A. Gimpel et al.
in a multivariate response data matrix (Greenacre and Primicerio, 2013).
The response matrix was individual diet composition according to the 12
nain prey categories. The explanatory variables included fish length class
factor) and the study area (factor), which is consistent with the classifica-
:jons in Fig. 3. Both explanatory variables were used as single-effect-terms
and two-way interaction. Significance tests of terms were done using
ANOVA-like permutation test. The results of the CCA model are presented
n terms of pseudo-F values, the ratio of constrained and unconstrained
:otal inertia indicating the explained variance by each explanatory term.
A biplot of the two leading CCA dimensions further demonstrated the rela-
Gjonship between diet composition, fish length and study area.
We collected muscle samples for stable isotope analysis to investigate
site fidelity and draw conclusions on the long term diet compositions inside
and outside the OWF. We selected 15 comparable individuals (5 samples
‚Tom the OWF in summer and winter respectively and 5 samples from the
GB in summer; Appendix A, Table A.2). These were all male, showing a
nean age of 1.3 y (+ 0.5), length of 31.4 cm (+ 4.0) and weight of
301 g (+ 115.7) for stable isotope analyses (5'°C: Ratio of the stable carbon
isotopes !°C and ??C and 8*°N: Ratio of the stable nitrogen isotopes !°N and
'*N) (Hünerlage and Buchholz, 2013). We extracted white muscle tissue
samples from the nape (the skin was removed) at sea. In the laboratory,
‘he samples were further lyophilized for 24 h and ground to powder
using a glass tissue grinder, The homogenate was weighed in tin capsules
(30-40 ug) (Hünerlage and Buchholz, 2013) and used for stable isotope
analysis at the Stable Isotope Analysis Laboratory at the Thünen Institute
of Climate-Smart Agriculture (https://www.thuenen.de/en/institutes/
climate-smart-agriculture/laboratory-department/stable-isotope-analysis-
aboratory). The analyses of stable isotopes allowed to determine possible
differences between regions, therefore we carried out pairwise comparisons
with unpaired t-tests. The significance level was set at P < 0.05.
Due to a limited number of samples in comparable size classes (Fig. 2)
we calculated commonly used indices reflecting stomach fullness or body
condition, which could be related to diet composition and food (Table 1,
Appendix B, Figs. B.1 and B.2). However, the sample size did not allow
for robust conclusions.
2.3. Reproduction potential
In order to investigate if offshore wind farms offer suitable conditions or
even enhance reproductive efforts for cod, we collected information on
adults and early life stages during winter spawning season. We defined the
sex and maturity stages of adult cod in order to assess potential spawning ac-
ävity and spawning locations. Further, we identified the maturity stage
Jased on a 6-point maturity scale for Atlantic cod including six gonadal de-
velopmental stages (1: juvenile/Immature; 2: maturing; 3: spawning; 4:
spent; 5: resting/skip of spawning; 6: abnormal) (ICES, 2014). Because
catch rates in the OWF were low during spawning season in winter, only
i2 individuals were available for assessing maturity stages (Fig. 1,
Table 1). As such a low number of samples limited the further use of param-
eters such as the sex ratio (Appendix B, Fig. B2), we decided to focus solely
on the presence of fish in spawning condition to draw conclusions on a po-
rential enhancement of reproduction through offshore wind farms.
from the ichthyoplankton hauls we presorted cod-like eggs on board
vased on size (> 1 mm and < 2 mm) and absence of an oil-droplet (Fox
et al., 2008; Russell, 1976). The age of each cod egg was approximated by
visual inspection of their developmental stage (Thompson and Riley,
1981). These eggs were fixed and preserved in 95 % ethanol (EtOH) for sub-
sequent species identification by genetic barcoding (Fox et al., 2008; Lewis
et al., 2016). Based on statistical relationships with temperature measure-
nents taken in relation to ambient water temperature at each station and
the developmental stage (Appendix C, Table C.1) the approximate age of
each egg (in hours) was estimated using the formula of Thompson and
Riley (1981):
D—94- „(AT+B)
Science of the Total Environment 878 (2023) 162902
where D represents the duration from fertilization to the end of each of the
5 cod egg stages (1A, 1B, 2-5), and A and B the regression coefficients for
2ach stage in relation to temperature (T) in degrees Celsius against log, of
:ime in hours. We analyzed a total of 214 cod-like fish eggs using DNA
yarcoding (according to standard procedures as described in Rossel et al.
