals during the 120 h following application ranged from 0.0 in 1999 to 59 mm in 2001 (Figure 1). The variability in ambient meteorological and soil water conditions influenced the pesticide volatilization, as shown by the differences in the maximum metolachlor volatilization fluxes and the cumulative losses for each year (Table 3). Metolachlor Volatilization. The metolachlor volatilization and Rs fluxes for each year are presented in Figure 2. The volatilization data were plotted on a logarithmic scale to accommodate the largest fluxes (factor of 10) observed in 2001 and to maintain consistency in comparison among the years. In all of the years, the volatilization fluxes were highest during the first 12-24 h after application and then decreased by an order of magnitude after 48-72 h as the source strength gradually diminished over time (Figure 2). The largest initial flux rates were observed in 2001 (1550 ng m-2 s-1), and the lowest were observed in 2002 (68 ng m-2 s-1). In 2000, the volatilization fluxes remained relatively constant until approximately 100 hours after application. Volatilization in all of the years occurred during the day and night. Diurnal oscillations were generally in phase with Rs; that is, daily peak fluxes coincided with the daily peak Rs (Figure 2). There
FIGURE 1. Precipitation amounts during the first 120 h after application from 1998 to 2002.
FIGURE 3. Metolachlor fluxes and relative humidity from 1998 to 2002 during the first 120 h after application.
FIGURE 2. Metolachlor and solar radiation fluxes from 1998 to 2002 during the first 120 h after application. were a few notable exceptions to this trend (1998, 1999) when in the early part of the sampling period peak volatilization fluxes occurred during the night. This suggests that other factors alone or in combination influence volatilization trends. The diurnal volatilization trends during the first 48 h after application for 1998-1999 were more correlated with RH than Rs, particularly in 1999, whereas for the other 3 years this trend was considerably less evident (Figure 3). During the study period in 1998 and 1999, the daytime RH value ranged from 40 to 55% and from 75 to 100% at night. We
reasoned that for metolachlor to volatilize from a dry soil it is required that the metolachlor molecules be desorbed from a soil surface in order to diffuse into the boundary layer of the atmosphere. The driving mechanism for desorption would be soil water in either the liquid or vapor phase. The soil water content in 1998 and 1999 (Table 3) indicated that the soil surface was nearly air dry (3.8-4% gravimetric soil water content), and yet in both years distinct periods can be observed when peak volatilization was occurring at night. We first considered water vapor condensation onto the soil surface as a source of water to displace metolachlor molecules from the soil surface. For condensation to occur (dew formation), the air in contact with the soil surface must reach saturation; that is, the ratio of the actual water vapor pressure must equal the saturation water vapor pressure at a given Ta. The meteorological data (Ta, RH) for the times when the metolachlor fluxes were increasing with RH did not support this hypothesis. We considered an alternative mechanism, specifically, preferential adsorption of water vapor by the soil relative to metolachlor. Ambient air in the boundary layer (next to the soil surface) that is increasing in RH (nighttime conditions) over a rapidly cooling and dry soil surface can result in adsorption of water vapor molecules that effectively compete with and displace the metolachlor molecules, resulting in increased availability of metolachlor for diffusion into the boundary layer. An additional source of water vapor to the soil surface can also occur by upward diffusion of water vapor from below the soil surface. Jackson (42) and Jackson et al. (33) demonstrated that during nighttime conditions the soil temperature and water content gradients reverse direction, resulting in water vapor flow upward toward the drier soil surface. Spencer and Cliath VOL. 39, NO. 14, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5223
(43) compared the dieldrin volatilization rates from a dry soil surface before and after exposure to high RH and found significantly greater volatilization losses after exposure to increasing RH. Pennell et al. (44) demonstrated that as RH increased, water vapor effectively competed with organic vapors for mineral adsorption sites, resulting in the suppression of VOC sorption and thus increased VOC diffusion. Chen and Rolston I & II (45, 46) compared the volatilization rates of diazinon when dry N2 and air enriched with water vapor were alternatively used as sweep gases over a dry soil surface treated with diazinon. Their results showed increased diazinon volatilization when moist air was swept over the surface and a significant reduction in volatilization when dry N2 was swept over the surface. Thus, during the first 48 h after application in 1998 and 1999, we speculate that the combination of dry soil surface conditions, increasing RH (atmosphere), and potential upward transport of soil water vapor from below the surface may explain volatilization fluxes that correlated with increasing RH (Figure 3). In 1998, after correlating with RH for approximately 2 days, volatilization rates were observed to track with Rs for the last 3 days of the study. This was attributed to periodic light precipitation events (
