Article pubs.acs.org/est
Assessment of PCDD/F Source Contributions in Baltic Sea Sediment Core Records Anteneh T. Assefa,*,† Mats Tysklind,† Anna Sobek,‡ Kristina L. Sundqvist,§ Paul Geladi,∥ and Karin Wiberg⊥ †
Department of Chemistry, Umeå University, SE-901 87 Umeå, Sweden Department of Applied Environmental Science (ITM), Stockholm University, SE-106 91 Stockholm, Sweden § ÅF AB, Umestan Företagspark, SE-903 47 Umeå, Sweden ∥ Forest Biomaterials and Technology, Swedish University of Agricultural Sciences (SLU), SE-901 83 Umeå, Sweden ⊥ Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences (SLU), SE-750 07, Uppsala, Sweden ‡
S Supporting Information *
ABSTRACT: Spatial and temporal trends of sources of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/ Fs) in the Baltic Sea were evaluated by positive matrix factorization (PMF) and principal component analysis (PCA). Sediment cores were sampled at eight coastal, one coastal reference, and six offshore sites covering the northern to the southern Baltic Sea. The cores, which covered the period 1919− 2010, were sliced into 2−3 cm disks among which 8−11 disks per core (in total 141 disks) were analyzed for all tetra- through octaCDD/Fs. Identification and apportionment of PCDD/F sources was carried out using PMF. Five stable model PCDD/F congener patterns were identified, which could be associated with six historically important source types: (i) atmospheric background deposition (ABD), (ii) use and production of pentachlorophenol (PCP), (iii) use and production of tetra-chlorophenol (TeCP), (iv) high temperature processes (Thermal), (v) hexa-CDD-related sources (HxCDD), and (vi) chlorine-related sources (Chl), all of which were still represented in the surface layers. Overall, the last four decades of the period 1920−2010 have had a substantial influence on the Baltic Sea PCDD/F pollution, with 88 ± 7% of the total amount accumulated during this time. The 1990s was the peak decade for all source types except TeCP, which peaked in the 1980s in the northern Baltic Sea and has still not peaked in the southern part. The combined impact of atmospheric-related emissions (ABD and Thermal) was dominant in the open sea system throughout the study period (1919−2010) and showed a decreasing south to north trend (always >80% in the south and >50% in the north). Accordingly, to further reduce levels of PCDD/Fs in the open Baltic Sea ecosystem, future actions should focus on reducing atmospheric emissions.
■
INTRODUCTION During the last century, there was considerable input of polychlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs) to the Baltic Sea, a polluted semienclosed sea.1 PCDDs and PCDFs (PCDD/Fs) are unintentionally produced in a variety of industrial processes, including metal smelting, waste incineration, and chlorine bleaching.2 The PCDD/Fs are included in the 12 original legacy persistent organic pollutants (POPs) of the Stockholm Convention. The levels of PCDD/Fs in most environmental compartments, including sediments in the Baltic Sea, have decreased considerably since the 1970s in coastal and the 1990s in offshore areas.1,3 The decrease in PCDD/F concentrations in the environment coincided with the introduction of a series of local and international measures and restrictions, which aimed at limiting the environmental pollution and human exposure of PCDD/Fs. Important ambitions and restrictions include bans on chlorophenol use, for example refs 3 and 4, the fifth EU Action Program’s goal to © 2014 American Chemical Society
reduce industrial atmospheric PCDD/F emissions by 90% before 2005,5 the intention of the UNEP Stockholm Convention on Persistent Organic Pollutants,6,7 and the Helsinki Commission (HELCOM) Baltic Sea Action Plan, the latter focusing particularly on the Baltic Sea. Previous studies on source apportionment of PCDD/Fs in the Baltic Sea have largely been restricted to either surface sediments8 or a limited area of the Baltic Sea9 (Gulf of Finland). However, these studies identified and apportioned a number of significant sources of PCDD/Fs, including use and production of tetra- and penta-chlorophenol (TeCP and PCP), chlorine bleaching, chloralkali production and other processes in pulp and paper production, as well as incineration and other Received: Revised: Accepted: Published: 9531
May 12, 2014 July 14, 2014 July 29, 2014 August 8, 2014 dx.doi.org/10.1021/es502352p | Environ. Sci. Technol. 2014, 48, 9531−9539
Environmental Science & Technology
Article
