Validation of a simple and robust multi-residue gas chromatography-mass spectrometry method for the analysis of polycyclic aromatic hydrocarbons, phthalates and biocides in rooﬁng material leachate and roof runoﬀ

A multi-residue method for the determination of 16 polycyclic aromatic hydrocarbons (PAHs), biocides diuron (DIU), octylisothiazolinone (OIT) and mecoprop (MCPP), and phthalates bis(2-ethylhexyl) phthalate (DEHP) and diisodecyl phthalate (DIDP) in material leachate and roof runoﬀ is presented. The method aims to keep sample pretreatment as simple as possible, not only to minimize sample contamination and sample losses, but also to have an environmental friendly and cost eﬃcient method with low solvent consumption, in line with the principles of Green Analytical Chemistry. Solid phase extraction (SPE, C 18 ) in combination with GC-MS was used for this purpose. A good separation and accurate detection for the majority of included analytes was obtained, resulting in overall low detection limits ranging between 0.1 ng/L and 18 μg/L in aqueous matrices. Furthermore, the addition of a surrogate standard containing three deuterated PAHs prior to SPE extraction was found to increase the method’s repeatability. For the higher molecular weighted PAHs and phthalates, recovery was aﬀected by the concentration level as higher recoveries were observed at lower spiking levels. Furthermore, also the matrix was found to aﬀect the recovery of these higher molecular weighted components as the recoveries in roof runoﬀ and material leachate were higher in comparison to the ones found in bidest water. Water solubility and the aﬃnity to adsorb onto particles were hypothesized to play an important role. The multi-residue SPE-GC-MS method was found to be suitable for the quick and reliable analysis of material leachates and roof runoﬀ samples.


Introduction
Urban stormwater can be contaminated by a wide range of pollutants including inorganic but also numerous organic pollutants such as polycyclic aromatic hydrocarbons (PAHs), pesticides, phthalates, alkylphenols, polybrominated diphenyl ethers (PBDEs) and polychlorinated biphenyls (PCBs).A significant contributor to this contamination is roof runoff [ 1 , 2 ], which conveys pollutants as a result of material leaching and atmospheric deposition.Since information on the leaching behavior of many roofing materials is scarce, further research on this topic is essential to estimate the effect of roofing materials on runoff quality and to determine the implications for rainwater harvesting and utilization.A common strategy is the use of laboratory leaching experiments [3][4][5] .However, such tests are not directly representative for the outdoor leaching behavior, hence the collection of roof runoff from real constructions remains of key importance [3] .
As numerous organic pollutants were found in urban stormwater, a selection of model components can be made based on roofing material compositions (e.g. the use of additives such as phthalates and biocides) and/or their expected occurrence in deposition.Examples of these model components include (1) polycyclic aromatic hydrocarbons (PAHs), which are present in bituminous materials, found in deposition and can be present in other industrial substances, (2) phthalates such as diisodecyl phthalate (DIDP) and bis(2-ethylhexyl) phthalate (DEHP), which are used as plasticizers and (3) biocides, such as diuron (DIU), mecoprop (MCPP) and octylisothiazolinone (OIT) which are used as additives to material formulations in order to prevent the growth of algae, lichens and to prevent root penetration of the materials.https://doi.org/10.1016/j.jcoa.2021.100007

Table 1
Overview of the applicable principles of Green Analytical Chemistry as defined by Ga ł uszka et al. [6] on the presented SPE-GC-MS method in this work.
Principles of Green Analytical Chemistry as defined by Ga ł uszka et al. [6] Principle applied in the developed and validated method?How the principle is applied in the developed method Direct analytical techniques should be applied to avoid sample treatment

No
The direct analysis and quantification of the selected components in water samples is not possible.However, if insights on material leaching and its effect on roof runoff become clear, it might be possible to replace part of the analytical methods by surrogate measurements (e.g. chemical oxygen demand, UVA 254 , fluorescence measurements, etc.) or modelling.

2
Minimal sample size and minimal number of samples are goals Yes Due to the integration of different analytical components in a multi-residue method, the required sample volume can be lowered.This mainly as a consequence of the fact that different analytical techniques require different sample pretreatments.This results in less sample volume to be collected, transported and stored.Furthermore, care was taken to analyze samples as soon as possible, without the need of long storage times in fridges or freezers.

3
In situ measurements should be performed No See explanation principle 1 4 Integration of analytical processes and operations saves energy and use of reagents Yes Due to the use of a multi-residue method, the same sample pretreatment and same instrumental analysis can be used, resulting in an efficient use of reagents, solvents and machine run-time.

