Environmental Science & Technology 
pubs.acs.org/est 
KL 
ww & Hi£h Rag 
WW 
zz 
-ı GPNS 
Sample 
blanks removal 
Identified features 6 2425 | 
_ —— 6 1501 
OÖ 17454 . 275 
® 14687 
2 1252 Oo 8704 \ Ö 3673 
& 8679 6 2748 
873 131 | 
Lab blank Sample 
removal blanks removal 
Legend 
5 LC-HRMS ESI + 
$ LC-HRMS ESI - 
/)GC-MS 
"DNS 
Aa 
OWF GE 
0 2298 
e &$ 2051 
<D 100 A \ 
, 8829 
1245 
x 
ship Re; 
a 
zz 
* 
1871 
1721 
90 
, 
4454 
0 2786 
5.214 
2 
ba 
0 
% 0623 
\ 0474 
34 
mn nn y 
/OWFA OWFB 
Ö 185 Ö 89 0790 * 
von |00, 002 
‚on \o4] 28 
#. 
4 
*Q 
po 
High Reit 
var 
Figure 3. Overview of the number of features detected using GC-MS and LC-HRMS in ESI + modes after blank correction. The remaining features are 
divided into different categories for both the Belgian (lower) and German (upper) parts of the North Sea (BPNS and GPNS, respectively) using a score 
“unction. 
he National Institute of Standards and Technology (NIST) databank 
“version 2.4)” and the Kovats index. A compound was assumed to be 
.dentified when the R-match was >800 and the Kovats index did not 
differ by more than 50. 
For the LC-HRMS, different active libraries (Environmental 
Contaminant Biotransformation Pathway Resource (enviPath),“® 
Aggregated Computational Toxicology (ACToR),** EPA Distributed 
Structure—Searchable Toxicity (DSSTox),** EPA Toxicity Forecasting 
“Toxcast),** NIST,”” Phenol-Explorer**) were used with a precursor 
:olerance of 10 ppm. A score is calculated by the software based on the 
‘ragmentation, mass error, isotope similarity, and retention time error. 
T'he identification with the highest score is accepted as long as the score 
>40 (out of 60), the isotope similarity >95%, and the mass error <5 
ppm. For all identified compounds, an AI search was done with 
Microsoft Copilot (accessed June 2025) to predict potential industrial 
applications and identify the chemical family. 
2.7. Data Availability 
The RAW chromatographic data are available on Zenedo for the GC- 
MS,“ LC-HRMS in ESI positive,“ and ESI negative modes.” 
3. RESULTS AND DISCUSSION 
3.1. Chemical Space 
GC-amenable compounds were extracted using hexane and a 
hexane:acetone (2:1) mixture, whereby the eluents of the 
1exane fraction on a silica column were discarded. This 
approach resulted in the elution (with hexane:acetone (2:1)) 
of a nonpolar to semipolar fraction (compounds with Kovats 
index of approximately 1000—3300), which was kept for 
analysis, while excluding very nonpolar compounds typically 
interfering with the GC-MS analysis, such as lipids, alkanes, and 
waxes, but also a part of the highly hydrophobic SVOCs 
potentially present in the sample. LC-amenable compounds 
were extracted with methanol, targeting polar to semipolar 
compounds, with some overlap between the GC- and LC- 
amenable fractions in the semipolar region. Surrogates were 
selected based on differences in their physicochemical proper- 
ties and were detected at retention times approximately located 
at the beginning, middle, and end of the chromatogram (see 
Supporting Information S4.4). However, the number of 
surrogates could have been increased to better represent the 
covered chemical space. 
3.2. Feature Selection 
Cor GC-MS, a total of 1252 features were identified after peak 
picking. For the LC-HRMS screening, 17,454 and 14,687 
features were detected in ESI positive (ESI +) and ESI negative 
‘ESI —) modes. After lab and field blank correction, 248 (GC- 
MS), 5077 (LC-HRMS ESI +), and 3,260 (LC-HRMS ESI —) 
features were detected in the BPNS and 304 (GC-MS), 6325 
“LC-HRMS ESI +), and 4507 (LC-HRMS ESI —) features were 
detected in the GPNS, with the majority of the features present 
in both study areas (Figure 3). 
3.3. NTS-Driven Source Apportionment 
For both study areas, the features were divided, using a score 
function, into impact (OWF A/B for Belgium and OWF GE for 
Germany), background (for both the BPNS and GPNS), and 
high ref and ship ref (Figure 3 and Table S1). For the BPNS, a 
distinction was made between OWF A and B, each of which 
counted over eight samples per area. However, because the 
GPNS counted six different OWFs, each with two to four 
samples per farm, no distinction was made between the different 
OWFSs, and the results were summarized as OWF GE. The 
German high and ship refs were, for the same reason, also 
merged into one category. 
Jor the GC-amenable compounds detected in the BPNS, 186 
in total were omnipresent in the study area (background 
compounds); 53 were detected in at least one OWF, with 21 
features in OWF A, 28 features in OWF B, and four features 
detected in both OWFs. Of the remaining features, eight were 
more abundantly detected in the high ref site, whereas only one 
nttps://doi.org/10.1021/acs.est.5c17939 
Environ. Sci. Technol. XXXX, XXX, XXX—-XXX