Defining habitats suitable for larval fish in the German Bight (southern North Sea): An IBM approach using spatially- and temporally-resolved, size-structured prey fields

Abstract

We employed a coupled biological–physical, individual-based model (IBM) to estimate spatial and temporal changes in larval fish habitat suitability (the potential for areas to support survival and high rates of growth) of the German Bight, southern North Sea. In this Lagrangian approach, larvae were released into a size-structured prey field that was constructed from in situ measurements of the abundance and prosome lengths of stages of three copepods (Acartia spp., Temora longicornis, Pseudocalanus elongatus) collected on a station grid repeatedly sampled from February to October 2004. The choice of prey species and the model parameterisations for larval fish foraging and growth were based on field data collected for sprat (Sprattus sprattus) and other clupeid larvae. A series of 10-day simulations were conducted using 20 release locations to quantify spatial–temporal differences in projected larval sprat growth rates (mm d− 1) for mid-April, mid-May and mid-June 2004. Based upon an optimal foraging approach, modeled sprat growth rates agreed well with those measured in situ using larval fish ototliths. On the German GLOBEC station grid, our model predicted areas that were mostly unsuitable habitats (areas of low growth potential), e.g. north of the Frisian Islands, and others that were consistently suitable habitats (areas that had high growth potential), e.g. in the inner German Bight. In some instances, modelled larvae responded rapidly (~ 5 days) to changing environmental characteristics experienced along their drift trajectory, a result that appears reasonable given the dynamic nature of frontal regions such as our study area in the southern North Sea.

Full text

This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright

Author's personal copy
Defining habitats suitable for larval fish in the German Bight (southern
North Sea): An IBM approach using spatially- and temporally-resolved,
size-structured prey fields
Wilfried Kühn
a,
⁎, Myron A. Peck
b
, Hans-Harald Hinrichsen
c
, Ute Daewel
b
, Andreas Moll
a
,
Thomas Pohlmann
a
, Christoph Stegert
a
, Susanne Tamm
a
a
Institut für Meereskunde, Universität Hamburg, Bundesstr. 53, 20146 Hamburg, Germany
b
Institut für Hydrobiologie und Fischereiwissenschaft, Universität Hamburg, Germany
c
Leibniz-Institut für Meereswissenschaften, Universität Kiel, Germany
article info abstract
Article history:
Received 27 March 2007
Received in revised form 7 February 2008
Accepted 12 February 2008
Available online 4 March 2008
We employed a coupled biological–physical, individual-based model (IBM) to estimate spatial
and temporal changes in larval fish habitat suitability (the potential for areas to support survival
and high rates of growth) of the German Bight, southern North Sea. In this Lagrangian approach,
larvae were released into a size-structured prey field that was constructed from in situ
measurements of the abundance and prosome lengths of stages of three copepods (Acartia spp.,
Temora longicornis,Pseudocalanus elongatus) collected on a station grid repeatedly sampled
from February to October 2004. The choice of prey species and the model parameterisations for
larval fish foraging and growth were based on field data collected for sprat (Sprattus sprattus)
and other clupeid larvae. A seriesof 10-day simulations were conducted using 20 release locations
to quantify spatial–temporal differences in projected larval sprat growth rates (mm d
−1
)for
mid-April, mid-May and mid-June 2004. Based upon an optimal foraging approach, modeled
sprat growth rates agreed well with those measured in situ using larval fish ototliths. On the
German GLOBEC station grid, our model predicted areas that were mostly unsuitable habitats
(areas of low growth potential), e.g. north of the Frisian Islands, and others that were
consistently suitable habitats (areas that had high growth potential), e.g. in the inner German
Bight. In some instances, modelled larvae responded rapidly (~5 days) to changing
environmental characteristics experienced along their drift trajectory, a result that appears
reasonable given the dynamic nature of frontal regions such as our study area in the southern
North Sea.
© 2008 Elsevier B.V. All rights reserved.
Keywords:
Coupled transport-growth model
IBM
Larval sprat
German Bight
North Sea
Habitat suitability
1. Introduction
Models employing ecophysiological principles (e.g., Neill
et al., 1994) have been developed to provide spatially-explicit
predictions of potential growth of specific life stages of var-
ious fish species. For example, estimates of the spatial varia-
bility in the growth rates of juvenile and adult striped bass
(Morone saxatilis) and Chinook salmon (Oncorhynchus tsha-
wytscha) were obtained using bioenergetics-based models
(Brandt and Kirsch, 1993; Mason and Brandt,1996). Spatially-
explicit growth predictions for earlier (larval) life stages of
marine fish have been gained using individual-based models
(IBMs) developed for such species as Atlantic cod (Gadus
morhua)(Werner et al., 1996; Hinrichsen et al., 2002; Lough
et al., 2005). These models can be effective tools to identify
critical habitats and processes when the appropriate temporal
and spatial scales are captured, (e.g., when models depict
scales of interactions between larval fish and their prey re-
sources). For example, an IBM developed for Baltic cod larval
drift, feeding and growth was able to predict, with a high
degree of confidence, the relatively strong recruitment period
Journal of Marine Systems 74 (2008) 329–342
⁎Corresponding author.
E-mail address: kuehn@ifm.uni-hamburg.de (W. Kühn).
0924-7963/$ –see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmarsys.2008.02.002
Contents lists available at ScienceDirect
Journal of Marine Systems
journal homepage: www.elsevier.com/locate/jmarsys

Author's personal copy
from 1986 to 1991 as well as the rapid decline in recruitment
starting in 1993 (Hinrichsen et al., 2002).
Within the North Sea, most previous studies utilizing
an IBM approach have focused on physical processes affecting
marine fish early life stages; the potential influence of changes
in prey resources on larval vital rates has largely been ignored.
