Available via license: CC BY 4.0
Content may be subject to copyright.
MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Advance View
https://doi.org/10.3354/meps14513 February 29, 2024
1. INTRODUCTION
Small pelagic fish (SPF) such as anchovy, sardine,
and herring are a major resource for both predators
and fisheries and can act as an important ecosystem-
structuring agent among diverse ecosystems, from
polar to tropical and from oligotrophic open ocean to
eutrophic upwelling systems. They are the target spe-
© L.C., M.C., S.G., J.T., H.M., K.R., G.R., L.S., A.S., D.S., S.W., and, out-
side the USA, The U.S. Government 2024. Open Access under Creative
Commons by Attribution Licence. Use, distribution and reproduction
are un restricted. Authors and original publication must be credited.
Publisher: Inter-Research · www.int-res.com
*Corresponding author: james.ruzicka@noaa.gov
REVIEW
The role of small pelagic fish in diverse ecosystems:
knowledge gleaned from food-web models
James Ruzicka1,*, Luciano Chiaverano2, Marta Coll3, Susana Garrido4, Jorge Tam5,
Hiroto Murase6, Kelly Robinson7, Giovanni Romagnoni8,13, Lynne Shannon9,
Alexandra Silva4,10, Dorota Szalaj10,11, Shingo Watari12
1Pacific Islands Fisheries Science Center, Honolulu, HI 96818, USA
2Instituto Nacional de Investigación y Desarrollo Pesquero (INIDEP), Mar del Plata, B7062HSA, Argentina
3Institut de Ciències del Mar (ICM-CSIC) & Ecopath International Initiative (EII), Barcelona 08003, Spain
4Instituto Português do Mar e da Atmosfera (IPMA), Lisbon 1749-077, Portugal
5Instituto del Mar del Perú, Lima 07021, Perú
6Tokyo University of Marine Science and Technology, Tokyo 108-8477, Japan
7University Louisiana, Lafayette, LA 70503, USA
8Leibniz-Zentrum für Marine Tropenforschung, Bremen 28359, Germany
9University of Cape Town, Cape Town 7701, South Africa
10Marine and Environmental Sciences Centre (MARE), Lisbon 1749-016, Portugal
11Institut de Ciències del Mar (ICM-CSIC), Barcelona 08003, Spain
12Japan Fisheries Research and Education Agency, Yokohama 236-8648, Japan
13Present address: Center for Ocean and Society (CeOS), University Kiel, Kiel 24118, Germany
ABSTRACT: Small pelagic fish (SPF) are important forage species and a target of major fisheries
within diverse ecosystems. SPF are a critical link between plankton and higher trophic levels. Under-
standing the network of dependencies among species and fisheries supported by SPF is required for
effective resource management and assessment of risks posed by environmental and anthropogenic
stressors. Food-web models represent a synthesis of knowledge of these dependencies and are a
platform for evaluating the consequences of change in SPF productivity. From Ecopath food-web
models archived within EcoBase (www.ecobase.ecopath.org) and from peer-reviewed literature, we
compiled physiological parameters, biomasses, diets, and fishery catch rates that define SPF charac-
teristics. From 199 models, metrics characterizing demand on ecosystem production, contribution to
predators and fisheries, and sensitivities to changes in SPF were calculated. Across all models,
globally, SPF represented 43% of total fish production and were supported by 8% of total primary
production (14% in open ocean and 10% in upwelling models). In turn, SPF represented 18% of total
fish and invertebrate catch (53% in upwelling models). From a services perspective, considering all
direct and indirect trophic pathways, SPF were major contributors to predators and fisheries. On aver-
age, SPF supported 22% of seabird production, 15% of mammal production, and 34% of total fisheries
catch. Support to upper trophic levels was greater in upwelling models (33% of seabird, 41 % of mam-
mal, and 62% of fishery production). These analyses show the importance of accounting for direct
and indirect support by SPF to predators and fisheries when making management decisions.
KEY WORDS: Forage fish · Mesopelagic fish · Food-web · Ecosystem services · EcoBase · Ecopath ·
ECOTRAN
O
PEN
PEN
A
CCESS
CCESS
Contribution to the Theme Section ‘Small pelagic fish: new research frontiers’
Mar Ecol Prog Ser · Advance View
cies in many of the world’s largest fisheries (Tam et al.
2008, Van Voorhees 2012, Pikitch et al. 2014, Watari
et al. 2019), and they can act as important structuring
agents in pelagic ecosystems. SPF are major prey spe-
cies for higher trophic level (TL) fish, seabirds, and
marine mammals (Cury et al. 2011, Ruzicka et al.
2013, Ouled-Cheikh et al. 2022) and are commonly
referred to as ‘forage fish’ in recognition of this role.
They can themselves be a major consumer of phyto-
plankton and zooplankton production (Cury et al.
2000, Smith et al. 2011). They serve as a critical
energy transfer node in pelagic food-webs, linking
lower and upper TLs (Bakun et al. 2010), and they can
exert control in limiting both lower and upper TL
production (Rice 1995).
The ecological role and importance of SPF can vary
between ecosystem types. In oligotrophic oceanic sys-
tems, small mesopelagic species (e.g. Myctophidae)
are not only important forage for predatory fish but
also play an important role in global carbon cycles as
a carbon pump to sub-surface waters via their diel
migration behavior (Davison et al. 2013). In highly pro-
ductive upwelling ecosystems, the mid-TLs are often
occupied by very few SPF species (Cury et al. 2000).
In such ‘wasp-waist’ food-webs, where energy flow
through mid-TLs is channeled through a small number
of species, SPF have enhanced influence on the large-
scale dynamics of the ecosystem. Well-documented
changes in SPF dynamics have been coincident with
large-scale and persistent changes in ecosystem struc-
ture (e.g. Anderson & Piatt 1999, Roux et al. 2013)
SPF share common behavioral and physiological
characteristics distinct from other consumer groups
that define their own sensitivities to local con-
ditions. They are short-lived, grow rapidly, tend to
form schooling aggregations, and can respond more
rapidly to environmental change than longer-lived
species (Peck et al. 2021, Baez et al. 2022b). They are
sensitive to climate variability (Alheit & Hagen 1997,
Chavez et al. 2003, Alheit & Niquen 2004) and to
changes in competition for plankton resources, such
as during jellyfish blooms, with negative consequences
to higher TLs and fisheries (Robinson et al. 2014,
Chiaverano et al. 2018, Baez et al. 2022a).
The purpose of this study is to evaluate the ecologi-
cal role of SPF in diverse ecosystem types. We first
aim to quantify the importance of SPF in terms of the
demands they place upon lower TL production and in
terms of their contribution as a resource to support
higher trophic consumer and fishery production. We
next aim to quantify the role of SPF as an ecosystem
structuring agent in pelagic ecosystems in terms of
their importance as a link between lower and higher
TLs. Identification of critical trophic links and knowl-
edge of how the dynamics of these groups propagate
through the ecosystem is critical to the development
of effective management strategies to maintain resili-
ent ecosystems and to better understand the con-
sequences of potential management actions. The
development of ecosystem models and model analy-
sis tools is critical to this task, particularly in the
context of increasing demand for fishery resources
and changing environmental conditions.
Our approach is to analyze the characteristics of
SPF and quantify the effects they have on other con-
sumer groups through a suite of metrics derived from
food-web models representing diverse ecosystems
around the globe. From publicly available Ecopath
food-web models (www.ecopath.org; Christensen &
Walters 2004) archived within the EcoBase repository
of food-web models (www.ecobase.ecopath.org; Col-
léter et al. 2015) and additional models from the liter-
ature, we compiled metrics summarizing SPF physio-
logical characteristics, trophic position, productivity,
contribution to fisheries, rates and sources mortality,
direct and indirect demands placed upon food-web
resources, direct and indirect contributions to higher
TL production, and the sensitivities of different food-
web components to changes in SPF abundance. For
each of these metrics, we identify how the ecological
role of SPF differs among major ecosystem types
(eastern boundary upwelling, continental shelf, open
ocean, bay/fjord/lagoon, estuary, and coral reef) and
across latitudes. We include additional analyses to
consider key SPF groups individually and to compare
4 major eastern boundary current upwelling systems
(the Humboldt, Benguela, California, and Canary
Current ecosystems).
2. METHODS
We analyzed 199 food-web models to compare the
roles of SPF in terms of their functional trophic rela-
tionships to other living groups within diverse ecosys-
tem types and latitude zones (Table 1). Table 2 de fines
27 metrics characterizing SPF within each model ana-
lyzed within this study. Each metric and its derivation
is detailed below, and the statistical analysis of these
metrics is described in Section 2.4.
2.1. Selection of food-web models
Ecopath (www.ecopath.org) is a publicly available
and popular modeling platform for the construction
2
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
and analysis of food-web models (Christensen &
Walters 2004). Ecopath calculates rates of biomass
transfer between each taxonomic group or functional
group defined in the model given their diets, intrinsic
(weight-specific) production rates, production effi-
ciencies, catch rates within defined fisheries, and
biomass densities. Data required for each para -
meterare typically assembled from a wide variety of
sources that may span months to decades of observa-
tion. Ecopath models, therefore, represent a broad
synthesis of abundance, diet, physiology, and fishery
knowledge.
We primarily relied upon Ecopath food-web models
available in the EcoBase model repository (www.
ecobase.ecopath.org). EcoBase is an open-access
repository of 205 published Ecopath model parameter
sets and metadata available for download (Colléter et
al. 2015). We developed a script within Matlab (www.
matworks.com) to query, parse parameters, and
further analyze models from EcoBase. EcoBase was
accessed in September 2022. Of 205 marine food-
web models, 181 models were successfully processed
through all steps to confirmation of mass-balance.
EcoBase models were excluded if they represented
freshwater or beach ecosystems (6), could not be ver-
ified to be in mass balance (12), or there remained
undiagnosed processing errors by our script (6). We
added an additional 18 models not yet included in the
EcoBase repository for a total of 199 models ana-
lyzed in this study. Model name, citation, ecosystem
type, and latitude information are provided in Table S1
in the Supplement at www.int-res.com/articles/
suppl/m14513_supp.xlsx (see Table S29 for a list
of ex cluded models). Model locations are shown in
Fig. 1.
2.2. Ecosystem type classifications
Models were classified by ecosystem type and by
latitude zone: polar (>60°), temperate (24°–60°), and
tropical (0°–24°). We used EcoBase metadata ecosys-
tem type assignments as guidance, but in spected and
then re-classified each model into 6 types based on
geographic location, domain area, and community
composition: eastern boundary current upwelling,
continental shelf, open ocean, bay/fjord/lagoon,
estuary, or coral reef. Upwelling models were further
sub-divided into 4 large marine ecosystems: (1) Hum-
boldt Current, (2) California Current, (3) Benguela
Current, and (4) Canary Current. Model counts
within each ecosystem type and latitude zone are
given in Table 1.
