Content uploaded by Jay Willis
Author content
All content in this area was uploaded by Jay Willis on Jan 21, 2016
Content may be subject to copyright.
MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 513: 51–69, 2014
doi: 10.3354/meps10922 Published October 22
INTRODUCTION
The model presented in this paper has been devel-
oped assuming the validity of the following basic
statements about the Antarctic ecosystem:
(1) Krill Euphausia superba abundance in the South-
ern Ocean used to be enough to support the whales
(and other krill predators) that existed before exploita-
tion and was close to its pristine ecosystem carrying
capacity.
(2) Krill abundance is now lower than it was before
the start of commercial whaling.
(3) Godlewska’s hypothesis is true: krill have
changed their habitat use because of the lack of
whales (more correctly, the lack of the perceived risk
of predation from whales).
(4) Whales enhance the surface feeding opportunity
for krill through nutrient recycling.
The aim of this study is to show how a simple model
that encompasses these assumptions leads to a pre-
diction about the current state of the ecosystem, and
likely future scenarios, which is more plausible than
other models that have been presented to date. The
first statement is hopefully uncontested given that
the ecosystem has existed long enough for the partic-
ipants to have evolved and that it is not contradicted
by any observation or theory; it also follows the eco-
logical theory around mature ecosystems (Odum
1969). Evidence to support the subsequent state-
ments is outlined below. The most important contri-
bution of this study is to provide evidence to support
these basic assumptions.
© Inter-Research 2014 · www.int-res.com*Corresponding author: jkwillis@gmail.com
Whales maintained a high abundance of krill;
both are ecosystem engineers in the Southern Ocean
Jay Willis*
Turnpenny Horsfield Associates Ltd., Ashurst Lodge, Ashurst, Southampton, Hampshire SO40 7AA, UK
ABSTRACT: Krill abundance was predicted to rise after the end of commercial whaling in the
Southern Ocean due to the release of predatory pressure from 2 million whales that were killed
between 1915 and 1970, but contrary to expectations, there has been a substantial decline in
abundance of krill since the end of whaling. I presented a model 7 yr ago which explained how
krill behaviour, in response to the threat of predation by whales, may provide an answer to this
paradox. The original model contained a speculative link: a mechanism by which krill could detect
the presence of whales over a wide area, and therefore could behave in response to a credible
threat. Recently, iron has been implicated in a positive feedback cycle between whales, krill and
primary production. The cycle depends on the buoyant faeces of whales fertilising surface layers.
This is both a plausible way for krill to detect whales over a wide area and an explanation for
enhanced feeding at the surface, but this was not incorporated in the original model. Thus, nutri-
ent retention and behavioural control are probably an example of niche construction and ecosys-
tem engineering by both krill and whales. In this paper I revisit and update the simple model of
krill mentioned above. The model is calibrated against known system states and is used to imply
the ecosystem level changes caused by commercial whaling. This improved model may explain
the reduction in krill abundance after the end of commercial whaling. Untested hypotheses which
can be falsified in designed experiments are listed.
KEY WORDS: Individual-based model · Ecosystem model · Cetacean
Resale or republication not permitted without written consent of the publisher
Mar Ecol Prog Ser 513: 51– 69, 2014
Krill abundance is now lower than before
commercial whaling started
Commercial whaling removed 2 million large
baleen whales from the Antarctic marine ecosystem
in less than 70 yr, starting around 1900 (Clapham &
Baker 2001). These included blue whales Balaeno -
ptera musculus, fin whales B. physalus, sei whales B.
borealis, humpback whales Megaptera novaengliae,
and Antarctic minke whales B. bonaerensis, here-
inafter referred to collectively as ‘whales’. These
whales were almost entirely sustained by krill. It was
assumed that the krill population would increase
tremendously as a result of the release in predatory
pressure due to commercial whaling (Laws 1977, Bal-
lance et al. 2006). This effect of predatory release
was expected to have been exacerbated by the com-
mercial harvesting of fur seals Arctocephalus gazelle
and ground fish, both krill predators, which occurred
around the beginning and throughout the period of
whaling, respectively (Ballance et al. 2006, Kock et
al. 2007). The krill population has not increased, and
rather the opposite is true. It is difficult to be certain
about krill abundance in the Antarctic marine eco-
system due to its size and inaccessibility; neverthe-
less, after 10 yr of searching, with equipment specifi-
cally designed for the purpose (Everson 2000),
scientists were unable to find enough krill to feed the
whales that were known to have been killed. Nicol et
al. (2000) calculated krill population abundance
around the year 2000 to be between 60 and 155 mil-
lion t. This figure was later revised to 133 million t
global biomass (Atkinson et al. 2009). Stomach con-
tents of >500 whales were used to estimate that,
before 1900, whales consumed a total of 175 to 190
million t of krill per year (Mackintosh 1973, 1974,
Ross & Quetin 1988) and this was used to infer that
the difference between total whale consumption of
krill before and after whaling was 147 million t (Laws
1977). This estimate was made under the assumption
that there were 975 000 whales in the pristine state
(Laws 1977) whereas now, with the benefit of Russian
catch records which were not previously available,
the best estimate is 1.697 million (Christensen 2006).
So, using the Laws (1977) multipliers (his Table 2) on
Christensen’s (2006) population estimates (her Table
3), the difference in krill consumed by whales before
and after whaling is now estimated at 276 million t.
This observed feeding requirement for the whales
that are known to have been killed is more than
twice the observed global standing stock of krill (133
million t mentioned above). Whales, however, are not
the only predators of krill: in 1985 it was estimated
that 470 million t of krill were eaten each year by
whales, seabirds, seals, squid and fish (Ross & Quetin
1988). This estimate of 470 million t required to main-
tain the system after exploitation of whales should be
compared with the present best modeled estimate of
342 to 536 million t per year gross larval production
(in 2000) and the consequent estimate of 128 to 470
million t per year available to predators (Atkinson et
al. 2009). The entire additional requirement of the
whales, if they existed now in their pristine abun-
dance, for ca. 276 million t of krill, would be unavail-
able at the higher bound of present modelled pro-
duction estimates. Hence, the production of krill
must have fallen by at least this amount between the
start and finish of commercial whaling (a decrease of
50% at the lowest proportional estimate). A huge
surplus of krill, and a tremendous rise in the abun-
dance of other predators of krill (compensatory pre-
dation) was expected (Ballance et al. 2006), but the
opposite has happened. Krill abundance has fallen,
and continues to fall (Atkinson et al. 2004). There
have been moderate changes in the abundance of
other krill predators; the fur seal population recov-
ered very strongly until around 1985 and has since
declinded, crabeater seals (Lobodon carcinophaga or
carcinophagus) have shown no long-term changes,
and most avian predators have declined in abun-
dance (Ballance et al. 2006). These changes do not
provide an adequate compensatory predator-based
explanation for the changes in krill abundance (Bal-
lance et al. 2006). The logical hypothesis is that
presently observed krill abundance is a fraction of
what was required to maintain the ecosystem before
whaling.
Godlewska’s hypothesis of krill behavioural
changes
Godlewska (1996) presented the results of sonar-
based krill surveys and suggested that krill had
changed their behaviour after the end of commercial
whaling. She suggested that krill had spent most
days at the surface before the end of whaling, but
after whaling spent the daylight hours at 100 m to
200 m depth and migrated each night to the surface
(Godlewska 1996). During the early period of whal-
ing, observers consistently reported a high abundan -
ce of krill and constant oc currence of surface schools
day after day (Marr 1964, Mackintosh 1973). Surface
schools of krill were described as being thick like
pea-soup, and extending continuously like immense
pastures over areas as large as 150 square miles
52
Willis: Whales and krill
(Marr 1964, Mackintosh 1973). Such was the impor-
tance, regularity and clear association with whaling,
that daylight surface schools of krill were mapped
across the entire area of commercial whaling by
observers from each of the main whaling countries
(Arsenev 1958, Marr 1964, Ozawa & Sato 1967). Early
scientists did not ignore the possibility of vertical
migrations; they searched for, and occasionally
recorded, this behaviour for adult and developmental
stages of krill (Mackintosh 1934, Hardy & Gunther
1936, Fraser 1937, Marr 1964). During the early
period of whaling, Mackintosh (1934) sampled the
vertical position of krill and concluded that for the
majority of day and night they stayed near the sur-
face, especially when in schools with only a minor
degree of diurnal variation. Marr (1964) reported that
krill only inhabited depths within 10 to 40 m of the
surface, day and night. This conclusion was derived
from a literature review and a comprehensive survey
based on net hauls. (Like more recent studies, this
study accounted for the fact that krill are adept at
avoiding nets at the surface in daylight; Marr 1964,
Nicol et al. 2010). After the end of commercial whal-
ing, there were no records of krill seen at the surface
in daylight despite continued observation. At this
time, a commercial krill fishery was developing and
it would have been convenient for the fishers to see
krill at the surface. However, sonar, echo sounders
and knowledge of previous abundance were the only
ways in which krill were found by the commercial
krill fishery during daylight (Ichii 2000). No large sur-
face schools were recorded as visible at the surface in
daylight in the scientific literature between 1970 and
2000. Vertical daily migration is now considered
standard behaviour for krill (Everson 2000). Typical
scientific samples during this period were taken
between the surface and 160 m in depth using sonar
for school location (Ross & Quetin 1983). Where super
swarms were reported in daylight in this period after
whaling (e.g. Higgin bottom & Hosie 1989, God lews -
ka 1996) they were detected with sonar; no krill were
seen at the surface. Thus there is strong evidence to
suggest Godlewska’s hypothesis is more a fact estab-
lished through observation rather than a hypothesis.