For example, models developed for early life stages (eggs and
larvae) of Atlantic cod, haddock (Melanogrammus aeglefinus)
and Atlantic mackerel (Scomber scombrus) employed simple
larval growth functions based upon mean length-at-age
(Heath and Gallego, 1997; Bartsch and Coombs, 2004). How-
ever, significant correlations between time-series of
Fig. 1. A) Model area for the hydrodynamic model HAMSOM with bottom topography. The rectangle indicates the area investigated in this modelling study;
B) GLOBEC–Germany zooplankton sampling stations in the German Bight in 2004. Open circles denote stations where multinet hauls were made and filled circles
denote stations where length measurements of zooplankton were available from bongo net hauls.
330 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
zooplankton abundance and recruitment strength of fish in
the North Sea (Rothschild, 1998; Heath et al., 1997; Beaugrand
et al., 2003) suggest that variability in zooplankton prey fields
can contribute to variability in rates of survival and growth
during fish early life stages. For example, a significant amount
of the inter-annual variability in the growth rate of North Sea
herring (Clupea harengus) was explained by inter- and intra-
annual differences in prey abundance and temperature (Heath
et al., 1997). Marked spatial and temporal variability in poten-
tial prey for larval fish (i.e., copepods) was evident from the
data compiled during the GLOBEC–Germany North Sea field
sampling campaign in 2004. The impact of this prey field
variability on the potential survival and growth of larval fish
was unknown and formed the impetus for the present
research.
In the present study, we utilized an individual-based
modelling approach to investigate early life stage dynamics of
sprat (Sprattus sprattus), an abundant small pelagic fish spe-
cies inhabiting the southern North Sea. Modelled abiotic
factors such as temperature and advection were combined
with spatially and temporally variable prey fields that were
developed from repeated (monthly) in situ zooplankton sam-
pling on a station grid. The biological–physical model pro-
vided a tool to investigate potential spatial and temporal
changes in the habitat suitability of the southern North Sea for
sprat early life stages. In this case, optimal (most suitable)
Fig. 2. A) Maximum (upper decile of) prey lengths versus standard length for larval sprat (Sprattus sprattus); B) Distribution of preylengths found in the gut of larval
sprat in the Baltic and North Sea (data from Voss et al., 2003; Dickmann, 2005). The box height, line lengths and points represents the 5th, 10th, 25th, 75th, 90th and
95th percentiles. The median is indicated within each box.
331W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
habitats were defined as areas with prey resources and water
temperatures that supported relatively high rates of larval
growth. Specifically, we simulated the potential survival and
growth rates of 10–14 mm standard length (SL) larvae, the size
range where growth variability tends to be most marked
during the larval period in sprat (Munk,1993; Lee et al., 2006).
2. The coupled transport-growth model
The model we used consists of two submodels that can be
run either separately or in a coupled mode. The first one is a
Lagrangian transport model that allows forward or backward
trajectories of particles, like larval fish, to be calculated. The
second submodel is an IBM designed to describe the foraging,
growthand mortality (due to starvation) of larvalsprat. The IBM
was similar to other general models (e.g., Letcher et al., 1996)in
that it contains modules for calculating the encounter rate,
ingestion of food, metabolic losses and the resulting growth in
dry weight (DW, µg) and SL (mm) of larvae. Our model was also
parameterised for (non-feeding) egg and yolk–sac phases.
The transport model used daily means of current velocities
(averaged over two M
2
tide cycles) and water temperatures
calculated with the hydrodynamic model HAMSOM (Backhaus,
1985; Pohlmann, 2006). The short-term tidal variations were
neglected, since this study focuses on large-scale transport.
The model area comprised the region of the North Sea south of
57 °N. The drift experiments in the present study were per-
formed in a smaller area, the German Bight, where intensive
field sampling was conducted during GLOBEC–Germany
(s. Fig. 1A and B). The horizontal resolution of the model grid
was 3 km; in the vertical a z-coordinate system with 21 layers
was used, the upper 50 m were resolved by 5 m-layers. The
model was forced by meteorological data provided by NCEP. For
the open boundary at 57 °N hydrodynamic fields from a
HAMSOM run for the whole North Sea were used.
Young sprat larvae are typically captured in the upper
10 m of the water column (Conway et al., 1997). In the present
study, sprat larvae were fixed at 2.5 m depth and did not have
any vertical movements (due to neither physical nor bio-
logical mechanisms). Given the relatively short duration of
the runs used here (10 days) and the lack of knowledge on the
vertical behaviour of sprat larvae in the NorthSea, we chose to
use this somewhat “over-simplified”method. Future model-
ing studies are planned to examine drift in greater detail (and
over longer periods of time) that will employ various vertical
behaviour scenarios to examine the influence of changes in
depth on transport of fish larvae in the southern North Sea,
Table 1
Mean lengths Lm (in µm) of the nauplii stages (N1 to N6) of the three key
species, Acartia clausi,Pseudocalanus elongatus and Temora longicornis,
(according to Hays et al., 1988) and the corresponding model size classes
(SC) given in µm
Acartia clausi Pseudocalanus
elongatus
Temora longicornis
Lm [µm] SC [µm] Lm [µm] SC [µm] Lm [µm] SC [µm]
N1 126 200 166 200 115 200
N2 152 201 300 178
N3 179 274 222 300
N4 205 300 339 400 271
N5 237 403 500 332 400
N6 271 447 430 500
Fig. 3. Size class-resolved abundance of copepods (Temora longicornis,Pseudocalanus elongatus and Acartia spp.) sampled in April, 2004 using multinet gear at eight
different stations in the German Bight.
332 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
since vertical behaviour can be an important factor influen-
cing modeled larval vital rates and transport (Kristiansen
et al., 2007; Vikebø et al., 2007).