high temperature processes.8,9 Technical formulations based on TeCPs (e.g., Ky-5) and PCP (e.g., Dowicide) were previously used as wood preservatives in saw mills, as well as in pulp and paper industries.9,10 Receptor modeling is a multivariate data analysis technique that can be applied for back-tracing and apportionment of sources of pollutants by using multiple-variable data from receptor sites (e.g., sediments and air).11−14 The basic assumption of receptor modeling is that emissions from a specific source type exhibit a characteristic chemical fingerprint that is a unique profile of the relative abundance of the chemicals being emitted.15 PCDD/F fingerprints are presumed to be conserved in sediment cores due to high persistence in this environment.16−18 Chemical mass balance (CMB) and positive matrix factorization (PMF) are two of the most frequently used receptor model types in environmental forensics in recent times. The main difference between the two methods is that CMB requires prior knowledge about the sources and their emission profile, whereas PMF reconstructs the profiles of potential sources. PMF is a robust technique but requires a large number of receptor samples (usually >100 samples) for reliable results. It is also sensitive to changes of model parameters, mainly the number of factors, which has to be defined prior to modeling. To date, only a few studies have successfully applied PMF for identification and apportionment of sources of PCDD/Fs in sediments.8,19 In spite of decreasing levels of PCDD/Fs in Baltic Sea sediments, there is no clear decreasing trend in Baltic biota.20 Levels in fatty fish occasionally exceed EU limits for food and feed, leading to marketing restrictions of fish, great concern and an intensive search for significant sources.8,21−27 Current sources have been identified and apportioned by analysis and modeling of PCDD/F patterns in surface sediments.8 However, up until now, historically important sources have not been assessed. In the current study, we identified historically important PCDD/F sources and assessed their trends in the open sea sediments and coastal areas of the Baltic Sea by analyzing 15 dated sediment cores covering the last century and applying PMF (receptor-to-source approach) and principal component analysis (PCA) modeling. This study thus provides new knowledge on the relative importance of various PCDD/F sources to the Baltic Sea both over time and space.
Figure 1. Locations of sampling sites: offshore stations (O1−O6) and coastal stations (C1−C9).
cores were selected for the complete analysis. For PCDD/F and TOC analysis, the cores were sliced into 2 or 3 cm thick disks, among which 141 (8−11 disks per core) were selected for analysis. TOC was homogeneous throughout the cores (average 4.0% ± 2.0%). The samples were Soxhlet extracted and the extracts cleaned-up using multilayer silica columns and copper (for sulfur removal) and fractionated using activated carbon. Congener-based (all tetra- through octa-substituted) quantification of PCDD/Fs was made by GC-HRMS using a DB-5 column (60 m × 0.25 mm). The analytical protocol for PCDD/F analysis has been described in Sundqvist et al.28 Dating of the disks was made using primarily 137Cs but also 210 Pb methods. First, the 137Cs activities were measured for all sediment disks and the depth profiles of the cores were examined. To support the interpretation of 137Cs activities, lamination analysis was also used. This approach was successfully used for all stations except C1−C3, for which the 210 Pb method had to be employed by applying a constant rate of supply model.29 Positive Matrix Factorization (PMF). PMF is a factor analysis technique that includes non-negativity constraints so that all factor elements become non-negative.30,31 The nonnegativity constraint is required because neither the concentration of a chemical emitted from a source nor the contribution of a source to a receptor can be 20 km from shore) were selected to cover the open seawater system of the sea along a north to south gradient. Stations C1 and C3−C9 were classified as industrial sites due to their proximity to ongoing and past industrial activities, whereas station C2, which was located in a remote area without known point sources, was classified as a coastal reference site (Table S1, Supporting Information (SI)). Below, a short summary of sampling, samples, quantification of PCDD/Fs, and sediment dating is given. Detailed descriptions are reported in Assefa et al.1 The sampling was performed using a Gemini corer from a sampling ship (S/V Ocean Surveyor; offshore) or a sampling vessel (R/V Sunbeam; coastal). At each site, several cores were taken for use in (i) lamination analysis (photographed with a digital technique) and visual inspection (longitudinal cut), (ii) dating analysis (137Cs), and (iii) PCDD/ F and total organic content (TOC) analysis. Only undisturbed 9532
dx.doi.org/10.1021/es502352p | Environ. Sci. Technol. 2014, 48, 9531−9539
Environmental Science & Technology
Article
Figure 2. Model fingerprints of PMF Factors 1−5, where Factor-1 is the average of two original factors. Data are shown as percentages of total PCDD/F concentration.