Whales enhance surface prey abundance for krill
through iron recycling and fertilisation
Krill fertilise the ocean in situ with their excrement
as they feed; a high krill biomass causes enhanced
primary production in Antarctic waters (Lehette et al.
2012). Krill bodies also contain high levels of iron rel-
ative to the surrounding waters, and thus the popula-
tion of krill represents a carrier for a significant pro-
portion of all the iron flux present in the surface
ocean of this region (Tovar-Sanchez et al. 2007).
Whale bodies and whale faecal material are similarly
rich in iron (Smetacek 2008). Whale faeces are buoy-
ant and fluid with a high water content, and whales
have been photographed defecating at the surface
(Smetacek 2008). This implies an iron and nutrient
storage and recycling mechanism that leads to a pos-
itive feedback in the food chain between the abun-
dance of primary production, krill and whales
(Smetacek 2008, Nicol et al. 2010). Krill may also
have engineered their environment and enhanced
primary production through turbulence (Kerr 2006,
Smetacek 2008, Willis 2013). Turbulence can be a
greater determinant of phytoplankton growth than
either salinity or temperature (Margalef 1978). Tur-
bulence as a determinant of phytoplankton growth
was largely overlooked until the energetic strength
of turbulence caused by fish and krill schools was
found to be on the same scale as that caused by wind
and waves (Huntley & Zhou 2004). It was found that
this turbulence is unlikely to contribute to overall
ocean mixing in the same way as wind, due to the
comparatively short mixing scales (Visser 2007).
However, turbulence with such energy and short
mixing characteristics is a very rare event in the
ocean outside of swimming animal schools (Jimenez
1997). Krill-induced turbulence may constitute eco-
system engineering through at least 3 mechanisms:
(1) by directly enhancing plankton productivity
(Margalef 1978), (2) by facilitating the sorting of prey
sizes and types by krill, which allows krill to engineer
the size and age composition of their prey (Willis
2013), and (3) by mixing the surface ocean to redis-
tribute physical substances (nutrients) and properties
such as salinity and temperature (Kerr 2006). The
impact of krill-induced turbulence on primary pro-
duction as a type of ecosystem engineering, or
energy fertilisation, process may well be significant.
Retention of nutrients and complex symbiotic inter-
actions are also consistent with the characteristics of
a mature ecosystem (Odum 1969). It is also interest-
ing to consider the higher density caused by com-
pression of the krill population into the top 10 m as
opposed to the deeper zone between 10 and 200 m.
Mackintosh (1934) suggested krill form denser
schools when at the surface. Compression and higher
density patches would potentially lead to more accu-
rate targeting of fertilisation of nutrients by whales
and would enhance the effects of turbulence caused
by the swimming action of krill (Willis 2013). In sum-
53
Mar Ecol Prog Ser 513: 51– 69, 2014
mary, there is a nutrient retention and recycling
mechanism which includes whales, krill and primary
production.
Ecosystem implication of the iron cycle
The existence of a nutrient and fertilisation mecha-
nism induced by whales and krill in the food chain
implies that the throughput of the entire Antarctic
ecosystem has been reduced due to decreased nutri-
ent retention caused by commercial whaling. This
means that the carrying capacity of krill is variable
and is dependent on the abundance of whales,
among other factors. In terms of the theory of krill
behaviour and abundance, outlined in this study, the
theory of an iron cycle provides 2 important explana-
tions: (1) the mechanism by which krill are aware of
whales and thus are able to react to the threat of pre-
dation could be odour in faeces, and (2) the iron cycle
linked through whales and krill defecation at the sur-
face coupled with high turbulence, temperature and
sunlight would provide a good explanation for higher
availability of food for krill at the surface in daylight.
In an earlier study (Willis 2007), and the original
model of krill on which it was based (Alonzo & Man-
gel 2001), the preferential conditions of the near-sur-
face habitat relating to higher prey availability for
krill were assumed to be present whether or not
whales were around.
Niche construction and ecosystem engineering
The basic questions of this study are (1) whether
whales and krill are mutual niche constructors which
engineer their environment and (2) whether this has
a significant impact on the abundances of both spe-
cies which that environment supports. The niche
construction theory explicitly recognises environ-
mental modifications (ecosystem engineering when
abiotic) by organisms and their legacy over time
(Odling-Smee et al. 2013). Ecosystem engineering is
not a new concept. Darwin (1881) showed that
worms were ecosystem engineers on geological
scales. Nevertheless it is often overlooked in stan-
dard behavioural ecological theory (for instance
Krebs & Davies 1993). Standard theory, although
recognising the physical environment as a source of
selection, tends to focus on the interactions of pheno-
types in different species, for example in food webs
(Odling-Smee et al. 2013). The model presented in
this paper may support niche construction theory by
giving a quantitative numerical example of the popu-
lation impact resulting from individual reproductive
success caused by niche construction. Ecosystem
engineering and niche construction as evolutionary
processes are possibly overlooked because they are
assumed to be of relatively minor impact and to hap-
pen over time scales much shorter than evolutionary
time scales (Odling-Smee et al. 2013). This is particu-
larly true of pelagic marine science where the eco-
system approach has tended to focus exclusively on
the food-web interactions of species, and concepts
such as mass-balance between species (Pauly et al.
2000). The theory around fisheries management has
given minimal regard to even the ecosystem level of
standard ecological theory and is generally based on
single species exploitation (for instance Hilborn &
Hilborn 2012). This lack of theoretical treatment is
challenged by the fact that pelagic marine species
can directly influence their physical environment,
especially through turbulence as mentioned above,
and through nutrient retention (Smetacek 2008). The
scale of niche constructor influences may be pro-
found (Laland et al. 1999), and it is certainly not out-
side the range of possibilities that marine animals
may impact their physical environment or even the
climate (Kerr 2006).
Mutualism and niche construction
Mutualism is important (Hay et al. 2004). Odum
(1969) summarised the importance of parasitism, pre-
dation, commensalism, mutualism and other forms of
symbiosis as a defining character of mature ecosys-
tems. It is not clear if the relationship between
whales and krill is simply mutualistic, in which both
cooperate to gain net survival or net reproductive
benefit, or is better described as manipulative (Krebs
& Davies 1993) or exploitative mutualism in a similar
way to which human farming exploits other species
(Smetacek 2008). For instance whales may influence
krill to a mode of behaviour which, on average,
increases the abundance of krill, whereas individual
krill may increase their reproductive success by
alternative strategies in the absence of whales (Willis
2007). Therefore niche construction is perhaps a bet-
ter fundamental concept with which to start than
mutualism because it encompasses both the wide
range of possibilities in which 2 species interact with
each other, as well as the range of ways they interact
with other species and their physical environment.
Indeed, the way in which krill and their prey interact
is also likely to be just as complex (Kawaguchi &
54
Willis: Whales and krill
Takahashi 1996). While mutualisms are fundamental
to all ecosystems, the dynamic balances between
exploitation, cooperation, benefits and disadvan-
tages are often very challenging to understand from
an evolutionary perspective (reviewed in Herre et al.
1999).
The model in this study
I have employed a krill life-history model which
was developed to target the behavioural choices (i.e.
between deep and shallow habitats) made by krill
under different predation pressures (Alonzo & Man-
gel 2001). This model was previously used for a simi-
lar task (Willis 2007) but here is extended to target
ecosystem-level changes. The environmental cycles
are based on Laws (1977) depiction of the annual
cycles of krill, whales and seals. I have used Chris-
tensen’s (2006) whale population estimates to model
changes caused by human exploitation. The object of
this study was to capture the basic patterns involved
in the ecosystem throughout the period of whaling to
provide a rational explanation for the apparent
changes in krill abundance that is consistent with (or
at least not falsified by) contemporary reports, con-
temporary data, and generally accepted ecological
theory. I have chosen to use an individual-based
modelling (IBM) approach, because these models are
usually simpler to understand than population mod-
els based on differential equations (a comprehensive
introduction to, and review of, IBMs in ecology is in
DeAngelis & Gross 2009).
METHODS
Design of the model
The fundamental driver of the model is individual
lifetime reproductive success of krill in a simple
model based on the balance of risks and availability
of nutrients. The model is calibrated to 2 overall sce-
narios: (1) whales present, and (2) whales absent.
There are 2 krill habitat options: (1) surface all the
time, and (2) surface at night, deep in day (standard
diurnal vertical migration [DVM]). At each time step,
krill feed, grow and develop fecundity and are sub-
ject to predation risk, and may die (Fig. 1). The
fecundity and growth of each krill using various
zones (surface or deep) is calculated using a set of
physiological parameters derived from earlier scien-
tific studies (summarised in Willis 2007 and also pre-
sented in Alonzo & Mangel 2001). Variables such as
feeding rate, length of day, temperature at depth,
and energy expenditure when moving between sur-
face and deep, are used to calculate weekly growth
and allocation of fecundity (Fig. 1). The equations
used in this model for these calculations are sum-
marised in Table 1 of Willis (2007). The changes in
habitat use throughout the year are predefined based
on Laws’s cartoon of the annual cycles of the Antarc-
tic ecosystem (Laws 1977) and daylight proportion is
calculated on a weekly basis using a standard math-
ematical model of sun elevation (NOAA Sunrise/
Sunset calculator, www.esrl.noaa.gov/) (Fig. 2). Each
year, in the annual loop (Fig. 1), the total population
fecundity is used to determine the number of new
recruits to the model. As the model progresses, the
abundance of krill changes (determined by risks and
population fecundity) along with the overall biomass
of krill, and the ‘production’ which is the sum of bio-
mass of krill which die at each time step and are
assumed to be eaten be predators (Fig. 1). The sum of
all surviving individual fecundity in the population
each year determines the number of new entrants
recruited into the population. Fecundity used in this
spawning process is therefore a direct measure of
individual lifetime reproductive success. Thus the
55
Feed
Grow
Calculate
fecundity
Survive
Die
Record as production
Assign
habitat
Weekly
loop
Krill Yearly
loop
Calculate total
population
fecundity
Risk
Reward
Add more
krill objects
Record individual lifetime
reproductive success (used fecundity)
Depends on
time of year
and whales
Aggregate all survivors
Record
length
and mass
Record used
fecundity in
survivors
Fig. 1. Flowchart of steps (functions in programming terms)
that occur during the running of the model in this study.