The IBM used in this study was thoroughly described
elsewhere (Peck and Daewel, 2007; Daewel et al., submitted
for publication). Most of the parameter estimates were
derived from laboratory studies on larval Atlantic herring
(metabolism and functional forms of prey capture success)
and field data on larval sprat (growth rates and gut contents).
Only the main features of the foraging and growth are
presented here. Larval growth rate (G, µg DW d
−1
) was cal-
culated as the difference between net dry weight (DW) of
consumed food and metabolic losses:
G¼C⁎AE ⁎1RSDA
ðÞRð1Þ
where the rate of consumed prey (C,µgDWd
−1
) was reduced
by an assimilation efficiency (AE, %) and metabolic losses (R,
µg DW d
−1
) which were divided into several components to
account for standard (R
S
), feeding (specific dynamic action,
R
SDA
) and active (R
A
) rates of energy loss. In Eq. (1), Rrepre-
sented R
S
at night and R
A
during daylight foraging hours.
Effects of larval DW and temperature on Rwere taken from
work on larval herring (Almatar, 1984; Kiørboe et al., 1987).
The amount of prey consumed daily was calculated as a
function of encounter rate (N
L
s
,i
), prey mass (m
i
, µg), capture
success (CS
L
s
,i
), and handling time (HT
L
s
,i
)(Letcher et al., 1996):
C¼Pimi⁎NLS;i⁎CSLS;i
1þPiNLS;i⁎HTLS;i
ð2Þ
An optimal foraging approach was used in which different
prey types were ranked according to their DW, capture
success and handling time. Prey items were sequentially
included into the diet based on ranking until profitability
decreased (see Letcher et al., 1996 and references therein).
Capture success depended upon prey length (pl
i
, mm) and
larval SL and was calculated using the functional form
reported by Munk (1992) for larval herring:
CSLS;i¼1:1a⁎pli
SL
 ð3Þ
where the parameter awas based upon the maximum prey
size pl
imax
:
a¼1:1⁎SL
plimax
ð4Þ
determined from the upper decile (90 to 100%) of prey sizes
found in the gut contents of 5.0 to 23 mm SL Baltic and North
Sea sprat (Fig. 2A). The handling time was calculated follow-
ing an empirically derived equation from Walton et al. (1992):
HTLS;i¼e0:264⁎107:0151 pli
SL
 ð5Þ
but re-parameterised using field data on larval sprat gut
contents (Dickmann, 2005). Overfeeding by larvae was elimi-
nated by employing a C
max
function:
Cmax ¼1:315⁎DW 0:83⁎2:872 T15ðÞ
10
½ ð6Þ
Fig. 4. Horizontal distribution of prey abundance (ind m
−3
) for two different copepod size classes in mid-April, mid-May and mid-June 2004. In the upper left panel,
triangles represent release locations, circles represent areas where zooplankton measurements were made.
333W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
yielding larval DW- and temperature (T, °C)- specific limits to
food consumption rate (µg dry weight of prey d
−1
) that ba-
lanced in situ estimates of DW- and T-specific sprat growth
(Munk, 1993; Ré and Gonçalves,1993; Huwer, 2004; Baumann
et al., 2006). In the model, Gwas partitioned between DW (µg)
and SL (mm) depending upon the condition factor (/):
U¼1000⁎DW
SL5:022 :ð7Þ
If growth in DW was positive and if /≥1.0, SL increased
according to equations provided by Peck et al. (2005).Ifgrowthin
DW was continuously negative, the larva died (was removed
from the simulation) when /b0.75 based upon the lowest values
of /calculated for field-caught larval sprat (see data in Peck et al.,
2005).
3. Environmental conditions (prey and temperature)
As part of the GLOBEC–Germany project, in situ zooplank-
ton sampling was conducted on a station grid in the German
Bight, North Sea during several periods from February to
October 2004. At all stations, standard hydrodynamic data
Fig. 5. Surface temperature in the German Bight as obtained from the hydrodynamic model HAMSOM for spring/early summer 200 4. A) mid-April, B) mid-May, C) mid-June.
334 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
were collected with vertically-averaged zooplankton concen-
trations (ind. m
−3
) using Multinet gear (mouth opening
0.25 m
2
, 50 µm mesh size). In this study we concentrated
on the following time periods: i) mid-April, ii) mid-May,
iii) mid-June 2004.
Examination of larval fish gut contents in the North Sea
indicated that larval sprat mainly fed on the nauplii and
copepodite stages of three copepods: Acartia spp., Temora
longicornis and Pseudocalanus elongatus (Dickmann, 2005).
Therefore, prey fields in the present study were calculated as
the sum of the abundances of these three copepods. In con-
trast to the species/stage approach used in many IBMs (e.g.,
Letcher et al., 1996; Hinrichsen et al., 2002; Lough et al.,
2005), we refined the prey field by using a series of 100 µm
prey size classes based upon measurements of size-at-stage
made on field samples. Utilizing size classes allowed us to
develop a prey field that better reflected the changes in prey
items found with increasing body size in sprat gut contents
(Dickmann, 2005, see Fig. 2B). Moreover, utilizing prey size
classes allowed us to parameterize the optimal foraging ap-
proach described previously.
Since no stage or length data on copepod nauplii were
collected in multinet samples, a scheme to distribute the total
nauplii abundance within 100 µm prey length classes was
devised. The method was based upon the assumptions of
1) an exponential stage distribution, 2) daily mortality rates of
0.1 d
−1
(C. Möllmann, IHF University of Hamburg, pers. com-
munication), and 3) steady-state conditions during which the
succession of stage abundance should mirror the frequency
distribution of stages observed in the field. Mean, species-
specific copepod naupliar length values were used to distri-
bute the nauplii stages to the prey size spectrum employed in
Fig. 6. Drift trajectories of larval sprat in the German Bight for A) mid-April, B) mid-Mayand C) mid-June 20 04. The colour denotes the dayaf ter release (blue =day 0,
red=day 10). The same release locations (n= 20) were used in each month.
