The side streams of Kraft pulp mills could provide an abundant source of sustainable raw materials in the future. Besides the mainly organic side streams, such as wastewater treatment sludge, some inorganic side streams are also generated in the mills in relatively large amounts. A review article on the separation, treatment and utilization of the inorganic side streams has been recently published by the present authors (Kinnarinen et al., 2016). Green liquor dregs, abbreviated as GLD in this paper, are the most important of these side streams, because they are not recycled back to the chemical recovery cycle after solid-liquid separation and washing, but are typically disposed of in the landfill site of the mill. A review by Sanchez (2007) gives a good overview of the green liquor separation equipment, such as clarifiers and dregs filters. The disposal of GLD has to be performed to remove non-process elements from the recovery cycle, i. e. a more closed circuit would cause serious difficulties in the process, for instance in the evaporators (Ulmgren, 1987; Wannenmacher et al., 2005; Tran and Vakkilainen, 2007) and the recovery boiler (Jaretun and Aly, 2000; Taylor, 2006). The amount of dregs produced has been reported to be approximately 4–20 kg/ton pulp in most mills (Nurmesniemi et al., 2005), and only 3–4 kg/ton in some mills having modern recovery boilers (Tikka, 2008). Dregs are composed of a wide variety of elements (oxygen not reported), such as Ca, Na, Mg, S, C, Fe, Mn, K, P, Si, Al, and Cu (Sanchez and Tran, 2005), the main minerals being calcite and calcium sulfate (Martins et al., 2007; Jia et al., 2013; Taylor and McGuffie, 2007; Taylor, 2013).
Green liquor dregs cannot be utilized very well for value-added purposes in the circular economy, which is due to their adverse properties, such as causticity and high concentrations of hazardous trace elements (HTE), such as cadmium, nickel and lead. Causticity is a smaller one of these problems, because it can be reduced by improved washing, as well as mixing GLD with acidic materials. Examples of this include the use of GLD as a neutralizing agent (Pöykiö et al., 2006) or as a sludge composting additive (Zambrano et al., 2010), and in the mining industry for preventing the environmental impacts caused by acidic mine waters (Ragnvaldsson et al., 2014; Mäkitalo et al., 2014, 2015, 2016). The considerable amounts of base cations, such as Ca, K, Mg and Na present in GLD makes them a suitable source of nutrients for the soil (Österås et al., 2015; Greger et al., 1996). Monovalent metals, such as Na and K are present in the form of soluble salts, while Ca and Mg are found in species having limited solubility (Nurmesniemi et al., 2005; Pöykiö et al., 2006; Mahmoudkhani et al., 2004). A major problem regarding the utilization, however, is that the HTE concentrations in the dregs may be tens of times higher than the concentrations allowed by the law when the residue is intended to be used for the production of fertilizers. For instance, the limit for cadmium in Finland is 1.5 mg/kgsolids, but the cadmium concentrations in GLD of Finnish pulp mills, without a lime mud precoat, vary typically between 10–50 mg/kgsolids, according to the unpublished results of a comprehensive analytical work performed in the project funding the present research (see Funding).
The difficulty with the removal of HTE from dregs arises from the fact that these elements are present in mineral phases that are practically insoluble in water (Nurmesniemi et al., 2005; Golmaei et al., 2017; Manskinen et al., 2011). Therefore, mechanical separation on the basis of size and density of the GLD particles could be a useful way to reduce the HTE levels in GLD. In this case, the fraction where HTE are concentrated has to be disposed of as well, and the cleaner part of the stream can be considered as utilizable material. However, depending on the separation efficiency obtained in the mechanical treatment stage, further removal of HTE e. g. by chemical treatment methods may be required, which will decrease the profitability of the treatment.
The aim of this study was to remove HTE from green liquor dregs, or in other words, to concentrate them in a smaller stream, by using three mechanical separation methods: dry sieving, gravitational settling, and centrifugation. All these separation methods were tested by using different sample preparation procedures. The experiments were performed only with the green liquor clarifier underflow samples, i. e. lime mud precoat was not present in the suspensions. All the samples used in the study were obtained from two large Finnish pulp mills.