Model krill (n = 1000) are introduced and every model week
they feed, grow and allocate fecundity. Each week they are
also subject to risk of predation, and some die. Those that do
not die contribute to the population additions in the annual
loop. All the fecundity in the population is added together
and used to determine how many new recruits appear.
When a model krill dies, it is recorded how much fecundity
it actually used in procreation. This is its individual lifetime
reproductive success, which is the driver for evolution
through natural selection
Mar Ecol Prog Ser 513: 51– 69, 2014
model aims to explore the balance between risk,
growth, and fecundity in various habitat selections
under various risk assumptions, all in the context of
what is known about krill physiology. The model is
written in a MATLAB ® (Mathworks) or GNU Octave
compatible script and is available by contacting the
corresponding author.
The model can be represented in a flowchart
(Fig. 1) and as pseudocode, as follows:
1. Initialise krill
2. Burn-in krill population using a calibrated steady-
state population (50 yr)
3. Initialise habitat parameters including presence of
whales
4. Loop through 5200 steps (52 wk in each of 100 yr)
a. Update environmental parameters based on time
of year
b. Update whale presence parameters based on year
c. Krill behave (choose habitat selections for week,
based on presence of whales)
d. Krill feed (dependent on habitat use)
e. Krill grow (dependent on existing size and food
gained)
f. If week is spawning week (Fig.2)
i. Calculate fecundity
ii. Record used fecundity for each individual
(lifetime reproductive success)
g. If week is recruitment week (Fig. 2)
i. Recruit new krill (based on total fecundity
the year before)
h. Krill die (of predation dependent on habitat
risks)—record details
5. End loop
6. Output results
Use of the model
The model is used here in 2 ways; (1) the risk bal-
ance of the habitats is calibrated to produce a result
56
02 Aug 21 Sep 10 Nov 30 Dec 18 Feb 09 Apr 29 May
0
0.2
0.4
0.6
0.8
1
A
Baleen whales
Crabeater s eals
Pack i ce
Krill zone
Krill uncovered
02 Au
g
21 Sep 10 Nov 30 Dec 18 Feb 09 Apr 29 May
0
0.2
0.4
0.6
0.8
1Recruit
Proportion of time Relative proportion
Spawn
B
Sunlight lat. –58° to –65°
Sunlight lat. –51°
Summer season
Fig. 2. Annual cycles in the Antarctic ecosystem used in the model. (A) lines are derived from Laws’ (1977) cartoon of the eco-
system and show (proportionally) when the pack ice recedes and when whales (red line) and krill (grey solid line) feeding
activity occurs. The pack ice cover as a proportion of maximum in any one year (grey dot-dash line) subtracted from the total
potential area for krill (dotted grey line) is used to estimate the changes to the area in which krill become available throughout
the year (solid grey line). Green line is crabeater seal feeding as a proportion of maximum feeding from Laws (1977). (B) The
proportional length of daylight is shown for various latitudes (the latitude indicated by the central blue dashed line was used
in this study). The summer season was defined in this model as when the krill activity was above 1.15, and weekly time for
recruitment and spawning (calculation of used fecundity) are shown as vertical dashed lines. Note that the summer season for
whales is characterised by an almost continual decline in daylight. This suggests that the starting time for whale feeding is
critical to their success, as they are thought to be visual predators
Willis: Whales and krill
that is consistent with what is known about the
behaviour of krill pre- and post-whaling, and (2) the
calibrated model is used to estimate the changes in
the abundance of krill before, during and after whal-
ing. Since risk is not associated with any physiologi-
cal characteristic it is a free variable in the model and
any value is equally plausible, in itself, in any habitat.
The same values as in Willis (2007) were used for rel-
ative food availability, but now the enhanced feeding
availability in the surface layers is only represented
in the scenario with whales present. All other values
are similar, representing no difference in food avail-
ability between scenarios and habitats and thus no
implied knowledge. Therefore the 4 risk variables
(summer surface and summer deep in each scenario
[whales and no whales]) are the tuning variables of
the model.
Initialisation and model running
The model time step is 1 wk and the usual run
length is 150 yr (Fig. 1). The model was initialised
with 1000 model krill. The physiological parameters
of krill and their life history, including the energy
assimilation from food, and the energetic require-
ments of growth and fecundity at various tempera-
tures, under various lighting conditions, and under
different travelling regimes are all held constant and
are all similar to earlier studies based on field and lab
observations (Alonzo & Mangel 2001). Reward is a
proportional instantaneous feeding rate between 0
and 1 which is used to define the comparative feed-
ing opportunity for krill between locations. Risk is a
survival probability between 0 and 1. Krill can die
each week; those that die are recorded as ‘produc-
tion’ and thus all are assumed to have died through
predation. New krill are added (as age 1 juvenile
recruits) at an annual interval and the number of new
recruits is based on a linear function of population
fecundity. So the number of krill in the model
changes throughout the model run. The summer sea-
son is defined as when Laws (1977) (normalised) car-
toon of krill activity is greater than 0.15 and spawn-
ing and recruitment weeks were likewise chosen to
be approximately at the appropriate time of year
(Fig. 2).
Calibration
There are 2 scenarios on which the model can be
calibrated; (1) pre-whaling and (2) post-whaling.
The assumed characteristics of these states are as
follows:
Pre-whaling state
• Roughly stable populations of krill and whales
• Population abundances close to the system carrying
capacity
• Krill abundance adequate to support pristine whale
population and populations of seals, flying birds,
penguins, squid and fish (approx. 700 million t).
• Krill are selected to use the surface zone in day-
light, therefore individual lifetime reproductive
success (termed ‘used fecundity’ in the model) is
expected to be higher for surface users.
Post-whaling state
• Stability is not assumed
• Population biomass of krill was around 150 million t
at some point
• Krill are selected to use the deep zone in daylight,
therefore ‘used fecundity’ higher for deep users in
daylight.
Operation
In the operation phase the model is run for 150 yr.
The first 50 yr are a ‘burn in’ period designed to allow
the model to stabilise. Since the model employs a
feedback mechanism to maintain a population of krill
(based on fecundity as described above) it is ini-
tialised with an estimated value of total population
fecundity. As the model progresses past the first few
years the influence of the initial estimated fecundity
is eliminated and the model reaches an independent
steady state dependent only on the risk parameters.
The operation stage of the model then progresses
from Year 50 to Year 150 and is designed to model
the impact of commercial whaling between 1900 and
2000. The abundance of whales in each year is calcu-
lated as a proportion that ranges between 1 (pristine)
and 0 (complete extinction) and is based on the total
biomass of the 5 main species of whale in this study.
The proportion is applied to the krill population to
determine if a krill is in an environment with or with-
out whales for the summer season. For example, if
there are 1200 krill at the start of a year, and, that
year, whale population biomass is at 50% of pristine,
on average (actual allocation is determined using a
57
Mar Ecol Prog Ser 513: 51– 69, 2014
random number generator) 600 krill will be in a
whale-free environment and 600 in an environment
with whales. In this concept a single krill represents
an isolated group of krill that either encounters or
doesn’t encounter whales during the year in ques-
tion. Those that encounter whales use the surface
habitat, and those that do not use the deep habitat
(surface at night, deep in day; DVM).
Abundance of whales
The abundances of whales during the period of
1900 to 2000 in the Southern Ocean were taken from
Christensen’s (2006) graphs of population abundance
(Fig. 3). The average adult mass of each species was
used to calculate the population biomass of each spe-
cies. The sum of species biomasses was used as the
estimate of the total population biomass used to
determine the proportional abundance of whales
throughout the period in the model (Fig. 3).
RESULTS
Calibration
The results of the calibration are
summarised in Tables 1 & 2, and
Figs. 4 & 5.
The key result in the calibration is
in the ‘used cumulative fecundity’
(lifetime reproductive success)
(Fig. 5A). Table 1 shows how in the
scenario with whales present, the
used lifetime fecundity was signifi-
cantly higher in the surface habitat
choice, which would suggest that
krill would be evolved to use this
habitat when they can detect whales.
With no whales the position is
reversed; the used fecundity was sig-
nificantly higher for the deep habitat,
suggesting that krill are evolved to
use the deep habitat when whales
are not detected. Table 2, giving
instantaneous risk (survival rates),
also shows this same pattern and
reversal, which is logically consistent
but not necessary for a plausible cali-
bration. That is; it is plausible that a
krill would trade a higher risk for
greater lifetime reproductive success,
but in this parameterisation that is
unnecessary. It is logically consistent
that the calibrated risk levels are also higher in both
zones when whales were present.
The weekly production in the surface habitat
when whales were around was 4 to 8 times as
much as the production in the deep habitat when
whales were not present (Fig. 4C), and the total
population biomass shows a similar pattern (Fig.