5
Automated an miniaturized methods should be selected Yes SPE, which was used here as sample pretreatment can be automated by the use of a sample preparation robot (fully automated) or vacuum manifold (semi-automated).6 Derivatization should be avoided Yes The presented method does not involve a derivatization step.7 Generation of large volume of analytical waste should be avoided and proper management of analytical waste should be provided Yes Due the low solvent consumption and low sample volume, the total analytical waste was limited.Furthermore, as a vacuum manifold was used, the conditioning solvents and first loading water could be separately collected, resulting in lower organic waste volumes.
No organic modifier was used.8 Multi-analyte or multi-parameter methods are preferred versus methods using only one analyte at a time

Yes
In the presented methodology, the EPA 16 PAHs, two phthalates and three biocides are integrated in a multi-residue method.

9
The use of energy should be minimized Yes By using the same sample pretreatment and same instrumental analysis, the energy use was minimized.10 Reagents obtained from renewable source should be preferred

No
No special attention was given on the origin of the used solvents and reagents.Further improvements of the methodology could focus on this aspect 11 Toxic reagents should be eliminated or replaced No For SPE conditioning, methanol and dichloromethane were used.However in comparison to other methods, the solvent volume used for each sample was limited.Nevertheless, further improvements of the methodology could focus on this aspect.