modeling, “pre-cleaning” of the data is beneficial. In cases where the concentration was below the detection limit (DL), half of the DL (DL/2) was used as the concentration value. Objects (the individual sediment disks) and variables (the levels of congeners) that had >40% of the data below DL were excluded from the PMF modeling. For identification of outliers based on Hotelling T2 and Q residual statistics, principal component analysis (PCA) was used. After this two-tiered precleaning procedure, 130 objects and 64 variables out of the original 141 objects and 78 variables were modeled using PMF. Prior to PMF, the data were normalized to the total concentration of all congeners to reflect fingerprint patterns. The uncertainty (U) of each data point was calculated as follows:32
model fingerprint of a source. Several factors are usually extracted using PMF. To obtain unique fingerprints for individual factors, a large number of variables are required. In the case of PCDD/Fs, there are >200 congeners that can theoretically be quantified and used for modeling. A total of 78 chromatographic peaks (variables) representing one or several PCDD/F congeners were analyzed in the current study. The solution for eq 1 was calculated by simultaneous iteration to minimize the value of Q (eq 3), which is the residual sum of squares weighted by the uncertainty of the data points (U), as explained below. p
X=
∑ Gi × Fi + E i=1
i. U = 5/6 × DL (if concentration is ≤DL) ii. U =DL + 0.1 × Concentration (if concentration > DL)
(1)
p
E=X−
∑ Gi × Fi i=1
⎡ E ⎤2 Q=⎢ ⎥ ⎣U ⎦
Number of Factors and Model Diagnostics. Choosing the optimal number of factors (source types) is a critical step in PMF modeling. In the current study, the number of factors was chosen to be the same as the number of significant principal components (PC) in the PCA, as this number is a good indicator of the number of unique patterns in the data set. A suitable number of factors is supported if the Q (true) value (eq 4) agrees well with the Q (robust) value of the model.32 The Q
(2)
(3)
where i = 1, 2, 3,..., p sources Evaluation of Measurement Data. Equation 3 is sensitive to outliers and noise in the data. Therefore, prior to PMF 9533
dx.doi.org/10.1021/es502352p | Environ. Sci. Technol. 2014, 48, 9531−9539
Environmental Science & Technology
Article
Table 1. Summary of PCDD/F Factor (Source Type) Characteristicsa factor
source type
markers
typical congeners
examples of typical sources
Factor-1
ABD
HpCDDs OCDD
1234679-HpCDD 1234678-HpCDD OCDD
Factor-1
PCP
1234678-HpCDF 1234689-HpCDF OCDF 1234679-HpCDD 1234678-HpCDD OCDD
Factor-2
Thermal
HpCDDs OCDD HpCDFs OCDF High abundance of PCDFs
Factor-3 Factor-4
TeCP HxCDD
HpCDFs OCDF HxCDD
Factor-5
Chl
TCDF TCDD
1234678-HpCDF 1234689-HpCDF OCDF 123679-HxCDD (other HxCDDs and PeCDDs) 1278-TCDF 2378-TCDF 12378-PeCDF 2378TCDD
air mass from rural and pristine areas, global background air, longrange air transport (LRAT) wood preservatives with pentachlorophenol formulation metal smelters, municipal solid waste incinerators (MSWI), relatively recent air emissions (indicating local and regional impact) wood preservatives with tetrachlorophenol formulation tall-oil distillation and pulp and paper production chloralkali production using graphite electrodes, chlorine gas bleaching in pulp and paper industry
a
ABD: Atmospheric background deposition. PCP: Pentachlorophenol. Thermal: High temperature processes. TeCP: Tetrachlorophenol. HxCDD: HxCDD-related sources. Chl: Chlorine-related source.