2.3. Functional group definitions
There were >3800 unique functional groups de -
fined among the 199 analyzed models. Our analyses
required the development of a common set of defini-
tions for all functional groups. SPF were restricted to
planktivorous, schooling species. When resolved by
the models, 9 separate SPF sub-groups were consid-
ered: (1) anchovy (Engraulidae); (2) sardine, (3) her-
ring, (4) menhaden, and (5) sprat (Clupeidae); (6)
smelt (Osmeridae); (7) shad (Alosidae); (8) bonga
shad (Dorosomatide); and (9) surface-associated fly-
ingfish (Exocoetidae), saury (Scomberesocidae), and
halfbeaks (Hemiramphidae). The classifications of
ambiguously named groups, such as ‘small pelagics’
and ‘fish planktivorous’, were evaluated individually
based on taxonomic details provided by published
model descriptions. Hybrid groups that included
both SPF and non-SPF were classed as non-SPF pela-
gic fish. Counts of models with aggregated SPF and
resolved SPF groups are provided in Table 1. We also
analyzed the role of small mesopelagic fish but we did
not pool them with other SPF species, which are pre-
dominately epipelagic. All other groups in each
model were assigned to a major lower TL producer
class (primary producers, zooplankton, benthic inver-
3
Total Unique Models
models ecosystems with SPF
Ecosystem type
Upwelling 27 10 27
Continental shelf 97 65 89
Open ocean 23 16 15
Bay/fjord/lagoon 33 28 22
Estuary 8 7 6
Coral reef 11 10 5
Total 199 136 164
Upwelling system
Humboldt Current 8 3 8
Benguela Current 11 2 11
California Current 3 2 3
Canary Current 5 3 5
Total 27 10 27
Latitude zone
Polar 13 10 10
Temperate 112 74 96
Tropical 74 53 58
Total 199 137 164
Table 1. Sample size of analyzed models and models that
in clude small pelagic fish (SPF) by ecosystem type and lati-
tude zone
Mar Ecol Prog Ser · Advance View
tebrates) or to a major higher TL consumer class
(squid, non-SPF pelagic fish, demersal fish, seabirds,
marine mammals, fisheries) (see Tables S27 & S28 for
functional group definitions).
All metrics, with the exception of tMTI (see Sec-
tion 2.5.5), were derived from individual food-web
models at their original level of taxonomic resolution.
Metrics from individual functional groups were
binned into SPF and SPF sub-groups via summation
(biomass, catch, production rates, footprint, reach,
and food-web modification-scenario response metrics)
or as production-weighted means (TL, omnivory index
[OI], physiological rate, and mortality metrics). Pro-
duction rates were calculated from the product of the
biomass and production/biomass ratio as provided by
each model. When present, multi-stanza groups were
aggregated in the same manner (e.g. juvenile and
adult sardines were aggregated into a single sardine
group and this aggregated sub-group also contributes
to the larger, aggregated SPF group).
4
Metric Definition Units
Physiological rate and trophic position metrics
P/B Production rate to biomass ratio yr–1
P/Q Production efficiency: production to consumption ratio Ratio
TL Trophic level: number of trophic steps preceding consumption of prey by SPF, referenced from –
primary producers or detritus at TL = 1 and weighted by the contribution of each prey type to
the SPF diet
OI Omnivory index: TL variance of prey types in an SPF group’s diet, weighted by the contribution –
of each prey type to the diet
Biomass, production, and catch metrics
B/Bfishes SPF biomass to total fish biomass ratio %
P/PPrimProd SPF production to total primary production ratio Ratio
P/Pfishes SPF production to total fish production ratio %
C Catch: rate that an SPF group is caught by a fishery. C = retained landings + discarded biomass t km–2 yr–1
C/Ctotal Ratio of an SPF group’s catch to total catch of all fish + invertebrate groups %
Cscaled Catch rate of SPF scaled by the geographic area of the model t yr–1
Mortality metrics
M2 Predation mortality rate: total rate that an SPF group is eaten by all consumer groups (biomass yr–1 ), yr–1
divided by the SPF group’s biomass
F Fishing mortality rate: total rate that an SPF group is caught by a fleet (biomass yr–1 ), divided by yr–1
the SPF group’s biomass
M2fishes Mortality rate due to predation by all fishes yr–1
M2seabirds Mortality rate due to predation by all seabirds yr–1
M2mammals Mortality rate due to predation by all mammals yr–1
Metrics of group demands upon (footprint) and contributions to (reach) the food-web
fPrimProd Gross footprint on primary production: fraction of total primary production that contributes to %
the production of an SPF group, via all direct and indirect trophic pathways. The gross footprint
includes the costs of metabolism and feces losses at each trophic step but excludes senescence
losses
θfishes Reach of SPF group to fishes: fraction of all fish production that is supported by an SPF group via %
all direct and indirect trophic pathways
θseabirds Reach to all seabird production %
θmammals Reach to all mammal production %
θfisheries Reach to all fishery production %
θlanded groups Reach to all fishery-targeted groups %
Food-web sensitivity metrics
ΔPpelagic fishes Scenario response of pelagic fishes: relative change in production of all non-SPF pelagic fishes % Change
following a forced 20% reduction in SPF biomass
ΔPdemersal fishes Scenario response of all demersal fishes % Change
ΔPseabirds Scenario response of all seabirds % Change
ΔPmammals Scenario response of all mammals % Change
ΔPlanded groups Scenario response of all fishery-targeted groups % Change
ΔPfisheries Scenario response of all fisheries % Change
tMTI Total mixed trophic impact: net bottom-up and top-down impact that a change in an SPF Percentile
group’s biomass will have on all other groups. Expressed as a percentile ranking compared to rank
the impacts caused by change to each group in the model
Table 2. A brief description of each metric evaluated in this analysis. SPF: small pelagic fish
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
2.4. Statistical analyses
Because taxonomic resolution varies widely be -
tween models, our statistical analyses consider SPF as
an aggregated functional group, giving the greatest
possible global coverage among ecosystem types and
latitude zones. The distribution of SPF parameters,
derived metrics, and sensitivity analyses were com-
pared across ecosystem types and latitudes using
generalized linear mixed effect models (GLMMs).
GLMMs are used to describe the relation between a
response variable Y (i.e. food-web metrics) and one or
more independent or interacting fixed-effect terms.
GLMMs also account for random-effects terms that
are additional sources of variability in the food-web
metric beyond error due to random sampling (Zuur et
al. 2013). Ecosystem type (categorical) and latitude
(continuous) are fixed-effect terms. Individual food-
webs (i.e. specific ecosystems) are treated as a ran-
dom effect, controlling for the fact that each ecosys-
tem may be represented by multiple food-web models.
The main GLMM model structure was Y ~ 1 + eco-
system type + latitude + (1|individual food web). A
second GLMM analysis was conducted to focus on
the differences between individual upwelling eco -
systems: Y ~ 1 + upwelling system + (1|individual
food web).
Most food-web metrics were normally distributed,
following square root or log transformation, and were
modeled with the ‘fitglme’ function of Matlab (www.
mathworks.com) using the default identity link func-
tion. The reach contribution of SPF to landed groups
targeted by fisheries (θlanded groups; see Section 2.5.4)
best matched a zero-inflated beta distribution and was
modeled with the ‘glmmTMB’ function in R (Brooks et
al. 2017). Mortality rates due to predation by seabirds
and mammals (see Section 2.5.3) followed Tweedie
distributions and were modeled using the ‘glmmPQL’
function within the ‘MASS’ package in R (Venables
& Ripley 2002). Significance of differences of most
metrics among ecosystem types and among lati-
tudezones, i.e. those modeled with the ‘fitglme’ and
5
Fig. 1. Locations and ecosystem types of 199 analyzed Ecopath food-web models. Blue: polar models; green: temperate models;
red: tropical models
Mar Ecol Prog Ser · Advance View
‘glmmPQL’ functions, were determined at the α = 0.05
level using t-statistics. Significant differences of
θlanded groups, modeled with the ‘glmmTMB’ function,
were determined at the α = 0.05 level using z-statistics.
2.5. Definition and derivation of food-web metrics
2.5.1. Trophic level and omnivory index
We surveyed SPF diets in terms of trophic level (TL)
and omnivory index (OI), which are both dimension-
less. TL represents a group’s position in the food web
as the number of trophic steps that biomass and energy
pass through before being consumed by the group.
TLs were calculated by Ecopath (Christensen et al.
2005) as the biomass-weighted mean of the TL of each
prey type consumed by the group, referenced to TL =
1 for primary producers and detritus. The OI is the
variance of the TL of all prey items consumed by a con-
sumer group, weighted by the contribution of each
prey type to the consumer’s diet. OI was calculated as
in Ecopath (Christensen et al. 2005). OI = 0 indicates
that the group eats only one prey type, and a large OI
means that the group has a varied diet and feeds across
several TLs. OI is robust against differences in tax-
onomic resolution among the food-web models.
2.5.2. Production, biomass, and catch
We surveyed SPF productivity in terms of intrinsic
rates of production (P/B), production efficiencies
(P/Q), standing stock biomass, rates of biomass pro-
duction, and rates of fishery catch. P/B is a group’s
defined intrinsic rate of production relative to bio-
mass (yr–1), and P/Q is the dimensionless ratio of a
group’s production rate relative to its consumption
rate. To minimize variability among models due to
differences in area of geographic coverage and differ-
ences in ecosystem primary production, we consid-
ered SPF production relative to total primary produc-
tion (P/PPrimProd) and relative to total fish production
(P/Pfishes). We considered SPF biomass relative to
total fish biomass (B/Bfishes) and fishery catch rates
relative to total catch rates of all fish and inverte-
brates (C/Ctotal).
2.5.3. Natural mortality and fishery mortality
We surveyed the natural mortality (M2; yr–1) and
fishing mortality rates (F; yr–1 ) of all models, and we
resolved the main drivers of predation mortality
(M2fishes, M2seabirds, and M2mammals). The consumption
matrix Qgc describes the predation pressure on each
producer, g, (rows) in terms of biomass consumed per
period of time by each consumer, c, (columns) and is
calculated via Ecopath or Rpath algorithms (Chris-
tensen & Walters 2004, Lucey et al. 2020):
Qgc = Dgc · q·c (1)
where Dgc is the diet matrix defining the fraction of
each prey type g in the diet of each consumer c, and
q·c is a horizontal vector defining the consumption
rate of each consumer. The mortality rates of an SPF
group g were calculated from the elements of Qgc,
summing down rows corresponding to specific con-
sumer or fishery groups c and dividing by the stand-
ing stock biomass of the SPF group (equations repre-
sent element-wise operations unless noted by bracket
notation [], in Eqs. 3 & 7). We assumed F = 0 among
models that did not include a fishing fleet (n = 8).