4B). This pattern is dimensionally similar to the
recorded changes in the real ecosystem discussed
in the introduction, where 1 g of model production
is equivalent to 1 million t of real krill. Individual
model krill growth shows the periodic shrinkage
which is thought to happen during the winter
period (Alonzo & Mangel 2001) (Fig. 4D).
Operation
The results of the operation of the model are sum-
marised in Fig. 6. The model krill abundance in -
creased over the initial 50 yr as the whale abundance
58
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
0
2
4
6
8x 105
AFin
Blue
Sei
Minke
Humpback
1900
Total population biomass (t) Abundance
1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
0
1
2
3
4x 107
B
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
0
2
4
6
8x 107
C
Fin
Blue
Sei
Minke
Humpback
Fig. 3. Abundance, biomass and total population biomass of the 5 species of
whales in this study in the Southern Ocean from 1900 to 2000 from Chris-
tensen (2006). The whales are: blue Balaenoptera musculus, fin B. physalus,
sei B. borealis, humpback Megaptera novaengliae, and Antarctic minke B.
bonaerensis. The abundance values from (A) Christensen (2006) have been
multiplied by the average whale mass from the same study to calculate (B)
species biomass, and these are summed to produce (C) the total whale bio-
mass. Total whale biomass as a proportion of the maximum is used in this
study to calculate the number of krill using each habitat selection in the model
Willis: Whales and krill 59
Whales No whales
Risks (annual survival rates)
Surface daylight 0.9534 0.96
Deep daylight 0.94 0.974
Night time 0.99 0.99
Under ice (winter) 0.995 0.995
Feeding opportunity (day and night time, proportion of
maximum)
Surface 1 0.7
Deep 0.7 0.7
Under ice (winter) 0.5 0.5
Table 2. Model krill risk and reward values for surface and
deep zones in calibration scenarios. Values in bold are those
that impact the comparative results between the scenarios.
Risk is expressed in annual survival proportion; thus a high
value indicates a lower risk
Whales No whales
Krill abundance (from an initial population of 1000)
Surface habitat 925 ± 164 1 ± 1 (collapse)
Deep habitat 1 ± 1 (collapse) 447 ± 126
Weekly production (g)
Surface habitat 7.1 ± 1.2 0
Deep habitat 0 1.2 ± 0.4
Lifetime used fecundity of a group of 300 model krill
tracked from 50 yr until death (arbitrary units)
Surface habitat 874 ± 271 389 ± 123
Deep habitat 400 ± 60 792 ± 54
Table 1. Model krill population parameters after 50 yr in the
calibration scenario, using the 2 habitat selections (surface
[all the time] and deep [surface at night, deep in day; i.e.
diurnal vertical migration] during the summer season) for
the 2 scenarios (whales and no whales), mean values ±1 SD
of 5 similar runs of 1000 model krill
0
A
10 20 30 40 50 60 70 80 90 100
0
500
1000
1500
2000
No. of yr
No. of yr
No. of yr
Total model krill abundance
Deep whales
Surf whales
Deep no whales
Surf no whales
010 20 30 40 50 60 70 80 90 100
0
100
200
300
400
500
600
Total krill population bioass (g)
010 20 30 40 50 60 70 80 90 100
0
2
4
6
8
10
12
Total krill production wk–1 (g)
1 2 3 4 5 6 7 8 9 10 11
10
20
30
40
50
60
70
A
g
e (yr )
Length-at-age (mm)
B
CD
Fig. 4. Results from the model in the calibration phase, showing the various implications of the risk and reward balances in the
4 habitat scenarios through 100 yr on model run time. ‘Deep’ means krill are at the surface in the nighttime and deep in day;
‘surf’ means krill are at the surface day and night. (A) Mean (± SE) total krill abundance of model krill after 5 model runs, show-
ing how abundance decreases for all scenarios except surface when whales were around. Mean ± SE (B) krill population bio-
mass and (C) production showing similar patterns. (D) Average length of a batch of 300 model krill after 50 yr. The lines are
truncated at different points as no krill lived past these ages. Krill are thought to shrink during periods of low food, and this
effect is also shown
Mar Ecol Prog Ser 513: 51– 69, 2014
60
Deep w Surf w Deep noW Surf noW
200
A
400
600
800
1000
1200
Used cumulative fecundity (sub-group, n = 300)
Arbitrary fec undity units
Habitat choices
Deep w Surf w Deep noW Surf noW
2.25
2.5
2.75
Years
Habitat choices
Age at deat h of group of 300
B
Fig. 5. Differences in (A) used cumulative fecundity (a direct measure of individual lifetime reproductive success) and (B) age
at death of a sub-group that entered the krill population after 50 yr in the calibration phase, based on the presence/absence of
whales in the model. The vertical lines indicate ±1 SD of 5 replicates. (A) When whales were around (deep w and surf w), the
used fecundity was significantly higher for the surface habitat, and that the position was reversed when whales were not pres-
ent (deep noW, surf noW). (B) Krill live longer on average in the deep habitat when whales were not present. Hence, ‘deep
noW’ is a live-long grow-slow habitat in contrast to ‘surf w’, which was a grow-fast die-young environment. This is the key
result of the calibration of the model. Krill evolved to go deep when whales were not present but to stay at the surface when
they were present
1900 1920 1940 1960 1980 2000
400
A
600
800
1000
1200
1400
1600
1800
Total population abundance
1900 1920 1940 1960 1980 2000
0
100
200
300
400
500
600
Total population biomass
(g)
Krill
Whales
1900 1920 1940 1960 1980 2000
1
2
3
4
5
6
7
8
9
Weekly production mass (g)
1900 1920 1940 1960 1980 2000
0
1
2
3
4
5
6
7
Weekly production mass (g)
Surface
Deep
Ice
B
CD
Fig. 6. The krill model in this study in relation to the proportion of whales present (in terms of biomass) throughout the period
from 1900 to 2000. The whale biomass, and therefore implied presence, is external information which drives the model results.
Means ±1 SD of 5 replicates are shown. (A) The total abundance of krill increases as the decline in whales begins. (B) The pop-
ulation biomass of krill remains reasonably stable during this period as more krill change habitat, which is similar to (C) pro-
duction. Average weekly production is shown (7 g would relate to ~371 g yr−1). By 1960, however, the abundance, biomass and
production of krill all diminish in the same pattern as the abundance of whales. (D) Krill production according to habitat, show-
ing how krill production at the surface follows a pattern very similar to the whale biomass, as krill are leaving the surface habi-
tat in that same proportion (red line). The krill in the deep habitat builds up but begins to decline after 1970. There is no evi-
dence from the real ecosystem which contradicts this pattern. Production during the winter when krill are under the pack ice
is shown in green
Willis: Whales and krill
fell. This is caused by the longer life of krill in the
deep habitat (Fig. 5B), which causes a very slightly
higher percentage of additional older krill to begin
each year, and thus increase the population fecun-
dity as spawning happens soon after the start of the
new season (Fig. 2). This is supported by the rela-
tively unchanged krill population biomass during
this period (1900–1950, Fig. 6B). This effect was bal-
anced and reversed after 50 yr as the whale popula-
tion became so low as to put more krill into the deep
habitat, and the pattern of population decrease in the
habitat became dominant. Overall production was
un changed during the initial period, which implies
that there was more food available during this period
for other surface predators of krill (as whales’ re -
quirements diminished). Both whales and non-whale
predators are assumed to make a numerical or be -
havioural response to the additional resource and so
maintain the risks as constant. The production switched
from availability in the surface zone to the deep zone
after ca. 50 yr and finally ends up predominantly in
the deep zone (Fig. 6C). Krill production in the ice
remained relatively constant throughout, only slightly
diminishing toward the end of the period (Fig. 6D). If
under-ice production of krill is the determinant for
the crabeater seal population (which are assumed to
be a large proportional predator on krill in the win-
ter) then it would explain why their population abun-
dance has remained relatively un changed, until a
slow decline started ca. 1960.
DISCUSSION
The calibrated model shows that there are plausi-
ble balances of risk that produce results consistent
with what is known about the ecosystem and krill
behaviour before and after whaling. The calibrated
model linked to the decrease in whales produces a
reasonably consistent pattern of decreased krill
abundance and biomass that has been observed in
the actual system. This includes the decrease in sur-
face production (and thus visibility of krill) matched
with the decrease in whale abundance (Fig. 6).
Therefore it is reasonable to assume that whales did
maintain a high abundance of krill, until any of the
main assumptions of this study are falsified.
The mutualism between whales and krill (ex -
pressed in terms of population abundance) is clearly
shown by the model as the reduction in whales leads
to a delayed response in the abundance of krill
(Fig. 6A,B). The individual lifetime reproductive suc-
cess of the model krill would be a plausible cause for
their likely habitat selection behaviour (Fig. 5). The
instantaneous risk (Table 2) would also drive this pat-
tern. This type of lagged response is typical of niche
construction theory (Laland et al. 1999). The lagged
response is an emergent property of the model and is
caused by multiple factors, including simple preda-
tory release, mixed life histories (krill developing
quickly in the surface layers and subsequent years
living in the less risky habitat) and the rate of decline
of population abundance in the deep habitat. The
decrease in population biomass of krill lags that of
whales by around 20 to 40 yr (Fig. 6), which is 5 to
10× the average lifespan of krill (Fig. 5) which is
indicative of the concept of niche construction pro-
viding legacy impacts to future populations beyond
the lifetime of a single animal (Odling-Smee et al.
2013).