335W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
the IBM and the relative frequencies of nauplii within the size
range 200–500 µm, for each of the three key species were
obtained (Table 1,Fig. 3).
On each of four sampling dates, the abundance of prey
within each 100-µm length class was interpolated onto the
more highly spatially-resolved hydrodynamic model grid by
objective analysis (Bretherton et al., 1976). The technique is
based on the Gauss–Markov theorem, a standard statistical
method that yields an expression for the linear least-square
error estimate of variables based on data sets containing only
a limited number of measurements (in this case, stations). A
total of 11,100-µm length classes (between 200 and 1300 µm)
was included in the calculation of prey fields to account for
the sizes of prey utilized by sprat larvae N10 mm SL. In order
Fig. 7. Temperatures (panel A) and prey abundances in three size classes (Panels B–D) experienced by larvae as well as modelled larval weight (Panel E) versus time
for the 10-day drift period in mid-April 2004. The black symbols always denote selected larvae, the starting position of which is given in the bottom panel.
336 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
to get an approximate estimate of the temporal changes in
prey abundance, daily linear interpolation of prey length
categories was performed for each model grid point. In this
manner, we were able to assign spatially- and temporally-
resolved, length-specific prey abundances (e.g., see Fig. 4)at
each point along the drift trajectory of each larval sprat.
Since the ability of areas to support high rates of growth
not only depends upon prey resources but on the interaction
between prey and temperature, daily temperature fields for
the model domain were provided by HAMSOM (Fig. 5). The
ability of HAMSOM to accurately reproduce the annual tem-
perature cycle in the North Sea has been previously demon-
strated by Pohlmann (1996).
4. Simulations
In this study, we examined habitat suitability by simulat-
ing the daily drift, feeding (by prey size class) and growth of
sprat released at a size of 10 mm SL. Maximum prey size of
sprat larvae b10 mm SL changes little and prey size increases
rapidly for 10 to 14 mm SL larvae. By releasing 10-mm larvae,
we attempted to model a life stanza that would respond most
dynamically (compared to smaller and larger larvae) to varia-
bility in copepod population demographics (e.g., changes in
both prey abundance and in stage/size distribution). As
release locations, we chose 20 of the German GLOBEC sam-
pling stations (Fig. 4). As release periods, we chose mid-April
(day 100), mid-May (day 130) and mid-June (day 160). These
periods overlapped with our in situ sampling and were cha-
racterized by very different abiotic (temperature, prevailing
current velocities) and biotic (prey densities) conditions
(Fig. 4). Larvae were restricted to the upper water layer,
since young sprat larvae tend to prefer the upper part of the
water column during daytime (Kloppmann, 1991; Conway
et al., 1997). Larval sprat are visual feeders (Dickmann, 2005)
and were, therefore, only allowed to feed during daylight
hours (photoperiods corresponding to the day of the year).
Prey fields based upon Multinet samples collected at the same
station were depth-averaged. Simulations lasted 10 days, a
time period determined in pilot simulations to be long
enough to allow sprat to grow through 14 mm in warmer
months.
5. Results
Model results indicated that, in mid-April 2004, the larvae
passively drifted mostly in a north-eastern direction, except
those released in the eastern part of the German Bight, which
drifted northward (Fig. 6A). Surface water temperatures in the
German Bight were between 6.5 and 8.0 °C, except in the
somewhat warmer, coastal region of the Wadden Sea (Fig. 7A).
During the 10-day simulation, water temperature increased by
about 1.5 °C. Larvae fed on prey within 300, 400 and 500 µm
size classes and concentrations of copepods in these size
classes were very low, generally b2 ind. l
−1
(Fig. 7B–D).
Nevertheless, in most areas, larvae were able to grow, albeit
slowly (growth rates of ~ 0.2 mm d
−1
, reaching a maximum SL
of ~12 mm and DW of 275 µg). However, in some regions, prey
concentrations were much lower (close to zero) and larvae
released in these areas (i.e., southern part of the German Bight,
north of the Frisian Islands between 5 °E and 7.5 °E), lost
weight (had negative growth rates) and suffered from star-
vation (Fig. 7E). At the 20 release locations in mid-April, none
of the larvae died although 3 of them (i.e. 15%) were in areas
that did not support positive growth rates in weight over most
(8 to 10 days) of the period examined.
In mid-May the drift pattern of the larvae was more
complex than in April: larvae at the northern boundary drifted
in south-eastern direction, whereas larvae released at the
southern stations drifted mainly eastward; larvae at the
north-eastern part of the region remained more or less in
the same location (s. Fig. 6B). Compared to mid-April, mean
surface temperature of the German Bight was several degrees
warmer (10–12 °C) and increased further by about 1.5 °C
during the 10-day simulation (Fig. 8A). During the first 5 days
of the simulation, prey concentrations were not much higher
than in April, but concentrations increased during the second
5-day period in most areas (Fig. 8B–D). With respect to larval
growth, two distinct groups could be discriminated. The best
growth conditions (with growth rates between 0.22 and
0.42 mm d
−1
) occurred within the inner region of the German
Bight, near the Schleswig–Holstein coast, where larvae
reached maximum weights of about 400 µg (13 mm SL).
Distinctly worse conditions (prey concentrations b0.5 ind. l
−1
)
prevailed north of the Frisian Islands (between 5.5 °E and 8 °E).
