Environ. Sci. Technol. 2004, 38, 4176-4186
Simulating the Influence of Snow on the Fate of Organic Compounds GILLIAN L. DALY AND FRANK WANIA* Department of Physical and Environmental Sciences and Department of Chemistry, University of Toronto at Scarborough, 1265 Military Trail, Toronto, Ontario, Canada M1C 1A4
Snow scavenging, a seasonal snowpack, and a dynamic water balance are incorporated in a non-steady-state generic multimedia fate model in order to investigate the effect of snow on the magnitude and temporal variability of organic contaminant concentrations in various environmental media. Efficient scavenging of large nonpolar organic vapors and particle-bound organic chemicals by snow can lead to reduced wintertime air concentrations and incorporation in the snowpack. The snow cover functions as a temporary storage reservoir that releases contaminants accumulating over the winter during a short melt period, resulting in temporarily elevated concentrations in air, water, and soil. The intensity of these peaks increases with the length of the snow accumulation period. Organic chemicals of sufficient volatility (log KOA < 9; e.g., light polychlorinated biphenyls) can volatilize from the snowpack, resulting in springtime concentration maxima in the atmosphere. The behavior of fairly water-soluble chemicals during snowmelt depends on their relative affinity for the newly formed liquid water phase and the rapidly diminishing ice surfacesquantitatively expressed by their interface-water partition coefficient (KIW). Chemicals with a preference for the dissolved phase (low KIW; e.g., pentachlorophenol) can become enriched in the first meltwater fractions and experience a temporary concentration peak in lakes and rivers. Organic chemicals that are neither volatile enough to evaporate from the snowpack nor sufficiently water soluble to dissolve in the meltwater (e.g., polybrominated diphenyl ethers) sorb to the particles in the snowpack. These particles may be sufficiently contaminated to constitute the major input route to the terrestrial environment upon release during snowmelt. Because wintertime deposition to the snowpack may be higher than to a non-snow covered surface, this can result in higher soil concentrations of persistent organic contaminants in the long term. The potential ecotoxicological significance of peak exposures demands a better understanding of the role of snow in the fate of organic contaminants.
Introduction Snow is an integral part of cold environments, particularly in areas of high latitude or altitude, affecting the fate of organic contaminants in a variety of ways (1). Organic chemicals in the atmosphere are scavenged by falling snow as vapors or associated with particulate matter (2, 3). A seasonal or permanent snow and ice cover prevents diffusive gas * Corresponding author telephone: (416)287-7225; fax: (416)2877279; e-mail:
[email protected]. 4176
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004
exchange of organic vapors between the atmosphere and lakes, polar seas, and the terrestrial environment. Contaminants stored in the snow or ice are delivered to soils, vegetation, and water bodies during melt, resulting in the possibility of temporarily elevated concentrations. This may be ecotoxicologically relevant because snowmelt often coincides with the onset of biological activity, when some organisms may be at life stages particularly vulnerable to contaminant exposure (4). In the Arctic, sea ice has been identified as an important vector of contaminants (5). Contaminants enter sea ice by deposition from the atmosphere or inclusion by shelf sediments. Most of the sea ice melt, and therefore contaminant release, occurs in the biologically important marginal ice zone, where it can be taken up by ice fauna (6). As ice-associated amphipods form the base of the Arctic food chain, the contaminants can be passed on to higher trophic levels (i.e. fish, seabirds and seals). Finally, the large sensitivity of snow and ice to temperature changes could amplify the effect of climate change on organic contaminant transport and deposition (7). The failure to include falling snow and a seasonal snow cover may compromise the accuracy of models that seek to describe organic chemical fate in cold environments, including global models that comprise polar regions. To our knowledge, no multimedia mass balance model presently accounts for the presence of snow and ice. The current work attempts to address this gap by developing a dynamic multicompartmental fate model that accounts for scavenging by falling snow and includes a seasonal snow compartment. A quantification of the chemical fate processes associated with snowmelt necessitates a dynamic description of the movement of water in the model environment. This work builds on earlier work simulating the fate of organic chemicals in an aging snowpack (8, 9). It further makes use of the results of recent research into quantifying the specific surface area of snow (10-12) and chemical-specific interfacial partition coefficients (13, 14). Hypothetical simulations with the new model are used to examine how falling snow and a seasonal snow cover affect the pathways, major reservoirs, and seasonal variability of organic contaminants in the environment. Specifically, it is used to determine the relative importance of the processes by which a chemical enters and is lost from a snowpack. For example, the percentage of chemical that enters the snowpack by vapor or particle scavenging is evaluated and linked to the compound’s physical-chemical properties. Similarly, the relative amount of chemical that is lost by degradation, volatilization, and runoff with meltwater (both dissolved and with particles) is estimated. The new model is also used to assess the effect of snow scavenging, snow cover, and snowmelt on the magnitude and seasonal variability of contaminant levels in air, water, and soil. Calculations are repeated with different snowmelt rates, length of snow accumulation, and snow surface area to evaluate model sensitivity to these key parameters.