Origin, use and properties of the GLD slurry samples.
|Mill||Experimental procedures (Figure 2)||Density (kg/m3)||Total solids cont. (wt.%)||Dissolved solids cont. (wt.%)||D50 (µm)|
|1||A, B, F, G||1230||26.0||21.2||17.9|
|2||C, D, E||1180||21.7||18.6||40.4|
* The sample was ground by using a stirred media mill: see particle size distribution (Figure 1) and description below
Initial concentrations of the target metals Cd, Pb and Zn in the suspended solids content of the slurry samples.
|Mill||cCd (mg/kgD.S.)||cPb (mg/kgD.S.)||cZn (mg/kgD.S.)|
Materials and methods
The slurry samples for the research were taken from two Finnish pulp mills. The sampling point in both cases was the underflow pipe of the green liquor clarifier. Table 1 summarizes the most relevant properties of the samples regarding the experimental study. In addition to the slurry properties. Table 1 shows also in which experiments each slurry was used, referring to symbols A–G in Figure 2.
The elemental compositions of the solid and liquid phases of the slurries are not presented in this section in closer detail. However, the initial concentrations of the three target metals are presented in Table 2 to illustrate the amounts of these metals and the differences between the slurries. A previous study by the authors (Golmaei et al., 2017) gives a good notion about the distribution of the elements between the solids and the liquid. To summarize this briefly, it can be stated that Al, Ba, Ca, Cd, Cu, Fe, Mg, Mn, and Zn are present almost exclusively in the suspended solids. Highly soluble elements in the alkaline liquor are B, K, Na, and S. According to Golmaei et al. (2017), the distribution of phosphorus between the solid and liquid phase seems to depend on the mill where the slurry sample is taken from.
The particle size distributions of suspended solids in the slurries are presented in Figure 1. In addition to the particle size distributions of the non-treated slurry samples, the size distribution for the solids after stirred media milling (Figure 2, procedure G, and Table 3) is shown as well.
Conditions in the stirred media milling in procedure G (Figure 2).
|mGLD (g)||mH2O (g)||mbeads (g)||dbeads (mm)||cbeads (wt.% of slurry)||Grinding speed (rpm)||Grinding time (h)|
The experimental procedures for the three separation techniques including dry sieving, settling, and centrifugation, are illustrated in Figure 2. The aim of the experiments was to classify the solids in two or more fractions and to get the hazardous trace elements concentrated in a small solid fraction to avoid loss of utilizable material.
Dry sieving experiments were carried out according to two different procedures (A, B), either with the cake with solubles or with the cake from which the readily washable solubles had been removed by washing with water, using a volumetric wash ratio of approximately 20. A Retsch vibratory sieve system and sieves manufactured by Retsch were used, the sieving time being 3 h in both cases. The aperture sizes of the sieves were the same (45–125 µm) in both experiments. The aim of carrying out the sieving in two different ways was to find out whether the solubles present in the mother liquor have an influence on the separation result obtainable by dry sieving.
The first series of settling experiments was performed in the mother liquor in a graduated glass cylinder, using a constant volume of slurry (500 cm3), the weight of which was 610 g. The settling experiments were performed in this case with the received GLD slurry without preliminary treatment (procedure C in Figure 2).
The second series of settling experiments was performed after separating the solids from the mother liquor by centrifugation until the supernatant was clear, and forming a new batch of slurry from the cake (procedure D). In these experiments, the slurry (
The objective of using the different procedures C and D was to investigate the influence of slurry composition, and especially the dissolved solids content of the liquid phase on the classification by settling. The settling time was constant 1 h in both settling experiments. Sampling in the end of settling was performed as follows: a certain volume or mass fraction of the slurry, > 30 % in all cases (Figure 2), was sucked out from the surface of the suspension with the help of a large syringe. Separation of the suspended solids from the liquid was then implemented by using a Jouan GT 422 bucket centrifuge as described in Figure 2 (points C, D).
Centrifugation was also used as one classification method, utilizing the above-mentioned bucket centrifuge and plastic test tubes. The main difference in the resulting solid bed between settling and centrifugation was that settling resulted in the formation of a loose solid bed with a slurry-like structure by the end of the experiment, whereas centrifugation resulted in the formation of a packed bed of particles, which could be treated like solid material and scraped off mechanically, layer by layer, in order to divide the bed into fractions of equal mass (see the Sampling steps in Figure 2 E–G). The rotational speed of the centrifuge and the centrifugation time were constant 2500 rpm and 120 s, respectively. The radius of the centrifuge rotor was 172 mm, so the maximum driving force of sedimentation, when the centrifuge was rotating at full speed, was 1200 G.
