ArticlePDF Available

Warming temperatures and smaller body sizes: Synchronous changes in growth of North Sea fishes

December 2013
Global Change Biology 20(4)

DOI:10.1111/gcb.12514

Source
PubMed

Authors:

Alan Baudron

Marine Scotland Science

Coby Needle

Marine Scotland Science

Adriaan D. Rijnsdorp

Wageningen University & Research

Decreasing body size has been proposed as a universal response to increasing temperatures. The physiology behind the response is well established for ectotherms inhabiting aquatic environments: as higher temperatures decrease the aerobic capacity, individuals with smaller body sizes have a reduced risk of oxygen deprivation. However, empirical evidence of this response at the scale of communities and ecosystems is lacking for marine fish species. Here we show that over a 40-year period six of eight commercial fish species in the North Sea examined underwent concomitant reductions in asymptotic body size with the synchronous component of the total variability coinciding with a 1-2°C increase in water temperature. Smaller body sizes decreased the yield-per-recruit of these stocks by an average of 23%. Although it is not possible to ascribe these phenotypic changes unequivocally to temperature, four aspects support this interpretation: (i) the synchronous trend was detected across species varying in their life history and life style, (ii) the decrease coincided with the period of increasing temperature, (iii) the direction of the phenotypic change is consistent with physiological knowledge and (iv) no cross-species synchrony was detected in other species-specific factors potentially impacting growth. Our findings support a recent model-derived prediction that fish size will shrink in response to climate-induced changes in temperature and oxygen. The smaller body sizes being projected for the future are already detectable in the North Sea. This article is protected by copyright. All rights reserved.

Content uploaded by Alan Baudron

Content may be subject to copyright.

For Review Only

Warming temperatures and smaller bo

dy sizes:

synchronous changes in growth of North Sea fishes

Journal:

Global Change Biology

Manuscript ID:

Draft

Wiley - Manuscript type:

Opinion

Date Submitted by the Author:

n/a

Complete List of Authors:

Baudron, Alan; University of Aberdeen, Zoology

Needle, Coby; Marine Scotland - Science, Marine Laboratory

Rijnsdorp, Adriaan D; IMARES, Fisheries

Marshall, Tara; University of Aberdeen, Zoology

Keywords:

climate change, temperature size rule, ectotherms, fish growth, von

Bertalanffy, Dynamic Factor Analysis, fisheries

Abstract:

Decreasing body size has been proposed as a universal response to

increasing temperatures. The physiology behind the response is well

established for ectotherms inhabiting aquatic environments: higher

temperatures decrease the aerobic capacity of individuals giving smaller

body sizes a fitness advantage through reduced risk of oxygen deprivation.

However, empirical evidence of this response at the scale of communities

and ecosystems is lacking for marine fish species. Here we show that over

a 40-year period six of the eight commercial fish species in the North Sea

examined underwent a synchronous reduction in asymptotic body size that

coincided with a 1-2ºC increase in water temperature. Smaller body sizes

decreased the yield-per-recruit of these stocks by an average of 23%.

Although it is not possible to ascribe these phenotypic changes

unequivocally to temperature, four aspects support this interpretation: (i)

the synchronous trend was detected across species varying in their life

history and life style, (ii) the decrease coincided with the period of

increasing temperature, (iii) the direction of the phenotypic change is

consistent with physiological knowledge and (iv) no synchrony was

detected in other species-specific factors potentially impacting growth. Our

findings support a recent model-derived prediction that fish size will shrink

in response to climate-induced changes in temperature and oxygen. The

smaller body sizes being projected for the future are already detectable in

the North Sea.

Global Change Biology

For Review Only

Warming temperatures and smaller body sizes: synchronous changes in growth of

North Sea fishes

Running head: Warming temperatures and smaller fish sizes

Alan R. Baudron

, Coby L. Needle

, Adriaan D. Rijnsdorp

and C. Tara Marshall

Institute of Biological and Environmental Sciences, University of Aberdeen, Tillydrone

Avenue, Aberdeen, AB24 2TZ, Scotland, UK.

Marine Scotland - Science, Marine Laboratory, PO Box 101, 375 Victoria Road, Aberdeen,

AB11 9DB, Scotland, UK.

IMARES, Institute of Marine Resources, and Ecosystem Studies and Aquaculture and

Fisheries Group, Haringkade 1, 1976 CP IJmuiden, The Netherlands.

Author to whom correspondence should be addressed

E-mail: alan.baudron@abdn.ac.uk, Telephone: +44 (0)1224-272648, Fax: +44 (0)1224-

272396

Key words: climate change, temperature size rule, ectotherms, fish growth, von Bertalanffy,

Dynamic Factor Analysis, fisheries.

OPINION

Page 1 of 40 Global Change Biology

For Review Only

Abstract

Decreasing body size has been proposed as a universal response to increasing temperatures.

The physiology behind the response is well established for ectotherms inhabiting aquatic

environments: higher temperatures decrease the aerobic capacity of individuals giving

smaller body sizes a fitness advantage through reduced risk of oxygen deprivation. However,

empirical evidence of this response at the scale of communities and ecosystems is lacking for

marine fish species. Here we show that over a 40-year period six of the eight commercial fish

species in the North Sea examined underwent a synchronous reduction in asymptotic body

size that coincided with a 1-2ºC increase in water temperature. Smaller body sizes decreased

the yield-per-recruit of these stocks by an average of 23%. Although it is not possible to

ascribe these phenotypic changes unequivocally to temperature, four aspects support this

interpretation: (i) the synchronous trend was detected across species varying in their life

history and life style, (ii) the decrease coincided with the period of increasing temperature,

(iii) the direction of the phenotypic change is consistent with physiological knowledge and

(iv) no synchrony was detected in other species-specific factors potentially impacting growth.

Our findings support a recent model-derived prediction that fish size will shrink in response

to climate-induced changes in temperature and oxygen. The smaller body sizes being

projected for the future are already detectable in the North Sea.