2.5.4. Demands upon and contributions to the
food-web (footprint and reach)
The demands upon and contributions to other
groups in the food-web by SPF are expressed with
footprint (f) and reach (θ) metrics, respectively. We
calculated the footprints of SPF and other major con-
sumer classes upon primary producers and the reach
of these groups to predatory fish, seabirds, marine
mammals, and fisheries.
The importance of SPF as energy transfer nodes was
evaluated within the ECOTRAN framework (Steele &
Ruzicka 2011). ECOTRAN is a modeling platform that
can be used to describe trophic interactions between
multiple species and fishing fleets, the recycling of
detritus and nutrients, and the exchange of material
between sub-regions and depth strata via sinking,
physical flux, and migration (Ruzicka et al. 2016).
ECOTRAN models can be run as time-dynamic sim -
ulations; however, in the context of this study, physi-
cal fluxes and migration are not considered, and all
metrics and simulations were evaluated as steady-
state expressions of the food-web. ECOTRAN is
based on the transformation of a food-web expressed
as a matrix of predation pressures upon each pro-
ducer g by each consumer c (Qgc) into a donor-driven
trophic matrix (Acg) describing the fraction of pro-
duction flowing from each producer g (columns) to
each consumer c (rows) (Steele 2009, Steele & Ruzicka
2011):
6
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
(2)
where term Σc(Dgc · q·c) is the total predation pressure
upon group g. Each element of Acg represents a pro-
portion of total biomass input to a producer group
and has a value between 0 and 1. Expression of the
food-web as trophic matrix Acg is convenient for
quantifying the role of any functional group in terms
of its energy demand on lower TLs and its con -
tribution to higher TLs. (The ECOTRAN platform
code base used for this review is archived at https://
github.com/JimRuzicka-NOAA/SmallPelagicFish_
EwEreview).
The footprint of consumer c upon producer g (e.g.
SPF as consumer c and phytoplankton as producer g)
is calculated as the fraction of g production that sup-
ports the production of c via all direct and indirect
trophic pathways:
fc· =
[
[diag(1/te·g)] – [Acg ]
]
–1 · [dg·] (3)
The footprint of a specific consumer c on producer g
is vector element fc·. Term te·g is a horizontal vector of
transfer efficiencies for each producer g to the next
TL that accounts for losses to metabolism, feces pro-
duction, and senescence. The footprint is calculated
for all consumers by driving the food-web with the
external input of 1 unit of the individual producer
group g of interest. Term dg· is the vertical food-web
driver vector with all elements = 0 except dg· = 1 for
the producer group of interest. To prevent double
counting the contribution of group g to group c, feed-
back loops and detritus recycling pathways within the
trophic matrix are deactivated. Detritus uptake is set
to zero (except flow between detritus pools). Ele-
ments of vector fc· are recalculated for each group c
consecutively with predation upon group c set to
zero. Losses due to metabolic costs and feces produc-
tion at each trophic step are included in the footprint
calculations by setting all transfer efficiencies te·g = 1
except for the terminal detritus group, which is
defined as 0.1. Senescence losses do not contribute to
group c production or to the footprint of c upon pro-
ducer g, but are directed to detritus pools in Acg. This
calculation of the gross footprint is analogous to the
primary production required (PPR) calculated by
Ecopath (Christensen et al. 2005, Essington 2006)
when g is a primary producer. As descriptive short-
hand in the text, the footprint of SPF on primary pro-
duction is given as fSPF,PrimProd.
The reach of SPF group g is the fraction of any con-
sumer group c production that is supported by the
SPF group. From Ruzicka et al. (2012), the reach of g
is calculated by iteratively multiplying the contrib-
ution of g to the diets of each consumer through diet
matrix Dgc. Matrix Tgc represents the fraction of bio-
mass passing through each trophic linkage in the
food-web that originated with g. Tgc is estimated
through iteration as:
Tgc = θc· · Dgc (4)
Vertical reach vector θc· is the fractional contrib-
ution of g to the diet of each consumer. θc· is initialized
as θc· = Dg·’, the vertical transpose of row g in the diet
matrix. In each iteration of Eq. (4), Tgc represents the
contribution of g to the diet of each consumer through
direct and indirect trophic pathways up to length l =
iteration count. The total contribution of g to each
consumer is recalculated after each iteration by sum-
ming Tgc down all rows:
θc· = (Σg Tgc)’ (5)
The reach of g (e.g. SPF) to any specific consumer c
is then element c in the final reach vector θc· . Before
calculating the reach, the diet composition of each
consumer (D.c) was renormalized to sum to 1 after set-
ting all cannibalism elements Dcc = 0, and the contrib-
ution of g to itself was set to θg· = 1 in each iteration.
We performed a maximum of l = 1000 iterations with
progression halted once no element of θ differed
bymore than 0.0001 from the previous iteration. As
descriptive shorthand in the text, the reach of an
SPF group to a specific consumer group is given as
θSPF,consumer.
2.5.5. Sensitivity analyses (abundance scenarios
and mixed trophic impact)
The effects of changes in SPF abundance on all
other elements of each food-web were investigated
using the methodology developed by Steele (2009). A
set of forced changes to food-web structure repre-
sents a ‘scenario’. Energy flow through SPF was mod-
ified by reducing the availability of SPF to predators
by an arbitrary but standardized 20% within each
trophic matrix Acg (SPF rows in Acg were reduced by
20%). Surplus prey production no longer consumed
by the reduced SPF group was distributed propor-
tionally among all other consumers so that total pre-
dation pressure on each group remained unchanged.
The production rates of all functional groups were
calculated for the unmodified base model and the
Acg =Qgc
/cDgc ·q·c
^h
7
Mar Ecol Prog Ser · Advance View
8
modified sensitivity model using Eq. (3). Resulting
changes to functional group productivities represent
the consequences of a linear reapportioning of avail-
able prey among consumers (Collie et al. 2009).
Sensitivities of each major consumer class c (SPF,
mesopelagic fish, non-SPF pelagic fish, demersal fish,
seabirds, mammals, fisheries) were calculated as the
change in production relative to the base model:
ΔPc = (Pc base – Pc modified)/Pc base (6)
The sensitivity of other groups in the food-web to
changes in SPF abundance was also evaluated with
mixed trophic impact indices (MTI). MTI is a dimen-
sionless metric that quantifies the net direct and indi-
rect impact that a hypothetical change in biomass of
impactor group g would have on every other model
group c. MTI was calculated as described by Ulano -
wicz & Puccia (1990):
MTIgc =
[
[I] – [(Dgc) – (Acg)]
]
–1 – [I] (7)
where I is the identity matrix, and flow to detritus
pools c in trophic matrix Acg is set to 0. The overall
sensitivity of the food-web to changes in each model
group was summarized as the total MTI (tMTI) (Pra-
novi et al. 2003, Coll et al. 2007). The tMTI of a group
is the sum of all its impacts weighted by the inverse
ofthe biomass of each impacted group. To compare
SPF groups across different food-web models, we
used the percentile rank of SPF tMTI among the
tMTI of all other living groups in each model. Ag -
gregations of SPF, seabirds, mammals, and fleets into
pooled functional group classes were made prior
tothe calculation of MTI by adding appropriate ele-
ments of the consumption matrix and deriving
the new diet and trophic matrices from the aggre-
gated Qgc.
3. RESULTS
3.1. TL and OI
Across all models, the global median TL of SPF
is3.10 (TL interquartile range [IQR]1,3 = 2.88–3.30;
Table S2). For context, SPF feed at a lower TL than
doother pelagic fish, squid, demersal fish, seabirds,
or mammals (Fig. 2a). SPF TLs differ significantly
among ecosystem types (t157 = 2.66, p < 0.01) and are
lowest in upwelling and reef system models (Table 3).
TLs also differ significantly by latitude (t157 = 5.68,
p < 0.01), with the lowest TLs in tropical systems
(Table S4; ad ditional detailed results are included in
Tables S2–S26).
The OIs are similar for SPF (median OI = 0.14,
IQR1,3 = 0.06–0.27), squid, other pelagic fish, seabirds,
and mammals, but they feed across a narrower range
of TLs than do demersal fishes (Fig. 2b). The OIs of SPF
do not differ significantly among ecosystem types,
though OI trends higher in coral reef models and lower
in the open ocean (Table 3). OIs do differ significantly
by latitude (t157 = –5.30, p < 0.01; Table S4).
3.2. Physiology
SPF have higher P/B rates than other fish, seabirds,
and mammals (median P/B = 1.50, IQR1,3 = 1.07–
2.11, n = 164), but SPF rates are only half that of ceph-
alopods (Fig. 2c). P/B rates differ significantly among
ecosystem types (t157 = 2.46, p = 0.02) and latitude
zones (t157 = –6.20, p < 0.01), being highest in coral
reef and tropical models and lowest in upwelling and
polar models (Table 3, Table S4).
Growth efficiencies (i.e. P/Q) are similar to those of
other fish groups (median P/Q = 0.16, IQR1,3 = 0.11–
0.23) but are substantially lower than those of cepha-
lopods and much higher than those of warm-blooded
seabirds and mammals (Fig. 2d). While P/Q do not
differ significantly among latitude zones, they do
differ among ecosystem types (t157 = 3.03, p < 0.01)
and are highest in continental shelf and estuary
models and lowest in upwelling models (Table 3).
3.3. Biomass, production, and fishery catch
Across all models, SPF represent 35% of the median
total fish biomass (B/Bfishes IQR1,3 = 20–47 %). This is
comparable to demersal fishes but nearly twice the
biomass of non-SPF pelagic fishes (Fig. 3a). The SPF
proportion of total fish biomass is highest in upwel-
ling models (Fig. 4a) and differs significantly among
ecosystem types (t157 = –2.84, p = 0.01; Table 3). The
SPF biomass contribution to total fish biomass does
not differ significantly across latitudes (Table S4). In
terms of production, SPF are as productive as all
other non-SPF pelagic and demersal fish combined,
whether scaled relative to total primary production or
to total fish production. Across all models, the global
median SPF production is 0.2% the scale of total pri-
mary production (P/PPrimProd IQR1,3 = 0.1–0.6 %) and
represents 43% of total fish production (P/Pfishes
IQR1,3 = 24–64 %; Fig. 3b,d). SPF production rates rel-
ative to total fish production are not significantly re -
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
lated to latitude but did differ significantly among eco-
system types (t157 = –3.16, p < 0.01; Fig. 4d, Table 3).
The highest median production rates relative to total
fish production are in shelf and upwelling models and
lowest in estuary and coral reef models.
The proportion of SPF in the total catch of fish and
invertebrates across all models, globally, is compa-
rable to that of non-SPF pelagic fishes, and the
median proportion of SPF in the total catch is only
slightly lower than that of demersal fish (median
C/Ctotal = 18%, IQR1,3 = 2–48 %; Fig. 3c). The relative
contribution of SPF to total catch is significantly dif-
ferent among ecosystem types (t157 = –2.71, p =
0.01), with the greatest contribution in upwelling sys-
tems (median C/Ctotal = 53%, IQR1,3 = 8–75%; Fig. 4c,
Table 3). Actual catch rates SPF are markedly higher
in upwelling models (median C = 1.7 t km–2 yr–1 ,
IQR1,3 = 0.3–5.6 t km–2 yr–1; Table 3), with 6 outliers
ranging from 14 to 90 t km–2 yr–1 all representing the
Humboldt Current system in Chile and Peru.