12
The safety of the operator should be increased Yes As no evaporation step was included in the methodology, the losses of volatile organic compounds (VOCs) and solvents during sample pretreatment was limited.In order to minimize the exposure of the operator, all sample pretreatment steps were done in a fume hood.Furthermore, also the handling of low solvent volumes and the cooling of the solvents contributed to the minimization of the exposure of the operator to VOCs and solvent fumes.
Currently, the applied analytical techniques used for the identification and quantification of the previously mentioned model components are often not in line with the principles of Green Analytical Chemistry (GAC) [6] ( Table 1 ).For example, the methods used to measure the contaminants of interest in aqueous matrices are not integrating the different component groups in a single analytical methodology.This not only results in more analytical run-time but also in more sample pretreatment, which results in higher costs and higher environmental impact.Furthermore, this fragmented approach also requires more sample volume, which might be a limiting factor especially during laboratory leaching tests.
In addition, both under laboratory and outdoor conditions, the number of samples can rise quickly as numerous factors are involved such as different roofing materials, temperature and rainfall conditions, first flush effects etc.As such, the development of a robust, reliable, simple and fast analytical methodology in line with the principles of GAC, is favorable for the analysis of organic pollutants in rainwater, roof runoff and material leachate.
Finally, such methodology should be capable to measure accurately in a wide concentration range.Typical concentrations of individual PAHs and biocides in roof runoff/rainwater are in the lower ng/L range ( < 100 ng/L), but might be as high as several μg/L under certain conditions.Likewise, for phthalates DEHP and DIDP, the observed concentrations in runoff range from below the detection limits (e.g.< 100 ng/L) up to almost 100 μg/L ( Table 4 ).The levels in laboratory leachate samples might be even higher than under outdoor conditions, which further supports the requirement of a wide measurement range.
A first step in the development of an analytical methodology, is the selection of an instrument.GC-MS has several advantages such as its wide availability in many laboratories, its affordability, the possibilities to identify non-targeted organic contaminants when operated in full scan mode, its good sensitivity, selectivity and reproducibility.In addition, for all of the model components, an analytical method was found that uses GC-MS for the identification and quantification (PAHs: [7] , MCPP: [8] , DIU: [9] , OIT: [10] , DEHP and DIDP: [11] ). Hence, GC-MS seems a suitable technique for the multi-residue analysis of the components of interest which is in line with the principles of GAC [6] ( Table 1 ) Next, a sample pretreatment should be chosen to extract and concentrate the components of interest.Although liquid-liquid extraction (LLE) was found to be used for the analysis of several of the components of interest (see appendix A), the large solvent volumes required are not in line with the principles of GAC and the technique was therefore disregarded.Greener alternatives include sorption based sample pretreatments such as solid-phase extraction (SPE), solid-phase microextractions (SPME), stir bar sorptive extractions (SBSE) and fabric phase sorptive extractions (FPSE) [ 12 , 13 ].In this study, solid phase extraction (SPE) was chosen as pretreatment step.It has several advantages including simplicity, robustness, rapidity, a relatively low solvent consumption, a high recovery for many organic contaminants and the ability to obtain high preconcentration factors as the analytes can be extracted exhaustively [ 7 , 14 ].Furthermore, SPE using an octadecyl modified (C 18 ) silica phase is recommended for the analysis of a wide range of organic contaminants in water under US EPA method 525.1/525.2conditions [15] .
Several analytical methods are also using a derivatization step in order to transform the analytes to more volatile derivates or add an organic modifier (e.g.methanol) to aqueous samples prior to SPE.The latter has several benefits.For example, the addition of an organic modifier can minimize the deactivation of the SPE cartridge [16][17][18] and can increase the recovery of the less soluble analytes as sorption onto the used SPE equipment can be minimized [ 17 , 19 , 20 ].A disadvantage of organic modifiers is the fact that the eluotropic strength of the sample increases, which results in smaller breakthrough volumes and which in turn can lower the recovery of lower molecular weighted/more soluble analytes, such as 2-and 3-ring PAHs [ 16 , 18 ].In literature, no real consensus was found on the addition of organic modifiers: some authors reported higher recoveries for all included analytes [ 16 , 21 , 22 ], whereas other studies also described negligible [ 17 , 19 , 20 ] or negative effects such as a lower recovery for other components and higher relative standard deviations [18] .Therefore, it was decided not to use an organic modifier in this study.This is also more compatible with the GAC philosophy to use an environmentally friendly analytical sample pretreatment with a low solvent consumption [6] .
In summary, a new multi-residue method for the simultaneous quantification of three biocides, 16 PAHs, and two phthalates in roof runoff and roofing material leachate is presented here.The method aimed to keep sample pretreatment as simple as possible, without the addition of an organic modifier and without the use of a derivatization and evaporation step.This in order to minimize sample contamination, to avoid sample losses and to have a low solvent consumption.Furthermore, the developed methodology implemented as much as possible principles of GAC as defined by Ga ł uszka et al. [6] ( Table 1 ).The method was successfully optimized, validated and applied on real water samples.
From the purchased standards, stock solutions were prepared in DCM for the instrument (GC-MS) optimization and validation.The concentration of these stock solutions was 200 mg/L for PAHs, and 2 g/L for the other components.Multi-analyte working solutions in DCM at lower concentrations were prepared from these stock solutions on a monthly basis.
To validate the entire SPE-GC-MS method, aqueous solutions of the selected components were prepared by spiking the components in bidest water.For the spiking of biocides and phthalates, a mixed stock solution in MeOH was prepared (DIDP: 400 mg/L, other components: 40 mg/L).For PAH spiking, a stock solution of 1 mg/L in acetonitrile was prepared, starting from a certified PAH mix standard in acetonitrile (PAH Calibration mix, Sulpelco, Sigma Aldrich, 10 μg/mL).This certified reference standard also contained the sixteenth US EPA PAH benzo[ b ]fluoranthene (BbF).Furthermore, an internal standard (EPA 525, 525.1 Internal Standard Mix, Sigma-Aldrich) was purchased and used as surrogate to correct for analytical and sample pretreatment fluctuations.This surrogate standard contained a mixture of the deuterated PAHs acenaphthene-d 10 (Ace-d 10 ), chrysene-d 12 (Chry-d 12 ) and phenanthrene-d 10 (Phe-d 10 ).
All standards and dilutions were stored at 4°C in brown bottles to prevent photodegradation of the components.
Ammonia and formic acid, used for glassware cleaning, were supplied by Chem-Lab.Ammonia was of a technical grade (25 weight % solution) and formic acid had a purity of more than 98%.
Anhydrous sodium sulfate, used to dehydrate the SPE extracts, was purchased from Sigma-Aldrich.

Solid-phase extraction procedure
As SPE sorbent, an octadecyl modified (C 18 ) silica phase was used as recommended under US EPA method 525.1/525.2[15] and as used in many other studies ( table 3 and table A.1,appendix A).More specifically, Chromabond C 18 ec (6 mL/ 500 mg, Macherey-Nagel) cartridges were placed on a vacuum SPE manifold (Vac Elut, Agilent) which was connected to a vacuum pump (Laboport N816.3 KT.18, KNF).The cartridges were conditioned with 4 mL DCM, 8 mL of methanol and 4 mL of bidest water.Next, 400 mL of aqueous standard solution or sample was loaded on the columns at a flow rate of 10 mL/min.Care was taken to avoid cartridge dry out during the conditioning and loading phase.After sample loading, the cartridge was rinsed with 10 mL bidest water and subsequently dried by keeping the vacuum on for 20 minutes.The retained components on the cartridges were eluted with 4 mL DCM.The effluents were collected in vials and dehydrated by anhydrous sodium sulfate.