Page 2 of 40Global Change Biology

For Review Only

Introduction

Aquatic environments pose inherent challenges for ectothermic organisms respiring

underwater (Graham, 2006). Warming temperatures compound these challenges by

increasing anabolic oxygen demand while decreasing oxygen solubility. Any imbalance

between oxygen demand and oxygen supply will constrain aerobic scope thereby impairing

individual performance (Pörtner & Knust, 2007). In warming environments, smaller-sized

individuals are better able to balance demand and uptake because of their larger surface area

to volume ratio

(Pauly, 2010). These physiological constraints lead to the expectation that

individuals experiencing higher temperatures will have smaller body sizes, an outcome

known as the temperature-size rule (TSR) (Atkinson, 1994). The physiological basis

underpinning the TSR (Pörtner & Knust, 2007; Forster et al., 2011) combined with cross-taxa

support (Gardner et al., 2011; Forster et al., 2012; Edeline et al., 2013) has led to smaller

body size being proposed as a universal outcome of warming temperatures

(Daufresne et al.,

2009). In marine ecosystems which include a high proportion of ectothermic species, the

implications of the TSR are profound. A recent simulation integrating this ecophysiological

understanding with temperature projections predicted that by 2050 the assemblage-averaged

maximum body weight of fish species would shrink by 14-24% globally due to the combined

impacts of smaller-sized species replacing larger-sized species and the TSR (Cheung et al.,

2013). This conclusion, which garnered global press coverage upon publication, has been

challenged on the grounds that the scale and the speed of the change are not credible (Brander

et al., in press). Criticism of the projection model was refuted by the authors (Cheung et al.,

in press). This debate highlights the need for an ecosystem-level test of whether body sizes of

fishes have synchronously decreased in regional seas that have undergone warming. While

the importance of TSR has been shown in laboratory conditions (Forster et al., 2012),

Page 3 of 40 Global Change Biology

For Review Only

empirical evidence is needed (Ohlberger, 2013). Statistical analyses of long term data series

are required to provide the most direct evidence for changes in body size caused by climate

warming (Daufresne et al., 2009).

Over the past 30 years water temperatures in the North Sea have increased by 0.2-0.6 °C per

decade with the rates of warming being rapid relative to other regional seas (Belkin, 2009).

During this period, declining body sizes have been observed in haddock (Melanogrammus

aeglefinus) (Baudron et al., 2011), herring (Clupea harengus) (Brunel & Dickey-Collas,

2010) and plaice (Pleuronectes platessa)

(van Walraven et al., 2010). The fact that three

species differing in their life histories, trophodynamics and vertical distribution in the water

column (Supplementary Table S1) exhibited smaller body sizes concomitant with a warming

environment is consistent with the claim that the TSR is a universal response. A complication

in establishing direct causality between warming temperatures and decreasing body sizes in

commercial stocks is that size-selective fishing mortality may select for genotypes affecting

growth (Enberg et al., 2012) and reduction in body size

could therefore be the result of non-

random genetic selection. Furthermore, commercial species experience particularly large

fluctuations in abundance that could introduce variability in growth rates via density-

dependent competition for resources (Taylor & Stefánsson, 1999). Unlike temperature, it is

difficult to see how these two factors could impact growth uniformly across species. The

scale and speed of an evolutionary response would be unique to each stock given that the

pattern and degree of selection varies across stocks and stocks differ in the life history traits

(e.g., age at maturity) that determine how quickly a phenotypic trait evolves (Supplementary

Table S1). Similarly, the mechanisms responsible for generating density-dependent growth

would also likely be species-specific given the variety in habitat and diet of North Sea fish

species (Supplementary Table S1). Although evolutionary and density-dependent changes in

Page 4 of 40Global Change Biology

For Review Only

growth seem unlikely to vary synchronously across species both factors must be examined

for synchronicity. If no synchronous trends across species are detectable in these two factors

then logically they cannot be responsible for generating phenotypic changes that are

synchronous across species. Establishing that declines in body size are, firstly, synchronous

across a range of species, and secondly, concurrent with temperature would strongly imply

the “omnibus” effect of temperature.

Our aim was to test whether the North Sea fish assemblage exhibited synchronous declines in

asymptotic body size that were concurrent with increases in temperature, and consistent with

TSR. We used the von Bertalanffy growth function (VBGF)

(Pauly, 2010) to estimate L

∞

, the

asymptotic body length, on a cohort-by-cohort basis for eight North Sea fish species for

which age and size data were available at least annually over the past four decades. Statistical

analyses were performed to test for a synchronous trend in L

∞

across species and compare

this trend with the temperature trend. Annually resolved indices of fishing mortality and

density were also examined for synchronicity across species to determine the potential for

these factors to generate a synchronous trend in L

∞

. The implications for yield were examined

for species exhibiting a synchronous component in variability of L

∞

Page 5 of 40 Global Change Biology

For Review Only

Methods

Sea bottom temperatures (SBT) from 1970 to 2008 were obtained from the NORWECOM

model (Skogen & Søiland, 1998) which gives monthly mean values averaged over 0.25

degree latitude by 0.5 degree longitude rectangles. SBT values were averaged per roundfish

area (Supplementary Fig. S1a) and per year in order to match the spatial and temporal

resolution of the biological data. SBT time series showed the same trend in northern (areas 1

and 2) and southern (areas 5 and 6) regions with a ca. 3°C gradient (Supplementary Fig. S1b).

Age-length keys (ALKs) generated from data collected during annual International Bottom

Trawl Surveys (IBTS) of the North Sea are available for demersal and pelagic species from

the DATRAS database (http://datras.ices.dk) maintained by the International Council for the

Exploration of the Sea (ICES). ALKs for benthic flatfish species are estimated from a

combination of commercial samples, survey samples and otolith back-calculations (Rijnsdorp

et al., 2010). Our analysis was restricted to commercial species having long time series of

otolith-derived age estimates which are essential for modelling growth. Demersal (distributed

nearer bottom) species were haddock, cod (Gadus morhua), whiting (Merlangius merlangus)

and Norway pout (Trisopterus esmarkii). Pelagic (distributed nearer surface) species were

herring and sprat (Sprattus sprattus). Benthic (distributed on the bottom) flatfish species were

plaice and sole (Solea solea). Collectively, these eight species (hereafter referred to as stocks)

span a range of habitats, body sizes, and life history traits that are representative of the North

Sea fish assemblage.

ALKs give the number of sampled fish in a given 1-cm length class that were assigned

through otolith reading to age t in year y. For IBTS surveys conducted in quarter 1 age values

remained as integers (t) while ages used for quarter 2, 3 and 4 were t + 0.25, t + 0.5 and t +

Page 6 of 40Global Change Biology

For Review Only

0.75, respectively. Because ALKs are generated through length-stratified sampling of the

catch, they do not accurately represent the true length distribution-at-age. To correct for this

bias, ALKs for each ICES roundfish area were raised by the catch-per-unit-effort per length

class for the area except for the two flatfish species where ALKs were raised by length

distributions instead. To account for the strong spatial gradient in temperature

(Supplementary Fig. S1), data for cod, whiting and herring were split into northern and

southern sub-stocks and the raised ALKs were combined for ICES roundfish areas 1 and 2

and for areas 5 and 6 to represent the northern and southern North Sea, respectively. Haddock

and Norway pout are found in the north, whereas sprat are found in the south. Plaice and sole

are found in the south but were split by sex to account for known differential growth that

gives rise to large phenotypic differences between males and females (Rijnsdorp et al., 2010).

