Environ. Sci. Technol. 2009, 43, 657–662
Organic Contaminant Release from Melting Snow. 1. Influence of Chemical Partitioning T O R S T E N M E Y E R , †,‡ Y I N G D U A N L E I , †,‡ IBRAHIM MURADI,‡ AND F R A N K W A N I A * ,†,‡ Department of Chemical Engineering and Applied Chemistry and Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada, M1C 1A4
Received July 21, 2008. Revised manuscript received November 16, 2008. Accepted November 19, 2008.
A melting snowpack can deliver organic contaminants to terrestrial and aquatic ecosystems in the form of short and concentrated pulses. The mechanisms and kinetics of the underlying processes need to be understood to successfully integrate them into contaminant and water quality models. Controlled laboratory-based snowmelt experiments using artificially produced snow spiked with organic target contaminants reveal how chemical behavior during melting is dependent on the partitioning between the different phases within the bulk snow. Behaving similar to inorganic ions, water soluble organic chemicals, such as atrazine, are preferentially released at an early stage of melting, because such chemicals, accumulated at the snow grain surface, dissolve in the downward percolating meltwater front. Hydrophobic substances attached to particles, such as the larger polycyclic aromatic hydrocarbons, are often released at the very end of the melt period, because particle coagulation and snow densification render the melting snowpack an efficient filter trapping the particles. A notable fraction of volatile chemicals, such as naphthalene, will transfer from the melting snowpack to the lower atmosphere due to evaporation. Organic pollutants with intermediate partition properties, such as lindane, can easily switch between the bulk snow phases and their elution behavior is therefore more sensitive to varying snow and melt characteristics.
Introduction Amplification processes occurring within bulk snow prior to and during melting can lead to the release of organic contaminants from melting snow in peak loads (1). Concentration peaks of organic contaminants during spring snowmelt have been measured repeatedly in rivers and lakes. Similarly elevated concentrations of organic pollutants have been observed within the lower atmosphere during snowmelt. A comprehensive literature review related to the snowmelt behavior of organic chemicals is provided elsewhere (1). Arctic and subarctic regions are especially susceptible to global climate change processes, and even small temperature differences not only influence the extent and duration of a * Corresponding author phone: +1-416-287-7225; e-mail:
[email protected]. † Department of Chemical Engineering and Applied Chemistry. ‡ Department of Physical and Environmental Sciences. 10.1021/es8020217 CCC: $40.75
Published on Web 12/23/2008
2009 American Chemical Society
seasonal snow cover but also determine whether precipitation takes the form of snow or rain. The efficiency by which organic contaminants are scavenged from the atmosphere can differ greatly between snow and rain (2). It is therefore of interest to study how climate change will influence the snow-related fate of organic contaminants in cold regions (3). Whereas investigations of the snowmelt behavior of inorganic chemicals have been quite common in both field and laboratory, the fate of organic contaminants in melting snow is far less well studied. Most field studies are limited to reporting residue concentration in bulk snow and meltwater samples taken in urban, rural, or remote areas (4). Very few studies have investigated the temporal elution sequences of organic chemicals during snowmelt (5, 6). In an early laboratory study, Scho¨ndorf and Herrmann (6) melted natural snow in a glass cylinder (length 1 m; inner diameter 14 cm) while monitoring physical and chemical snow properties. Whereas concentration peaks of watersoluble organic contaminants were observed at the beginning of the melt period, the bulk of the more hydrophobic substances was released at the end of melting along with the particles (6) (Figure S1 in the Supporting Information). In the field study conducted by Simmleit et al. (5), however, most of the particle-associated organic compounds were released at the beginning of the melt period along with the dissolved fraction. The contaminant pulse load at the early stage of a snowmelt period is referred to as type 1 enrichment and is due to the uptake of chemicals from snow grain surfaces by the downward moving meltwater front (1). Particles and associated chemicals are efficiently held back from being washed out early, due to filter-like processes as a result of particle coagulation and an increasing snow density during melting. The release of particle-bound organic pollutants at the end of melting is referred to as type 2 chemical enrichment (1). Building on the pioneering work of Scho¨ndorf and Herrmann (6), this study investigated the behavior of organic chemicals during the melting of artificially produced snow in a temperature-controlled cold room using a method previously developed and described (7). The objective of these experiments was to gain a mechanistic understanding of the processes determining the release of organic chemicals from a melting snowpack, and in particular to decipher how they are influenced by a chemical’s partitioning properties between the different bulk snow phases. Chemical equilibrium partitioning maps are used to illustrate how an organic chemical’s distribution characteristics, in concert with the physical snow properties, determine in which bulk snow phase a chemical is likely to reside and how the chemical is therefore released during the melt. A comparison of the presented results of this study with the findings from Scho¨ndorf and Herrmann (6) can be found in the Supporting Information (Figure S1).