Methods Multimedia Fate Model. Snow was incorporated into CoZMo-POP, an existing dynamic multimedia organic chemical fate model (15). This model is formulated in terms of fugacity using Z values to describe equilibrium partitioning and D values to describe transport and transformation processes (16). Prior to the addition of snow, the model had compartments describing the atmosphere, forest canopy, forest soil, agricultural soil, freshwater, freshwater sediment, 10.1021/es035105r CCC: $27.50
2004 American Chemical Society Published on Web 06/29/2004
TABLE 1. Equations Used To Describe Water Content in Model Compartments and Water Fluxes between Them snowpack soil freshwater parameter wG wGa-w wGa-e wGa-sp wGw-a wGe-a wGsp-a wGsp-w wGsp-e wGe-w wGw-c U3 Ax Vx vfwe vfswe-sp
Water Balance Equations dhsp/dt ) (wGa-sp - wGsp-a - wGsp-w - wGsp-e)/(Aspvfswe-sp) dvfwe/dt ) (wGae + wGsp-e)/Ve - (kevap-e + krunoff-e)vfwe dVw/dt ) wGaw + wGsp-w + wGew - (kevap-w + krunoff-w)Vw equation
U3Aw U3Ae U3Asp Vwkevap-w vfweVekevap-e vfswe-sphspAspksub vfswe-sphsp-maxAwkmelt vfswe-sphsp-maxAekmelt vfweVekrunoff-e Vwkrunoff-w
hsp hsp-max kevap ksub kmelt krunoff-e krunoff-w
coastal water, and coastal sediment and allowed for seasonally variable temperature, wind speed, and hydroxyl radical concentrations. Details will only be given for expressions relating to snow and the dynamic water balance. A complete model description can be found in Wania et al. (15). Dynamic Water Balance. Whereas CoZMo-POP previously used annually averaged water fluxes, accounting for the impact of a rapidly melting snowpack demands a nonsteady-state description of the water balance. The water content in soils, the snowpack, and freshwater are now functions of time, determined dynamically by the balance of water flowing in and out of these compartments (Table 1). The fluxes are calculated as the product of user-defined rate constants k in units of reciprocal time and a term describing the size of the compartmental water reservoirs (e.g., soil moisture, freshwater depth). Snowmelt is an exception in that it is assumed to occur at a constant rate, independent of the size of the snow compartment. This is obviously a simplification, as the melting rate will vary in response to diurnal and longer-term temperature fluctucations. Using a temporally averaged melting rate appears justified considering the added model complexity that would be required to resolve such short-term variability. Evaporation k values increase with temperature, while k values describing runoff from soils are functions of soil moisture. All k values have a physical basis, and their actual values are chosen such that the fluxes, seasonal trends, and relative amount of water in each compartment are reasonable and describe a seasonal cycle typical for south central Canada. Details on the k values are provided in the Supporting Information. Three seasons are defined based on temperature: a snow accumulating season (starting when the temperature drops below 0 °C), a snow melting season (starting when the temperature rises above 0 °C, and lasting a user-defined fixed length), and a summer season (starting when the snow has completely melted). No melting occurs during snow accumulation, and snow coverage is always complete (i.e., there is no partial snow cover). The snowpack is assumed to be