Figure 3 illustrates the principle of the centrifugation experiments and sampling from the sediments. The starting material in case E was obtained from Mill 2, and the material in cases F and G was obtained from Mill 1. In case E, the slurries for the centrifugation experiments were prepared by dispersing the wet centrifuge cake in deionized water. The resulting total solids concentrations were 10, 7, and 4 wt.%. The slurry in case F was prepared from an oven-dry filter cake which had been crushed carefully prior to forming the slurries for centrifugation. When the dry filter cake was used, the slurries were let to stabilize for 2 h under intermittent manual mixing to obtain appropriate wetting of the particles before starting the centrifugation.
The idea of performing centrifugation experiments by using these three procedures was to investigate the effect of total solids concentration (4–20 wt.%) and particle size distribution of the solids (D50 = 17.9 µm (F) or 1.9 µm (G)). Reduction of particle size (Figure 1) by stirred media milling was carried out to enhance the liberation of the minerals carrying hazardous trace elements, aiming thus at improvement of their separation efficiency. Additionally, the sampling from the sediment layer was performed in different ways, dividing the sediment into 2-5 layers, depending mainly on the height of the layer, which determined the availability of a sample for analysis.
Details of the particle size reduction by stirred media grinding are summarized in Table 3. The dimensions of the grinding vessel and the applied pin stirrer are presented in Kinnarinen et al. (2015). The stress energy of the grinding media, calculated according to Stender et al. (2004), was approximately 50 µJ.
The particle size distributions of the suspended solids in the samples were measured with a Malvern Mastersizer 3000 laser diffraction particle size analyzer equipped with a Hydro EV wet dispersion unit. The measurements were performed without ultrasonication, using a stirrer speed of 3000 rpm, and after five or more parallel measurements, the data was averaged to form the average volumetric distributions presented in Figure 1.
In the centrifugation experiments, each test was performed twice, to have a large enough sample for the elemental analyses and for the determination of the solid contents. The solid contents were measured in all cases by drying a pre-weighed sample in an oven: samples with high suspended solids contents were dried at 105 °C, and samples containing mainly dissolved solids were dried at 180 °C until the weight became constant.
The concentrations of the elements in all dregs samples were measured with a Fisher Scientific ICAP6500 Duo inductively coupled plasma optical emission spectrometer (ICP-OES). The method 3051 by US EPA was followed in the acid digestion stage of sample preparation.
Results and discussion
The results of this research have been divided into three sub-sections according to the separation technique: sieving, settling and centrifugation. Cadmium, lead and zinc have been selected as the target elements, the removal of which is evaluated. The results are presented with respect to the mass balances, so the absolute concentrations of the elements are not discussed in closer detail.
Sieving experiments were performed by using GLD sampled from the underflow of the green liquor clarifier at Mill 1. Washing the solubles off prior to drying, crushing and sieving was carried out, in addition to performing a test with a non-washed sample. Figure 4 summarizes the mass-based distributions of the particles and the distributions of target elements between the sieves.
The deposition of solids on the sieves with aperture sizes of 45, 50, 90, and 125 µm is presented in Figure 4a, showing a comparison between the washed and non-washed dregs. Independent of the washing, the two finest fractions (< 50 µm) form approximately two thirds of the total weight of the dregs. The proportion of the largest particle size class (> 125 µm) was considerably higher in the case of the washed dregs.
The distribution of the target elements was evaluated with relation to the mass distribution of the solids, i. e. the solids fraction in each size class (Figure 4b, c). According to the results presented in Figure 4b, the target metals appeared to be present slightly more abundantly in the two finest fractions when the non-washed dregs were sieved. When the dregs were washed prior to sieving, the elemental distribution was observed to change slightly, showing increased relative deposition of the target elements on the three coarsest sieves (Figure 4c). It is difficult to find the reasons behind the different separation properties in the dry sieving experiments, due to the high number of factors which are not readily measurable, including for instance electric interactions between the particle-particle and sieve-particle interfaces when the vibration of the sieve stack is on, and possible blocking of the sieve apertures caused by agglomeration of fine particles.