Materials and Methods The experimental design, the methods used to produce artificial snow spiked with target substances and to monitor the physical snow properties, as well as the chemical analytical methods are described in detail elsewhere (7). Only the main features of the methods and any changes from the previously published procedure are described herein. A rectangular stainless steel vessel (0.24 m3; height 40-50 cm, width 100 cm, depth 55 cm) with a conical shaped bottom disemboguing in a meltwater outflow, was filled with artificial snow and exposed to different melt conditions in a walk-in cold chamber. VOL. 43, NO. 3, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
657
Changes in physical snow properties were measured online by means of time domain reflectometry (TDR). In particular, the density of dry snow and the water content of the snow during melting were measured at various snow depths. The travel time of an electromagnetic pulse along aluminum rods embedded in snow is proportional to the relative dielectric permittivity of the rods’ surrounding. The latter is different for the individual bulk snow phases air pore space, ice, and meltwater. Based on the measured permittivity variations and the length of the rods, the density of dry snow can be calculated directly. By interpolating the density of melting snow from two subsequent dry snow periods at subfreezing temperatures, the water content during a melt phase can be calculated. The average density of the bulk snow was additionally recorded by measuring the snow volume and the remaining snow-water equivalent (SWE) at different stages of melting. To further study the physical snow properties the snow microstructure was tracked at different melt stages by means of macrophotography. The approximate size and shape of the snow grains provide clues about the hydraulic properties and the storage capacity of the snow. By combining the snow classification scheme by Colbeck et al. (8) with the studies by Legagneux et al. (9) and Domine´ et al. (10), relationships between snow density and the specific surface area (SSA) could be established. The SSA is a crucial parameter in estimating the sorptive capacity of the internal snow surface for organic chemicals. By means of a “snow gun” artificial snow was produced exhibiting pellet-shaped snow grains with diameter of approximately 100 µm, a density of 0.16 ( 0.01 g/cm3, and a specific surface area (SSA) of 580 ( 50 cm2/g. Pellet-shaped snow grains represent one of many possible natural snow grain forms. However, melt metamorphism quickly eliminates the effects of grain shape on water permeability (11, 12). Bulk snow densities were measured using gravimetric and volumetric methods and were compared to values obtained by TDR. The SSA was determined by measuring the adsorption isotherm of Krypton on snow at 77.15 K (liquid N2 temperature) (adapted from ref 9). Mixtures of solutions containing six different organic target substances either in the dissolved phase or attached to particles were injected into the water stream leading to the snow gun. The spiked artificial snow contained average concentrations of 1500 ng/L SWE of four polycyclic aromatic hydrocarbons (PAHs) (naphthalene (NAP), phenanthrene (PHE), pyrene (PYR), benzo(ghi)perylene (BghiP)), of atrazine (ATR), and of lindane (LIN). The concentrations of PHE, PYR and BghiP are of the same order of magnitude as previously measured concentrations in natural road-side snow and snow near an industrial area (see refs in ref 13). The partition coefficients describing the equilibrium distribution between the internal snow surface, meltwater, air pore space, and organic matter within snow range over several orders of magnitude for this group of six compounds (Table S1). The water used for making the artificial snow was further spiked with 3.5 mg of humic acid per L SWE (Aldrich Corp., St. Louis, MO). A cooling liquid circulating through the double-walled bottom of the snowmelt vessel was ensuring natural temperature gradients within the bulk snow. Melting was enforced either by elevated cold room temperatures while keeping the snow bottom at 0 °C or by six 175 W infrared lights whose heights above the snow surface was adjustable. Light in this wavelength range keeps photodegradation to a minimum (6). The 2 ( 0.2 L meltwater fractions were sampled at the bottom of the vessel, measured for their specific conductivity, filtered, extracted, and eventually both the dissolved and particulate phase were analyzed separately for organic contaminants. At the end of melting a large fraction of the particles adhered to the bottom of the snow 658