Especially, larvae released at positions 9 and 10 did not
develop successfully (Fig. 8E). In mid-May, larvae within 6 of
the release locations experienced starvation conditions (lost
weight) for at least some portion (1 to 4 days) of the simulated
period.
In mid-June the drift pattern of the larvae was similar to
that of May, with a slightly stronger eastward component
(Fig. 6C). The mean surface water temperature in the German
Bight was about 14.5 °C (range: 13–16 °C), remaining more or
less the same during the 10-day simulation, except for a slight
cooling on the tenth day in the eastern part (Fig. 9A).
Compared to mid-May, the concentration of available food
in mid-June was drastically increased in some places, espe-
cially for the prey size class of 400 µm, that reached
concentrations between 5 and10 ind. l
−1
in the inner German
Bight (Fig. 9B–D). Similar to mid-May, spatially-explicit esti-
mates of growth rates indicated distinctly different habitats
experienced by larvae. At release locations near the Schles-
wig–Holstein coast, larvae experienced very favourable
growth conditions (growth rates N0.5 mm d
−1
) , where maxi-
mum weights of 680 to 880 µg (~15 mm SL) were reached
after 10 days. In contrast to mid-May, larvae in the southern
part of the German Bight (north of the Frisian Islands) in June
also experienced good growth conditions having growth rates
of 0.4 mm d
−1
. Larvae released at the northern edge of the
model domain, however, experienced only moderate growth
conditions, and extremely bad conditions existed for the
larvae released at three location at the south-western edge of
the study area (at about 4 °E) where starvation conditions
prevailed for several days (Fig. 9E). Spatially-explicit esti-
mates of growth rates (mm d
−1
) in June clearly depicted areas
that were expected to be good, suitable habitats where
growth rates were relatively high and areas that were poor,
unsuitable habitats where growth rates were relatively low
(Fig. 10). An indexof habitat suitability was assigned to ranges
in growth rates according to otolith-based growth rates of
larval sprat captured in the field (see Discussion section).
337W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
6. Discussion
During early life stages of marine fish, rates of survival are
often positively correlated with rates of growth (Houde, 1987;
Cowan and Shaw, 2002) and the latter are most influenced by
differences in temperature and food availability (Heath, 1992).
Optimal habitats, therefore, are areas with high prey resources
and adequate temperatures that support relatively high rates
of growth. According to our modelling study there was
considerable regional and temporal variability with respect
to the habitat suitability of the southern North Sea for larval
sprat. This conclusion is based upon spatially-explicit esti-
mates of potential survival and growth of sprat larvae during
spring/early summer 2004 in the German Bight. The prey
Fig. 8. Temperatures (panel A) and prey abundances in three size classes (PanelsB–D) experienced by larvae as well as modelled larval weight (Panel E) versus time
for the 10-day drift period in mid-May 2004. The black symbols always denote selected larvae, the starting position of which is given in the bottom panel.
338 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
concentrations in the size classes 300–700 µm were highly
variable in space and time which determined, to a large extent,
the growth potential of the 10 mm larvae. During this period,
spatial variability in surface water temperatures also con-
tributed to between-station differences in projected larval
growth rates.
One remarkably constant result of our study was that larvae
released within the inner (eastern) German Bight during each of
the 3 months we simulated encountered the highest food
concentrations and temperatures and, therefore, had optimal
conditions for survival and growth. This tends to support the
idea of the German Bight as an important spawning location for
Fig. 9. Temperatures (panel A) and prey abundances in three size classes (Panels B–D) experienced by larvae as well as modelled larval weight (Panel E) versus time
for the 10-day drift period in mid-June 2004. The black symbols always denote selected larvae, the starting position of which is given in the bottom panel.
339W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
adult sprat (Alheit et al.,1987) and nursery area for young sprat
larvae (Valenzuela et al., 1991). For the whole German Bight,
however, sucha general conclusion cannot be drawn because of
the high variability in prey resources.
The present study was the first to utilize both observed
(in situ)preyfields along with (simulated) meso-scale hydro-
graphic features to examine short-term variability in larval
fish growth rates in the southern North Sea. In thisLagrangian
approach, larvae experienced prey fields that were both
spatially-resolved (via objective analysis) and temporally-
resolved (via linear interpolation between in situ sampling
dates). To minimize the influence of such “averaging”,we
focused on short-term (10-day) growth simulations and re-
leased larvae on dates that correspond to in situ zooplankton
sampling. Using model-derived prey fields, Daewel et al.
(submitted for publication) reported on the impact of spatial
and temporal differences in prey concentrations and tem-
peratures on rates of survival and growth of sprat eggs and
young (b12 mm SL) larvae over the entire North Sea. In that
study, a simulation for 1993 projected clear differences
between the central and southern areas of the North Sea.
Larvae in southern areas were projected to have higher rates
of survival and growth compared to the central and northern
areas of the North Sea, especially during the warmer summer
months (June and July). Our estimates tended to agree with
those of Daewel et al. (submitted for publication) and sug-
gested that northern regions of the German Bight, at times,
can be poor areas for survival and growth of young fish larvae.
These modelling results agree well with the field distribution
and timing of sprat spawning which encompasses broad areas
of the southern North Sea from April through June (Alheit
et al., 1987). The multiple batch-spawning strategy of sprat
and other clupeids (Alheit et al., 1987)) is thus a necessary
life history strategy to cope with the temporal (and spatial)
match-mismatch dynamics of suitable habitats (e.g., driven
by changes in prey availability estimated in the present study)
and to promote early life stage survival and growth (Cushing,
1990).