Splitting gave a total of 13 sub-stocks and accounted for known sources of variation in

temperature (by region) and growth (by gender) that would otherwise have confounded the

analysis. For each sub-stock the growth of a cohort spawned in year y was modelled by fitting

the VBGF to the length distribution-at-age represented by the raised ALKs:

)1(

)(

ttK

eLL

−

∞

−=

where L

is the length (cm) at age t, K is the Brody growth parameter (year

-1

), and t

is the

hypothetical age (year) at length equal to 0. As cohort-specific values of L

∞

and K are

negatively correlated (Pauly, 2010) (Supplementary Fig. S2) examining temporal variation in

one of the two parameters is sufficient to describe growth. Examining variation in L

∞

cohort assumes that the growth trajectory of a cohort is established in the early stages of life,

an assumption supported experimentally

(Forster et al., 2011; Scott & Johnston, 2012). For

each sub-stock the VBGF was fit for cohorts 1970 to 2008; cohorts 1970 to 2001 were

modelled with data for ages 1 to 10 while cohorts 2002 to 2008 were modelled with nine to

three years of data. Outliers in the L

∞

time series (> 2 * average or L

∞

having standard errors

Page 7 of 40 Global Change Biology

For Review Only

>10% of the value) were omitted from subsequent analysis (see Supplementary Table S2).

For all other cohorts, estimated parameters (L

∞

, K and t

) had p-values <0.05. A Dynamic

Factor Analysis (DFA) (Zuur et al., 2003) was used to objectively estimate common trends in

the L

∞

time series of the sub-stocks considered. DFA is a multivariate extension of structural

time series which can analyse short, non-stationary time series containing missing values.

The aim of DFA is to model as few common trends as possible while giving a reasonable

model fit. Prior to inclusion in the DFA model, L

∞

time series for all sub-stocks were

standardized by subtracting the mean and dividing by the standard error

(Zuur et al., 2003).

The 13 time series were each modelled as a combination of common trends (x), factor

loadings (Z) plus some offset (a) as follows:

tsstisitss

axZxZtL

,,,,1,1,

...)(

∞

where ε

~ MVN(0, R) with MVN standing for Multivariate Normal and R being the error

covariance matrix, s is the considered sub-stock and i is the number of common trends. The

magnitude and sign of Z indicate to what extent the common trends are related to the original

times series. DFA models with 1 to 6 common trends and with either a diagonal and equal or

a diagonal and unequal error covariance matrix were tested. The best of the twelve candidate

models was selected using the Akaike’s information criterion (AIC). Correlation tests

between the predominant common trend (Trend 1) and SBT were performed using different

time windows for temperature impacts: SBT experienced at age 0 (no lag), at age 1 (lag 1), at

age 2 (lag 2), during the first two years of life (average SBT from age 0 to 1) and during the

first three years of life (average SBT from age 0 to 2). Correlation tests between Trend 1 and

the common trend in density were performed to test for intra-cohort (no lag), inter-cohort (lag

1 and lag 2), and cumulative (sum of densities at lags 0 to 1, and lags 0 to 2) density-

dependent growth. Since, the purpose of these tests was to assess the correlation between

underlying trends rather than short-term, high frequency variations, the P-values presented do

Page 8 of 40Global Change Biology

For Review Only

not account for autocorrelation as this would have involved detrending the time series.

Correlations were summarized by Pearson product-moment correlation coefficients. As

multiple tests were conducted, a sequential Bonferroni correction was applied to adjust the

level of significance of the multiple inferences.

To control for synchronicity in fishing mortality and density across species, the DFA and

correlation tests were repeated for both factors. Average fishing mortalities were obtained

from ICES (http://www.ices.dk/) 2012 assessment reports for the Working Group on the

Assessment of Demersal Stocks in the North Sea and Skagerrak (WGNSSK) and the Herring

Assessment Working Group (HAWG). Assessment data were available from 1970 to 2011

for all species apart from whiting, Norway pout and sprat which assessments began

respectively in 1990, 1983 and 1991. For stocks distributed across the northern and southern

North Sea (whiting and herring), a survey-based assessment (SURBA) model (Beare et al.,

2005) was used to obtain local estimates of total mortality in order to capture spatial gradients

in fishing pressure. Assuming a constant natural mortality, total mortality times series for

these two stocks were used as proxies for fishing mortalities. Abundance at age 1 indices

were used as a proxy for density (no sex-specific abundance index were available for plaice

and sole sub-stocks). For stocks distributed across the northern and southern North Sea, the

survey abundance at age 1 indices were split by area. For other sub-stocks, XSA abundance

at age 1 indices given in the 2012 assessment reports were used. For Norway pout North and

sole South, recruitment time series from the assessment were used as no abundance at age 1

indices were available. Data were available from 1970 to 2011 for all species apart from

whiting, Norway pout and sprat for which indices were available from 1990, 1983 and 1984

respectively.

Page 9 of 40 Global Change Biology

For Review Only

Comparative yield-per-recruit analyses were performed as a proxy for changes in yield prior

to and after changes in individual body size. Yield-per-recruit was approximated by

simulating a fishery on a single cohort composed of eleven year classes (from age 0 to age

10), and with an initial recruitment (abundance at age 0) of 10000 individuals. 5-year mean of

∞

and K prior and after changes in growth were used to compute length-at-age values,

assuming t

=0 for all sub-stocks. Length values were converted into weights-at-age using

length-weight relationships obtained from Marine Scotland and IMARES. Fishing mortality-

at-age was estimated by the mean over the last three historical years (2009 to 2011) while

natural mortality-at-age and proportion mature-at-age were assumed to be constant. For

plaice and sole the fishing mortality was assumed to be equal for both sexes. All estimates

were obtained from the 2012 assessment reports. The cumulative contribution to yield of the

successive year classes of the cohort were then summed and divided by the original number

of recruit to obtain approximations of yield-per-recruit.

All statistical analyses were performed using the R software (version 2.15.1; http://www.r-

project.org/). DFA was performed using the MARSS package (Holmes et al., 2012).