It seems counter-intuitive that whales evolved to
maintain a high abundance of krill. The al ternative is
even more counter-intuitive. The broad implication
of the importance of niche construction for mature
ecosystems are often mirrored elsewhere (Laland et
al. 1999, Odling-Smee et al. 2013). Regularly remov-
ing large quantities of what is essentially 99.9%
grassy biomass from a domestic lawn to maintain a
grass-based ecosystem is a direct analogue to the
fundamental mechanism of this study. Grass and her-
bivores evolved in lowland ecosystems, and removal
of large grass-eating herbivores leads to less grass,
rather than more grass (Vera 2000). In exactly the
same way, niche construction theory suggests that
the removal of whales from the ecosystem to which
that had evolved was bound eventually to lead to less
krill, their only prey, rather than more, and any other
outcome would be illogical and unexpected.
The predictive potential of the theory presented
here, based on niche construction, may be no better
than the previous, least daring, theory. For example,
expectations of an experiment involving the reintro-
duction of wolves in Yellowstone National Park
(Beschta & Ripple 2013) were not met. The ecosystem
did not return to its previous state after the reversal of
the removal of wolves. Irreversibility such as this is
called hysteresis, a common ecosystem phenomenon
(Mangel 2006). Species that were thought unimpor-
tant may in fact have a pivotal role; others held
footholds (constructed niches) that they had previ-
ously been denied. For example, krill preferentially
eat salps, which leads to potentially complex link-
ages between krill and their prey (Kawaguchi &
Takahashi 1996). Salps and copepods both are filter
feeders, like krill, and both have a far higher overall
biomass in the Southern Ocean than krill (Voronina
61
Mar Ecol Prog Ser 513: 51– 69, 2014
62
1998). Salps appear to have increased in abundance
while krill have decreased (Atkinson et al. 2004).
Prey at one life stage may be a competitor at another
stage, potentially leading to a further level of niche
construction which is very common in aquatic eco-
systems (Hay et al. 2004). Furthermore, large mobile
species which have been either invisible or at appar-
ently low abundances may play a pivotal role; espe-
cially fish and have cephalopods which predate krill
using the deep habitat and which have the potential
for a rapid numerical response to environmental per-
turbation (Ainley et al. 2010).
Objection to Godlewska’s hypothesis
The most compelling argument against Godlew -
ska’s hypothesis (i.e. that krill have changed their
behaviour after the end of commercial whaling) is
that it is hard to explain in terms of evolution by
natural selection, particularly in light of how quickly
the change is supposed to have occurred (Smetacek
2008). In response to this argument it is worth exam-
ining behaviours of closely related species, because
if they have been shown to exhibit similar behaviour
to that proposed for krill, it suggests that such behav-
iour is plausible for krill. In fact, there are many
examples of Crustacea changing their daily vertical
migrations in response to predatory threat. A partic-
ularly convincing example of this behavior is the
variations in patterns of DVM in the water flea Daph-
nia magna when the concentration of fish odour
increases or decreases (Loose & Dawidowicz 1994).
High concentrations of fish odour in the environment
cause an instantaneous switch to more intense DVM
and less exposure to light (Loose & Dawidowicz 1994,
Giske et al. 1998). There are many other examples of
marine and freshwater Crustacea changing DVM in
response to the perception of threat from visual pred-
ators (e.g. Stich & Lampert 1981). These similar
examples have been derived from field and labora-
tory observations and are cued through olfaction,
mechanical stimulation or vision (Stich & Lampert
1981, Gliwicz 1986, Bollens & Frost 1991, Fiksen &
Giske 1995, Fiksen 1997, Fiksen & Carlotti 1998).
Mackintosh (1934), Marr (1964) and Hardy & Gun-
ther (1936) all independently concluded that most
krill stayed at the surface but a minority exhibited
DVM. So, an instant switch in DVM due to avoid-
ance of whales as visual predators, cued by odour,
mechanical stimulus or vision, is not only plausible; it
is a very strong candidate for an explanation of DVM
of krill. Arsenev (1958) explains how whales and krill
are most often observed together at the surface, but
that deviations from this regularly occur (whales
without krill and vice versa). Arsenev (1958) explains
that it appeared that such phenomena (deviations)
are only observed in the course of a single day; on the
following day whales appeared where there was only
krill before, and vice versa. Arsenev’s (1958) obser-
vations suggest that krill will switch DVM locally
within one daily migratory cycle in response to the
threat of predation by whales, but also that whales do
not have to be physically present to invoke the
response. A similar rapid switch of DVM has been
shown experimentally for the marine copepod Acar-
tia hudsonica, again only due to the credible percep-
tion of predatory pressure rather than the presence of
actual predators (Bollens & Frost 1991). Added to
these examples of closely related species, theoretical
considerations of krill physiology and habitat selec-
tion support the plausibility of Godlewska’s hypothe-
sis (Alonzo & Mangel 2001). Although Godlewska
(1996) considered the behavioural change to be
potentially an evolutionary change, that is not a nec-
essary requirement of the hypothesis. The change
was most likely due to behavioural plasticity of krill,
however, the developing theory around rapid niche
evolution would suggest either is a possibility (Kozak
& Wiens 2010). So, in conclusion, it is likely that all
krill regularly switched between both behavioural
states in the pristine ecosystem. When the abun-
dance of whales was high, the majority of krill
remained at the surface day and night, but all krill
retained the ability to switch daylight habitat based
on the immediate perception of predatory threat
from whales. This study examines what this conclu-
sion implies for the changes to abundance in krill as
a result of commercial whaling.
Behavioural trigger
The linkage between odour and behaviour has not
been shown in krill but there is some evidence that
this is a plausible hypothesis (which is worth testing).
Dogs (Canis canis) can be trained to pilot boats to lo-
cate whale faeces from several kilometres away using
smell (Rolland et al. 2006). Whale faeces are buoyant
and remain at the surface for days (D. Ainley pers.
comm.). In turn, krill are sensitive to olfactory cues.
The highest krill olfactory sensitivity was shown to
be to Newcastle Brown Ale, where experimental sub-
jects needed to be prised off the pipette used to intro-
duce the substances into the test apparatus (Hamner
& Hamner 2000). Perhaps not coincidently Newcastle
Willis: Whales and krill
Brown Ale, which is a dark bottled beer, has a rela-
tively high concentration of soluble iron (Sancho et al.
2011). Whales in the Arctic have been observed pref-
erentially attacking krill at depth in daylight (Laidre
et al. 2010) which supports the concept that deep in
the day was higher risk for krill when whales were
around, but which also suggests that whales are used
to finding krill at depth as well as at the surface and
adds credibility to the theory that krill may have
evolved this behaviour as a response to the presence
of whales.
Alternative explanation:
a fisheries management approach
Mori & Butterworth (2004, 2006) attempted to
explain the changes in krill abundance due to com-
mercial whaling by fitting a fisheries management
model to whale catch data. This is a prime example of
standard ecological theory which is has been con-
trasted unfavourably to niche construction theory in
other examples (Laland et al. 1999). Their model led
them to hypothesize that a krill surplus did happen
due to release of predatory pressure, and that krill
abundance peaked in 1950, well before the end of
whaling (which ended around 1970), and has de -
creased since then. Mori & Butterworth (2006) con-
cluded that krill biomass under unexploited coexis-
tence with whales was around 150 million t, which
then gradually increased to about 700 million t dur-
ing the first half of the 20th century, after which it
declined again to around 200 to 300 million t around
2006. The main challenge in the Mori & Butterworth
(2006) model was to explain how the high abundance
of krill caused by the predator release from whaling
was subsequently reduced through alternative pred-
ators. There are similarities in the suggested abun-
dance pattern of krill in the Mori & Butterworth
(2006) model and the model presented here (Fig. 4),
the principle difference between the studies lies in
the explanation for the decline following the peak
abundance. The Mori & Butterworth (2006) model
fails to provide an explanation for the observed
changes in krill abundance, for 2 principal reasons;
(1) the main hypothesis has been falsified, and (2) the
initialisation is implausible. These points are ex -
plained in more detail in the paragraphs following.
The Mori & Butterworth (2006) model was conven-
ient for those wanting to support a fishery for krill or
minke whales; it suggested that the abundance of
both are higher now than they were in the pristine
state.
Contrary evidence to the conventional model
Mori & Butterworth (2006) proposed that minke
whales and crabeater seals increased in abundance
as a result of increased krill availability between
1920 and 1950. This then is supposed to have
caused compensatory predatory pressure which led
to a decrease in krill abundance since 1950, even as
whale abundance continued to be reduced through
whaling until 1970. They confirm that their model
would not fit with whales alone and so crabeater
seals are employed as the additional factor which
drives the krill abundance down after 1950. In order
for this to be a partially plausible scenario, the pop-
ulation abundance of minke whales would have
needed to have risen very quickly before the 1950’s
from a previously low level (they suggest from
300 000 to 1 200 000) and crabeater seal abundance
would also have needed to grow from very low
(~1 000 000) to high abundance (~20 000 000) during
the last 20 yr of commercial whaling. In addition to
no other evidence to suggest minke whales in -
creased in abundance, these hypotheses have been
falsified (1) through genetic studies of the minke
whale population (Ruegg et al. 2010), and (2) by
contemporary reports. For instance, crabeater seals,
sometimes in high abundance, were reported by
every major expedition that passed through the
pack-ice in the southern summer before 1962 or
stayed over winter in the pack ice, as a Belgian
expedition did in 1898; estimated abundances were
consistently be tween 2 and 5 million (Marr 1964).