description water fluxes in m3/h air-freshwater (w) air-soil (e) air-snow (sp) water-air soil-air snow-air snow-freshwater snow-soil soil-freshwater freshwater-coastal (c) precipitation rate in m/h surface area of compartment x in m2 volume of compartment x in m3 volume fraction of water in soil volume fraction snow water equivalent in snowpack height of snowpack in m maximum height of snowpack in m evaporation rate in h-1 sublimation rate in h-1 melting rate in h-1 soil runoff rate in h-1 freshwater outflow rate in h-1
homogeneous with respect to temperature, physical properties, and chemical concentrations. Physical snow characteristics (specific snow surface area, porosity) are userdefined, time-variant functions. The particle content in the snowpack is determined from the rates of particle scavenging with falling precipitation and dry particle deposition. Contaminant Fate. The inclusion of a snow compartment requires the description of contaminant fate processes that deliver contaminant to and from the snowpack (Figure 1). However, it also requires a consideration of the fact that other contaminant fate processes cease to operate during the presence of a snow cover. The occurrence of contaminant transfer processes is defined based on the three seasons outlined above (Figure 2). During snow accumulation, there is no air-surface exchange except with the snowpack, and neither is there contaminant exchange between the snowpack and the surface compartments. Also, runoff from soil to freshwater is assumed to cease when temperatures drop below 0 °C. During the snowmelt period, air-surface exchange resumes with the forest canopy, and contaminant can be transferred from snow to surface media with meltwater and particles. During the summer season, the model is identical to the original CoZMo-POP before the introduction of snow. The mass balance equation for the snow compartment and the expressions used to calculate the snow-related Z and D values are given in Table 2. The description of contaminant partitioning in the snowpack is adopted from refs 8 and 9 and assumes that the ice-air interface, organic matter, and liquid water constitute the only significant reservoirs for organic chemicals in snow. The capacity of the snow surface for organic chemicals is expressed with the help of an interfacial partition coefficient (KSA in m). The organic matter in snow is assumed to have the same partitioning properties as the organic matter present in atmospheric particles (17). Gaseous air-snow exchange of chemical occurs by sequential molecular diffusion through the snowpack’s airVOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4177
FIGURE 1. Processes involved in the delivery and loss of organic contaminants in a seasonal snow cover and described in the modified CoZMo-POP model. nants. Particles and the contaminants they contain can also be deposited to the snowpack in the absence of precipitation, which is referred to as dry particle deposition. Rain can also scavenge contaminant by dissolution in the aqueous phase or particle scavenging. To evaluate which deposition mechanism is dominant, knowledge of the chemical’s phase distribution as a function of temperature, phase composition (snow specific surface area, particle content), kinetic parameters (deposition velocities, precipitation rate, particle scavenging ratio), and temperature is required (18). Degradation and volatilization contribute to chemical loss from the snowpack. Chemical is also lost from the snowpack by dissolution in the draining meltwater and sorption to organic particles. The model assumes that particles, and thus also chemicals sorbed to them, remain in the snowpack until the moment that the snow completely disappears, at which point they are transferred to the underlying soil and freshwater.