Page 10 of 40Global Change Biology

For Review Only

Results and discussion

A decrease in L

∞ (

expressed as difference between average 1973-1977 and average 1993-

1997) of 29%, 13%, 29%, 10%, 19%, 16%, 13%, 1% and 12% (average 16%) was observed

for haddock North, whiting North, whiting South, herring North, Norway pout North, sprat

South, male sole South, female sole South and male plaice South, respectively (Fig. 1c-f and

h-l). These nine sub-stocks also exhibited narrow 95% confidence intervals (95%CI) around

∞

estimates. The four remaining sub-stocks (cod North, cod South, herring South and female

plaice South) showed divergent trends in L

∞

(Fig. 1a-b, g and m). Cod North, cod South and

herring South time series showed an increase in L

∞

but included high proportions of outliers

(Supplementary Table S2). The wide 95%CI for the two cod sub-stocks indicate that the

VBGF was a poor fit to the data and the high (>200 cm) values of L

∞

reflect near-linear

growth rather than the asymptotic growth assumed by the VBGF. Herring South showed a

sudden decline in the late 1970s (Fig. 1f) while L

∞

for female plaice South showed an

increase from 1970 to 1990 followed by a recent decrease (Fig. 1m).

The best model identified by DFA to describe temporal variation in L

∞

included two common

trends (Supplementary Table S3). Fitted values for all sub-stocks showed that the model

succeeded in describing the overall trends in L

∞

(Fig. 1). The DFA model captured the

decrease in L

∞

for the nine sub-stocks identified and exhibited narrow 95%CI apart from the

most recent cohorts which have fewer sampled age classes (Fig. 1c-f and h-l). The first

common trend (Trend 1) showed a steep decline from 1978 to 1993 after which it remained

stable (Fig. 2a). The second common trend (Trend 2) increased slightly until the early 1990s

after which it showed a sharp decline (Fig. 2c). A DFA model fitted with one common trend

only returned a trend almost identical to Trend 1 (Fig. 2a), which indicates that Trend 1 is

Page 11 of 40 Global Change Biology

For Review Only

predominant in describing the common trend in L

∞

(Zuur et al., 2003). The nine sub-stocks

showing a decrease in L

∞

from the mid-1970s to the mid-1990s (Fig. 1) were all positively

related to Trend 1 (Fig. 2b). Six of these sub-stocks (haddock North, whiting North, whiting

South, herring North, Norway pout North and male sole South) had similar factor loading

values (Z) on Trend 1, indicating that the stock-specific trends in L

∞

were equally well

described by Trend 1 (Fig. 2b). These sub-stocks also showed small Zs on Trend 2 (Fig. 2d).

Female sole South and male plaice South, although positively related to Trend 1, showed

larger Zs on Trend 2 (Fig. 2d).

Cod North, cod South, and female plaice South were negatively related to Trend 1 but

positively related to Trend 2 (Fig. 2b and 2d). Herring South was negatively related to Trend

2 and showed a different temporal trend (Fig. 1g). The high Z of female plaice South on

Trend 2 (Fig. 2d) suggested that Trend 2 was mainly driven by this sub-stock. Because the L

∞

time series for the cod sub-stocks and herring South included many outliers (Supplementary

Table S2) it is difficult to describe decadal-scale trends for these sub-stocks with confidence.

Flatfish sub-stocks showed mixed trends, with males of both species conforming to Trend 1,

whereas, the decrease in L

∞

was small in female sole South and absent in female plaice

South. Differences between males and females in their growth response to temperature have

been detected in fish (Kuparinen et al., 2011) although the physiological basis is unknown.

The common trend in L

∞

represented by Trend 1 describes the synchronous component of

variability in the L

∞

time series. Residual variation around Trend 1 for a given stock reflects

the influence of stock-specific factors and estimation error, neither of which are beyond the

scope of this study. Trend 1 showed concurrent and opposite variation with sea bottom

temperature (SBT) and the temperature increase through the 1980s was concurrent with the

Page 12 of 40Global Change Biology

For Review Only

decline in L

∞

described by Trend 1 (Fig. 2e). This is consistent with the TSR prediction that

higher temperatures result in smaller body sizes. Furthermore, this cross-stock synchronicity

was detected when growth was examined by cohort suggesting that temperatures experienced

early in the life of the cohort are critical to determining L

∞

, a result also consistent with

current physiological understanding of growth (Forster et al., 2011; Scott & Johnston, 2012).

Trend 1 was significantly, negatively correlated (p ≤ 0.05) with the average monthly SBT

experienced at age 0, age 1, age 2 as well as during the first two years and the first three years

of life (Table 1). The nine sub-stocks exhibiting the synchronous decline in L

∞

described by

Trend 1 represent six species with different life-history characteristics and asynchronous

trends in fishing mortalities (Supplementary Fig. S3 and S5) ruling out the possibility that

non-random genetic selection is responsible for inducing cross-stock synchronicity in L

∞

Although a common trend in density was detected by DFA, it was not equally strongly

supported by all nine sub-stocks (Supplementary Fig. S4 and S5). No significant correlations

were observed between Trend 1 in L

∞

and the common trend in density (Supplementary

Table S4). Therefore, the TSR is a plausible and parsimonious explanation for the

synchronous reduction in L

∞

detected in the majority of North Sea sub-stocks and species

examined here.

The nine sub-stocks exhibiting a detectable degree of synchrony had fewer outliers and less

uncertainty in L

∞

estimates than the four sub-stocks that did not conform to Trend 1. They

included fast-growing/early-maturing species as well as slow-growing/late maturing ones and

their habitats encompassed pelagic, demersal and benthic regions. This suggests a uniformity

of the response to warming temperature which is consistent with current physiological

(Pörtner & Knust, 2007; Forster et al., 2011; Scott & Johnston, 2012) and ecological

(Daufresne et al., 2009; Gardner et al., 2011; Sheridan & Bickford, 2011; Ohlberger, 2013)

Page 13 of 40 Global Change Biology

For Review Only

understanding. It is not possible to infer direct causality from our analysis due to the

comparative shortness of the L

∞

time series used here (<40 years) and the intrinsically

“uncontrolled” nature of ecosystems. Support for inferring a causal relationship between

temperature and growth would be provided if ecosystems showing strong warming

consistently showed evidence of synchrony in growth rates across species while ecosystems

with little to no warming were asynchronous. For example, juvenile growth rates

(proportional to the VBGF parameter K) were correlated with temperature for six of eight

long-lived commercial fish species in the temperate southwestern Pacific (Thresher et al.,

2007). The intrinsically negative correlation between K and L

∞

(Pauly, 2010) (Supplementary

Fig. S2) suggests that there may have been a corresponding reduction in asymptotic body size

in the Pacific species, a hypothesis that is difficult to test due to their longevity. The two

studies, undertaken in temperate regions of the northern and southern hemisphere, confer a

degree of verisimilitude on the inference that temperature is responsible for imposing a

detectable, synchronous signal on temporal variation in individual growth rates of fish that is

consistent with TSR.