3.4. Mortality rates
Across all models, M2 in the pooled SPF group is
about 14 times greater than F: median M2 = 1.01 yr–1
(IQR1,3 = 0.62–1.38 yr–1 , n = 164) versus F = 0.07 yr–1
(IQR1,3 < 0.01–0.21 yr–1 ; Table S8). The median pre-
dation mortality on SPF is twice that suffered by non-
SPF pelagic fish and demersal fish but half that suf-
fered by squid. The major contributors to predation
mortality are piscivorous fish, which impose much
higher mortality than does fishing (median M2fish =
0.73 yr–1 , IQR1,3 = 0.41–1.11 yr–1; Table S8). The
9
Fig. 2. Global distribution of small pelagic fish metrics from food-web models by consumer class. (a) Trophic level, (b) omnivory
index, (c) production to biomass ratio, (d) production to consumption ratio (P/Q). In each boxplot, the notch and center bar
represent the median, the shaded area represents the interquartile range (IQR) between the 1st and 3rd quartiles, whiskers
represent highest and lowest observations within 150% of the IQR, and dots represent outliers outside the IQR
Mar Ecol Prog Ser · Advance View
combined predation mortality due to seabirds and
marine mammals is less than that of the rate of F
(median M2seabirds = 0.01 yr–1 , IQR1,3 < 0.01–0.04 yr–1;
M2mammals = 0.02 yr–1 , IQR1,3 < 0.01–0.07 yr–1 ;
Table S8). Predation mortality by fishes differs signif-
icantly by latitude (t157 = –6.66, p < 0.01; Table S10)
and between ecosystem types (t157 = 2.28, p = 0.02),
and M2fishes was highest in the estuary and coral reef
models (Table 4). Mammal-driven mortality also
differs significantly by latitude (t157 = 4.26, p < 0.01;
Table S10) and between ecosystem types (t157 = 2.52,
p = 0.01) and is greatest in open ocean (M2mammals =
0.03 yr–1 , IQR1,3 = <0.01–0.15 yr–1) and continental
shelf models (M2mammals = 0.03 yr–1 , IQR1,3 = <0.01–
0.01 yr–1 ; Table 4). Seabird-driven predation mortal-
ity does not differ significantly among ecosystem
types or by latitude (Table 4, Table S10).
Across all models, the F on SPF was comparable
to F on non-SPF pelagic and demersal fish groups
(median F = 0.07 yr–1 , IQR1,3 = <0.01–0.21 yr–1, n =
164; Table S8). However, the median global F rate is
not a good indicator of fishing pressure in specific eco-
systems. F on SPF differs significantly among ecosys-
tem types (t157 = –2.31, p = 0.02) and by latitude (t157 =
–2.05, p = 0.04). Fishing pressure is highest in coral
reef (F = 0.16yr–1 , IQR1,3 = 0.09–0.75 yr–1) and up -
welling models (F= 0.11 yr–1 , IQR1,3 = 0.03–0.17 yr–1)
and lowest in the open ocean and polar ecosystems
(F < 0.01 yr–1, IQR1,3 = <0.01–0.01 yr–1 ; Table 4,
Table S10).
3.5. Demands upon and contributions to the
food-web (footprint and reach)
The demands that SPF impose upon the ecosystem
can be expressed with footprint metrics, the propor-
tion of a lower TL group’s production that supports
SPF production via all direct and indirect trophic
pathways. The global median SPF footprint on total
net primary production is 8.4% (fPrimProd IQR1,3 = 3.3–
19%; Fig. 5a). This is comparable to the high-end
estimates of the demands that non-SPF pelagic fish
(12%) and demersal fish place on the ecosystem (13%;
Table S14). The SPF footprint on phytoplankton does
not differ significantly among ecosystem types or by
latitude but tends to be higher in open ocean models
(Fig. 5c, Table S15).
The contribution of SPF to the ecosystem is ex -
pressed by reach metrics, the proportion of a con-
sumer group’s production that is supported by SPF
via all direct and indirect trophic pathways. The
global median contribution of SPF to other fish is
θfishes = 4.7% (IQR1,3 = 1.9–8.2 %), to seabird produc-
10
Upwelling Shelf Ocean Bay/fjord Estuary Coral reef Sig.
(n = 27) (n = 89) (n = 15) (n = 22) (n = 6) (n = 5)
TL 2.90 3.12 3.22 3.10 2.95 2.74 **
2.60–3.20 2.98–3.35 3.05–3.28 2.78–3.17 2.82–3.22 2.65–3.02
OI 0.18 0.12 0.09 0.13 0.22 0.28 NS
0.13–0.27 0.03–0.27 0.03–0.20 0.08–0.27 0.09–0.31 0.16–0.44
P/B 1.34 1.51 1.43 1.51 2.17 3.00 *
(yr–1 ) 1.24–1.65 0.97–2.11 0.88–2.81 0.87–1.86 1.99–3.41 1.97–4.05
P/Q 0.10 0.20 0.16 0.12 0.21 0.14 **
0.10–0.14 0.14–0.26 0.10–0.22 0.07–0.20 0.16–0.25 0.10–0.18
B/Bfishes 40.31 32.99 26.20 28.97 13.29 19.65 **
(%) 35.7–59.3 21.1–48.9 7.30–47.3 16.5–39.7 6.7–23.2 8.75–28.5
P/PPrimProd 0.27 0.18 0.30 0.18 0.12 0.35 NS
(%) 0.17–0.88 0.08–0.54 0.04–0.72 0.11–0.35 0.03–0.22 0.26–0.66
P/Pfishes 45.94 43.85 38.10 39.09 25.21 26.98 **
(%) 37.0–67.8 26.8–66.1 4.25–65.0 18.4–53.8 11.9–40.5 18.4–35.1
C/Ctotal 53.41 25.22 6.94 8.86 7.26 16.52 **
(%) 8.47–75.0 3.91–48.2 0.09–13.0 <0.01–43.3 2.14–17.6 4.45–22.0
C 1.70 0.25 0.03 0.25 0.10 0.50 ***
(t km–2 yr–1 ) 0.31–5.57 0.04–0.95 <0.01–0.13 <0.01–1.52 0.02–0.52 0.30–2.67
Cscaled 282 509 29 725 15975 127 28 18 290 –
(t yr–1 ) 16 038–914 776 3672–161146 9–72 232 <0.01–1753 20–1119 2214–27786
Table 3. Distributions of food-web metrics for small pelagic fishes (SPF) within different ecosystem types. See Table 2 for defini-
tions of each metric. Values shown are the 50th percentile and the 25th–75 th percentiles. Significance determined by general-
ized linear mixed effect analysis: NS, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001. Total catch C and geographically scaled
catch Cscaled omit models with currencies other than tons live weight: shelf n = 83, bay/fjord n = 21, estuary n = 5
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
tion is θseabirds = 22% (IQR1,3 = 0.2–48 %), and to mam-
mal production is θmammals = 15% (IQR1,3 = 1.2–41 %;
Fig. 5b). These contributions to higher TLs are much
higher than those of cephalopods or other fish (Fig. 5b,
Table S14). SPF reach to other fish was slightly higher
in continental shelf and upwelling models but did not
differ significantly between ecosystem types. Reach to
seabirds and mammals were both significantly higher
in upwelling models (θseabirds = 33%, IQR1,3 = 24–80%,
t157 = –2.08, p = 0.04; θmammals = 41 %, IQR1,3 = 35–
55%, t157 = –4.84, p < 0.01; Fig. 5d, Table S15).
Across all models, the reach contribution of SPF
to fisheries is greater than their contribution to
otherconsumer groups (global median θfisheries = 34%,
IQR1,3 = 12–59 %; Fig. 5b, Table S14). The SPF con-
tribution to fisheries is greater than either demersal
fish or non-SPF pelagic fish, but less than all non-SPF
fish combined (Table S14). The global median SPF
contribution to fisheries is nearly 10 times greater
than that of squid. SPF reach to fisheries differs sig -
nificantly among ecosystem types (t157 = –2.88, p <
0.01) but not among latitude zones. SPF reach to fish-
eries is greatest in upwelling models (upwelling
median θfisheries = 62%, IQR1,3 = 25–80 %; Fig. 5d).
Across all models, the contribution of SPF to the
landed groups targeted by fisheries is much lower
than the SPF contribution to fleets themselves
(global median θlanded groups = 1.8%, IQR1,3 = 0.46–6.3%;
Table S14) and is comparable to the SPF reach
to fishes in general. SPF reach to groups targeted
by fisheries differs significantly by ecosystem type
(z153 = –1.98, p = 0.05), and the median is
slightlyhigher in upwelling systems (upwelling median
θlanded groups = 4.0%, IQR1,3 = 2.0–18%; Table S15).
11
Fig. 3. Global distribution of metrics of biomass, production, and fishery catch for small pelagic fish (SPF), mesopelagic fishes,
small carangid mackerels, small scombrid mackerels, non-SPF pelagic fishes, and demersal fishes. (a) Group biomass to total
fish biomass ratio; (b) group production to primary production ratio, (c) group catch in fisheries to total fish and invertebrate
catch, (d) group production to total fish production. Boxplot parameters as in Fig. 2
Mar Ecol Prog Ser · Advance View
3.6. Comparison of major upwelling systems
Table 5 provides the distributions of SPF metrics
among the 4 major eastern boundary current upwel-
ling ecosystems. SPF in the Humboldt Current are
parameterized with higher median P/B rates than
other upwelling systems, but the California Current
models have the highest growth efficiencies (P/Q).
SPF in the Humboldt Current represent a signifi-
cantly higher proportion of the total fish biomass
(median B/Bfishes = 69 %, t23 = 2.77, p = 0.01), are sig-
nificantly more productive than other upwelling sys-
tems relative to total fish production (median P/Pfishes
= 80%, t23 = 2.87, p = 0.01), and represent a signifi-
cantly higher proportion of the total catch (median
C/Ctotal = 84%, t23 = 2.46, p = 0.02). F also differs
significantly among the 4upwelling systems and is
highest in the Humboldt Current (F = 0.36 yr–1, t23 =
3.29, p < 0.01). In the Humboldt Current, the median
F rate is nearly one-third the predation mortality rate
(M2 = 1.29 yr–1 ; Table S11).