There was no mention of a 20-fold increase, and a
44 yr time series in the Antarctica Peninsula region
suggests no long-term trend in crabeater abundance
after 1950 (Ballance et al. 2006). In comparison, the
model presented in this study suggests that krill
abundance, available to predators in the deep habi-
tat, was relatively high during the 20–40 yr around
the end of commercial whaling (1940–1980) but
then decreased (Fig. 6D). This may correspond to an
increase in crabeater seals (which is difficult to
quantify), and the population patterns of fur seals
(Ballance et al. 2006) and gentoo penguins (Emslie
et al. 2013).
Initialisation of the conventional model
The other difficulty with Mori & Butterworth’s
(2006) model was that it was based on implausible
initial assumptions about the pristine state of the
system. In short, they assumed too few whales eat-
63
Mar Ecol Prog Ser 513: 51– 69, 2014
64
ing too small an amount of krill. This then allowed
for the model to fit a massive proportional increase
in krill abundance to levels consistent with con-
temporary reports in the 1970’s. Their estimate of
the krill eaten by the whales which were killed
during whaling was roughly 50 million t as op -
posed to the commonly quoted estimate of 147 mil-
lion t (Laws 1977). The Mori & Butterworth (2006)
estimate was an implausibly low number; whales
alone weighed roughly 60 to 70 million t (see ear-
lier analysis of Christensen 2006). Mori & Butter-
worth (2006) excluded 50% of fin whales and all
the sei whales (although, for instance, Harrison
Matthews (1938) suggested sei whales only ate
krill). There is no reason, or precedent, for a sepa-
rate fin whale population, rather the potential sep-
aration of feeding locations is more consistent with
the theory of mature ecosystems and niche spe-
cialisation (Odum 1969). Their fitted model esti-
mated the pristine abundance of blue and fin
whales (187 000 and 235 000 respectively) to be
much less than estimates in other studies; an alter-
native study based on a reasonable life history
model estimated 327 000 and 625 000 respectively
(Christensen 2006). Laws (1977) had suggested
200 000 and 400 000, without the benefit of Russian
catch reports that were then not available. It is
likely the pristine abundances of these species
were more than these highest estimates, as has
been suggested through genetic studies of other
similarly exploited whales (Ruegg et al. 2013). The
Mori & Butterworth (2006) model further relied on
whales in the pristine state being continually
starved through ‘overfishing’ krill. This scenario
would simply have led to smaller or fewer whales
over time in either a behavioural or function
response to the limitation of food (Holling 1959).
The concept of whales over-harvesting the krill
below the maximum sustainable yield at a fraction
of the carrying capacity (~30%) is an anthropo-
morphic concept contrary to ecological theory.
Ecological theory would suggest that the amount
of standing crop biomass supported by the avail-
able energy flow is expected to increase to a max-
imum in the mature or climax stages of an ecosys-
tem (Odum 1969). The idea that whales were
‘half-starved’ as a result of overfishing the krill
before the start of whaling is also falsified by con-
temporary reports based on stomach contents.
Whales examined during the early whaling period
usually had stomachs full of krill (Hardy &
Gunther 1936). Mackintosh & Wheeler (1929)
examined 519 blue and fin whales and all but 68
had full stomachs of krill. Ruud (1932) found that
of 300 whales examined in factory ships, only 2
immature whales had empty stomachs, and all oth-
ers had krill in large quantities. Mori & Butterworth
(2006) make several other assumptions which are
inconsistent with ecosystem ecological theory. As
mentioned above, they suggest that there was a
high level of interspecific competition which caused
a suppression of the abundance of minke whales
and crabeater seals in the pristine state; apart from
the fact there is no evidence for this, and no expla-
nation for how it might have happened (baleen
whales do not attack each other or seals and vice
versa), it is inconsistent with the characteristics of
mature ecosystems which have high levels of
niche specialisation and mutualism (Odum 1969).
The only example they give to suggest this might
be plausible involves 2 species of odontocetes
fighting over fish hooked on a longline, again an
unnatural situation, involving whales which are
very different from the mysticetes discussed else-
where in this study. On the other hand, ecosystem
theory is again borne out through detailed reports
of separation between minke, fin, sei and blue
whales feeding times and/or places (Kemp & Ben-
nett 1932, Marr 1956, Laws 1977, Santora et al.
2010). Mori & Butterworth (2006) also suggest a
relatively high range of natural mortality for the
larger whales, allowing their model to fit to large
whales with a low average natural lifespan and
thus quicker population turnover. They modelled
large whales, some of which were the largest ani-
mals that have ever existed, as if they were
smaller animals (bigger animals live longer than
smaller ones; Speakman 2005). For instance, they
used a limit of 0.03 to 0.06 in annual mortality for
blue whales (Laws [1997] had estimated 0.027 to
0.033). A range of 0.01 to 0.03 or lower would have
been more consistent with the facts (Branch et al.
2004, Ramp et al. 2006), our uncertainty, and the
theory of longevity of apex predators in mature
ecosystems (Odum 1969).
Other explanations: climate change
Flores et al. (2012) suggested that the long-term
decline in krill abundance was the result of climate
change through the reduction of sea ice and other
environmental changes. Sea ice extent has been
shown to be a factor in krill recruitment (Kawaguchi
& Satake 1994). Later studies showed that sea ice had
not reduced as Flores et al. (2012) had thought, and
Willis: Whales and krill
rather that it had an inverse relationship to sea tem-
perature (Shu et al. 2012). Therefore sea ice reduc-
tion is not a plausible explanation for the long-term
decline in krill abundance. In general, climate change
cannot yet be implicated as the major cause of krill
abundance changes since the end of whaling (Trathan
et al. 2012); the present study does not falsify any
hypotheses derived from that theory.
Hypotheses generated by this study
The model presented here, or more correctly the
modelling process, helps develop hypotheses that
would be valuable to test (Table 3). The model sug-
gests that krill respond to the threat of whales. The
smell of faeces is a likely cue. Krill can change their
daily migrations within a single cycle. All these
65
Condition Predictions or hypothesis
1. Surplus of krill in the deep habitat from about 1950 to Predators which are able to exploit krill in these areas 2000
(see Fig. 6). are likely to have had higher abundance during this period,
for instance: crabeater seals, fur seals, fish, squid, some pen-
guins and other species (Ballance et al. 2006).
2. Krill population abundance has decreased Krill population abundance may continue to decrease or be
since ca. 1970. approaching a low asymptote (Atkinson et al. 2004).
3. Non-whale surface predators may have benefited from Previous abundances of surface krill predators may provide
a cooperative relationship with whales. indication of previous abundance patterns of whales.
4. Whales may have developed mutualistic relationships The abundance patterns and numbers of fin, blue, minke, sei,
between species. and similar species may have been mutually interrelated. For
instance, they may have inhabited the same areas in succes
sion throughout the summer season as day length reduced.
Successively smaller feeding requirements could have main-
tained the nutrient retention cycle for surface krill.
5. Whales may have developed mutualistic relationships The abundance patterns of whales are likely to have been
between individuals. dispersed or grouped dependent on the availability of
surface-inhabiting krill. Where krill are at the surface, whales
may develop mutually beneficial fertilisation and strong
aggregation behaviour, whereas whales may disperse to
search for deep inhabiting krill.
6. Krill will react strongly to whale faeces. Krill are attracted to substances rich in soluble iron (Hamner
& Hamner 2000). They should react by changing their
behaviour.
7. Krill on the surface in daylight indicates whales are Observations already confirm this (Arsenev 1958) and more
around, will be around or were around the day before. advanced instruments should further confirm.
8. Abundance patterns of krill will be dependent on the Krill may swim laterally to areas of high concentration of
deposition of whale faeces. whale faeces and thus congregate in patterns that are differ-
ent from passively advected particles. The retention of krill in
the Antarctic Peninsula Plume (Smetacek 2008) may be an
example involving a major ocean feature. It is otherwise un-
known how krill navigate.
9. Krill size will vary less through the annual cycle: they If winter feeding opportunity is lower than summer, then the
may now be smaller and live longer than when whales lack of additional nutrients due to the breakage of the
were abundant. nutrient retention cycle means that krill body size fluctuation
will be less (Fig. 5), however the lowered risk may mean krill
live longer on average.
10. Whales will defecate in patterns (spatially and temp- If an area is saturated (over-fertilised) or unproductive it is
orally) which will serve to preserve the patterns of max- likely a whale will travel to an alternative location to defe-
imum feeding opportunity, and maximum retention of cate. This depends on digestive rate and swimming speed
faeces in the surface layers. and on the whale’s cognitive and sensory appreciation of the
hydrodynamic environment. Top predators target Lagrangian
coherent structures (Tew Kai et al. 2009) which are eddies,
fronts, convergences or diverges which can serve to concen-
trate substances, small animals and plants at the ocean sur-
face. Whales may sense these directly or indirectly.
Table 3. Summary of predictions and hypotheses derived from the modelling and discussed in the text
Mar Ecol Prog Ser 513: 51– 69, 2014
behaviours have been the subject of extensive
research on other closely related species, as outlined
in the introduction, and much further research with
krill in the laboratory and field is warranted. The fer-
tilisation process derived from whale faeces must
take some time, perhaps a few days or weeks, and
there must be a process of dilution and transport
related to water movements and thus weather or
hydrodynamics. There is great opportunity to model
the implications of fertilization and turbulence in bio-
geochemical ecosystem models. Mackintosh (1934)
mentioned the hydrodynamic associations with
whales in convergences, vortices, etc. and assumed
that krill passively aggregated in these areas, but the
situation is potentially more to do with the aggrega-
tion of planktonic prey of krill. There may also be
local feedbacks where a group of whales fertilising
the same area leads to an overall increase in feeding
opportunity for krill, and whales and other predators.