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 3, 2009
vessel. Deviating from the procedure described in Meyer et al. (7), those particles were rinsed with additional drinking water and wiped into the exit funnel with a brush. The vessel bottom was subsequently rinsed with acetone and both rinses were extracted and analyzed. Separation between the dissolved and the particulate phases of each sample was achieved by suction filtration using 0.45 µm nylon hydrophilic membrane filters (Whatman, Brentford, UK). After leaving the snowmelt vessel and during storage and filtering, the chemicals may have partially switched between the dissolved and the particulate phase. Therefore, the presented relative concentrations in both phases within one sample may not exactly mirror the partitioning at the time when released from the snowpack. The contaminants in the dissolved phase were extracted by means of C18 SPE cartridges (Supelco, Sigma-Aldrich, St. Louis, MO). The extract was eluted with 5 mL each of ethyl acetate, dichloromethane/ethyl acetate (1:1), and dichloromethane (Caledon Laboratories Ltd., Georgetown, ON) (14). The extraction of the particulate fraction was achieved by ultrasonification (VWR Aquasonic) in acetone, filtering and solvent exchange to iso-octane. Both dissolved and particulate fractions were analyzed using gas chromatography-mass spectrometry with electron impact ionization and selected ion monitoring mode. Surrogate contaminants were added prior to analysis to obtain recovery rates and mass balances (7). All chemical standards came from Cambridge Isotope Laboratories, Inc. (Andover, MA). The concentrations of particles within the meltwater were measured by filtering 1 L of each sample through a glass fiber filter (GF/F, Whatman, Brentford, UK) with an approximate cutoff size of 0.7 µm. The latter differs from the cutoff size used for the separation of particle-bound organic contaminants. The bulk of the air-borne particles have diameters in the range between 0.4 and 0.7 µm (15). Although a notable fraction of the particles will form large coagulates that are retained on GF/F, the particle concentrations in the meltwater samples can not be linked quantitatively with the concentrations of particle-bound organic substances. Most of the particles in the artificial snow originate from indoor air that is sucked into an air compressor which provides compressed air for snow production. Those particles presumably exhibit organic carbon content similar to that of the surrounding outdoor air (16, 17). The particle concentration within the artificial snow amounted to approximately 30 mg/L SWE. Combined with the injected amount of humic acid, the organic matter content was estimated to be 10 mg/L SWE, which reflects natural conditions (18, 19). A mass balance was calculated encompassing the chemical’s fate from the deposition of the snow into the vessel to the sampling of the meltwater, whereby the average meltwater concentrations were compared to the concentrations in bulk snow samples collected just after snowmaking. The approximate percentage loss was 8% ATR, 17% LIN, 85% NAP, 35% PHE, 26% PYR, and 20% BghiP. This implies that a notable fraction of NAP and, to a smaller extent, PHE, PYR, and LIN evaporated prior and during melting which is corroborated by the inverse correlation between the percentage loss of the chemicals (excluding BghiP), and their respective log KIA values (Figure S2). Ice-air partitioning controls the extent of evaporation in dry snow and to a large part also in melting snow. The loss of BghiP is likely due to irreversible sorption to the vessel bottom and sample containers (see refs in ref 7). The elution sequences of the target chemicals are presented as relative meltwater concentrations of a particular chemical within one experiment (7). The last particulate fraction in each of the elution sequences refers to chemicals that were adsorbed to the particles remaining at the vessel
FIGURE 1. Chemical space plots of the phase distribution of chemicals in a melting snowpack as a function of the snow surface/air sorption coefficient (log (KI/A/m), the humic acid/water (log KHA/W), and the air/water partition coefficients (log KAW). KI/A, KHA/A, and KAW values were determined based on equations in refs 22 and 23. bottom. The results presented here refer to the moderate to fast melting of an aged snowpack.