SPF footprints on primary production do not differ
significantly among upwelling systems but tend to be
higher among Humboldt Current models (median
fPrimProd = 23%). The median contribution of SPF to
higher TL consumers, the reach (θ), is substantially
higher in the Humboldt Current. Among Humboldt
Current models, 13% of total fish production, 90% of
seabird production, 69% of marine mammal produc-
tion, and 90% of fishery production is supported by
SPF. By comparison, the median SPF contributions to
fisheries in the other 3 upwelling ecosystems range
from θfisheries = 25 % in the California Current to 62%
in the Benguela Current, contributions to seabird
production range from θseabirds = 10% in the Canary
Current to 32% in the Benguela. SPF contributions
12
Fig. 4. Distribution of metrics of biomass, production, and fishery catch for small pelagic fish by ecosystem type. (a) Group bio-
mass to total fish biomass ratio, (b) group production to primary production ratio, (c) group catch in fisheries to total fish and
invertebrate catch, (d) group production to total fish production. Boxplot parameters as in Fig. 2. Significant differences
between ecosystem types determined by generalized linear mixed effect analysis: NS, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001
Ruzicka et al.: Small pelagic fish—knowledge from food-webs 13
Upwelling Shelf Ocean Bay/fjord Estuary Coral reef Sig.
(n = 27) (n = 89) (n = 15) (n = 22) (n = 6) (n = 5)
M2 1.01 0.92 1.04 0.78 1.18 1.41 *
(yr–1 ) 0.64–1.21 0.65–1.39 0.75–1.81 0.36–1.30 1.02–1.74 1.16–2.21
F 0.11 0.08 0.01 0.06 0.05 0.16 *
(yr–1 ) 0.03–0.17 0.01–0.26 <0.01–0.01 <0.01–0.28 <0.01–0.21 0.09–0.75
M2fishes 0.79 0.62 0.65 0.57 1.07 1.20 *
(yr–1 ) 0.50–0.95 0.36–1.10 0.35–1.20 0.35–1.03 0.95–1.74 0.82–1.72
M2seabirds 0.03 0.01 0.03 0.01 0.02 0.02 NS
(yr–1 ) 0.01–0.09 <0.01–0.03 <0.01–0.04 <0.01–0.06 <0.01–0.10 0.01–0.10
M2mammals 0.02 0.03 0.03 <0.01 <0.01 0 *
(yr–1 ) 0.01–0.07 <0.01–0.09 <0.01–0.15 <0.01–0.02 <0.01–0.03 –
Table 4. Distributions of mortality rates for small pelagic fishes within different ecosystem types. See Table 2 for definitions of
each metric. Values shown are the 50th percentile and the 25th–75 th percentiles. Significance determined by generalized linear
mixed effect analysis: NS, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001
Fig. 5. Group demands on primary production (footprint) and contributions to higher trophic level consumers (reach). (a) Foot-
print of small pelagic fish (SPF), mesopelagic fishes, small carangid mackerels, small scombrid mackerels, and cephalopods on
primary production. (b) Reach of SPF, mesopelagic fishes, small carangid mackerels, small scombrid mackerels, and cephalo-
pods to all fishes, seabirds, mammals, landed groups targeted by fisheries, and fisheries production. (c) Footprint of SPF on pri-
mary production in different ecosystem types (upwelling, continental shelf, open ocean, bay/fjord/lagoon, estuary, and coral
reef). (d) Reach of SPF to all fishes, seabirds, mammals, landed groups targeted by fisheries, and fisheries production in different
ecosystem types. Boxplot parameters as in Fig. 2. Significant differences between ecosystem types determined by generalized
linear mixed effect analysis in (c) and (d): NS, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001
Mar Ecol Prog Ser · Advance View
to marine mammal production range from median
θmammals = 36% in the California Current to 41 % in the
Benguela Current. However, only the reach to fish-
eries differs significantly between the upwelling sys-
tems (t23 = 3.10, p = 0.01).
3.7. Differences among SPF types
Table 6 summarizes the metrics of biomass, catch,
and reach contributions to the ecosystem of 9 major
SPF groups. Sardine, herring, and anchovy have the
greatest representation among all models analyzed.
The most productive groups relative to total fish pro-
duction are smelt (median P/Pfishes = 23%), anchovy
(19%), sardine (14%), and menhaden (14%). In terms
of contribution to higher TLs, the greatest support to
seabirds is from smelt (θseabirds = 18%), anchovy
(9.3%), bonga shad (7.8%), and menhaden (6.6%). The
greatest support to marine mammals is from smelt
(θmammals = 18%), sardine (6.3%), and anchovy (5.7%).
The most heavily harvested groups relative to total
fish catch are menhaden (median C/Ctotal = 33%),
bonga shad (20%), sardine (11 %), and herring (5.1%). F
rates are highest for bonga shad (F = 0.20 yr–1), men-
haden (0.17 yr–1 ), and sardine (0.14 yr–1). The most
important groups supporting fisheries
via direct and indirect trophic path-
ways are menhaden (θfisheries = 36%),
bonga shad (22%), and sardine (17%),
matching the groups with the highest
harvest and fishery mortality rates.
Anchovy and herring support 10 and
8.1% of the total fishery catch, respec-
tively.
Small mesopelagic fishes, small ca -
rangid mackerels, and small scombrid
mackerels are also well-represented
among food-web models but are con-
sidered separately from the pooled
SPF group. These 3 groups feed at
slightly higher TLs than most other
SPF groups (median TL = 3.4, 3.5, and
3.6, respectively; Table S6). Mesopela-
gic fish production relative to total
fish production is on average, globally
lower than that of the SPF group as a
whole (P/Pfishes = 15%; Fig. 3d) but
comparable to the median production
rates of anchovy (19 %) and sardine
(14%), individually. The relative pro-
duction rates of small carangid
(P/Pfishes = 6.7%) and scombrid mack-
erels (2.3%) are also lower than those of most SPF
groups (Fig. 3d). Mesopelagic fish contribute very lit-
tle to the total catch (Fig. 3c), and the reach contrib-
ution of mesopelagics to fishery production is much
lower than the contribution of most other SPF groups
(θfisheries = 1.6%; Fig. 5b). Small mackerels also con-
tribute less to support fishery production than most
other individual SPF groups (θfisheries = 6.9 and 3.7%
for carangids and scombrids, respectively; Fig. 5b,
Table S18).
Table 7 identifies the top 3 individual SPF groups
in each ecosystem type in terms of their footprint
demands on primary production and their reach con-
tribution to fishery production. Anchovy, sardine,
and herring are among the most important groups in
terms of both footprint and reach in most ecosystem
types. Carangid mackerels also appear among the 3
most important groups in several ecosystem types.
Mesopelagic fishes, smelts, and carangid mackerels
are the most important groups in open ocean models
in terms of their resource demands and in terms of
their contribution to fishery production. Mesopelagic
fishes have a moderately larger footprint on primary
production in coastal bay/fjord/lagoon models than
in oceanic models, though this is estimated from only
2 bay/fjord/lagoon models.
14
Humboldt Benguela California Canary Sig.
(n = 8) (n = 11) (n = 3) (n = 5)
P/B 1.93 1.28 1.09 1.30 **
(yr–1 ) 1.76–2.21 1.24–1.34 1.02–1.36 1.05–1.32
P/Q 0.13 0.10 0.23 0.14 ***
0.10–0.15 0.10–0.10 0.18–0.25 0.10–0.14
B/Bfishes 68.9 37.0 26.2 42.6 *
(%) 58.7–78.7 35.7–39.3 21.0–38.2 33.0–50.2
P/Pfishes 79.5 39.1 48.3 49.0 *
(%) 65.1–89.5 35.3–43.3 35.4–67.6 39.5–56.4
C/Ctotal 83.6 53.4 9.56 28.1 *
(%) 76.8–88.9 <0.01–60.2 2.39–22.4 18.2–52.5
F 0.36 0.05 0.03 0.14 ***
(yr–1 ) 0.19–0.50 <0.01–0.11 0.01–0.06 0.11–0.15
fPrimProd 22.81 8.61 9.65 7.14 NS
(%) 13.0–39.2 6.71–27.2 6.16–14.6 6.14–10.8
θseabirds 89.5 31.6 23.7 9.77 NS
(%) 85.6–93.3 29.2–57.1 5.93–55.2 8.66–22.8
θmammals 69.1 41.4 35.8 37.5 NS
(%) 44.1–95.1 20.0–47.0 13.9–51.4 19.3–38.4
θfisheries 90.3 61.7 25.0 35.9 **
(%) 83.3–95.1 12.3–69.1 20.4–29.3 23.2–56.9
Table 5. Comparison of food-web metrics for small pelagic fishes (SPF) among
4 major eastern boundary current upwelling zones. See Table 2 for definitions
of each metric. Values shown are the 50th percentile and the 25th–75 th percent-
iles. Significance determined by generalized linear mixed effect analysis: NS,
p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001
Ruzicka et al.: Small pelagic fish—knowledge from food-webs 15
Sardine Herring Anchovy Smelt Sprat Menhaden Bonga shad Shad Flyingfish
(n = 56) (n = 52) (n = 50) (n = 27) (n = 11) (n = 9) (n = 9) (n = 8) (n = 14)
P/B 1.29 1.03 1.80 1.43 1.21 1.55 0.87 0.60 2.28
(yr–1 ) 0.95–1.65 0.74–1.42 1.40–2.49 0.89–1.78 1.05–1.31 1.18–2.25 0.86–1.22 0.43–1.22 1.24–2.81
P/Q 0.11 0.15 0.15 0.22 0.14 0.12 0.06 0.21 0.16
0.10–0.16 0.11–0.21 0.10–0.22 0.16–0.29 0.12–0.24 0.08–0.22 0.06–0.10 0.14–0.22 0.11–0.23
B/Bfishes 9.37 10.38 10.51 10.34 3.69 9.44 7.86 1.31 2.00
(%) 4.04–23.3 2.99–16.9 4.82–26.7 3.59–27.0 3.06–4.37 4.45–15.8 6.28–16.6 0.59–2.65 0.72–3.20
P/Pfishes 14.44 8.55 19.03 25.11 5.53 13.99 6.90 1.67 1.95
(%) 4.09–24.9 3.69–18.2 8.39–33.3 6.47–47.8 4.15–7.74 6.14–20.8 3.63–21.2 0.66–2.47 0.99–4.06
C/Ctotal 11.40 5.07 2.93 0.05 1.35 33.49 19.54 0.39 0.19
(%) 1.65–24.6 0.66–16.8 <0.01–22.5 <0.01–1.80 1.16–2.82 1.46–37.1 16.0–29.8 0.09–1.78 0.00–1.48
F 0.14 0.06 0.08 <0.01 0.08 0.17 0.20 0.06 0.01
(yr–1 ) 0.03–0.42 <0.01–0.21 <0.01–0.30 0–<0.01 0.02–0.16 0.08–0.59 0.10–0.31 0.01–0.15 <0.01–0.03
fPrimProd 3.70 3.07 5.75 5.79 1.02 2.41 1.13 0.44 0.44
(%) 0.68–7.83 1.04–6.77 1.62–9.68 0.88–13.2 0.52–3.38 0.89–3.31 0.57–3.10 0.10–1.10 0.23–1.00
θseabirds 5.70 2.48 9.26 17.63 3.43 6.61 7.80 1.31 0.16
(%) 3.74–25.9 0.13–3.77 1.38–20.8 5.24–44.3 0.60–10.4 1.50–13.0 6.66–9.66 0.31–2.01 <0.01–2.91
θmammals 6.29 3.87 5.71 13.32 3.95 1.27 1.88 0.26 0.26
(%) 2.28–17.5 0.42–8.11 0.74–31.3 1.73–29.5 1.41–6.96 0.00–6.58 1.39–6.48 0.12–1.25 0.00–1.51
θfisheries 16.51 8.05 10.21 5.93 2.95 35.81 21.90 0.47 1.13
(%) 7.26–29.5 2.28–22.6 1.58–33.4 0.95–21.3 2.17–4.85 3.75–40.0 16.9–34.5 0.16–1.75 0.36–5.86
Table 6. Comparison of food-web metrics among specific types of Small pelagic fishes (SPF) across all models, globally. See Table 2 for definitions of each metric. Values
shown are the 50th percentile and the 25th–75th percentiles
Upwelling Continental shelf Open ocean Bay/fjord/lagoon Estuary Coral reef
fPrimProd Anchovy 4.0 Anchovy 7.6 Mack. Carangid 50.9 Mesopelagic 16.3 Anchovy 6.5 Mack. Carangid 11.0
(%) Sardine 3.9 Sardine 4.5 Smelt 18.8 Herring 8.6 Mack. Carangid 3.7 Anchovy 8.7
Herring 3.1 Herring 3.1 Mesopelagic 12.5 Mack. Carangid 8.3 Sardine 2.3 Sardine 3.5
θfisheries Anchovy 30.5 Menhaden 37.3 Mack. Carangid 50.5 Herring 74.4 Sardine 16.3 Sardine 17.4
(%) Sardine 16.5 Bonga shad 21.9 Smelt 33.0 Bonga shad 17.2 Mack. Scombrid 6.5 Mack. Scombrid 16.1
Mack. Carangid 7.2 Sardine 17.6 Mesopelagic 10.6 Sardine 13.7 Anchovy 5.3 Mack. Carangid 13.8
Table 7. Identification of the 3 most important small pelagic fish groups in each ecosystem type in terms of their demands on primary production and their contribution to
fishery production via all direct and indirect trophic pathways. Mesopelagic fishes, small carangid mackerels, and small scombrid mackerels are included in the rankings.