Whales have the choice where to deposit the fertiliser
and this might lead to spatial patterns that are inde-
pendent of physical processes or enhanced by them,
i.e. whales may congregate or disperse to defecate,
or go to certain areas such as persistent eddies or
upwellings in order to make best use of the fertilisa-
tion process. There is much scope for observation
and modelling of whale activity in this respect. The
distribution of krill is associated with ice cover
(Kawaguchi & Satake 1994) and so there may be
other confounding factors.
Improvements to the model
The model is very simple and so there is an almost
endless number of improvements that could be made
if realism is the objective, although this may not be
the best objective for such an idealised model (Oden-
baugh 2005). Firstly, the model includes no limita-
tions on resources and no density-dependent release
of predation at low abundances. This does not impact
the calibrated states, but does limit the model as a
predictor outside these states. For instance, the
model calibration suggests it is plausible that the krill
population has diminished, but cannot be used in any
way to aid prediction about when the decrease will
be terminated through density-dependent effects,
i.e. through a lower risk which should happen at
some low abundance. As for other more minor poten-
tial improvements, there are a number of alternative
options for linking whale abundance to krill en -
counter and thus krill habitat use. Rather than an
annual allocation, the likelihood of presence of
whales for each krill could be re-determined on a
weekly, monthly or other temporal schedule. In this
case, however, it is unknown how long the period
between whale fertilisation and feeding advantage
takes. Presumably it takes some time (at least 1 wk)
and begins a positive feedback mechanism locally
and so may take several weeks or months to build up
to a threshold or maximum. Presumably once it
begins to build up it makes the area more attractive
to other whales and so may reinforce itself for
sequential groups of whales. Kemp & Bennett (1932)
suggested whales of different species (fin and blue)
frequented the same areas in succession. Once the
positive feedback mechanism has started it seems
therefore unlikely that a whale would move on from
a higher feeding opportunity to speculate about feed-
ing potential in other areas (Table 3). However, water
movement, ice, competition and weather patterns
may impact the effect. Or the whales may purposely
fertilise and return on a schedule based around a
number of locations. Since these factors are mostly
unknown it seemed most parsimonious to assume
linkage on an annual basis.
CONCLUSIONS
The mechanisms outlined in this study offer the
potential for development of a krill fishery based on
fertilisation and behavioural control of krill. It must
be possible to distribute an analogue of whale faeces
over small areas and bring krill to the surface,
thereby causing krill abundance and production to
increase in the same way humans farm, e.g. grain
and cows, through a mixture of protection and aggre-
gation with individuals compressed in artificially fer-
tile areas. This would be a more sustainable plan
than competing with recovering whale populations
for a diminishing residual resource and perhaps driv-
ing the krill and whales to extinction.
Krill abundance is less than one quarter of what it
used to be in the pristine ecosystem. Even in the face
of high-quality unequivocal scientific evidence, this
stark fact has been overlooked or obfuscated in the
scientific conversation around the exploitation of
krill. The precautionary approach would be to stop
all krill fishing until the krill and whale population
abundance pattern is clearly understood. Further-
more, the impact of whales on the ecosystem is likely
to be a lot more complex than the one feedback
mechanism in this study (Ainley et al. 2010). At very
least, a representative spatial and temporal protected
area should be implemented in order to avoid any
66
Willis: Whales and krill
further confusion about the state of the ecosystem in
the absence of exploitation. Maintenance of an eco-
logically relevant protected zone would be the mini-
mum indication that a precautionary approach exists
in action rather than solely as rhetoric. The failed
protection of the Ross Sea would have been such a
signal of intent (Blight et al. 2010).
Southern Ocean whale population biomass de -
creased during the 20th century in a strangely
smooth curve (Fig. 3) only interrupted by the Second
World War, demonstrating the complete futility of the
size-based rules (Marr 1956) and any other manage-
ment intervention. Experiences of whalers from the
Arctic moving on to work in the Antarctic meant that
serial depletion of whale populations on a global
scale was clearly common knowledge well before the
peak of whaling in the Southern Ocean (Villiers
1925). Likewise, we should accept our own limita-
tions in the face of very strong evidence. We should
admit that history suggests that a profitable krill fish-
ery, however small it is now, is a warning of the likely
expansion of the fishery until krill are commercially
extinct.
Acknowledgements. I thank the editor, anonymous peer
reviewers and Marc Mangel for suggesting many improve-
ments to this study.
LITERATURE CITED
Ainley DG, Ballard G, Blight LK, Ackley S and others (2010)
Impacts of cetaceans on the structure of Southern Ocean
food webs. Mar Mamm Sci 26: 482−498
Alonzo S, Mangel M (2001) Survival strategies and growth
of krill: avoiding predators in space and time. Mar Ecol
Prog Ser 209: 203−217
Arsenev VA (1958) Distribution of feeding grounds and the
accumulation of baleen whales in the Antarctic. VNIRO
33: 96−100
Atkinson A, Siegel V, Pakhomov E, Rothery P (2004) Long-
term decline in krill stock and increase in salps within
the Southern Ocean. Nature 432: 100−103
Atkinson A, Siegel V, Pakhomov EA, Jessopp MJ, Loeb V
(2009) A re-appraisal of the total biomass and annual
production of Antarctic krill. Deep-Sea Res I 56: 727−740
Ballance L, Pitman RL, Hewitt RP, Siniff DB, Trivelpiece WZ,
Clapham PJ, Brownell RL Jr (2006) The removal of large
whales from the Southern Ocean: evidence for long-term
ecosystem effects? In: Estes JA, Demaster DP, Doak DF,
Williams TE, Brownell RL Jr (eds) Whales, whaling and
ocean ecosystems. University of California Press, Berke-
ley, CA, p 215−230
Beschta RL, Ripple WJ (2013) Are wolves saving Yellow-
stone’s aspen? A landscape-level test of a behaviorally
mediated trophic cascade comment. Ecology 94:
1420−1425
Blight LK, Ainley DG, Ackley SF, Ballard G and others
(2010) Fishing for data in the Ross Sea. Science 330: 1316
Bollens SM, Frost BW (1991) Diel vertical migration in zoo-
plankton: rapid individual response to predators.
J Plankton Res 13: 1359−1365
Branch TA, Matsuoka K, Miyashita T (2004) Evidence for
increases in Antarctic blue whales based on baysian
modelling. Mar Mamm Sci 20: 726−754
Christensen LB (2006) Marine mammal populations: recon-
structing historical abundances at the global scale. Fish-
eries Centre, University of British Columbia. https: //
circle.ubc.ca/handle/2429/17967
Clapham PJ, Baker CS (2001) How many whales were killed
in the Southern Hemisphere in the 20th century? Inter-
national Whaling Commission, Cambridge
Darwin CR (1881) The formation of vegetable mould,
through the action of worms, with observations on their
habits. John Murray, London
DeAngelis DL, Gross LJ (2009) Individual-based models and
approaches in ecology: populations, communities, and
ecosystems. Chapman & Hall, New York, NY
Emslie SD, Polito MJ, Patterson WP (2013) Stable isotope
analysis of ancient and modern gentoo penguin egg
membrane and the krill surplus hypothesis in Antarctica.
Antarct Sci 25: 213−218
Everson I (ed) (2000) Krill: biology, ecology and fisheries.
Blackwell Science, Oxford
Fiksen O (1997) Diel vertical migration and allocation pat-
terns - modelling the optimal Daphnia. Ecology 78:
1446−1456
Fiksen O, Carlotti F (1998) A model of optimal life history
and diel vertical migration in Calanus finmarchicus.
Sarsia 83: 129−147
Fiksen O, Giske J (1995) Vertical distribution and popula-
tion dynamics of copepods by dynamic optimisation.
ICES J Mar Sci 52: 483−503
Flores H, Atkinson A, Kawaguchi S, Krafft BA and others
(2012) Impact of climate change on Antarctic krill. Mar
Ecol Prog Ser 458: 1−19
Fraser FC (1937) On the development and distribution of the
young stages of krill (Euphausia superba). Discov Rep
14: 1−192
Giske J, Huse G, Fiksen O (1998) Modelling the spatial
dynamics of fish. Rev Fish Biol Fish 8: 57−91
Gliwicz MZ (1986) Predation and the evolution of vertical
migration in zooplankton. Nature 320: 746−748
Godlewska M (1996) Vertical migrations of krill (Euphasia
superba Dana). Pol Arch Hydrobiol 43: 9−63
Hamner W, Hamner P (2000) Behavior of Antarctic krill
(Euphausia superba): schooling, foraging, and anti -
predatory behavior. Can J Fish Aquat Sci 57(Suppl 3):
192−202
Hardy AC, Gunther ER (1936) The plankton of the South
Georgia whaling grounds and adjacent waters. Discov
Rep 11: 1−377
Harrison Matthews L (1938) The sei whale, Balaenoptera
borealis. Discov Rep 17: 185−290
Hay ME, Parker JD, Burkepile DE, Caudill CC, Wilson AE,
Hallinan ZP, Chequer AD (2004) Mutualisms and aquatic
community structure: The enemy of my enemy is my
friend. Annu Rev Ecol Evol Syst 35: 175−197
Herre EA, Knowlton N, Mueller UG, Rehner SA (1999) The
evolution of mutualisms: exploring the paths between
conflict and cooperation. Trends Ecol Evol 14: 49−53
Higginbottom IR, Hosie GW (1989) Biomass and population
structure of a large aggregation of krill near Prydz Bay,
Antarctica. Mar Ecol Prog Ser 58: 197−203
67
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
Mar Ecol Prog Ser 513: 51– 69, 2014
Hilborn R, Hilborn U (2012) Overfishing — what everyone
needs to know. Oxford University Press, Oxford
Holling CS (1959) Some characteristics of simple types of
predation and parasitism. Can Entomol 91: 385−398
Huntley M, Zhou M (2004) Influence of animals on turbu-
lence in the sea. Mar Ecol Prog Ser 273: 65−79
Ichii T (2000) Krill harvesting. In: Everson I (ed) Krill: biol-
ogy, ecology and fisheries. Blackwell Science, Oxford,
p 228−262
Jimenez J (1997) Oceanic turbulence at millimeter scales.