If the synchronous decline in L

∞

observed in several species were driven by temperature, as

the reasoning above suggests, then this study can be considered to support the prediction that

the future assemblage-averaged maximum body weight of species will be substantially

smaller (Cheung et al., 2013). This model-derived projection of future shrinkages (14-24%

smaller by 2050) reflects both temperature impacts on individual growth and biogeographic

shifts towards communities having a higher proportion of smaller-sized species. Considering

only impacts on individual growth resulted in 10% shrinkage (Cheung et al., in press). Our

study showed that a relatively small increase in temperature (<2 ºC) over a 40-year period

was concurrent with reductions in L

∞

that were variable across species (1% to 29%) but

Page 14 of 40Global Change Biology

For Review Only

surprisingly large (average 16%) and comparable in scale to the Cheung et al. projection.

The speed and scale of Cheung et al.’s prediction about body sizes in future oceans appear

more reasonable than critics claim (Brander et al., in press). In our opinion the projections of

Cheung et al. have merit. In combination with our empirical observation of synchronous

growth trends in North Sea fishes, it is clear that a comprehensive analysis of available data

on individual growth is required.

Our results also provide empirical support for a recent simulation study showing that

shrinking body sizes impact fisheries yield (Audzijonyte et al., 2013). Comparing two years

before and after the decline in L

∞

(1978 and 1993, respectively) shows that, under several

assumptions, yield-per-recruit i.e., the catch in weight per recruit entering the fishery

(Beverton & Holt, 1957) for the affected North Sea stocks decreased by 3.1% to 48.1% with

an average reduction of 23.1% (Table 2). Despite the many assumptions required to estimate

yield-per-recruit, it is self-evident that smaller body sizes will decrease per capita estimates

of productivity. The magnitude of these declines seems both substantial and underappreciated

relative to the well-documented impacts of fishing over recent decades (Fernandes & Cook,

2013). Given that seasonal mean surface temperatures in the North Sea are predicted to

increase by 2.42-3.27ºC by the end of the century (MCCIP 2010) future synchronous

reductions in yield-per-recruit are probable.

Warming temperatures are generally associated with faster growth rates (higher K) for

temperate stocks (Thresher et al., 2007; Neuheimer & Grønkjaer, 2012). However, it has not

been fully appreciated that, by virtue of the negative relationship between L

∞

and K, the

downside of fast early growth is smaller adult body size. Consequences of smaller adult body

size include reduced per-capita reproductive rates (Rijnsdorp et al., 2010), decreased

Page 15 of 40 Global Change Biology

For Review Only

resilience (Hsieh et al., 2006)

and altered ecosystem function and services (Sheridan &

Bickford, 2011; Edeline et al., 2013). If individual growth rates change directionally in

response to warming then management strategies that assume productivity can be restored to

levels observed when temperatures were cooler must be re-considered.

Page 16 of 40Global Change Biology

For Review Only

Acknowledgements

Funding support was provided by Marine Scotland – Science. C. Millar and S. Palmer are

thanked for their help. The authors declare no conflict of interest.

Page 17 of 40 Global Change Biology

For Review Only

References

Atkinson D (1994) Temperature and organism size – a biological law for ecototherms?

Advances in Ecological Research, 25, 1–58.

Audzijonyte A, Kuparinen A, Gorton R, Fulton EA (2013) Ecological consequences of body

size decline in harvested fish species: positive feedback loops in trophic interactions

amplify human impact. Biology letters, 9, 20121103.

Baudron AR, Needle CL, Marshall CT (2011) Implications of a warming North Sea for the

growth of haddock Melanogrammus aeglefinus. Journal of Fish Biology, 78, 1874–

1889.

Beare D, Needle C, Burns F, Reid D (2005) Using survey data independently from

commercial data in stock assessment: an example using haddock in ICES Division VIa.

ICES Journal of Marine Science, 62, 996–1005.

Belkin IM (2009) Rapid warming of Large Marine Ecosystems. Progress in Oceanography,

81, 207–213.

Beverton RJH, Holt SJ (1957) On the dynamics of exploited fish populations. Fisheries

Investigations, London, Ser. 2. 19: 533 pp.

Brander K, Neuheimer A, Andersen KH, Hartvig M (in press) Overconfidence in model

projections. ICES Journal of Marine Science, doi:10.1093/icesjms/fst055.

Brunel T, Dickey-Collas M (2010) Effects of temperature and population density on von

Bertalanffy growth parameters in Atlantic herring: a macro-ecological analysis. Marine

Ecology Progress Series, 405, 15–28.

Cheung WWL, Sarmiento JL, Dunne J, et al. (2013) Shrinking of fishes exacerbates impacts

of global ocean changes on marine ecosystems. Nature Climate Change, 3, 254–258.

Page 18 of 40Global Change Biology

For Review Only

Cheung WWL, Pauly D, Sarmiento J (in press) How to make progress in projecting climate

change impacts. ICES Journal of Marine Science, doi:10.1093/icesjms/fst133.

Daufresne M, Lengfellner K, Sommer U (2009) Global warming benefits the small in aquatic

ecosystems. Proceedings of the National Academy of Sciences of the United States of

America. 106, 12788–12793.

Edeline E, Lacroix G, Delire C, Poulet N, Legendre S (2013) Ecological emergence of

thermal clines in body size. Global change biology, doi: 10.1111/gcb.12299.

Enberg K, Jørgensen C, Dunlop ES, et al. (2012) Fishing-induced evolution of growth:

concepts, mechanisms and the empirical evidence. Marine Ecology, 33, 1–25.

Fernandes PG, Cook R (2013) Reversal of fish stock decline. Current Biology, 23, 1432–

1437.

Forster J, Hirst AG, Atkinson D (2011) How do organisms change size with changing

temperature? The importance of reproductive method and ontogenetic timing.

Functional Ecology, 25, 1024–1031.

Forster J, Hirst AG, Atkinson D (2012) Warming-induced reductions in body size are greater

in aquatic than terrestrial species. Proceedings of the National Academy of Sciences of

the United States of America, 109, 19310–19314.

Gardner JL, Peters A, Kearney MR, Joseph L, Heinsohn R (2011) Declining body size: a

third universal response to warming? Trends in ecology & evolution, 26, 285–91.

Graham JB (2006) The physiology of fishes (eds Evans DH and Clairborne JB) pp. 85–117.

Taylor and Francis.

Holmes EE, Ward EJ, Wills K (2012) MARSS : Multivariate Autoregressive State-space

Models for Analyzing Time-series Data. The R Journal, 4/1.

Hsieh C-H, Reiss CS, Hunter JR, Beddington JR, May RM, Sugihara G (2006) Fishing

elevates variability in the abundance of exploited species. Nature, 443, 859–862.