FIGURE 2. Occurrence of contaminant transfer processes between the environmental compartments of CoZMo-POP is dependent on the seasons, which are defined based on temperature: snow accumulation season (T < 0 °C), snowmelt season (T > 0 °C), and summer season (T > 0 °C, snow has disappeared). filled pore space and the boundary layer above it. Snow ventilation may increase contaminant movement within the snowpack beyond the rate of molecular diffusion, which is parameterized through a wind-pumping factor. A detailed examination of how this process influences air-snow exchange is provided later. In addition to the dry gaseous deposition described above, organic vapors can be delivered to the snowpack by wet gaseous deposition (i.e., adsorption to the surface of falling snow flakes). Wet particle deposition is a result of snow scavenging particles containing contami4178
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004
Model Scenarios. To evaluate the impact of snow on chemical fate, the modified CoZMo-POP model was used to carry out simulations with and without a seasonal snow cover. Two hypothetical temperature scenarios were defined (Figure 3). They are identical, except that in one scenario the temperature never drops below 0 °C and no snow cover develops. This scenario thus resembles a calculation with the original CoZMo-POP. The other scenario has a seasonal snow cover lasting for about 5 months from the end of September to early April. Whereas OH concentrations were allowed to vary seasonally, wind speed and precipitation were assumed to be constant to isolate the seasonality stemming from the presence of the snowpack. For the same reason, the model simulations were completed without a forest compartment. We further eliminated the coastal model compartments to better reflect the continental climatic conditions of central Canada. A generic emission scenario with steady emissions into the atmosphere was used. Snow Properties. The model simulation employed generic environmental parameters. Only the parameters relating to the snow compartment are discussed here. All other pa-
TABLE 2. Model Equations Related to Contaminant in the Snowpack
Msp t fA, fsp parameter
Mass Balance Equation for Contaminant in Snowpack dMsp/dt ) Da-sp fA - (Dsp-a + Dsp-e/w + Dpf-e/w + DR)fsp amount of contaminant in snowpack in mol time in h fugacity in air and snowpack in Pa equation
DR Dsp-e Dsp-w Dpf-e Dpf-w Dsp-a Da-sp
kspVspZsp wGsp-eZw wGsp-wZw GpfZw + Ae/AspMo-maxZo/Fo GpfZw + Aw/AspMo-maxZo/Fo Asp/(1/US/A-blZA) + 1/(wpfUS/A-psZA)) Dsp-a + AspDDVspvfAZQ + wGa-spBZx
Za Zw ZI Zo Zsp BZrain BZsnow ksp Vsp wpf Mo-max Fo US/A-ps US/A-bl Ba vfA DDVsp KWA KOA KSA KIA R T Fpart, Foct, Fmw va, vw, vo Asnow Wp foa Gpf
1/(RT) KWAZa KIAZa KOA(Fpart/Foct)Za Zava + Zwvw + ZiAsnowFmw + Zovo Zw + Wp1vfAZo foa ZiAsnowFmw + Wp2vfAZo foa
Ba(va10/3/va + vw)2/(ln(2)hsp)
description
D Values in mol/Pa‚h reaction in snowpack meltwater runoff to soil meltwater runoff to freshwater particle flush to soil particle flush to freshwater snow to air transfer air to snow transfer (BZx ) BZrain at T > 0 °C, BZx ) BZsnow at T > 0 °C) Z Values in mol/m3‚Pa air water interface organic matter snowpack rain falling snow reaction rate of contaminant in snowpack in h-1 volume of snowpack in m3 wind-pumping factor maximum mass of organic carbon in snowpack in g density of organic matter in g/m3 mass transfer coefficient for air-filled pore space in m/h mass transfer coefficient for air boundary layer above snow in m/h molecular diffusivity in air in m2/h volume fraction of aerosols in air in m3 solid/m3 bulk phase dry particle deposition velocity to snow in m/h dimensionless partition coefficient between water and air dimensionless partition coefficient between octanol and air partition coefficient between snow surface and air in m partition coefficient between liquid water surface and air in m gas constant in J/(mol‚K) absolute temperature in K density of atmospheric particles, octanol, and snowmelt water in g/m3 volume fraction of air, liquid water, and organic matter in snowpack snow surface area in m2/g particle scavenging ratio for rain (1) and snow (2) mass fraction of organic matter in atmospheric particles water remaining in snowpack when particle flush occurs in m3
FIGURE 3. Temperature profiles used in the model simulations. The “no snow” temperature profile is the same as the “snow” temperature profile, except the temperature does not drop below 0 °C. rameters are described in ref 15. Specific snow surface area (Asnow in m2/g), the surface area accessible to gases for a given mass of snow, is a key parameter determining the capacity of the snow phase for organic contaminants. Asnow of falling snow controls the extent of vapor scavenging, and