GC-MS analysis
An Agilent 6890 gas chromatograph (GC) coupled to a Hewlett Packard 5973 mass spectrometer (MS) detector was used for the chromatographic separation and analysis of the analytes.The GC was equipped with a fused silica HP-5ms column (30m x 0.25mm x 0.25 m, Agilent).Helium (Air Products, He BIP®, > 99.9999%) was used as carrier gas, at a flow rate of 1.0 mL/min.For sample injection, an automated injector equipped with a 10 μL syringe was used.A sample volume of 1 μL was injected in splitless mode at 280°C.Between injections, the syringe and needle were automatically rinsed with DCM.A purge gas flow of 25 mL/min (He), followed by a gas saver flow of 10 mL/min was applied to avoid any carry-over effect.
The GC oven program was adapted from Zhang et al. [23] .After optimization, the following temperature program was obtained: an initial temperature of 40°C held constant for 2 min, followed by a first ramp of 12°C/min to 250°C, held constant for 1 min, and followed by a second ramp to 310°C at a rate of 5°C/min.This final temperature was held constant for another 5 min.
The temperature of the MS transfer line was set at 325°C.The MS was operated in electron impact (EI) mode (70eV) with an ion source temperature of 230°C and an MS quadrupole temperature of 150°C.A solvent delay of 6 minutes was implemented in order to avoid filament damage.For initial component identification, the MS was operated in full SCAN mode, scanning from 40 to 500 amu.For component quantitation, selective ion monitoring (SIM) with one target (T) and two qualifier ions (Q 1 and Q 2 ), was used (see also appendix C, Table C.1).The total time of one GC-MS run was 37min 30sec.
The obtained chromatograms were processed by the MSD ChemStation software (Agilent).

Inter-and intraday repeatability
The inter-and intraday repeatability was evaluated both for the instrument (GC-MS) and entire method (SPE-GC-MS).For the GC-MS method, pure standards of the analytes in DCM at three different concentration levels were prepared (n = 3 for each concentration level).The repeatability of the entire SPE-GC-MS methodology was validated by the analysis of aqueous solutions of the analytes in bidest water (3 concentration levels, triplicates for each level).The used concentration/spiking levels were a factor 10 apart, which was in accordance with the Flemish validation guidelines (WAC/VI/A/001) for water analysis and allowed to validate the method's repeatability and trueness for a wide concentration range.
The intraday repeatability was calculated for each concentration level as the relative standard deviation (RSD) of the consecutive analysis of the prepared standards and was analyzed by the same analyst on the same day.The interday repeatability is evaluated similar to the intraday repeatability, however here, the preparation and analysis of the samples is performed at three different moments in time (at least 5 days between the analyses).
In order to correct and/or compensate for analytical and sample pretreatment fluctuations, also the surrogate corrected repeatabilities for the developed SPE-GC-MS method were calculated.This was done by the addition of 100 μL of a surrogate standard which contained three deuterated PAHs (conc.standard 40 μg/L) to the aqueous solution prior to SPE.By calculating the surrogate corrected factors, which is the ratio between the obtained MS signal of the analyte and the signal of the corresponding deuterated PAH as listed in Table C.1 (appendix C), the surrogate corrected inter-and intraday-repeatabilities were calculated.Important to note is that due to a loss of sensitivity of the GC-MS, the MS ion source was entirely cleaned and a new HP5-MS column was placed prior to the injection of the SPE extracts for method validation (SPE-GC-MS) and method application on real samples.This resulted in a significant increase of GC-MS sensitivity.

Measuring range, linearity and detection limits
To evaluate the instrument measuring range (GC-MS), a 10-point calibration curve in DCM was prepared (further referred to as DCM calibration).The concentration ranges were between 0-1000 μg/L for PAHs, between 0-5000 μg/L for the biocides and DEHP, and between 0-50 mg/L for DIDP.To evaluate the measuring range of the entire method (SPE-GC-MS), a 9-point calibration curve by spiking of bidest water was prepared (further referred to as aqueous calibration).Aqueous concentrations ranged between 0-1000 ng/L for PAHs, 0-100 μg/L for biocides and DEHP, and 0-1000 μg/L for DIDP.For linearity assessment, the correlation between the MS signal and the concentrations was evaluated based on the R 2 of the regression curve.The measuring range was defined as the range between the limit of quantification (LOQ) and the highest concentration in the linear area of the regression curve.
The instrument limit of detection (LOD) and LOQ were calculated based on the signal to noise ratio (S/N-ratio) of the lowest DCM standard.For the calculation of the method's LOD and LOQ, the S/N-ratio of the lowest aqueous standard was used, being an S/N-ratio of 3 and 6 for LOD and LOQ respectively.