Page 19 of 40 Global Change Biology

For Review Only

Kuparinen A, Cano JM, Loehr J, Herczeg G, Gonda A, Merilä J (2011) Fish age at

maturation is influenced by temperature independently of growth. Oecologia, 167, 435–

443.

MCCIP (2010). Marine Climate Change Impacts Annual Report Card 2010–2011 (eds Baxter

JM, Buckley PJ, and Wallace CJ) Summary Report, MCCIP, Lowestoft, 12 pp.

Neuheimer AB, Grønkjaer P (2012) Climate effects on size-at-age: growth in warming waters

compensates for earlier maturity in an exploited marine fish. Global Change Biology,

18, 1812–1822.

Ohlberger J (2013) Climate warming and ectotherm body size - from individual physiology

to community ecology. Functional Ecology, 27, 991–1001.

Pauly D (2010) Gasping fish and panting squids: oxygen, temperature and the growth of

water-breathing animals. International Ecology Institute.

Pörtner HO, Knust R (2007) Climate change affects marine fishes through the oxygen

limitation of thermal tolerance. Science (New York, N.Y.), 315, 95–7.

Rijnsdorp AD, van Damme CJG, Witthames PR (2010) Implications of fisheries-induced

changes in stock structure and reproductive potential for stock recovery of a sex-

dimorphic species, North Sea plaice. ICES Journal of Marine Science, 67, 1931–1938.

Scott GR, Johnston IA (2012) Temperature during embryonic development has persistent

effects on thermal acclimation capacity in zebrafish. Proceedings of the National

Academy of Sciences of the United States of America, 109, 14247–14252.

Sheridan JA, Bickford D (2011) Shrinking body size as an ecological response to climate

change. Nature Climate Change, 1, 401–406.

Skogen MD, Søiland HA (1998) User’s guide to NORWECOM v2.0. The NORWegian

ECOlogical Model system. Tech.report, Fisken og Havet 18/98, Institute of Marine

Research, Pb.1870, NO-5024 Bergen, 42pp.

Page 20 of 40Global Change Biology

For Review Only

Taylor L, Stefánsson G (1999) Growth and Maturation of Haddock ( Melanogrammus

aeglefinus ) in Icelandic Waters. Journal of Northwest Atlantic Fishery Science, 25,

101–114.

Thresher RE, Koslow JA, Morison AK, Smith DC (2007) Depth-mediated reversal of the

effects of climate change on long-term growth rates of exploited marine fish.

Proceedings of the National Academy of Sciences of the United States of America, 104,

7461–7465.

van Walraven L, Mollet FM, van Damme CJG, Rijnsdorp AD (2010) Fisheries-induced

evolution in growth, maturation and reproductive investment of the sexually dimorphic

North Sea plaice (Pleuronectes platessa L.). Journal of Sea Research, 64, 85–93.

Zuur AF, Fryer RJ, Jolliffe IT, Dekker R, Beukema JJ (2003) Estimating common trends in

multivariate time series using dynamic factor analysis. Environmetrics, 14, 665–685.

Page 21 of 40 Global Change Biology

For Review Only

Supporting Information legends

Table S1. Data availability, distribution and life history traits of the species considered in the

analysis. Mean length-at-age 1 was calculated from the age-length keys obtained from the

DATRAS database (http://datras.ices.dk/Home/Default.aspx), except for plaice and sole

which calculated length-at-age 1 using the von Bertalanffy equation. A50 (age at 50%

maturity) values were averaged across time series and were estimated by fitting maturity

ogives to sex maturity age-length keys obtained from the DATRAS database except for

plaice where proportion of mature-at-age values were used. For sole, the value 2.5 was

chosen as A

is reached between age 2 and age 3. The selectivity (age classes targeted by the

fishery) values were obtained from ICES assessment working group reports and correspond

to the age ranges used to estimate the average fishing mortality (F). Main preys were

estimated from diet data given in Pinnegar et al. (2011) and Greenstreet (1996).

Table S2. Summary table of the cohorts considered as outliers for each sub-stock, with their

∞

values and associated standard errors (S.E.). Unrealistically high values of L

∞

reflect

growth trajectories that are more linear than asymptotic.

Table S3. Selection table of candidate models tested in the Dynamic Factor analysis

including log-likelihood, Akaike criterion (AIC) and the difference (∆

AIC

) between the AIC

of the considered model and the best candidate model (minimum AIC observed).

Table S4 Estimated correlations between Trend1 and the trend in density for the sub-stocks

related to Trend 1, with their corresponding P-values. The lag included in the estimation of

the density is indicated (see Methods). Significance was adjusted by a sequential Bonferroni

Page 22 of 40Global Change Biology

For Review Only

correction: the ordered P-values were compared with the inequality, P

≤ α(1 + k − i)

-1

, where

α is the confidence level to test for significance (0.05), K is the number of correlation tests

carried out and i is the rank of the correlation considered. Correlations for which the

inequality is met are significant (*).

Figure S1. a: the ICES standard roundfish areas for the North Sea used for the International

Bottom Trawl Surveys. b: Overall average annual sea bottom temperature (thick continuous

line) between the average of roundfish areas 1 and 2 (lower continuous line) and the average

of roundfish areas 5 and 6 (upper continuous line). The two lower dashed lines correspond to

areas 1 and 2, the two upper dashed lines to areas 5 and 6.

Figure S2. Log-scaled relationships between the K and L

∞

parameters for the sub-stocks

considered in the analysis (triangles: cod, straight crosses: haddock, circles: whiting, squares:

herring, diagonal crosses: Norway pout, stars: sprat, F and M: female and male plaice, f and

m (in grey): female and male sole). Filled symbols stand for sub-stock in northern North Sea,

empty symbols for sub-stocks in the southern North Sea. Lines correspond to linear models

fitted to the data points.

Figure S3. Fishing mortality (filled circles) for the sub-stocks related to Trend 1 (fishing

mortality was assumed to be equal for both male and female sole South) plotted along the

fitted values from the best Dynamic Factor Analysis model (line) and their corresponding

95% confidence intervals (a: haddock North, b: Norway pout North, c: Sprat South, d: plaice

South, e: sole South, f: whiting North, g: whiting South, h: herring North).

Page 23 of 40 Global Change Biology

For Review Only

Figure S4. Abundance at age 1 indices (filled circles) used as a proxy for density for the sub-

stocks related to Trend 1 (for both plaice and sole sub-stocks the abundance index stands for

the males and females together as no sex-specific abundance index were available) plotted

along the fitted values from the best Dynamic Factor Analysis model (line) and their

corresponding 95% confidence intervals (a: haddock North, b: Norway pout North, c: Sprat

South, d: plaice South, e: sole South, f: whiting North, g: whiting South, h: herring North).