Asnow in the snowpack has a strong influence on diffusive snow-atmosphere exchange, in particular the potential for evaporation from the aging snowpack. Several studies have attempted to quantify Asnow using sorption isotherms of methane, nitrogen, or a higher molecular weight organic compound on the surface of artificial or natural ice (10-12). Asnow has also been estimated using optical and scanning electron microscopy (19). The Asnow typically decreases over the time scale of days as the snow ages. Legagneux et al. (11) have suggested that Asnow for falling snow varies between 0.036 and 0.158 m2/g. In the model simulations, a seasonally dependent Asnow was adopted, with a value of 0.1 m2/g during the snow accumulation period and a linear decrease from 0.1 to 0.01 m2/g during the snowmelt period. The liquid water content in the snowpack was also defined based on season. During the snow accumulation season, the amount of liquid water contained in the snowpack is calculated from the specific surface area assuming an average thickness of the quasi-liquid layer of 10 nm (20). The volume fraction of liquid water in the snowpack increases rapidly during the snowmelt period, which was parameterized using a power function increasing from the initial value to a maximum of 0.40 at the end of snowmelt. A constant snow density of 0.433 g/cm3 was assumed to apply throughout winter (9). The density of VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4179
surface snow ranges from 0.01 g/cm3 for fresh snow to 0.5 g/cm3 for wind-packed snow (12), so this value represents fairly dense, aged snow. The snowmelt period is assumed to last 33 d. As mentioned above, the particle content in snow is determined dynamically from the amount of particles scavenged by falling snow and dry deposited to the snowpack. Particle scavenging ratios for snow (WP) and dry particle deposition velocities to a snowpack are highly variable. Franz and Eisenreich (2) reported a WP for snow of 5 × 106, and Wania et al. (1) suggest a range from 5 × 105 to 1 × 106. We chose a lower value of 105 for both snow and rain to avoid overestimating the effect of snow in the model. Such a value also results in realistic organic carbon concentrations in snow (21). Ibrahim et al. (22) determined dry particle deposition velocities to snow, and Davies et al. (23) listed several reported values. From a review of these data, a value of 1.8 m/h was deemed reasonable. Influence of Wind-Pumping on Air-Snow Exchange. For a chemical vapor to undergo snow to air exchange, it has to diffuse through the air-filled pore space to the top of the snowpack and from there through a boundary layer to the bulk atmosphere. Whereas the latter is expressed with the help of a snow-air boundary layer mass transfer coefficient (US/A-bl, units of m/h), the mass transfer coefficient for molecular diffusion through the air-filled pore space (US/A-ps, units of m/h) is a function of the molecular diffusivity in air, the volume fraction of air in the snowpack, and the mean diffusion path length, which is related to snow depth (8). As the snowpack gets deeper, the diffusion path length and thus the calculated resistance to molecular diffusive transport within the snowpack becomes so large that no gas exchange would take place at all. This is unrealistic for two reasons: A chemical in the surface layer of the snowpack only has to diffuse a small distance to the top of the snowpack, and wind-driven advective motion through the snowpack may greatly accelerate vapor transport in the snow pores. A windpumping factor is used to account for this increase in mass transfer, but the numerical value for such a factor is not known. In fact, it will greatly vary depending on wind exposure and snow permeability. The term wpf‚US/A-ps/US/A-bl indicates which resistance is controlling snow-air chemical transfer. When greater than 1, the boundary layer poses the greater resistance to chemical movement. When less than 1, transport within the snowpack is rate-limiting. We chose a value for the wind-pumping factor, where the resistance to air-snow diffusion posed by diffusion within the snowpack and within the boundary layer above the snowpack are of similar magnitude (wpf‚US/A-ps/US/A-bl ≈ 1). As US/A-ps is dependent on snow depth, it and thus also the term wpf‚US/A-ps/US/A-bl is a function of time, being larger at the beginning of snow accumulation and