The present modeling study utilized size-structured, in situ
prey fields consisting of the three copepods (Pseudocalanus,
T. longicornis and Acartia spp) found most frequently in the gut
contents of larval sprat (Dickmann, 2005). By resolving the
prey field into 100 µm size bins we were able to employ an
optimal foraging approach that provided estimates of the
changes in prey sizes utilized by the larvae as they grew from
10 mm to 15 mm SL. Optimal foraging theory suggests that
predators should attempt to maximize energy gain while
minimizing energy loss (Stephens and Krebs, 1986). Previous
studies on larvae of marine fish species such as Atlantic her-
ring and Atlantic cod have quantified optimal prey/predator
size ratios between 0.04 and 0.06, and noted that this ratio
generally increased with increasing larval age and/or size (e.g.,
Munk 1992). In the present study, prey transitions were
evident during the 10-day model runs in the fastest growing
larvae. For example, the fastest-growing larvae in April
utilized 300 µm prey for 9 of the 10 days whereas, in warmer
months (May and June), the fastest-growing larvae abandoned
feeding on the 300 µm size class after day six and day four,
respectively (see Figs. 7–9). More importantly, the fastest-
growing larvae started to utilize larger prey (e.g., 500 µm)
progressively sooner with increasing temperature; the switch
to 500 µm prey occurred on day six, day four, and day three
in April (7.5 to 9.5 °C), May (12 to 14 °C) and June (16.0 to
16.5 °C), respectively. Furthermore, threshold concentrations
were apparent, below which some larvae (e.g., crosses in
Figs. 7–9) were unable to grow fast enough to transfer onto
larger prey items. The importance of larger prey items for
larval sprat growth and survival has been previously inferred
from field studies (Voss et al., 2006; Dickmann et al., 2007).
Specifically, data available from 11 cruises between April and
July 2002 on stage-resolved abundance of copepods and the
abundance and condition of sprat larvae in different size
Fig. 10. Spatial distribution of larval growth rate (mm d
−1
) for mid-June 2004 obtained from the IBM simulation. The colours also denote habitat suitability (dark
colours: bad; light colours: good).
340 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
classes suggested sprat larvae N11 mm SL suffered high mor-
tality when concentrations of relatively large copepods (late
stage copepodites and adults) were ≤5–8 ind l
−1
(Voss et al.
2006). Moreover, Dickmann et al. (2007) suggested that
variability in prey fields experienced by medium-sized sprat
larvae likely contributes most to inter-annual variability in
survival of this species in the Baltic. They based this conclusion
on gut content analyses and narrow trophic niche breadths
calculated for sprat larvae ≥16 mm SL (Dickmann et al., 2007).
Our model results provide an estimate of how rapidly 10 mm
SL sprat larvae could transfer to larger prey sizes while opti-
mally foraging, and the importance of a spatial–temporal
match in both prey concentration and prey size to fueling
rapid growth in this larval sprat size range.
In order to gain more robust estimates of habitat suita-
bility, our IBM estimates of larval sprat growth rate need to be
viewed in light of in situ larval growth rates observed for this
species. Otolith-based growth rate estimates for sprat were
available from a number of studies conducted in the North,
Irish and Baltic Seas. In most cases, mean growth rates of
larvae in the southern North Sea were between 0.37 and
0.48 mm d
−1
(Valenzuela and Vargas, 2002, Ré and Goncalves,
1993; Huwer, 2004). Slightly higher mean growth rates were
reported for the Baltic and Irish Seas (Shields, 1989; Lee et al.,
2006; Dänhardt et al., 2007). While mean growth rates pro-
vided a good benchmark for model growth comparison, in-
formation on growth variability among larval sprat at specific
body sizes was also useful. The latter helped us to target the
size range of sprat used in our simulations of habitat suita-
bility. For example, sprat captured in May and June in the Irish
Sea (7–31 mm SL, 9–12 °C) had the greatest variability in
otolith-based growth rates (between ~ 0.3 and 0.8 mm d
−1
)
between 12 and 16 mm SL (Lee et al., 2006). In a second study,
otolith-based growth rates of individuals captured at different
sites along a frontal zone in the North Sea startedto diverge at
~10 mm SL and were distinctly different by 15 mm SL (Munk,
1993; his Fig. 10B). Thus, field observations on larval sprat size
and growth variability indicated that the changes in habitat
suitability might be most critical for growth rates of 10 to
14 mm SL larvae.
Variability in otolith-based growth rates of field fish also
enabled us to better interpret our IBM results. Our growth rate
projections were between 0.15 and 0.6 mm d
−1
depending
upon release location and date. The lowest growth rates
projected for 10–15 mm SL larvae in this study were lower
than the lowest rates observed for field-caught larvae in the
same size range. For example, of the otolith-based growth
rates of larval sprat captured in August (15–16 °C) north of the
German Bight (between 54.0–56.5°N and 6.5–8.0°E) only
10% were b0.25 mm d
−1
and no larvae had growth rates
b0.20 mm d
−1
. For sprat captured in May/June (9–12 °C) in the
Irish Sea, growth rates b0.25 mm d
−1
were not observed in 10
to 15 mm SL fish (Lee et al., 2006). It appears, therefore,
that for larval sprat experiencing spring/early summer water
temperatures, individuals with growth rates b0.25 mm d
−1
are removed from populations (perhaps via predation).
In our study, larvae in May and June with growth rates of
b0.25 mm d
−1
were considered to inhabit the poorest habitats
and larvae with the highest growth rates (0.3 to 0.58 mm d
−1
)
to be in optimal habitats. Using this approach, we were able to
generate maps of habitat suitability (e.g., see Fig. 10, June
2004). Intermediate habitats existed at the northern edge of
the area investigated and larvae clearly benefited from inha-
biting areas closer to the coast and within the German Bight.