‘2021)). After DNA extraction, all 214 eggs were subjected to PCR with spe-
öific COI barcoding primers for fish. The DNA sequences were compared
with a DNA reference data set for fish. Parallel to the DNA barcoding,
VIALDI-TOF MS was performed on the remaining homogenate for prote-
ome fingerprinting (Rossel et al., 2021). Of all fish eggs analyzed, 171
eggs were successfully identified by DNA barcoding and 190 eggs by prote-
ome analysis.
2.4. Egg drift simulations
The transport of cod eggs by ocean currents was studied with the La-
grangian particle tracking model ‘PELETS-2D’ (Callies et al., 2021). The
»rogram conducts offline simulations based on pre-calculated 2D current
äelds that have to be defined on an irregular triangular grid (Callies
et al., 2011). In this study, the pre-calculated 2D current fields were ob-
:ained from the 3D-finite difference ocean circulation model ‘BSHcmod’
hat is operational at the German Federal Maritime and Hydrographic
Agency (Dick and Kleine, 2007). The underlying model grid has a resolu-
.jon of 900 m in the inner German Bight and roughly 5 km around. In
order to meet the requirements for PELETS, the original 3D current fields
of BSHcmod were converted to 2D current fields by calculating vertical
neans. During the simulations, the current velocities were updated by lin-
aar interpolation each time a particle was transferred to another grid cell
or when the time step exceeded 15 min, respectively. No extra wind drift
and Stokes drift were taken into account.
In our study, two different PELETS setups were implemented. The first
;etup was used to localize the source areas of cod egg production. For this
Jyurpose, we initiated one tracer particle for each cod egg that had been
sampled during our winter campaigns in 2019 and 2020. Each particle
was released according to the exact time and position of the corresponding
2gg observation. Particle trajectories were integrated backward in time and
>»ositions recorded on an hourly basis. We referred to the location reached
after the calculated life span of the simulated egg as its potential “spawning
site” (Fig. 1 and Table 1). All identified spawning sites were further distin-
zuished between either inside the wind farm (OWF spawn), the wind farm
:luster (Wind Energy Cluster) or the German Bight (GB spawn) (Fig. 1 and
Appendix C, Table C.2).
We used a second PELETS setup to simulate the general dispersal pattern
of cod eggs under the weather conditions that prevailed during our winter
campaigns in 2019 and 2020. For this purpose, we initiated daily spawning
events during our two investigation periods. The first period comprised 32
events between 20 December 2018 and 20 January 2019. The second period
comprised 26 events between 3 January 2020 and 28 January 2020. Each
avent comprised an ensemble of 1000 particles that were released at random
ocations within the OWF. Each particle trajectory was integrated for a max-
mum of 480 h (20 days) forward in time. Particle locations were recorded
on an hourly basis and subsequently evaluated for their spatial distributions
in the investigation area (Appendix B Table B.2). Therefore, we counted on a
:aster with a 10 x 10 km resolution the proportion of particles that have
ever visited a grid cell within the 480 h simulation interval.
3. Results
3.1. Length frequencies and diet composition
{n summer, the mean length of fish caught in the OWF by angling was
almost twice as high as in the GB, where data were collected with bottom
rawls (Table 1). In contrast, comparing the mean length in the OWF and
che GB, a considerably higher proportion of large individuals was found
n the GB in winter (Table 1, Fig. 2A-D). This indicates seasonal variability
'n the distribution of cod around the OWF, with a concentration of larger