See Table 2 for definitions of each metric. Values shown are medians of all models in each ecosystem type
Mar Ecol Prog Ser · Advance View
16
3.8. Sensitivity analysis
The effects of a 20% reduction of SPF biomass on
other consumer groups are shown in Fig. 6a. The ef -
fects were usually smaller than the 20% forced
reduction of SPF. The net effect on predators and
fisheries was generally negative. Seabirds, mammals,
and fisheries declined in most ecosystem types. Pela-
gic fishes, demersal fishes, and fishery-targeted
landed groups potentially include both competitors
and predators of SPF and generally increased when
the availability of SPF was reduced. These fish
groups and fisheries were most sensitive in upwel-
ling models.
While the SPF reduction scenarios highlight the
bottom-up effects on higher TL consumers when the
availability of SPF as a trophic link between lower and
upper TLs is altered, the MTI analysis considers the
role of SPF as both a predator and as prey. Fig. 6b
shows the percentile rank of the total MTI (tMTI) of
SPF on all living groups in the food-web in relation to
the tMTI ranking of all other living groups in each
model. In general, SPF are among the most influential
groups in most ecosystem types, with median tMTI
rankings within the upper 75th percentile in upwel-
ling, continental shelf, and coral reef models and
within the upper 66th percentile in open ocean and
bay/fjord/lagoon systems. In estuary models, how -
ever, SPF were below the median in terms of net
effects on food-web groups. Also shown are the tMTI
rankings of mesopelagic fishes and small mackerels.
These latter groups show a lot more variability be -
tween ecosystem types than SPF but generally have
lower tMTI rankings. However, mesopelagic fishes in
open ocean models have comparable tMTI rankings
to SPF.
Fig. 6. (a) Effects of a 20 % reduction in small pelagic fish (SPF) biomass upon the productivity of major consumer classes (squid,
non-SPF pelagic fish, demersal fish, seabirds, marine mammals, and fisheries. Results presented as percent change in produc-
tion relative to an unaltered model. Results are arranged by ecosystem type (upwelling, continental shelf, open ocean,
bay/fjord/lagoon, estuary, and coral reef). (b) Distributions of the percentile ranking of total mixed trophic impact (tMTI) of
SPF on living groups in each ecosystem model by ecosystem type. Boxplot parameters as in Fig. 2
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
4. DISCUSSION
4.1. Approach of this review and analysis
The purpose of this study was to evaluate the role of
SPF as a resource to predators and fisheries and as an
ecosystem-structuring agent among diverse ecosys-
tems. Our approach was to take advantage of the large
body of peer-reviewed Ecopath food-web models
(Christensen & Walters 2004) to derive a suite of met-
rics that characterize SPF in terms of their trophic po-
sition, productivity, the demands they place on lower
TL resources, the contributions they provide to higher
TL consumers, and the sensitivity of different living
groups to variability in SPF abundance and biomass.
Our analyses were restricted to metrics that could be
derived from mass-balanced Ecopath food-webs rather
than from time-dynamic, environmentally driven eco-
system simulations such as those produced by Ecosim
and Ecospace (e.g. Coll et al. 2006), Atlantis (e.g. Ka-
plan et al. 2019), or OSMOSE (e.g. Travers et al. 2006).
We did this because of the large number and high eco-
system diversity represented by existing Ecopath
models, the relative uniformity of the parameterization
structure allowed for rapid analyses across models,
and so that our analyses focused on the im plications of
food-web structure rather than time-dependent vari-
ability in environmental conditions. Metrics and sensi-
tivity analyses derived from steady-state food-webs as
presented here are informative ofthe trophic relation-
ships and energy flow patterns averaged across sea-
sons and years, but they cannot capture the non-linear
dynamics of interacting functional groups with differ-
ing response rates tochanging conditions. We consid-
ered evidence of consistent differences in SPF charac-
teristics among ecosystem types to be informative of
the range of ecological roles that SPF can take within
ecosystems sub ject to different physical drivers and
environmental conditions.
We were not selective of the models analyzed and
made no attempt to evaluate the quality of any model.
Models within the EcoBase archive were only re jected
if there were persistent reading errors by our analysis
code or if the food-web model was not in thermody-
namic balance (i.e. predation demands ex ceeded pro-
duction for any group). Additional peer-reviewed
models were added to the analysis based on the au-
thors’ knowledge of work done within regions of their
personal expertise. We did not limit the coverage of
any ecosystem to a specific era or climate condition
(e.g. Pacific models representing either El Niño or La
Niña conditions). We considered all such models to
be valid representations of that ecosystem at different
points in time. Comparison of models by ecosystem
type was limited by model availability, and ecosystem
types were not necessarily evenly represented. Cover-
age of Indian Ocean ecosystems was particularly
poor, while coverage of the Humboldt and Benguela
Current ecosystems was particularly good. Our statis-
tical analyses used GLMMs to control for the fact that
individual ecosystems may be represented by multiple
food-web models. Finally, we did not explicitly in-
clude the pelagic juvenile stages of fish species as a
major consumer class. Juveniles of non-SPF species
are known to be both important planktivores and
prey; for example, juvenile hake in the Southern
Benguela (Shannon et al. 2003) and juvenile mackerel
in the Canary Current (Garrido et al. 2015).
4.2. The importance of SPF as a consumer group
SPF are a very productive mid-TL group. They have
high intrinsic production rates (Fig. 2c) and, given the
resources, their biomasses can grow faster than the
other major fish groupings defined in this study.
Across all models analyzed, SPF were about as pro-
ductive as all other pelagic and demersal fish com-
bined (Fig. 3d). However, SPF production efficiency
was not higher than other fish (P/Q; Fig. 2d), and high
SPF production rates place high demands on ecosys-
tem resources. As an average across ecosystem
models, we estimate that SPF used 8% of the total pri-
mary production (fPrimProd; Fig. 5a). SPF demands are
slightly higher in open ocean models, where SPF used
14% of the primary production on average and meso-
pelagic fishes used 13% (Fig. 5a). In upwelling
models, we estimate that SPF required 10% of the
total primary production on average (Fig. 5c). Jarre-
Teichmann & Christensen (1998) used a standard
Ecopath algorithm to estimate the PPR to support
anchovy and sardine production within the same 4
major upwelling ecosystems as studied here. Their
estimated range of 10–25% agrees with our upwel-
ling footprint estimates of 7–23% (Table S17). High
estimates of the PPR to support coastal fisheries of all
targeted species (24–35%) led Pauly & Christensen
(1995) to suggest that there is little unused plankton
production that could be exploited to expand produc-
tion of the world’s major fisheries.
4.3. The importance of SPF as a producer group
Reach metrics (θ) calculated in our analysis show
that SPF are major contributors to higher TL con-
17
Mar Ecol Prog Ser · Advance View
sumer and fishery production (Fig. 5d). Our work fol-
lows the analysis of 72 globally distributed food-web
models by Pikitch et al. (2014) to estimate the contrib-
utions of forage fish and euphausiids to support the
production of predators and fisheries. They charac-
terized the importance of SPF to predators in terms of
diet composition and found that SPF represented at
least 10% of the diets in nearly half of all predator
groups in their model set. The highest direct pred-
atory dependence on SPF was in upwelling models (if
euphausiids are excluded), and seabirds were par-
ticularly dependent on SPF. In our analysis, the
importance of SPF to seabirds, mammals, and fish-
eries was also highest in upwelling models (Fig. 5d)
and comparable to contributions of all other pelagic
and demersal fishes combined. A study of covariance
between observed seabird breeding success and SPF
abundance confirmed the high dependence of sea-
bird populations on SPF abundance in real-world eco-
systems (Cury et al. 2011).
Pikitch et al. (2014) presented support service met-
rics analogous to the reach metrics calculated here,
though they only accounted for direct trophic link-
ages from SPF to fisheries and predatory fish targeted
by fisheries. They estimated that SPF support to fish-
eries accounts for 20% of the monetary value of the
global catch of fish. Our reach estimates show that
SPF directly and indirectly supports, on average ac-
ross all models, 34% of the global catch. This is about
twice the fraction of total catch that is composed of
SPF (C/Ctotal = 18%; Fig. 3c). Pikitch et al. (2014) did
not provide rates of total catch in terms of weight to
allow for a direct comparison of their monetary value-
based support service estimate to our biomass-based
rach estimate. However, they did observe that the
value of the fisheries supported by SPF, but not com-
posed of SPF, was also about twice the value of the
SPF catch, indicating the importance of accounting
for the indirect role that SPF play in supporting fish-
eries when making management trade-off decisions.