Results and Discussion Phase Partitioning Maps. The fate of an organic chemical during melt is determined by its distribution among the various components making up the bulk snow. Chemicals can be present as gases in the pore space, attached to particles in the snow, adsorbed at the internal snow surface, or dissolved within the meltwater (1). Incorporation within the ice lattice of the snow grains is presumably very minor for most organic chemicals, because it would cause large defects in the crystal structure (6, 20, 21). The distribution depends on the chemical’s relative affinity for each of the snow phases and can be expressed quantitatively with equilibrium partition coefficients. The latter are different for different chemicals, but they are constant in an isothermal melting snowpack. The equilibrium phase distribution of organic chemicals within the melting snowpack can be plotted in a twodimensional chemical partitioning space defined by the chemical’s air/water partition coefficient log KAW and the humic acid/water partition coefficient log KHA/W for a specific snow surface/air sorption coefficient log (KI/A/m) (Figure 1), all at 0 °C (1). Within this partitioning space, regions of predominant presence at the snow surface (bright blue), within the interstitial pore space (red), sorption to organic matter in the snow (brown), or presence in the dissolved
aqueous phase (dark blue) can be delineated. The phase distribution of an organic chemical also depends on the relative size of the bulk snow phases (1). Figure 1 illustrates the partitioning of the six organic substances at an early stage of melting of an aged snowpack that was previously exposed to several melt-freeze cycles. The artificial snow of the associated experiment exhibited a density of 0.26 g/cm3, a specific surface area in the range between 100 and 200 cm2/g (10), and an estimated organic matter concentration of 2.6 µg/mL snow, which is typical for snow from less-polluted urban regions (18, 19). The water content was in the range between 5 and 7% of the bulk snow volume which refers to an intermediate to strong melt intensity (11) (Figure S3). To accommodate the different ice-air partition properties of the chemicals, the maps were drawn for five log (KI/A/m) values (Figure 1, Table S1). The target chemicals were placed on the partitioning maps based on their estimated distribution properties at 0 °C. The fate of those chemicals that are located in the transition regions between the bulk snow phases in Figure 1 is very sensitive to the capacity of the internal surface area, the organic matter content, the meltwater content, and the pore space. Chemicals that partition appreciably (>50%) into the liquid water phase of the wet snowpack, i.e., are located within the dark blue region toward the upper left of the diagrams in Figure 1, are likely to be subject to the type 1 chemical enrichment. When the melting becomes intense VOL. 43, NO. 3, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
659
FIGURE 2. Relative elution sequences of atrazine, lindane, naphthalene, and the relative specific conductivity during snowmelt. the dark blue areas in Figure 1 will expand and more chemicals will dissolve into the downward moving meltwater phase (1). In fresh, fine-grained and clean snow the interface is favored, whereas the chemical distribution in aged and dirty road-side snow will be shifted toward the particles. Semivolatile chemicals may favor the gaseous interstitial space if the snow exhibits a very low density such as in fresh snow or in depth hoar (1). The boundaries between the colored areas in Figure 1 constantly change during snowmelt because the phase composition of the snowpack also changes as snowmelt progresses. As the melting snowpack densifies and the snow grain size increases, the relative extent of meltwater and particles will increase at the expense of snow surface and pore space (1). Accordingly, the phase transition boundaries in Figure 1 will in most cases shift to the upper right as melting proceeds. This also implies that more chemicals that originally were sorbed to the snow surface will either start to dissolve in meltwater or sorb to particles, depending on their location in the chemical partitioning space and the amount of particles in snow. Release of Relatively Water-Soluble Organic Contaminants. According to its partitioning properties ATR is preferentially present in the aqueous snow phase (Figure 1D). The elution pattern