Figure S5. The common trends (black line) identified by the best-fitting Dynamic Factor

Analysis to describe temporal variation in fishing mortality model (panels a and c) and

density (panel e) for the eight sub-stocks that were positively related to Trend 1 (grey line)

and their corresponding factor loadings for each sub-stock (panels b, d and f respectively).

For fishing mortality, the best model identified by DFA included more than one trend

suggesting that there is no synchrony in the fishing mortality time series for these eight sub-

stocks. For density, although the best model identified by DFA included a single trend,

haddock North and sprat South did not conform to it while whiting North showed the highest

factor loadings of all sub-stocks, suggesting that the trend was mainly driven by this sub-

stock only.

Page 24 of 40Global Change Biology

For Review Only

Tables

Table 1 Estimated correlations between sea bottom temperature SBT and Trend 1 and their

corresponding P-values. The time period included in the estimation of mean temperature is

indicated (see Methods). Significance was adjusted by a sequential Bonferroni correction: the

ordered P-values were compared with the inequality, P

≤ α(1 + k − i)

-1

, where α is the

confidence level to test for significance (0.05), K is the number of correlation tests carried out

and i is the rank of the correlation considered. Correlations for which the inequality is met are

significant (*).

Time

period

Correlation

p-value α(1 + k − i)

-1

0 to 2 years -0.54 0.00064 0.010*

0 to 1 years -0.49 0.00182 0.013*

2 years

0.49

0.00200

0.017*

1 year

0.45

0.00510

0.025*

0 years -0.43 0.00640 0.050*

Page 25 of 40 Global Change Biology

For Review Only

Table 2 Yield-per-recruit (YPR) values (kg) prior (1978) and after (1993) the observed

decline in L

∞

, with corresponding individual yield loss in value (kg) and percentage. 1977

and 1997 were years in which the standardized common Trend 1 reached its maximum and

minimum values prior and after the decline in L

∞

. For sole and plaice the low natural

mortality estimates (0.1 at all ages) resulted in high YPR values for these two species

compared to other species.

Sub-stock YPR 1978 YPR 1997 Individual yield loss % loss

Haddock North

0.00473

0.00290

0.00183

38.7

Whiting North 0.00089 0.00086 0.00003 3.1

Whiting South 0.00116 0.00060 0.00056 48.1

Herring North 0.00514 0.00450 0.00063 12.3

N. Pout North

0.00171

0.00133

0.00038

22.2

Sprat South

0.00075

0.00072

0.00003

4.0

Sole male South 0.10458 0.08600 0.01858 17.8

Sole female South 0.14949 0.12571 0.02377 15.9

Plaice male South 0.12375 0.06664 0.05711 46.2

Average

23.1

Page 26 of 40Global Change Biology

For Review Only

Figure legends

Figure 1 Standardized L

∞

time series for the thirteen sub-stocks considered in the analysis

(filled circles) with their 95% confidence intervals (vertical segments), along with the fitted

values from the selected Dynamic Factor Analysis model (line) and their corresponding

confidence intervals (shaded areas). a) cod North; b) cod South; c) haddock North; d) whiting

North; e) whiting South; f) herring North; g) herring South; h) Norway pout North; i) sprat

South; j) male sole South; k) female sole South; l) male plaice South; m) female plaice South.

Figure 2 Common trends given by the best candidate model (panels a and c) to describe L

∞

time variations over time and the corresponding factor loadings for the thirteen sub-stocks

(panels b and d). In panel a the grey line corresponds to the common trend given by a model

fitted with one common trend. In panel e the Trend 1 is plotted along with the sea bottom

temperature (grey line) averaged across the roundfish areas 1, 2, 5 and 6.

Page 27 of 40 Global Change Biology

For Review Only

508x288mm (72 x 72 DPI)

Page 28 of 40Global Change Biology

For Review Only

237x282mm (72 x 72 DPI)

Page 29 of 40 Global Change Biology

For Review Only

Supporting Information

Table S1 Data availability, distribution and life history traits of the species considered in the analysis. Mean length-at-age 1 was calculated from

the age-length keys obtained from the DATRAS database (http://datras.ices.dk/Home/Default.aspx), except for plaice and sole which calculated

length-at-age 1 using the von Bertalanffy equation. A

(age at 50% maturity) values were averaged across time series and were estimated by

fitting maturity ogives to sex maturity age-length keys obtained from the DATRAS database except for plaice where proportion of mature-at-age

values were used. For sole, the value 2.5 was chosen as A

is reached between age 2 and age 3. The selectivity (age classes targeted by the

fishery) values were obtained from ICES assessment working group reports and correspond to the age ranges used to estimate the average

fishing mortality (F). Main preys were estimated from diet data given in Pinnegar et al. (2011) and Greenstreet (1996).

Cod Haddock Whiting Herring Norway

pout Sprat Plaice Sole

Time

period Time series 1971 - 2011 1970 - 2011 1970 - 2011 1970 - 2011 1972 - 2011 1972 - 2011 1970 - 2011 1970 - 2011

Region North * * * * *

South * * * * * *

Lifestyle

Demersal * * * *

Pelagic * *

Benthic * *

Body size Length-at-age 1

(mm) 229 194 181 156 125 98 124 103

Maturity A

3.2 2 1.3 2.6 1.1 1.0 2.87 2.5

Selectivity

F 2 - 4 2 - 4 2 - 5 2 - 6 1 - 2 1 - 2 2 - 6 2 - 6

Diet Main prey

Norway

pout

Sandeel

Benthos

Copepods

Sandeel Copepods Krill

Copepods

Fish eggs Benthos Benthos

Page 30 of 40Global Change Biology

For Review Only

Table S2. Summary table of the cohorts considered as outliers for each sub-stock, with their

∞

values and associated standard errors (S.E.). Unrealistically high values of L

∞

reflect

growth trajectories that are more linear than asymptotic.

Sub-stock Cohort L

∞

S.E.