Congeners that functioned as markers of the candidate sources are compiled in Table 1, and detailed discussions can be found in Sundqvist et al.8 The optimal number of model patterns (factors) was five. However, on the basis of the previous modeling8 and known emission source history for some of the coastal sites (see below), we suggest that these factors represent six source types. The six source types are atmospheric background deposition (ABD; Factor-1), use and production of pentachlorophenol (PCP; Factor-1), high temperature processes (Thermal; Factor-2), use and production of tetrachlorophenol (TeCP; Factor-3), a hexa-CDD-related source (HxCDD; Factor-4), and chlorine-related sources (Chl; Factor-5). The congener pattern of Factor-1 (here identified as ABD) has previously been suggested to represent “aged” air emissions as it appears in air and atmospheric deposition from rural and pristine areas.8,34 We argue that this factor also represents the PCP source type. As shown in Table 1, both source patterns (ABD and PCP) were dominated by 1234679-HpCDD, 1234678-HpCDD, and OCDD. However, in the case of PCP, congeners 1234678-HpCDF, 1234689HpCDF, and OCDF also contributed considerably.8,35−37 Therefore, interpretation of this factor for the individual station should be made by expert judgment, taking the pollution history and general pollution patterns into account. The congener pattern of Factor-2 (Figure 2) showed a high fraction of PCDFs and only a small contribution of PCDDs. None of the available real source fingerprints showed a complete match with this factor. Emissions of PCDD/Fs from high temperature processes do not always follow a standard pattern since the formation of PCDD/Fs is highly dependent on factors such as process design, temperature, type of fuel, raw materials, etc. However, several studies have shown that the pattern of high temperature processes is usually dominated by PCDFs.38−41 Although the PCDF-dominated patterns reported in literature are mostly based on homologue or 2378-PCDD/F concentrations, they are generally in agreement with Factor-2. Moreover, Sundqvist et al.8 demonstrated that although congener patterns of high temperature processes differ, the isomer patterns within homologue groups are usually conserved. In this aspect, the isomer pattern of Factor-2 (Figure S1, SI) is similar to that of the high temperature processes published in Sundqvist et al.8 Factor-2 was therefore identified as high temperature processes, including metal smelters, incineration and other combustion sources. Detailed discussion of the identification process of Factor-3, -4, and -5 can be found in Sundqvist et al.8 (see also SI). The average
(true) value is calculated using the total number of all components, whereas Q (robust) is calculated using the number of components that have scaled residuals between −4 and 4. A Q (true) value >1.5 times the Q (robust) value indicates a poor fit of the data. Q = nm − p(n + m)
(4)
where n = number of samples, m = number of components and p = number of factors. The PMF model was computed using the EPA PMF version 3 software available via the U.S. Environmental Protection Agency Web site.33 To ensure a good fit to the data and reliable results, variables (congeners) that fell under any of the following criteria were excluded and the model recalculated: (i) categorized as “bad” (built-in criteria based on signal/noise value), (ii) scaled residuals below −4 or above 4, and (iii) r2