4.4. SPF as an ecosystem-structuring agent
Upwelling ecosystems stand out in our analyses
because in these systems SPF are shown to be an
important energy-transfer node. The SPF play the
largest role as consumers of lower trophic production
and as producers supporting predators and fisheries
in upwelling systems. This is especially true in the
Humboldt Current (Table 5), which produces more
fish per unit area than any other region in the world’s
oceans (Chavez et al. 2008).
Marine food-webs are structured with high species
diversity at the bottom and top TLs, but in upwelling
ecosystems, the middle TLs are occupied by a small
number of planktivorous SPF. This food-web configu-
ration is termed a ‘wasp-waist’ structure (Bakun 1996).
In a wasp-waist food-web, SPF groups have been
hypothesized to play a pivotal role in ecosystems (e.g.
Travers-Trolet et al. 2014), exerting strong bottom-up
resource limits on the production of top TLs and
strong top-down predation pressure limiting the bio-
mass of bottom TLs (Cury et al. 2000). The implication
of wasp-waist trophic control is that changes in SPF
abundance would affect multiple TLs across the pela-
gic food-web. A reduction in SPF abundance would
lead to a reduction in predator production and an
increase in production among the mid-TL competi-
tors of SPF. The mid-TL groups that take advantage of
reduced competition do not themselves necessarily
constitute an efficient alternate energy transfer path-
way to higher TLs (e.g. jellyfish and mesopelagic fish)
and increased detritus production and benthic food-
web production may result (Shannon et al. 2009).
Our sensitivity analysis of steady-state food-webs
shows which groups would benefit and which suffer
from a reduction in SPF abundance. The especially
large footprint and reach metrics and the strong
effects that SPF have on seabirds, marine mammals,
and fisheries in our sensitivity analysis on upwelling
models are consistent with a wasp-waist structured
food-web with strong bottom-up control by SF on
predator populations. We also see that demersal fish
production increases when SPF are reduced, but we
did not distinguish whether this is due to increased
detritus production and increased benthic food-web
production or to lower competition for plankton
among those demersal species that include plankton
in their diets. At least in upwelling systems such as
the Benguela, this has been shown to be the former
(e.g. Shannon et al. 2009).
The form of the static food-web scenarios con-
ducted here only simulates the effects of a reappor-
tioning of available prey among consumers (i.e. no
prey switching by predators) and is not able to eval-
uate the importance of top-down control by SPF on
plankton. This is another necessary condition of a
true wasp-waist system. However, the tMTI analysis
does consider the net effects of both bottom-up pro-
cesses, where SPF as prey limit predator production
and top-down processes, where SPF as a predator
limit plankton production. We see that SPF do have a
relatively high impact on the food-web compared to
other groups (Fig. 6b), but their impact in upwelling
systems is not noticeably high compared to other eco-
18
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
system types. Other studies have cast into doubt the
importance of wasp-waist control in specific ecosys-
tems. In the northern California Current, a high
degree of omnivory and the importance of euphausi-
ids in the diets of consumers at multiple TLs creates
pathways of bottom-up control that bypass the limita-
tions of the narrow waist of SPF at mid-TLs (Miller et
al. 2010). High taxonomic but low species diversity at
mid-TLs can provide alternate trophic pathways
between lower and upper TLs in upwelling systems
(Fréon et al. 2009). Fréon et al. (2009) found a high
degree of variability in the amount of lower TL pro-
duction consumed by SPF and the amount of produc-
tion that SPF, in turn, transfer to higher TLs. Wasp-
waist control is a dynamic process, and evidence of
top-down control on prey is dependent on the time-
period and SPF abundance when the ecosystem is
evaluated (Coll et al. 2008). Whether or not wasp-
waist control is an accurate description of the tropho-
dynamics of upwelling ecosystems, this survey of
food-web models and that of Pikitch et al. (2014) show
the important role that SPF can play as an energy
transfer node across multiple ecosystem types.
Mesopelagic fish are also a mid-trophic SPF group,
though we considered them separately in our analy-
ses. Although previous modeling studies do show
mesopelagics as having notable impacts on competi-
tor and predator groups (e.g. Smith et al. 2011), we did
not find mesopelagic fish to be as important an
energy transfer node as other SPF groups. Though
they place a large demand on primary production in
open ocean models and bay/fjord/lagoon models
(Table S19) they do not in turn contribute as much as
other SPF species to predators or fisheries. Only in
their support of marine mammals in the open ocean
do they match the importance of other SPF groups.
However, the biomass and trophic role of mesopela-
gic fish are poorly quantified in most ecosystems.
These results may reflect our limited knowledge of
mesopelagic fish biomass and trophic ecology when
these models were constructed. Mesopelagic fish do
have an additional important ecological role. Their
daily vertical migration from the epipelagic in the
daytime to the mesopelagic at night makes this group
of SPF an important pathway for transferring plank-
ton production and carbon from the surface to deep
waters (Davison et al. 2013, Anderson et al. 2019).
4.5. Implications and next steps
The management of a single species considers the
physiological limits of its productivity, its demogra-
phy, and how it is likely to respond to a realistic range
of environmental conditions. Instead, an ecosystem-
based fishery management (EBFM) approach ‘takes
major ecosystem components and services — both
structural and functional —into account in managing
fisheries’ (Garcia et al. 2003, Lidström & Johnson
2020). EBFM is potentially more accurate because it
considers the competitive and predator–prey inter-
actions between multiple species within a changing
environment that limit species’ growth, thus ensuring
the integrity of the ecosystem and consequently the
sustainability of the target single species. EBFM is
also a more versatile approach than single-species
management because it allows consideration of the
cost and benefit trade-offs of management policies. In
ecosystems where SPF occupy a critical position for
energy transfer within the food-web, or are them-
selves both a targeted fishery and the forage-base
supporting other targeted fisheries, development of
EBFM tools and protocols ‘seems especially war-
ranted’ (Pikitch et al. 2014).
Though we did see some evidence of SPF playing a
central wasp-waist ecosystem-structuring role in the
upwelling models analyzed here, this role was not
particularly stronger in upwelling models than in
other models we examined. However, we did find that
SPF do act as important energy transfer nodes linking
lower and upper TLs and that this was particularly
evident in upwelling systems and also true in conti-
nental shelf and open ocean systems. SPF appear to
have a lesser structuring role in bay/fjord/lagoon,
estuary, and coral reef systems, as evidenced by
smaller footprint demand and reach contribution
metrics and smaller impact on other consumer groups
in food-web sensitivity analyses. These findings sup-
port those of Essington & Munch (2014), who con-
cluded that trade-offs between fisheries that target
SPF versus those that target predatory fish that prey
on SPF are highly variable and depend on the balance
between bottom-up control of SPF in food-webs and
top-down control exerted through predation pres-
sure. Our results may also reflect the lack of knowl-
edge of early life stages of SPF and of coastal areas,
where those early stages congregate.
SPF are traded internationally but global supply
and price are volatile because biomasses of individual
stocks depend on regionally highly variable climatic
conditions (FAO 2022). Ecosystem-structuring pro-
cesses are dynamic, and a next logical step following
this analysis would be to conduct analyses with these
same models within a time-dynamic framework. Nev-
ertheless, even without temporally dynamic compar-
ative modeling, this study underlines a central and
19
Mar Ecol Prog Ser · Advance View
often overlooked aspect of EBFM; namely, that the
complex inter-relations between SPF and their pred-
ators have variable and complex implications for
simultaneous management of multiple fisheries and
conservation, and that indicators such as reach and
footprint as examined in this study and as found to
vary with ecosystem type can be a useful tool in focus-
ing management considerations per ecosystem type.
For example, in the case of upwelling and coral reef
systems, reach and footprint patterns highlight pro-
ductivity-linked changes (climate change) and sea-
bird forage as key management considerations,
whereas in the open ocean processes that might
influence mesopelagic fish warrant stronger focus,
and in bay/fjord and coral reef systems, marine mam-
mal–SPF interactions seem less important. Even a
non-dynamic food-web model, informed by robust
trophic ecology studies and multi-species ocean sur-
vey data, can serve as valuable tools for EBFM, high-
lighting the interactions between SPF and other
species and revealing potential trade-offs between
multi-species management goals and conservation
objectives across diverse ecosystems. We believe this
review highlights the need to continue research on
network indicators, such as reach and footprint, for
the management of exploited ecosystems, particu-
larly within the context of climate change.
Acknowledgements. This review is a contribution to the
Working Group on Small Pelagic Fish started jointly by ICES
(WGSPF) and PICES (WG43) to continue world-wide col-
laboration to advance knowledge of the drivers of popula-
tions of small pelagics. M.C. acknowledges the Spanish
Research Project ProOceans (RETO-PID2020-118097RB-100)
and the ‘Severo Ochoa Centre of Excellence’ accreditation
(CEX2019-000928-S). The scientific results and conclusions,
as well as any views or opinions expressed herein, are those
of the authors and do not necessarily reflect those of the
home institutions of the authors. Any use of trade, firm, or
product names is for descriptive purposes only and does not
imply endorsement by the home institutions and govern-
ments of the authors. S. Ruzicka (US Forest Service) pre-
pared the map in Fig. 1. Dr. J. Suca (University of Hawaii,
CIMAR) gave valuable guidance and assistance with the sta-
tistical analyses. This manuscript benefitted from the sug-
gestions of 2 anonymous reviewers.
LITERATURE CITED
Alheit J, Hagen E (1997) Long-term climate forcing of Euro-
pean herring and sardine populations. Fish Oceanogr 6:
130– 139
Alheit J, Niquen M (2004) Regime shifts in the Humboldt
Current ecosystem. Prog Oceanogr 60: 201– 222
Anderson PJ, Piatt JF (1999) Community reorganization in
the Gulf of Alaska following ocean climate regime shift.