Generally speaking, the separation efficiency of dry sieving was not sufficient. This result could imply that the target metals cannot be concentrated unless the separation of the finest particles in the size class of some microns only is efficient.
The results of the settling experiments performed in a graduated glass cylinder can be evaluated on the basis of mass balance of the elements as shown in Figure 5. It can be seen in Figure 5 that the recoveries of metals in the overflow fractions were not high: it was possible to remove less than 18 % of Cd when 30 wt.% of the total batch weight was removed as the overflow stream. This poor result was due to the low separation efficiency: the Cd concentration in the dry solids was not any higher in the overflow, being 23 mg/kg in both the overflow and underflow. The reason why the recovery with the overflow was lower, compared to the underflow, was naturally the lower suspended solids content of the overflow. The same rule applied to all the investigated elements, although the results for phosphorus were less consistent with the highest two overflow fractions.
In comparison with settling in the mother liquor (Figure 5a), the settling results were even slightly improved when the settling experiments were carried out by using water as the liquid phase of the suspension (Figure 5b, c).
Experimental conditions, properties of the sediment bed and the fractions downwards from the top of the sediment. S and h represent the total solids content and height, respectively.
|Procedure||Sslurry (wt.%)||hbed (mm)||No. of fractions (-)||Sbed (wt.%)||Sfraction (wt.%)|
|F||7||33||3||42.2||32.5; 45.5; 47.5|
|F||20||61||5||48.4||24.6; 42.2; 54.1; 60.3; 60.9|
|G||4||13||3||47.6||32.0; 50.0; 60.9|
|G||7||18||3||40.1||32.1; 43.9; 44.2|
|G||10||23||3||43.1||38.8; 44.4; 46.2|
As a conclusion of the settling tests, it can be stated that the separation efficiency was poor and the technique, as simple as it would be to implement in the existing settling facilities at mills, cannot be recommended for the purification of GLD.
The centrifugation experiments were performed in different ways (Table 4) to have a diverse set of data. First, the sediment obtained by centrifugation was divided into two fractions of equal mass, the surface and the bottom. The next step was to divide the sediment to 3-5 fractions, and the last stage involved grinding the sample to reduce the particle size for improving the mineral liberation and division of the sediment into 3 fractions of equal mass. The experiments were carried out by varying the solids concentrations. The target elements in this part of the study were Cd, Pb, and Zn.
Division of the sediment into two fractions
The centrifugation experiments targeting at concentrating the target elements in either the surface or bottom fractions showed promising results, as can be observed in Figure 6. The experiments were performed at solids contents of 4, 7, and 10 wt. of the suspension. According to the results presented in Figure 6, the separation performance for the target metals was highest when dilute suspensions were used. By removing 50 % of the slurry mass, is was possible to remove 2/3 of the hazardous trace metals content of the slurry.
The suspended solids contents of green liquor clarifier underflow slurries are typically below 5 wt.% in Finnish pulp mills, so this is a good indication about the possibility to perform industrial-scale separation of HTE without making any changes in the upstream processes.
Division of the sediment into 3 or 5 fractions
The experiments where the sediment was divided into 3 or 5 fractions of equal mass showed also good separation results (Figure 7). High solids concentrations were used in these experiments, because it was easier to have a big enough sample for the analyses when the thickness of the sediment bed was high. However, the sharpness of separation became weaker with increasing concentration of solids. When the sediment was divided into 3 or 5 fractions, the results showed clearly that the surface fraction contained less solids, and the bottom fraction was the most concentrated with solids, which resulted from the compaction of the sediment bed. In other words, the suspended solids contents were increased towards the bottom of the sediment, and correspondingly the porosity of the bed increased towards the surface, as is usual in the case of sediments. Dividing the sediment formed from the 7 wt.% slurry into three equal mass fractions showed that the solids contents of the middle and the bottom fractions did not differ from each other much. Correspondingly, the solids content was almost constant in the two fractions closest to the bottom of the test tube when the solids content of the suspension was 20 wt.%.
The distribution of the target elements between the fractions changed clearly depending on the fraction, decreasing towards the bottom of the sediment. This result means that the target elements were the most concentrated in the surface fraction, according to the lower density and/or finer particle size of the particles carrying these elements. Unfortunately, the concentrations of the target elements were too low for the investigation of their exact mineral composition and the mineral phases where they were bound to.