in Figure 2A shows a pronounced type 1 enrichment. The first quarter of the released meltwater contained 74% of the ATR with an enrichment in the first sample by a factor of 5 relative to the average meltwater concentration. This elution behavior is closely mimicking that of inorganic ions which is expressed by the relative specific conductivity (Figure 2B). The melt behavior of ions has been extensively investigated in the past, enabled by the ease of measuring the conductivity in meltwater. An ion flush in the early meltwater has been reported throughout most of these studies (24, 25). However, ions accumulate at the snow grain surface mainly due to a gradual “out-freezing” during dry and wet snow metamorphism, after which they can be taken up by the first downward moving meltwater. The ion amplification during early melting thus depends largely on the extent of snow metamorphism prior to the melt period. Organic substances, on the other hand are presumably arranged around the snow grain surfaces already prior to deposition and during snow grain growth because of the chemicals’ relatively large molecular size (6, 20, 21). Differences in the snowmelt behavior of organic chemicals are likely due to chromatographic effects, i.e., a differential retention at the ice-water interface (26). 660
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 3, 2009
LIN has a smaller tendency to dissolve in water than ATR (Table S1) hence, it is placed in the transition region separating the chemical’s preferential presence in the meltwater phase and on the ice-air interface (Figure 1C). LIN partitions into both phases in approximately equal parts, leaving its melt behavior very sensitive to the changing snow properties. Figure 2C shows the relative meltwater concentrations of both the dissolved and the particulate fractions of LIN during melting. The first quarter of the released meltwater contained 49% of the LIN (excluding the last particulate sample). The amplification of LIN in the early meltwater is thus notably smaller than that of ATR. The tendency of LIN to adhere to the internal ice surface is simply larger, which is reflected by its higher log (KI/W/m) values (Table S1). Small parts of LIN will also sorb to organic matter (Figure 1). Approximately 15% of the LIN was found within the particulate fraction of the sampled meltwater (30%, if the last sample is included) compared to 10% as indicated in Figure 1. This discrepancy may be caused by an incorrect estimation of organic matter in the artificial snow, by the aforementioned phase exchange after melting, and/or by the increasing relative proportion of particles during melting. A large part of the LIN was associated with the particles that remained at the vessel bottom after melting and is represented by the last brown bar in Figure 2C. This fraction made up 20% of all analyzed LIN. During melting LIN was segregated mainly into two fractions experiencing a very different fate, resulting in both type 1 and type 2 enrichment. Such segregation should become stronger the higher the particle content in the snow and the higher the meltwater content per volume. The melt behavior of a substance like NAP is more difficult to predict because notable fractions of the chemical are present within all major phases of a melting snowpack (Figure 1A). The part of NAP that is dissolved in the aqueous meltwater phase is released at an early stage of melting, similar to what is observed for ATR and LIN (Figure 2D). The first quarter of the meltwater contained 45% of the analyzed NAP. However, most of the chemical (nearly 85%) evaporated from the snowpack prior to and during melting, consistent with the prediction that NAP is the only compound that partitions appreciably into the snowpack’s air-filled pore space (Figure 1). Soon after the onset of melting the snowpack becomes denser and some NAP may become trapped in interstitial pores. During melting the pore space presumably acts as a reservoir which steadily supplies the meltwater phase with NAP (Figure 2D). We have no explanation for the small local maxima appearing near the middle of the elution sequences in Figure 2. Melt Behavior of Hydrophobic Organic Contaminants. Melt/freeze cycles cause substantial coagulation of fine particles in snow (27). During the freezing phase and with increasing dehydration groups of particles are pushed together and the resulting coagulates are held together by van der Waals forces (28). During the