Recoveries in bidest water and real samples
The method's trueness was evaluated for the entire method and was based on the recovery of spiked bidest water samples (n = 3 for each concentration level, measured under interday repeatability conditions).The non-surrogate corrected recoveries were calculated based on the external calibration curve in DCM.Furthermore, also surrogate corrected recoveries were calculated.For this calculation, the surrogate corrected signals were weighted against the surrogate corrected aqueous calibration curves.To evaluate the method's trueness in real samples, the recovery (both surrogate and non-surrogate corrected) was determined in roof runoff and in roofing material leachate by spiking roof runoff and roofing material leachate samples at following aqueous concentration levels: 200 ng/L for the PAHs, 10 μg/L for DEHP, DIU and OIT, and 100 μg/L for DIDP.All analyses for recovery calculation in real samples were performed in triplicate under intraday repeatability conditions.

Application on real water samples: Roofing material leachate and roof runoff
In order to confirm the applicability of the method on real samples, material leachate and roof runoff were analyzed.Material leachate from bituminous roofing materials was collected analogous to the method used in De Buyck et al. [24] .Details on the test set-up can be found in Appendix B. Roof runoff originated from a house equipped with a clay tile roof with zinc gutters, located in a residential area at the periphery of a municipality (Wevelgem, Belgium), and in the direct vicinity of agricultural fields.Therefore, the sampling site was categorized as suburban -rural.For this experiment, the runoff was collected from an outdoor tap which was connected to the rainwater harvesting tank (15000 L).
For recovery calculation, one aliquot of 400 mL was directly analyzed, whereas another aliquot was spiked with the components of interest to evaluate the method's trueness.All analyses were performed in triplicate (under repeatability conditions) and none of the samples were filtered prior to analysis.The surrogate corrected calibration was used for the calculation of concentrations in these samples.By forcing the calibration plots through the calibration blanks, a correction for procedural contaminations was made.

Glassware cleaning
To avoid contamination, all glassware used for sample pretreatment, the preparation of stock solutions and for laboratory leaching tests, was rinsed according to following procedure: a first rinse with a 0.1% formic acid solution, subsequently a rinse with 0.1% ammonia solution, two times with methanol and a final rinse with bidest water.After the bidest rinse, the glassware was air-dried prior to usage.

SPE-GC-MS method validation using bidest water Instrumental (GC-MS) limitations
During instrument validation (see also appendix C), it was found that DIU completely decomposed into 3,4-dichlorophenylisocyanate (3,4-DCPI) and its corresponding amines under the applied chromatographic conditions as DIU is a thermally labile urea pesticide.Therefore, for DIU, the target and qualifier ions of 3,4-DCPI were included in the SIM program and used for further method development.Important to note is that other pesticides, such as linuron and neburon, might also decompose into 3,4-DCPI [25] .The instrument detection limits varied between 0.005 and 0.06 μg/L for PAHs, and for the biocides DIU/3,4-DCPI and OIT, an LOD of respectively 0.06 and 0.5 μg/L was found.DEHP and DIDP were found to have an LOD of respectively 0.3 μg/L and 10 μg/L.For MCPP, a higher instrumental LOD was obtained (1.5 mg/L).The lower sensitivity of the GC-MS for this component was not surprising as this chlorophenoxy acid has a higher polarity.To overcome this, these components can be converted in less polar, more volatile methyl esters with a higher thermal stability by derivatization [26] .However, as it was the purpose in this research to keep sample pretreatment as simple as possible, no derivatization step was used.As a consequence of the lower sensitivity of this component, the analyte was treated as a semi-quantitative component in the further method development.More details on the instrument validation can be found in appendix C.

Linearity, measuring range and detection limits of the SPE-GC-MS method in bidest water
As can be seen in Table 2 , the linearity as expressed by the R 2 of the aqueous calibration curve (surrogate corrected), was > 0.99 for the majority of analytes.In terms of the measuring range it should be noted that due to the lower aqueous solubility of DIDP, DEHP and several PAHs, the obtained MS signal was not linear through the entire range of included calibration points.The upper limit of the measuring range was limited therefore.
The detection limits varied between 0.1 and 3.9 ng/L for PAHs in water.For DIU/3,4-DCPI, the LOD was 0.4 ng/L and for OIT, 4.7 ng/L.For the phthalates DEHP and DIDP The LOD was 0.1 μg/L and 4.8 μg/L, respectively.Due to the lower sensitivity of the GC-MS for MCPP (see appendix C), the LOD for this analyte was 18 μg/L.For the lower molecular weighted PAHs and DIU/3,4-DCPI the obtained detection limits are relatively low in comparison to other methodologies ( Table 3 ).For the PAHs with a higher molecular weight, OIT and DEHP, the detection limits were found to be in the same range or slightly higher.For DIDP and MCPP, the detection limits obtained by the developed SPE-GC-MS methodology were higher than the ones found in the consulted literature ( Table 3 ).