Reduction of particle size by grinding and dividing the sediment into 3 fractions
Due to the promising findings in the previous centrifugation experiments, the solids were ground to a finer particle size (Figure 1), targeting at a better separation efficiency. The results are summarized below in Figure 8, 9. As the relative distributions of the target metals in the surface, middle and bottom fractions illustrate, the separation was less efficient, compared to the non-ground GLD in Figure 7. In this case, not even the use of the lowest solids content of 4 wt.% resulted in better separation performance than that obtained with the 7 wt.% slurry in Figure 7.
Another important finding was that the distribution of the target metals became more uniform as a result of particle size reduction: the largest differences between the mass fractions of Cd, Pb and Zn were only approximately 1 wt.%, which can be seen best in Figure 8. This was probably due to the better homogenization of the suspended solids as the particle size was reduced and the particle size distribution narrowed.
Figure 9 presents the distribution of total solids and the target metals in the form of curves, from a different point of view, as a function of the total solids content of the slurry. In order to understand what the results presented in Figure 9 mean, the following two remarks are made: 1) a large distance between the solids fraction curve and the Cd, Pb and Zn fraction curves indicates good separation performance, and 2) the increasing fraction of solids and elements in the surface fraction is caused by the more uniform solids content of the sediment layer as the solids content of the slurry is increased.
There are a few possible reasons why particle size reduction did not improve the separation performance in this research. The first and perhaps the most likely factor was the shape of the particle size distribution of the ground solids (Figure 1). Additionally, the settling of these fine particles may have been hindered, and particle aggregates with greater sizes may have formed. Migration of the fine particles in the sediment bed is also basically possible, but seemingly not the most important reason for the decreased efficiency of separation.
The removal of hazardous trace elements of green liquor dregs by using three mechanical separation methods, namely sieving, settling, and centrifugation, was investigated in this experimental study. The results implied that the separation efficiency that could be reached by dry sieving and settling in either mother liquor or water, was not enough for the removal of these elements. On the other hand, centrifugation turned out to be a useful technique for the purpose of concentrating Cd, Pb and Zn in the surface fractions of the sediment layer. The results of the centrifugation experiments showed that these target metals were consistently less concentrated in the bottom of the sediment, which means that this fraction could be most potentially considered for utilization. Increasing the solids concentration had a negative influence on the separation performance by centrifugation, which may not be a dramatic problem in plant scale, because the green liquor clarifier underflows contain typically contain less than 5 wt.% (< 2 vol.-%) of suspended solids. Reduction of the particle size prior to centrifugation did not result in improved separation, which was possibly due to the narrowed particle size distribution and increased co-sedimentation of the particles. However, the particle size reduction made the differences between the target metals very small, implying that some other separation systems, such as grinding circuits utilizing hydrocyclones, could have a greater separation efficiency for all the target metals. Continuously operating centrifugation equipment, such as decanter centrifuges, could also have potential in the application.
Golmaei, M., Kinnarinen, T., Jernström, E., Häkkinen, A. (2017) Study on the filtration characteristics of green liquor dregs. Chem. Eng. J. 317:471–480.
Greger, M., Hamza, K., Arvanitis, L. (1996): Mixed waste products from pulp and paper industry used as fertilisers in forest, AFR-Rep. Swed., 1996.
Jaretun, A., Aly, G. (2000) Removal of chloride and potassium from Kraft chemical recovery cycles. Sep. Sci. Technol. 35(3):421–438.
Jia, Y., Stenman, D., Mäkitalo, M., Maurice, C., Öhlander, B. (2013) Use of amended tailings as mine waste cover. Waste and Biomass Valorization 4:709–718.
Kinnarinen, T., Golmaei, M., Jernström, E., Häkkinen, A. (2016) Separation, treatment and utilization of inorganic residues of chemical pulp mills. J. Clean. Prod. 133:953–964.
Kinnarinen, T., Tuunila, R., Huhtanen, M., Häkkinen, A., Kejik, P., Sverak, T. (2015) Wet grinding of CaCO3 with a stirred media mill: Influence of obtained particle size distributions on pressure filtration properties. Powder Technol. 273: 54–61.
Mahmoudkhani, M., Richards, T., Theliander, H. (2004) Recycling of solid residues to the forest: experimental and theoretical study of the release of sodium from lime mud and green liquor dregs aggregates. Process Saf. Environ. Prot. 82: 230–237.