melt period those coagulates clog the pores and hydrophobic chemicals associated with the particles are efficiently held back from being eluted. When melting progresses and the snow surface subsides the particulate fraction of the chemicals is gradually accumulating in the upper snowpack until it is released at a late stage of melting. This peak release at the end of the melt period becomes more pronounced when the snow has undergone intense melt-freeze metamorphism and densification (6, 26). Dirt cones on snow can concentrate particleassociated chemicals literally at the top of the snow surface leading to their retention until the very end of melting. BghiP is either sorbed to organic matter or to the ice-air interface because of its high affinity for both phases (Table S1). During melting BghiP is likely to gradually transition from the snow grain surface to the particle phase. Thus, the
FIGURE 3. Relative elution sequences of phenanthrene, pyrene, and benzo(ghi)perylene during snowmelt. (A) Blue columns, dissolved phase; brown columns, particulate fractions. bulk of this chemical is released in the very end, together with most of the particles (Figure 3C). PYR and PHE are somewhat less hydrophobic than BghiP and have a higher affinity toward the aqueous meltwater phase (Figure 1, Table S1), whereby PHE in particular is more amenable to becoming dissolved (Figure 3). This behavior is consistent with the position of both chemicals on the chemical partitioning space in Figure 1. However, most of both chemicals finally becomes sorbed to particles and accumulates in the upper snowpack. Relationship between Contaminant Distribution in the Snow Pack and Amplification in Melt Water. The fate of organic contaminants during snowmelt is clearly correlated to their partitioning between the phases contained within the melting snowpack. The dissolved and the particulate fractions of a chemical tend to move into opposite directions of the elution order. Sufficiently water soluble substances are released early during melting because a large fraction is taken up from the snow grain surfaces by early meltwater. Organic contaminants adsorbed to particles tend to be released toward the end of a melt period. Melt-freeze cycles lead to particle coagulation and snow densification. As a consequence particle-associated chemicals are efficiently retained within the snowpack until the end of melting. The more exclusively a chemical is either dissolved within the aqueous phase or sorbed to particles, the more pronounced is the associated peak release. ATR easily dissolves in meltwater and most of the BghiP eventually becomes completely associated with the particulate phase during melting. Hence, those chemicals are exposed to the strongest amplification during snowmelt. On the other hand, the elution behavior of chemicals with intermediate partitioning properties such as LIN or NAP very much depends on the melt scenario, i.e., the meltwater content, snow porosity, SSA, and organic matter content. To aid in the assessment of whether a particular chemical is likely to show type 1 or type 2 chemical enrichment in a particular snowpack, we propose the calculation of an indicator value R, which quantifies the fraction of the chemical which is present in the liquid meltwater phase of the snowpack. The R value combines chemical properties and snowpack characteristics: R ) VW ⁄ (VW + KHA⁄W · VHA + KI⁄W · SSA + KAW · VAir)
(1)
where KHA/W, KI/W [m], and KAW are the equilibrium partition coefficients that describe the chemical’s distribution among the phases present in the melting snowpack, SSA [m2/m3] is the specific surface area, and VHA [m3/m3], VAir [m3/m3], and VW [m3/m3] are the volume fractions of humic acid, air pore space, and liquid water in the snowpack, respectively. Based on the results of the experiments in combination with the calculations of phase distribution (Figure 1) we propose two threshold values for R (Figure S4). If a melt scenario is characterized by an R value larger than 0.9, i.e., if the chemical is predicted to be present within the meltwater phase to >90%, the release pattern of this chemical is likely to resemble
that of the ions. In such situations the release profile of the organic contaminant from the melting snowpack can simply be approximated by the ion release, which is readily measured as electrical conductivity. With R values smaller than 0.1 the chemical is predicted to be present within the aqueous phase to