Cod North 1984 144.15 22.21

1987 201.43 32.54

1991 358.03 36.63

1996 499.24 75.95

2000 214.24 30.58

2001 190.90 38.39

2002 176.35 33.65

2003 161.43 23.16

2004 409.59 325.06

2005 192.89 23.77

2006 1508.37

0.54

2007 775.07 978.26

2008 753.37 0.40

Cod South 1971 142.77 16.39

1979 360.00 55.28

1982 168.17 17.39

1984 165.66 21.66

1988 251.69 29.54

1989 171.30 45.43

1992 4597.72

2.05

1995 245.78 28.18

1997 1579.30

0.56

1998 152.11 18.73

2000 262.16 70.70

2001 1255.88

0.29

2002 144.82 17.50

2005 483.12 423.22

2006 2111.77

0.46

2007 1413.29

0.35

2008 79.45 27.70

Haddock North 1975 132.33 13.81

Whiting North 1980 705.58 0.63

Herring South 1989 85.97 11.99

1992 407.45 0.38

1997 166.15 13.69

2003 261.36 0.57

2008 271.79 0.28

Page 31 of 40 Global Change Biology

For Review Only

Norway pout North 1976 44.55 2.01

Sprat South 1981 161.84 0.70

1985 18.87 0.62

1995 63.44 0.76

1997 -322.35 1.11

Sole male South 2005 43.17 4.45

2006 81.56 1.37

Sole female South 2006 234.83 1.12

2007 195.84 0.69

Plaice male South 1992 93.06 11.72

2005 169.96 3.02

2006 96.24 1.98

2007 52.55 13.51

Plaice female South 1995 384.98 69.40

2004 530.15 2.45

2005 690.01 1.61

2006 566.30 0.80

2007 364.75 0.52

Page 32 of 40Global Change Biology

For Review Only

Table S3. Selection table of candidate models tested in the Dynamic Factor analysis

including log-likelihood, Akaike criterion (AIC) and the difference (∆

AIC

) between the AIC of

the considered model and the best candidate model (minimum AIC observed).

Error covariance

matrix

Number of

trends

Log-

likelihood

AIC ∆

AIC

diagonal and equal 2 -542.67 1140.75

0.00

diagonal and unequal 3 -516.89 1144.33

3.59

diagonal and equal 3 -531.79 1144.58

3.84

diagonal and unequal 2 -533.06 1149.51

8.76

diagonal and equal 1 -563.05 1155.09

14.35

diagonal and equal 4 -525.32 1156.16

15.41

diagonal and unequal 4 -509.85 1156.34

15.59

diagonal and unequal 1 -554.61 1164.62

23.87

diagonal and unequal 5 -502.42 1166.14

25.40

diagonal and equal 5 -523.51 1175.69

34.94

diagonal and unequal 6 -500.78 1185.79

45.05

diagonal and equal 6 -523.25 1196.69

55.94

Page 33 of 40 Global Change Biology

For Review Only

Table S4 Estimated correlations between Trend1 and the trend in density for the sub-stocks

related to Trend 1, with their corresponding P-values. The lag included in the estimation of

the density is indicated (see Methods). Significance was adjusted by a sequential Bonferroni

correction: the ordered P-values were compared with the inequality, P

≤ α(1 + k − i)

-1

, where

α is the confidence level to test for significance (0.05), K is the number of correlation tests

carried out and i is the rank of the correlation considered. Correlations for which the

inequality is met are significant (*).

Lag Correlation

p-value α(1 + k − i)

2 years -0.31 0.06193 0.010

0 to 2 years -0.26 0.11790 0.013

1 year -0.23 0.15790 0.017

0 to 1 year -0.18 0.27780 0.025

0 year -0.15 0.36450 0.050

Page 34 of 40Global Change Biology

For Review Only

Figure S1. a: the ICES standard roundfish areas for the North Sea used for the International Bottom Trawl Surveys. b: Overall average annual

sea bottom temperature (thick continuous line) between the average of roundfish areas 1 and 2 (lower continuous line) and the average of

roundfish areas 5 and 6 (upper continuous line). The two lower dashed lines correspond to areas 1 and 2, the two upper dashed lines to areas 5

and 6.

Page 35 of 40 Global Change Biology

For Review Only

Figure S2. Log-scaled relationships between the K and L

∞

parameters for the sub-stocks

considered in the analysis (triangles: cod, straight crosses: haddock, circles: whiting, squares:

herring, diagonal crosses: Norway pout, stars: sprat, F and M: female and male plaice, f and

m (in grey): female and male sole). Filled symbols stand for sub-stock in northern North Sea,

empty symbols for sub-stocks in the southern North Sea. Lines correspond to linear models

fitted to the data points.

Page 36 of 40Global Change Biology

For Review Only

Figure S3. Fishing mortality (filled circles) for the sub-stocks related to Trend 1 (fishing

mortality was assumed to be equal for both male and female sole South) plotted along the

fitted values from the best Dynamic Factor Analysis model (line) and their corresponding

95% confidence intervals (a: haddock North, b: Norway pout North, c: Sprat South, d: plaice

South, e: sole South, f: whiting North, g: whiting South, h: herring North). For stocks

distributed across the northern and southern North Sea (whiting and herring), a survey-based

assessment (SURBA) model (Beare et al. 2005) was used to obtain local estimates of total

mortality in order to capture spatial gradients in fishing pressure. Assuming a constant natural

mortality, total mortality times series for these two stocks were used as proxies for fishing

mortalities. Fishing mortality as well as the stock weight-at-age and proportion of mature-at-

age required for the SURBA model were obtained from the International Council for the Sea

(http://www.ices.dk/) 2012 assessment reports for the Working Group on the Assessment of

Demersal Stocks in the North Sea and Skagerrak (WGNSSK) and the Herring Assessment

Working Group (HAWG). Assessment data were available from 1970 to 2011 for all species

apart from whiting, Norway pout and sprat which assessments began respectively in 1990,

1983 and 1991.

Page 37 of 40 Global Change Biology

For Review Only

Figure S4. Abundance at age 1 indices (filled circles) used as a proxy for density for the sub-

stocks related to Trend 1 (for both plaice and sole sub-stocks the abundance index stands for

the males and females together as no sex-specific abundance index were available) plotted

along the fitted values from the best Dynamic Factor Analysis model (line) and their

corresponding 95% confidence intervals (a: haddock North, b: Norway pout North, c: Sprat

South, d: plaice South, e: sole South, f: whiting North, g: whiting South, h: herring North).

For stocks distributed across the northern and southern North Sea (whiting and herring), the

survey abundance at age 1 indices were split by area. For other sub-stocks, XSA abundance

at age 1 indices given in the International Council for the Sea (http://www.ices.dk/) 2012

assessment reports for the Working Group on the Assessment of Demersal Stocks in the

North Sea and Skagerrak (WGNSSK) and the Herring Assessment Working Group (HAWG)

were used. For Norway pout North and sole South, recruitment time series from the

assessment were used as no abundance at age 1 indices were available. Data were available

from 1970 to 2011 for all species apart from whiting, Norway pout and sprat for which

indices were available from 1990, 1983 and 1984 respectively.

Page 38 of 40Global Change Biology