toward the end of the snowmelt, when the snowpack is shallow (Figure 4). Figure 4 shows that a wpf of below 10 leads to rate limitation by transport within the snowpack, whereas a wpf of 1000 results in air-snow exchange that is largely controlled by the boundary layer. Therefore, a wind-pumping factor of 100 was chosen for the model calculations. Physical-Chemical Properties. Model simulations were performed for 10 organic contaminants: the polycyclic aromatic hydrocarbons (PAHs) phenanthrene (PHE), pyrene (PYR), and benzo[a]pyrene (BaP); the polychlorinated biphenyl (PCB) congeners 28, 101, and 180; the polybrominated diphenyl ether (PBDE) congeners 47 and 209; pentachlorophenol (PCP); and R-hexachlorocyclohexane (R-HCH). To characterize the distribution behavior of these chemicals, the model requires several physical-chemical properties (Table 3), including equilibrium partition coefficients between air, octanol, and water (KOA, KAW, KOW) and their related energies of phase transfer. The volatility of the 10 substances 4180
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004
FIGURE 4. Relative importance of the resistance within the airfilled pore space of the snowpack and in the boundary layer above the snowpack for diffusive gas exchange between atmosphere and snowpack as a function of time, given various numerical values for the wind-pumping factor. ranges over 7 orders of magnitude, from the relatively volatile R-HCH and PHE (log KOA < 8), which are gas-phase chemicals even at wintertime temperatures, to BDE-209 (log KOA > 15), which is completely particle bound (Table 3). Within a group of related substances (e.g., within the PCBs), KWA, KOA, and KSA increase with molecular size (i.e., the larger representatives of a group partition more strongly from the gas phase to the condensed phases water, octanol, and the snow surface). However, KWA increases much less than KOA and KSA because water solubility also decreases with size. This is also reflected in an increasing KOW with increasing size. The PCBs cover a similar volatility range as the PAHs (7.5 < log KOA < 11) but are somewhat less water soluble (log KAW and log KOW are 1-2 orders of magnitude higher). The PBDEs have partitioning behavior comparable to that of the PCBs but are less volatile. R-HCH and PCP have a higher preference for dissolving in water than the PCBs and PBDEs. As no measured KSA values for the chemicals of interest are available, a linear free energy relationship to estimate the sorption coefficient of an organic compound to the snow surface by Roth et al. (13) was employed:
∑β + 3.53∑R - 6.85
log KSA (-6.8 °C) ) 0.639log L16 + 3.38
H 2
H 2
(1)
The required compound properties are the summation H hydrogen bond acidity (RH 2 ) and basicity (β2 ), as well as the logarithm of the gas-hexadecane partition coefficient (L16). For R-HCH, these have been presented in Abraham et al. (24). The values for the PAHs, PCBs, PBDEs, and PCP were calculated using the method from Platts et al. (25). The KSA at -6.8 °C was extrapolated to other temperatures using an enthalpy of sorption ∆HIA for the liquid water surface estimated from an empirical relationship given in Goss and Schwarzenbach (14):
∆HIA ) -5.07 ln KIA(15 °C) - 108
(2)
where KIA is the interfacial adsorption coefficient on the liquid water surface (in m). At temperatures close to the melting point, the surface of ice is formed by a quasi-liquid layer (26), and sorption to the ice surface was indeed found to be similar, though not identical, to that on the liquid water surface (13). Comparison of heats of adsorption to an ice
TABLE 3. Partitioning Coefficents between Octanol and Air (KOA), Air and Water (KAW), Octanol and Water (KOW), and Air and Snow Surface (KSA in m), Gas-Phase Reaction Rate with OH Radicals (kAref in 10-13 cm3/s), and Degradation Half-Lives in Water and Soil (t1/2 in h) for All Chemicals Used in Model Simulationsr
PCP R-HCH PCB-28 PCB-101 PCB-180 PBDE-47 PBDE-209 phenanthrene pyrene benzo[a]pyrene
log KOA 25 °C
log KAW 25 °C
log KOW 25 °C
log KSA -6.8 °C
kAref 25 °C
t1/2 water 25 °C
t1/2 soil 25 °C
Φ -10 °C
Φ 0 °C
10.02a 7.46f 7.85j 8.73j 10.16j 10.44m 15.27m 7.57n 8.80n 11.12q
-4.91b -3.53f -1.91j -2.01j -2.48j -3.35m -5.07m -2.92o -3.42o -4.35o
5.11c 3.94f 5.66j 6.33j 7.16j 6.39m 8.70m 4.65p 5.38p 6.77p
0.51d 0.23g -0.98d -0.10d 0.60d 1.50d 4.80d -1.03d 0.28d 2.01d
5.5e 4.9h 10.4k 3.0k 1.0k 10.0e 0.34e 130e 500e 500e
11 600e 8 760i 5 500l 31 000l 55 000l 14 300e 800 000e 6 500e 3 000e 8 100e
23 300e 2 190i 10 000l 100 000l 1 000 000l 28 600e 1 600 000e 13 100e 6 100e 16 200e
61