When used in the manner above, otoliths provide
information on lifetime growth rates of individuals. In the
case of 10–15 mm SL sprat, these growth rates arise from the
effects of abiotic and biotic factors interacting over a period of
~20 to 30 days. On the German GLOBEC station grid, our
model predicted areas that were consistently poor (unsui-
table habitats) and others that were consistently favourable
(suitable habitats) during the three months examined. How-
ever, in some instances, modelled larval growth rates
responded rapidly (~5 days) to changing environmental
characteristics experienced along their drift trajectory, a re-
sult that seems reasonable given the dynamic nature of
frontal regions such as our study region in the southern North
Sea (e.g., Munk and Nielsen, 1994). For this reason, future
model simulations will be compared to measurements made
on shorter-term proxies of in situ growth and condition (e.g.,
such as biochemical analyses) for sprat collected at the same
stations used here. Such an examination will enable us to
determine whether habitat suitability changes on time scales
reflected in our model results or whether spatial differences
in habitat suitability are more persistent features of the
southern North Sea as inferred, for example, from lifetime
growth rates of larval fish.
Acknowledgments
We would like to thank the crews and scientists that
collected the in situ zooplankton data that were used in this
study. This project was funded by the GLOBEC–Germany pro-
gram (FKZ # 03F0320E).
References
Alheit, J., Wahl, E., Cihangir, B., 1987. Distribution, Abundance, Development
Rates, Production and Mortality of Sprat Eggs. ICES, p. 45. C.M. H.
Almatar, S.M., 1984. Effects of acute changes in temperature and salinity on
the oxygen uptake of larvae of herring (Clupea harengus) and plaice
(Pleuronectes platessa). Mar. Biol. 80, 117–124.
Backhaus, J.O., 1985. A three-dimensional model for the simulation of shelf
sea dynamics. Deutsche Hydrographische Zeitschrift 38, 165–187.
Bartsch, J., Coombs, H., 2004. An individual-based model of the early life
history of mackerel (Scomber scombrus)intheeasternNorthAtlantic,
simulating transport, growth and mortality. Fish. Oceanogr. 13,
365–379.
Baumann, H., Gröhsler, T., Kornilovs, G., Makarchouk, A., Feldmann, V.,
Temming, A., 2006. Temperature-induced regional and temporal growth
differences in Baltic young-of-the-year sprat Sprattus sprattus. Mar. Ecol.
Prog. Ser. 317, 225–236.
Beaugrand, G., Brander, M., Lindley, J.A., Souissi, S., Reid, P.C., 2003. Plankton
effect on cod recruitment in the North Sea. Nature 426, 661–664.
Brandt, S.B., Kirsch, J., 1993. Spatially explicit models of striped bass growth
potential in Chesapeake Bay. Trans. Amer. Fish. Soc. 122, 845–869.
Bretherton, F.P., Davis, R.E., Fandry, C.B., 1976. A technique for objective
analysis and design of oceanographic experiment applied to MODE-73.
Deep-sea Res. 23, 559–582.
Conway, D.V.P., Coombs, S.H., Smith, C., 1997. Vertical distribution of fish eggs
and larvae in the Irish Sea and southern North Sea. ICES J. Mar. Sci. 54,
136 –147.
Cowan Jr., J.H., Shaw, R.F., 2002. Recruitment. In: Fuiman, L.A., Werner, R.E.
(Eds.), The Unique Contribution of Early Life Stages. Blackwell Publishing,
pp. 88–111.
Cushing, D.H., 1990. Plankton production and year class strength in fish
populations: an update of the Match/Mismatch Hypothesis. Adv. Mar.
Biol. 26, 249–293.
341W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342

Author's personal copy
Dänhardt, A., Peck, M.A., Clemmesen, C., Temming, A., 2007. Depth-
dependent nutritional condition of sprat larvae in the central Bornholm
Basin, Baltic Sea. Mar. Ecol. Prog. Ser. 341, 217–228.
Daewel, U., Peck, M.A., Kühn, W., St. John, M.A., Schrum, C., submitted for
publication. Coupling ecosystem and individual-based models to
simulate the influence of climate variability on potential growth and
survival of larval fish in the North Sea. Fish. Oceanogr.
Dickmann, M., 2005. Feeding ecology of sprat (Sprattus sprattus L.,) and
sardine (Sardine pilchardus W.) larvae in the Baltic Sea and in the North
Sea. PhD. Thesis, Institute für Ostseeforschung, University of Rostock.
Dickmann, M., Möllmann, C., Voss, R., 2007. Feeding ecology of Central Baltic
sprat larvae (Sprattus sprattus) larvae in relation to zooplankton
dynamics: implications for survival. Mar. Ecol. Prog. Ser. 342, 277–289.
Hays, S.J., Evans, G.T., Gamble, J.C., 1988. Birth, growth and death rates for
enclosed populations of calanoid copepods. J. Plankton Res. 10, 431–454.
Heath, M.R., 1992. Field investigations of the early life stages of marine fish.
Adv. Mar. Biol. 28, 1–174.
Heath, M., Gallego, A., 1997. From the biology of the individual to the
dynamics of the population: bridging the gap in fish early life studies.
J. Fish Biol.51 (Suppl. A), 1–29.
Heath, M., Scott, B., Bryant, A.D., 1997. Modelling the growth of herring from
four different stocks in the North Sea. J. Sea Res. 38, 413–436.
Hinrichsen, H.-H., Möllmann, C., Voss, R., Koester, F.W., Kornilovs, G., 2002.
Biophysical modeling of larval Baltic cod (Gadus morhua) growth and
survival. Can. J. Fish. Aquat. Sci. 59, 1858–1873.
Houde, E.D., 1987. Fish early life dynamics and recruitment variability. Am.
Fish. Soc. Symp. 2, 17–29.