Repeatability of the developed SPE-GC-MS method in bidest water
Regarding repeatability, only the relevant concentration levels which were within the defined measuring range, were evaluated.The concentration levels which were not within the defined measuring range are also included in Table 2 , but are rather indicative.For MCPP, the repeatability was only evaluated for the highest included concentration level as GC-MS was found to have a lower sensitivity for this component.The intraday variability (non-surrogate corrected), expressed as RSD, was for most analytes below 10% at the relevant concentration levels (InPy: 10%, OIT: 12%, DEHP: 13% and DIDP: 12%).After surrogate correction, the intraday repeatability of some components improved, whereas for other analytes, the intraday repeatability was slightly worse.Therefore it can be stated that the effect of the surrogate correction on the intraday repeatability was rather limited.However, when looking at the interday repeatability, the surrogate correction significantly improved the repeatability as the RSD for the included analytes was lower (only exception on this was for Nap at the highest spiking level).All surrogate corrected interday RSDs were good ( < 30%, [ 27 , 28 ]), and below 20% at the relevant concentration levels.For the majority of analytes, the interday RSDs were even below (or equal to) 15% at the relevant spiking levels (except for InPy, BghiP, OIT and MCPP).Without surrogate correction, the interday repeatabilities ranged between 8 and 62% (expressed as RSD).It can thus be stated that surrogate correction is favorable to increase the overall robustness of the developed method as a result of the better interday repeatability.

Method's trueness (recovery) in bidest water
To evaluate the method's trueness, both surrogate and non-surrogate corrected recoveries were calculated (interday repeatability conditions, n = 3 for each concentration level).For PAHs, the non-surrogate corrected recovery varied between 12% and 158% at the relevant concentration levels.For the lower molecular weighted PAHs Nap, Acy, Ace, Flu, Phe, Ant, Flt and Pyr, the recoveries were good (close to 100%, between 70% and 130% [29] ) and in the same range as the recoveries from other methods ( Table 3 ).However, for higher molecular weighted PAHs, which tend to have a lower water solubility and a higher log K OW (Chry, BbF, BaP, InPy, BghiP, DahA), the recoveries were lower in comparison to other studies.The use of different SPE sorbents, other elution conditions and the use of organic modifiers prior to extraction might explain the higher observed recoveries in those studies.In addition, also matrix effects in real environmental samples might play a role (see Recovery in roof runoff and material leachate).
Furthermore, when looking at the different spiking levels ( Table 2 ), it can be observed that the recoveries are decreasing with increasing spiking concentration and that this effect is more pronounced for the higher molecular weighted PAHs.For phthalates, a similar pattern could be observed ( Table 2 ).This effect of lower recoveries at higher spiking levels for PAHs was also reported by Wang et al. [21] and the effect of lower recoveries for DEHP at higher concentration levels, is in agreement with the findings of Jara et al. [20] (summarized in Table 3 ).Also here, the water solubility of DEHP was hypothesized to be the main cause of the lower recoveries at higher concentration levels as DEHP can form colloids, aggregates and micro drops at higher concentration levels.In addition, also sorption onto glasswork and the used SPE equipment in case of pure water, and sorption onto particles in case of real environmental samples, might explain the lower recoveries at higher concentrations (both for PAHs and phthalates).
The recoveries for DIU and OIT were slightly higher than the ones reported in literature ( Table 3 ).
The non-surrogate corrected recovery of BkF could not be determined as this analyte was not included in the purchased standard used to evaluate and optimize the GC-MS methodology.
It was previously shown that surrogate correction was beneficial for the (interday) repeatability of the developed SPE-GC-MS method.Therefore, also surrogate corrected recoveries were calculated.By surrogate correction of the recoveries, lower RSD were obtained for most of the included analytes.For the higher molecular weighted PAHs and phthalates at the highest concentration levels, the effect of surrogate correction on the recoveries RSD was limited.
At the relevant concentration levels, the surrogate corrected recoveries for the PAHs varied between 80% and 135%, between 96% and 176% for DIU, between 78% and 87% for OIT and between 76% and 92% for DIDP.For DEHP, even at the lowest spiking level (1 μg/L), the use of aqueous calibration standards in combination with surrogate correction was found to overestimate the actual concentration in the spiked samples, resulting in extremely high surrogate corrected recoveries for this components ( > 190%).Therefore, for DEHP, the use of non-surrogate corrected recoveries, in combination with calibration curves in DCM is favorable.
For MCPP no recoveries were calculated as this component was included as a semi-quantitative analyte in the method.