Manskinen, K., Nurmesniemi, H., Pöykiö, R. (2011) Total and extractable non-process elements in green liquor dregs from the chemical recovery circuit of a semi-chemical pulp mill. Chem. Eng. J. 166:954–961.
Martins, F.M., Martins, J.M., Ferracin, L.C., da Cunha, C.J. (2007) Mineral phases of green liquor dregs, slaker grits, lime mud and wood ash of a Kraft pulp and paper mill. J. Hazard. Mater. 147:610–617.
Mäkitalo, M., Lu, J., Maurice, C., Öhlander, B. (2016) Prediction of the long-term performance of green liquor dregs as a sealing layer to prevent the formation of acid mine drainage. J. Environ. Chem. Eng. 4:2121–2127.
Mäkitalo, M., Maurice, C., Jia, Y., Öhlander, B. (2014) Characterization of green liquor dregs, potentially useful for prevention of the formation of acid rock drainage. Minerals 4:330–344.
Mäkitalo, M., Mácsik, J., Maurice, C., Öhlander, B. (2015) Improving properties of sealing layers made of till by adding green liquor dregs to reduce oxidation of sulfidic mine waste. Geotech. Geolog. Eng. 33:1047–1054.
Nurmesniemi, H., Pöykiö, R., Perämäki, P., Kuokkanen, T. (2005) The use of a sequential leaching procedure for heavy metal fractionation in green liquor dregs from a causticizing process at a pulp mill. Chemosphere 61:1475–1484.
Pöykiö, R., Nurmesniemi, H., Kuokkanen, T., Perämäki, P. (2006) Green liquor dregs as an alternative neutralizing agent at a pulp mill. Environ. Chem. Lett. 4:37–40.
Ragnvaldsson, D., Bergknut, M., Lewis, J., Drotz, S., Lundkvist, A., Abrahamsson, K., Fernerud, S. (2014) A novel method for reducing acid mine drainage using green liquor dregs. Environ. Chem. Lett. 12:443–447.
Sanchez, D.R. (2007) Recausticizing - Principles and practice, Vector Process Equip. Inc. 2.1 2007, 1–30.
Sanchez, D., Tran, H. (2005) Treatment of lime slaker grit and green liquor dregs – current practice. In: TAPPI PRESS, Philadelphia. 2005.
Stender, H.-H., Kwade, A., Schwedes, J. (2004) Stress energy distribution in different stirred media mill geometries. Int. J. Miner. Process. 74S:S103–S117.
Taylor, K. Effect of non-process elements on kraft mill efficiency in: Proc. PAPTAC Western Branch Technical Mini-Conference, Prince George, BC, September 27–29, 2006 (2006).
Taylor, K. (2013) Detailed characterization of poor settling green liquor dregs. Tappi J. 12:29–36.
Taylor, K., McGuffie, B. (2007) Investigation of non-process element chemistry at Elk Falls Mill green liquor clarifier and lime cycle. Pulp Pap. Can. Ont. 108:27–32.
Tikka, P. (2008) Chemical Pulping Part 2, Recovery of Chemicals and Energy. second ed. Paperi ja Puu Oy, Jyväskylä, Finland.
Tran, H., Vakkilainen, E.K. (2007) Advances in the Kraft chemical recovery process. In: Source 3rd ICEP international colloquium on eucalyptus pulp, Belo Horizonte, Brazil, 2007. pp. 4–7.
Ulmgren, P. (1987) The removal of aluminium from recovery system of closed kraft pulp mill. Nord. Pulp Pap. Res. J. 2:4–9.
Wannenmacher, N., Kimura, S., Frederick, W.J. (2005) Solubility of aluminosilicates in kraft green and white liquors. Ind. Eng. Chem. Res. 44:9316–9322.
Zambrano, M., Pichún, C., Alvear, M., Villarroel, M., Velásquez, I., Baeza, J., Vidal, G. (2010) Green liquor dregs effect on Kraft mill secondary sludge composting. Bioresour. Technol. 101:1028–1035.
Österås, A., Sunnerdahl, I., Greger, M., (2005) The impact of wood ash and green liquor dregs application on Ca, Cu, Zn and Cd contents in bark and wood of Norway spruce. Water Air Soil Pollut. 166:17–29.