Huwer, B., 2004. Larval growth of Sardina pilchardus and Sprattus sprattus in
relation to frontal systems in the German Bight. Diplomarbeit. Christian-
Albrechts-Universität Kiel. 108 pp.
Kiørboe, T., Munk, P., Richardson, K., 1987. Respiration and growth of larval
herring Clupea harengus: relation between specific dynamic action and
growth efficiency. Mar. Ecol. Prog. Ser. 40, 1–10.
Kloppmann, M., 1991. Vertical Distribution of Sprat Larvae in the German
Bight. ICES C.M., p. L:26.
Kristiansen, T., Fiksen, Ø., Folkvord, A., 2007. Modelling feeding growth and
habitat selection in larval Atlantic cod (gadus morhua): observations and
model predictions in a macrocosm. Can. J. Fish. Aquat. Sci. 64, 136–151.
Lee, O., Danilowicz, B.S., Dickey-Collas, M., 2006. Temporal and spatial
variability in growth and condition of dab (Limanda limanda) and sprat
(Sprattus sprattus) larvae in the Irish Sea. Fish. Oceanogr.15, 490–507.
Letcher, B.H., Rice, J.A., Crowder, L.B., Rose, K.A., 1996. Variability in survival of
larval fish: disentangling components with a generalized individual-
based model. Can. J. Fish. Aquat. Sci. 53, 787–801.
Lough, R.G., Buckley, L.J., Werner, F.E., Quinlan, J.A., Edwards, K.P., 2005. A
general biophysical model of larval cod (Gadus morhua) growth applied
to populations on Georges Bank. Fish. Oceanogr. 14, 241–262.
Mason, D.M., Brandt, S.B., 1996. Effects of spatial scale and foraging efficiency
on the predictions made by spatially-explicit models of fish growth rate
potential. Environ. Biol. Fishes 45, 283–298.
Munk, P., 1992. Foraging behaviour and prey size spectra of larval herring
Clupea harengus. Mar. Ecol. Prog. Ser. 80, 149–158.
Munk, P., 1993. Differential growth of larval sprat Sprattus sprattus across a
tidal front in the eastern North Sea. Mar. Ecol. Prog. Ser. 99, 17–27.
Munk, P., Nielsen, T.G., 1994. Trophodynamics of the plankton community at
Dogger Bank: predatory impact by larval fish. J Plankton Res 16,
1225–1245.
Neill, W.H., Miller, J.M., van der Veer, H.W., Winemiller, K.O., 1994.
Ecophysiology of marine fish recruitment: a conceptual framework for
understanding interannual variability. Neth. J. Sea Res. 32, 135–152.
Peck, M.A., Daewel, U., 2007. Physiologically-based limits to food consump-
tion and individual-based modelling of foraging and growth of larval fish.
Mar. Ecol. Prog. Ser. 347, 171–183.
Peck, M.A., Clemmesen, C., Herrmann, J.-P., 2005. Ontogenic changes in the
allometric scaling of the mass and length relationship in Sprattus sprattus.
J. Fish Biol. 66, 882–887.
Pohlmann, T., 1996. Predicting the thermocline in a Circulation Model of the
North Sea–Part I: Model description, calibration and verification. Cont.
Shelf Res. 16, 131–146.
Pohlmann, T., 2006. A meso-scale model of the central andsouthern North Sea:
consequences of an improved resolution. Cont. Shelf Res. 26, 2367–2385.
Ré, P., Gonçalves, E., 1993. Growth of sprat Sprattus sprattus larvae in the
German Bight (North Sea) as inferred from otolith microstructure. Mar.
Ecol. Prog. Ser. 96, 139–145.
Rothschild, B.J., 1998. Year class strengths of zooplankton in the North Sea and
their relation to cod and herring abundance. J. Plankton Res. 20,
1721–1741.
Shields, R.J. 1989. Studies of growth and nutritional status in 0-group sprat,
Sprattus sprattus (Clupeidae), using otolith microstructure and lipid
analysis techniques. PhD thesis, University of Wales, Bangor.
Stephens, D.W., Krebs, J.R.,1986. Foraging Theory. Princeton University Press,
Princeton.
Valenzuela, G., Alheit, J., Coombs, S.H., Knust, R., 1991. Spawning Patterns of
Sprat and Survival Chances of Sprat Larvae in Relation to Frontal Systems
in the German Bight. ICES C.M., p. L:45.
Valenzuela, G.S., Vargas, C.A., 2002. Comparative growth rate of Sprattus
sprattus in relation to physical and oceanographic features in the North
Sea. Arch. Fish. Mar. Res. 49, 213–230.
Vikebø, F., Jørgensen, C., Kristiansen, T., Fiksen, Ø., 2007. Drift, growth and
survival of larval Northeast Arctic cod with simple rules of behaviour.
Mar. Ecol. Prog. Ser. 347, 207–219.doi:10.3354/meps06979.
Voss, R., Clemmesen, C., Baumann, H., Hinrichsen, H.-H., 2006. Baltic sprat
larvae: coupling food availability, larval condition and survival. Mar. Ecol.
Prog. Ser. 308, 243–254.
Voss, R., Köster, F.W., Dickmann, M., 2003. Comparing the feeding habits of
co-occurring sprat (Sprattus sprattus) and cod (Gadus morhua) larvae in
the Bornholm Basin, Baltic Sea. Fish Res 1497, 1–15.
Walton, W.E., Hairston, N.G., Wetterer, J.K., 1992. Growth-related constrains
on diet selection by sunfish. Ecology 73, 429–437.
Werner, F., Perry, R.I., Lough, R.G., Naimie, C.E., 1996. Trophodynamic and
advective influences on Georges Bank larval cod and haddock. Deep-sea
Res. II 43, 1793–1822.
342 W. Kühn et al. / Journal of Marine Systems 74 (2008) 329–342