Recovery in roof runoff and material leachate
To study the recovery and possible matrix effects in real samples, roof runoff and material leachate were spiked with the components of interest (PAHs 200 ng/L, 10 μg/L for DEHP, DIU and OIT, 100 μg/L for DIDP).Both the surrogate and non-surrogate recoveries were calculated and compared with the recoveries obtained during method development (at the intermediate spiking level in bidest water).

Non-surrogate corrected recoveries in roof runoff and material leachate
The non-surrogate corrected recoveries varied between 58% and 120% in roof runoff and between 25% and 127% in roofing material leachate ( Figure 1 a).For the lower molecular weighted PAHs, DIU and OIT, the recoveries were in the same range as determined during method validation in bidest water.
However, for the components with a lower water solubility, the recoveries in material leachate and roof runoff were found to be significantly higher than in bidest water.
As bidest water was used both for material leaching and for method validation, the introduction of the material must be responsible for the effect on the recoveries.The release/wash-off of small particles/dust from the materials during the leaching experiment and the subsequent sorption of the spiked analytes with a lower water solubility/higher log K OW is hypothesized to play a key role here.As the samples were not filtered prior to SPE sample loading, also particle adsorbed analytes could be retained in/on top of the SPE sorbent bed, which was also visually confirmed.When eluting the SPE cartridges, both C 18 -adsorbed and particle-adsorbed analytes will be desorbed, resulting in higher recoveries.For roof runoff, a similar effect is expected as dust and other particles are also present in these samples.
The capability of the developed method -where the samples are not filtered prior to SPE -to quantify (partially) particle adsorbed compounds in material leachate and roof runoff is favorable for screening purposes and will thus also allow the quantification of these substances.a NSC: non-surrogate corrected.b SC: surrogate corrected.c Non-surrogate corrected recovery: recovery as calculated based on the external calibration curve in DCM, not corrected by the corresponding deuterated surrogate, surrogate corrected recovery: recovery as calculated based on the aqueous calibration curve and corrected by the corresponding deuterated surrogate.The reported recoveries were measured under interday repeatability conditions (n = 3).d Due to sample contamination, the lowest spiking level for naphthalene could not be evaluated.e Spiking level was not within the defined measuring range.f Parameter not evaluated as MCPP was considered as a semi-quantitative compound.This is also the reason why SPE cartridge manufacturers advise to not filter the samples if components of interest can be adsorbed onto particles, unless the solid particles interfere with the method (cartridge blocking) [30] .

Surrogate corrected recoveries in roof runoff and material leachate
As can be seen in Figure1 b, the surrogate corrected recoveries in roof runoff and material leachate of the more soluble analytes were in line with the surrogate corrected recoveries as determined in bidest water during method validation.For the less soluble analytes, the surrogate corrected recoveries in roof runoff and material leachate were higher than 200% (see Appendix D).This confirms that these analytes interact with the matrix and the components present in these real samples.

Application to real samples: roof runoff and material leachate
The developed methodology was used to identify and quantify the target analytes in roof runoff and roofing material leachate.It was decided to use the aqueous calibration curves for this, in combination with the surrogate corrected recovery.From the included analytes, DEHP, DIDP, OIT, MCPP and higher molecular weighted PAHs were not detected in runoff nor in material leachate.

Roof runoff
In roof runoff, several PAHs, including Nap, Flu, Ant, Flt, Pyr and Chry were found in trace concentrations (max 38 ng/L).In comparison to the maximum PAH concentrations reported in literature ( Table 4 ), the observed PAH levels in the analyzed roof runoff samples are relatively low.As the roofing materials at the sampling site were clay tiles, the detected PAHs in the runoff are likely originating from atmospheric deposition rather than from material leaching [1] .Nevertheless, as the sampling site was located in a residential neighborhood and not in the direct vicinity of highways ( > 1.5 km) nor heavy industry, it is likely that residential heating by wood combustion is the main source of PAH deposition.This hypothesis is strengthened by the molecular diagnostic ratios of the detected PAHs (for details, see appendix D).
Next to PAHs, also traces of 3,4-DCPI were found.As the sampling site was located close to an agricultural area ( < 100 m), the detection of this component might be a consequence of the agricultural use of urea herbicides.However, it is important to note that not only DIU, but also several other herbicides (e.g.linuron, neburon) might thermally decompose into 3,4-DCPI during the gas chromatographic separation.Therefore, the identification of the exact herbicide is impossible.Nevertheless, as DIU was used as parental component for method validation, literature data on DIU in roof runoff was included in Table 4 for comparison.