Mar Ecol Prog Ser 189: 117– 123
Anderson TR, Martin AP, Lampitt RS, Trueman CN, Henson
SA, Mayor DJ (2019) Quantifying carbon fluxes from pri-
mary production to mesopelagic fish using a simple food
web model. ICES J Mar Sci 76: 690– 701
Baez JC, Pennino GM, Giraldez A, Albo Puigserver M, Coll
M, Bellido JM (2022a) Effects of environmental con-
ditions and jellyfish blooms on small pelagic fish and fish-
eries from the Western Mediterranean Sea. Estuar Coast
Shelf Sci 264: 107699
Baez JC, Pennino MG, Czerwinski IA, Coll M and others
(2022b) Long term oscillations of Mediterranean sardine
and anchovy explained by the combined effect of multi-
ple regional and global climatic indices. Reg Stud Mar
Sci 56: 102709
Bakun A (1996) Patterns in the ocean: ocean processes and
marine population dynamics. University of California
Sea Grant in cooperation with Centro de Investigaciones
Biológicas de Noroeste, San Diego, CA, and La Paz
Bakun A, Babcock E, Lluch-Cota S, Santora C, Salvadeo C
(2010) Issues of ecosystem-based management of forage
fisheries in ‘open’ non-stationary ecosystems: the exam-
ple of the sardine fishery in the Gulf of California. Rev
Fish Biol Fish 20: 9– 29
Brooks M, Kristensen K, van Benthem K, Magnusson A and
others (2017) glmmTMB balances speed and flexibility
among packages for zero-inflated generalized linear
mixed modeling. R J 9: 378– 400
Chavez FP, Ryan J, Lluch-Cota SE, Niquen MC (2003) From
anchovies to sardines and back: multidecadal change in
the Pacific Ocean. Science 299: 217– 221
Chavez FP, Bertrand A, Guevara R, Soler P, Csirke J (2008)
The northern Humboldt Current System: brief history,
present status and a view towards the future. Prog Ocean-
ogr 79: 95– 105
Chiaverano LM, Robinson KL, Tam J, Ruzicka JJ and others
(2018) Evaluating the role of large jellyfish and forage
fishes as energy pathways, and their interplay with fish-
eries, in the Northern Humboldt Current System. Prog
Oceanogr 164: 28– 36
Christensen V, Walters CJ (2004) Ecopath with Ecosim:
methods, capabilities and limitations. Ecol Model 172:
109– 139
Christensen V, Walters CJ, Pauly D (2005) Ecopath with Eco-
sim: a user’s guide. Fisheries Centre University of British
Columbia, Vancouver
Coll M, Palomera I, Tudela S, Sardà F (2006) Trophic flows,
ecosystem structure and fishing impacts in the South
Catalan Sea, Northwestern Mediterranean. J Mar Syst
59: 63– 96
Coll M, Santojanni A, Palomera I, Tudela S, Arneri E (2007)
An ecological model of the Northern and Central Adri-
atic Sea: analysis of ecosystem structure and fishing
impacts. J Mar Syst 67: 119– 154
Coll M, Palomera I, Tudela S, Dowd M (2008) Food-web
dynamics in the South Catalan Sea ecosystem (NW
Mediterranean) for 1978– 2003. Ecol Model 217: 95– 116
Colléter M, Valls A, Guitton J, Gascuel D, Pauly D, Chris-
tensen V (2015) Global overview of the applications of
the Ecopath with Ecosim modeling approach using the
EcoBase models repository. Ecol Model 302: 42– 53
Collie JS, Gifford DJ, Steele JH (2009) End-to-end foodweb
control of fish production on Georges Bank. ICES J Mar
Sci 66: 2223– 2232
Cury P, Bakun A, Crawford RJM, Jarre-Teichmann A, Qui-
nones R, Shannon LJ, Verheye HM (2000) Small pelagics
in upwelling systems: patterns of interaction and struc-
tural changes in ‘wasp-waist’ ecosystems. ICES J Mar Sci
57: 603– 618
20
Ruzicka et al.: Small pelagic fish—knowledge from food-webs
Cury PM, Boyd IL, Bonhommeau S, Anker-Nilssen T and
others (2011) Global seabird response to forage fish deple-
tion —one-third for the birds. Science 334: 1703– 1706
Davison PC, Checkley DM, Koslow JA, Barlow J (2013) Car-
bon export mediated by mesopelagic fishes in the north-
east Pacific Ocean. Prog Oceanogr 116: 14– 30
Essington TE (2006) Pelagic ecosystem response to a cen-
tury of commercial fishing and whaling. In: Estes JA,
DeMaster DP, Doak DF, Williams TM, Brownell RL Jr
(eds) Whales, whaling, and ocean ecosystems. University
of California Press, Berkeley, CA, p 38– 49
Essington TE, Munch SB (2014) Trade-offs between support-
ive and provisioning ecosystem services of forage species
in marine food webs. Ecol Appl 24: 1543– 1557
FAO (2022) The state of world fisheries and aquaculture
2022. Towards blue transformation. FAO, Rome
Fréon P, Arístegui J, Bertrand A, Crawford RJM and others
(2009) Functional group biodiversity in Eastern Bound-
ary Upwelling Ecosystems questions the wasp-waist
trophic structure. Prog Oceanogr 83: 97– 106
Garcia SM, Zerbi A, Aliaume C, Do Chi T, Lasserre G (2003)
The ecosystem approach to fisheries: issues, terminol-
ogy, principles, institutional foundations, implementa-
tion and outlook. FAO Fisheries Technical Paper No. 443.
FAO, Rome
Garrido S, Silva A, Pastor J, Dominguez R, Silva AV, Santos
AM (2015) Trophic ecology of pelagic fish species off the
Iberian coast: diet overlap, cannibalism and intraguild
predation. Mar Ecol Prog Ser 539: 271– 286
Jarre-Teichmann A, Christensen V (1998) Comparative mod-
elling of trophic flows in 4 large upwelling ecosystems:
global vs. local effects. In: Durand MH, Cury P, Mendels-
sohn R, Roy C, Bakun A, Pauly D (eds) From local to
global changes in upwelling systems. ORSTOM, Paris,
p423– 443
Kaplan IC, Francis TB, Punt AE, Koehn LE and others (2019)
A multi-model approach to understanding the role of
Pacific sardine in the California Current food web. Mar
Ecol Prog Ser 617– 618: 307– 321
Lidström S, Johnson AF (2020) Ecosystem-based fisheries
management: a perspective on the critique and develop-
ment of the concept. Fish Fish 21: 216– 222
Lucey SM, Gaichas SK, Aydin KY (2020) Conducting repro-
ducible ecosystem modeling using the open source mass
balance model Rpath. Ecol Model 427: 109057
Miller TW, Brodeur RD, Rau GH, Omori K (2010) Prey dom-
inance shapes trophic structure of the northern Califor-
nia Current pelagic food web: evidence from stable iso-
topes and diet analysis. Mar Ecol Prog Ser 420: 15– 26
Ouled-Cheikh J, Gimenez J, Albo Puigserver M, Navarro J
and others (2022) Trophic importance of small pelagic
fish to marine predators of the Mediterranean Sea. Mar
Ecol Prog Ser 696: 169– 184
Pauly D, Christensen V (1995) Primary production required
to sustain global fisheries. Nature 374: 255– 257
Peck MA, Alheit J, Betrand A, Catalán IA and others (2021)
Small pelagic fish in the new millennium: a bottom-up
view of global research effort. Prog Oceanogr 191: 102494
Pikitch EK, Rountos KJ, Essington TE, Santora C and others
(2014) The global contribution of forage fish to marine
fisheries and ecosystems. Fish Fish 15: 43– 64
Pranovi F, Libralato S, Raicevich S, Granzotto A, Pastres R,
Giovanardi O (2003) Mechanical clam dredging in Ven-
ice lagoon: ecosystem effects evaluated with a trophic
mass-balance model. Mar Biol 143: 393– 403
Rice J (1995) Food web theory, marine food webs, and what
climate change may do to northern marine fish popula-
tions. Publ Spec Can Sci Halieut Aquat 121: 561– 568
Robinson KL, Ruzicka JJ, Decker MB, Brodeur RD and
others (2014) Jellyfish, forage fish, and the world’s major
fisheries. Oceanography 27: 78–89
Roux JP, van der Lingen CD, Gibbons MJ, Moroff NE, Shan-
non LJ, Smith AD, Cury PM (2013) Jellyfication of marine
ecosystems as a likely consequence of overfishing small
pelagic fishes: lessons from the Benguela. Bull Mar Sci
89: 249– 284
Ruzicka JJ, Brodeur RD, Emmett RL, Steele JH and others
(2012) Interannual variability in the Northern California
Current food web structure: changes in energy flow path-
ways and the role of forage fish, euphausiids, and jelly-
fish. Prog Oceanogr 102: 19– 41
Ruzicka JJ, Steele JH, Ballerini T, Gaichas SK, Ainley DG
(2013) Dividing up the pie: whales, fish, and humans as
competitors. Prog Oceanogr 116: 207– 219
Ruzicka JJ, Brink KH, Gifford DJ, Bahr F (2016) A physically
coupled end-to-end model platform for coastal ecosys-
tems: simulating the effects of climate change and chang-
ing upwelling characteristics on the Northern California
Current ecosystem. Ecol Model 331: 86– 99
Shannon LJ, Moloney CL, Jarre A, Field JG (2003) Trophic
flows in the southern Benguela during the 1980s and
1990s. J Mar Syst 39: 83– 116
Shannon L, Coll M, Neira S, Cury P, Roux J (2009) Impacts of
fishing and climate change explored using trophic
models. In: Checkley D, Alheit J, Oozeki Y, Roy C (eds)
Climate change and small pelagic fish. Cambridge Uni-
versity Press, Cambridge, p 158– 190
Smith ADM, Brown CJ, Bulman CM, Fulton EA and others
(2011) Impacts of fishing low-trophic level species on
marine ecosystems. Science 333: 1147– 1150
Steele JH (2009) Assessment of some linear food web
methods. J Mar Syst 76: 186– 194
Steele J, Ruzicka JJ (2011) Constructing end-to-end models
using ECOPATH data. J Mar Syst 87: 227– 238
Tam J, Taylor MH, Blaskovic V, Espinoza P and others (2008)
Trophic modeling of the Northern Humboldt current
ecosystem, Part I: comparing trophic linkages under La
Niña and El Niño conditions. Prog Oceanogr 79: 352– 365
Travers M, Shin YJ, Shannon L, Cury P (2006) Simulating and
testing the sensitivity of ecosystem-based indicators to
fishing in the southern Benguela ecosystem. Can J Fish
Aquat Sci 63: 943– 956
Travers-Trolet M, Shin YJ, Shannon LJ, Moloney CL, Field
JG (2014) Combined fishing and climate forcing in the
southern Benguela upwelling ecosystem: an end-to-end
modelling approach reveals dampened effects. PLOS
ONE 9: e94286
Ulanowicz RE, Puccia CJ (1990) Mixed trophic impacts in
ecosystems. Coenoses 5: 7– 16
Van Voorhees D (2012) Fisheries of the United States 2012.
In: Lowther A (ed) Current fishery statistics. NOAA
National Marine Fisheries Service, Silver Spring, MD
Venables W, Ripley B (2002) Modern applied statistics with
S, 4th edn. Springer, New York, NY
Watari S, Murase H, Yonezaki S, Okazaki M and others
(2019) Ecosystem modeling in the western North Pacific
using Ecopath, with a focus on small pelagic fishes. Mar
Ecol Prog Ser 617– 618: 295– 305
Zuur A, Hilbe J, Leno E (2013) A beginner’s guide to GLM
and GLMM with R. Highland Statistics, Newburgh
21
Editorial responsibility: Isaac Kaplan (Guest Editor),
Seattle, Washington, USA
Reviewed by: L. Koehn and 1 anonymous referee
Submitted: April 30, 2023
Accepted: December 18, 2023
Proofs received from author(s): February 20, 2024