Sci Mar 61(Suppl 1): 47−56
Kawaguchi S, Satake M (1994) Relationship between
recruitment of the Antarctic krill and the degree of ice
cover near the South Shetland Islands. Fish Sci 60:
123−124
Kawaguchi S, Takahashi Y (1996) Antarctic krill (Euphausia
superba Dana) eat salps. Polar Biol 16: 479−481
Kemp S, Bennett AG (1932) On the distribution and move-
ments of whales on the South Georgia and South Shet-
land whaling grounds. Discov Rep 6: 167−190
Kerr RA (2006) Creatures great and small are stirring the
ocean. Science 313: 1717
Kock KH, Reid K, Croxall J, Nicol S (2007) Fisheries in the
Southern Ocean: an ecosystem approach. Philos Trans R
Soc Lond B Biol Sci 362: 2333−2349
Kozak KH, Wiens JJ (2010) Accelerated rates of climatic-
niche evolution underlie rapid species diversification.
Ecol Lett 13: 1378−1389
Krebs JR, Davies NB (1993) An introduction to behavioural
ecology, 3rd edn. Blackwell Science, Oxford
Laidre KL, Heide-Jorgensen MP, Heagerty P, Cossio A,
Bergstrom B, Simon M (2010) Spatial associations
between large baleen whales and their prey in West
Greenland. Mar Ecol Prog Ser 402: 269−284
Laland KN, Odling-Smee FJ, Feldman MW (1999) Evolu-
tionary consequences of niche construction and their
implications for ecology. Proc Natl Acad Sci USA 96:
10242−10247
Laws RM (1977) Seals and whales of Southern Ocean. Philos
Trans R Soc Lond B Biol Sci 279: 81−96
Lehette P, Tovar-Sanchez A, Duarte CM, Hernandez-Leon S
(2012) Krill excretion and its effect on primary produc-
tion. Mar Ecol Prog Ser 459: 29−38
Loose C, Dawidowicz P (1994) Trade-offs in diel vertical
migration by zooplankton: the costs of predator avoid-
ance. Ecology 75: 2255−2263
Mackintosh NA (1934) Distribution of the macroplankton in
the Atlantic sector of the Antarctic. Discov Rep 9: 67−158
Mackintosh NA (1973) Distribution of post-larval krill in the
Antarctic. Discov Rep 36: 95−156
Mackintosh NA (1974) Sizes of krill eaten by whales in the
Antarctic. Discov Rep 36: 157−178
Mackintosh NA, Wheeler JFG (1929) Southern blue and fin
whales. Discov Rep 1: 257−540
Mangel M (2006) The theoretical biologist’s toolbox: quanti-
tative methods for ecology and evolutionary biology.
Cambridge University Press, Cambridge
Margalef R (1978) Life-forms of phytoplankton as survival
alternatives in an unstable environment. Oceanol Acta 1:
493−509
Marr J (1956) Euphausia superba and Antarctic surface cur-
rent. Norsk Hvalfangst-Tidende: 127−134
Marr J (1964) The natural history and geography of the
Antarctic krill (Euphausia superba Dana). Discov Rep 32:
33−464
Mori M, Butterworth DS (2004) Consideration of multi-
species interactions in the Antarctic: a preliminary model
of the minke whale−blue whale−krill interaction. Afr
J Mar Sci 26: 245−259
Mori M, Butterworth DS (2006) A first step towards model-
ing the krill– predator dynamics of the Antarctic ecosys-
tem. CCAMLR Sci 13: 217−277
Nicol S, Constable A, Pauly T (2000) Estimates of circumpo-
lar Antarctic krill abundance based on recent acoustic
density measurements. CCAMLR Sci 7: 87−99
Nicol S, Bowie A, Jarman S, Lannuzel D, Meiners K, van der
Merwe P (2010) Southern Ocean iron fertilization by
baleen whales and Antarctic krill. Fish Fish 11: 203−209
Odenbaugh J (2005) Idealized, inaccurate but successful: a
pragmatic approach to evaluating models in theoretical
ecology. Biol Philos 20: 231−255
Odling-Smee J, Erwin DH, Palkovacs EP, Feldman MW,
Laland KN (2013) Niche construction theory: a practical
guide for ecologists. Q Rev Biol 88: 4−28
Odum EP (1969) The strategy of ecosystem development.
Science 164: 262−270
Ozawa K, Sato K (1967) Summer distribution of patches of
Euphausia superba. Mer (Paris) 5: 227−236
Pauly D, Christensen V, Walters C (2000) Ecopath, ecosim,
and ecospace as tools for evaluating ecosystem impact of
fisheries. ICES J Mar Sci 57: 697−706
Ramp C, Berube M, Hagen W, Sears R (2006) Survival of
adult blue whales Balaenoptera musculus in the Gulf of
St. Lawrence, Canada. Mar Ecol Prog Ser 319: 287−295
Rolland R, Hamilton P, Kraus S, Davenport B, Gillett R,
Wasser S (2006) Faecal sampling using detection dogs to
study reproduction and health in North Atlantic right
whales (Eubalaena glacialis). J Cetacean Res Manag 8:
121−125
Ross RM, Quetin LB (1983) Spawning frequency and fecun-
dity of the Antarctic krill Euphausia superba. Mar Biol
77: 201−205
Ross RM, Quetin LB (1988) Euphausia superba: a critical
review of estimates of annual production. Comp Biochem
Physiol 90B: 499−505
Ruegg KC, Anderson EC, Scott Baker C, Vant M, Jackson
JA, Palumbi SR (2010) Are Antarctic minke whales
unusually abundant because of 20th century whaling?
Mol Ecol 19: 281−291
Ruegg KC, Rosenbaum HC, Anderson EC, Engel M, Roth-
schild A, Baker CS, Palumbi SR (2013) Long-term popu-
lation size of the North Atlantic humpback whale within
the context of worldwide population structure. Conserv
Genet 14: 103
Ruud JT (1932) On the biology of southern Euphausiidae.
Hvalradets skrifter 2: 1−105
Sancho D, Blanco CA, Caballero I, Pascual A (2011) Free
iron in pale, dark and alcohol-free commercial lager
beers. J Sci Food Agric 91: 1142
Santora JA, Reiss CS, Loeb VJ, Veit RR (2010) Spatial asso-
ciation between hotspots of baleen whales and demo-
graphic patterns of Antarctic krill Euphausia superba
suggest size-dependent predation. Mar Ecol Prog Ser
405: 255−269
Shu Q, Qiao F, Song Z, Wang C (2012) Sea ice trends in the
Antarctic and their relationship to surface air tempera-
ture during 1979−2009. Clim Dyn 38: 2355−2363
Smetacek V (2008) Are declining antarctic krill stocks a
result of global warming or of the decimation of the
whales? In: Carlos M Duarte (ed) Impacts of global
68
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
Willis: Whales and krill 69
warming on polar ecosystems. Fundación BBVA Madrid,
p47–81
Speakman JR (2005) Body size, energy metabolism and life-
span. J Exp Biol 208: 1717−1730
Stich HB, Lampert W (1981) Predator evasion as an explana-
tion of diurnal vertical migration by zooplankton. Nature
293: 396−398
Tew Kai E, Rossi V, Sudre J, Weimerskirch H and others
(2009) Top marine predators track Lagrangian coherent
structures. Proc Natl Acad Sci USA 106: 8245−8250
Tovar-Sanchez A, Duarte CM, Hernandez-Leon S, Sanudo-
Wilhelmy SA (2007) Krill as a central node for iron
cycling in the Southern Ocean. Geophys Res Lett 32:
L11601, doi: 10.1029/2006GL029096
Trathan PN, Ratcliffe N, Masden EA (2012) Ecological driv-
ers of change at South Georgia: the krill surplus, or cli-
mate variability. Ecography 35: 983−993
Vera FWM (2000) Grazing ecology and forest history. CAB
International, Wallingford
Villiers AJ (1925) Whaling in the frozen south, being the
story of the 1923-24 Norwegian whaling expedition to
the Antarctic. Bobbs-Merrill, Indianapolis, IL
Visser AW (2007) Biomixing of the oceans? Science 316:
838−839
Voronina NM (1998) Comparative abundance and distribu-
tion of major filter feeders in the Antarctic pelagic zone.
J Mar Sci 17: 375−390
Willis J (2007) Could whales have maintained a high abun-
dance of krill? Evol Ecol Res 9: 651−662
Willis J (2013) Wake sorting, selective predation and bio-
genic mixing: potential reasons for high turbulence in
fish schools. PeerJ 1: e96
Editorial responsibility: Peter Corkeron,
Woods Hole, Massachusetts, USA
Submitted: February 26, 2014; Accepted: June 16, 2014
Proofs received from author(s): September 29, 2014
➤
➤
➤
➤
➤
➤
➤