Table 4
Typical concentration and detection frequency (DF) of the selected PAHs (ng/L), biocides (μg/L) and phthalates (μg/L) in roof runoff and rainwater.As for DIU and OIT, only limited information on their occurrence in roof runoff was found, also information on their concentration levels in façade runoff are included here.'-' no data given in the reference.

Material leachate
In bituminous roofing material leachates, Nap, Flu, Phe, Ant, Pyr and Chry were detected.Other studies investigating the leaching behavior of bituminous materials under laboratory conditions found the same PAHs ( Table 5 ).However, it should be noted that a direct comparison of the obtained concentrations in this study with the concentrations reported in literature remains difficult as different liquid-to-surface ratios, other contact times, and other bituminous materials were tested.Nevertheless, it can be seen that the PAH concentrations as detected in this study, are in line with the typical concentrations observed by others.Furthermore, as can be seen from Table 5 , none of the tested materials, both in this study and in other studies, found less soluble PAHs (with five or more benzene-rings) in the material leachates.
Despite the fact that mecoprop is sometimes used in the formulation of bituminous roofing materials, this biocide was not detected in the leachate of the present study [31] .However, it should be noted that not in all bituminous roofing materials biocides are added.Therefore, it is likely that no MCPP was used in the tested material.For DIU and OIT, no uses as biocide in the bituminous materials are known by the authors.Furthermore, no traces of DEHP and DIDP were detected in the material leachates, which is in line with the current available information on phthalate leaching from bituminous roofing materials [1] .

Conclusions
In this study, an SPE-GC-MS multi-residue method for the analysis of a selection of PAHs (EPA 16 PAHs), phthalates (DEHP and DIDP) and biocides (MCPP, OIT and DIU) in roof runoff and roofing material leachate was successfully validated.Sample pretreatment was kept as simple as possible in order to obtain a robust, but cost and resource efficient method without the need for a derivatization or the addition of an organic modifier.Furthermore, the validated method is in line with several principles of green analytical chemistry.The inclusion of MCPP as a semi-quantitative analyte is a compromise for the higher instrument detection limits of this component as no derivatization was used.
The use of a surrogate standard to correct for analytical fluctuations, sample pretreatment losses and matrix effects, was found to increase the reproducibility (interday) of the method.
During method validation, the effect of the spiked concentration on the recovery of phthalates and several PAHs could be observed.This effect was most pronounced for the components with a lower water solubility.Therefore, it seems that sorption phenomena or the formation of colloids, aggregates and micro drops, due to their lower water solubility, played a role here.These sorption phenomena also resulted in higher recoveries for these analytes in roof runoff and material leachate.The validated multi-residue SPE-GC-MS method was successfully used to identify and quantify the targeted components in real roof runoff and material leachate samples.

Fig. 1 .
Fig. 1. (a): Comparison of non-surrogate corrected recoveries in roof runoff, material leachate and bidest water.(b) Comparison of the surrogate corrected recoveries the more soluble analytes in roof runoff, roofing material leachate, and bidest water.* Due to a technical issue only for one replicate a signal could be registered for OIT.Bidest mid-level corresponds to the spiking concentrations used for method validation at the intermediate concentration level (aqueous concentration of 100 ng/L for PAHs, 10 μg/L for DIU, OIT and DEHP, and 100 μg/L for DIDP in bidest water).

Table 3
Comparison of non-surrogate corrected recoveries found in roof runoff and material leachate and detection limits of the validated SPE-GC-MS method with literature.
( continued on next page )

Table 3 (
continued ) Non-surrogate corrected recoveries.c Expressed as added concentration.d recovery corresponding with the respective concentration level as defined in the second column of the table.e Exact concentration level not specified.f Average recovery of three concentration levels (40 ng/L, 80ng/L and 400 ng/L).g Recovery range found at different sampling sites.h Environmental samples were filtered prior to spiking and an organic modifier (2-propanol, 10%v/v) was added prior to extraction.i Another type of LC-MS device was used.j PSDVB = polystyrene-divenyl benzene.k Recovery estimated from bar chart.l ESI = electron spray ionization.m Crosslinked polystyrene-divinyl benzene (PSDVB) resin with C 18 grafted chains was used as SPE sorbent.
a Expressed as aqueous concentration.b n LOQ.

Table 5
PAH concentrations (ng/L) found in bituminous roofing material leachate and comparison with other studies.