Temperature effects on thresholds of hypoxia for marine benthic
organisms.
Raquel Vaquer-Suner and Carlos M. Duarte
Abstract
Global warming will contribute to decrease the global average dissolved oxygen in
the oceans worldwide, and may also affect the oxygen requirements of marine benthic
macrofauna. The effect of warming on the oxygen requirements and the survival of benthic
organisms under hypoxia was tested using a meta-analysis of published experimental results
evaluating the effects of increasing temperature on the median lethal time and median lethal
concentration of benthic macrofauna under hypoxia. The meta-analysis confirmed that
survival times under hypoxia are reduced by an average of 74% and that median lethal
concentration increases by a mean of 16% in marine benthic organisms exposed to higher
temperatures. Warming reduced survival times of marine benthic macrofauna under hypoxia
by a median of 3.95 ± 1.67 hours ºC
-1
and increased the oxygen thresholds for hypoxia-driven
mortality by a median of 1.02 ± 0.15 % saturation (i.e. 0.07 ± 0.01 mg O
2
L
-1
ºC
-1
).
Assessment of the impact of the 4 ºC warming expected during the 21
st
Century on the
survival time and the threshold oxygen concentrations for mortality of benthic macrofauna
using the average Q
10
values for median survival time (3.01 ± 0.29) and median lethal oxygen
concentration (2.09 ± 0.20) derived here predict that survival times will decrease by a mean
of 35.6 % under hypoxia and that the threshold oxygen concentrations for high mortality to
occur will increase by, on average, 25.5 % in a 4 ºC warmer ocean. Hence, ocean warming is
expected to increase the vulnerability of benthic macrofauna to reduced oxygen
concentrations, increasing the mortality of benthic fauna and greatly extending the area of
coastal ecosystems affected by hypoxia-driven mortality.
Introduction
Global warming is forecasted to lead to increase the mean global temperature by 1.8
to 4 ºC by the end of the 21st Century (Meehl et al. 2007), with important consequences on
climate, hydrology, biodiversity, and biogeochemical cycles. Impacts from global warming
will combine with those derived from other human pressures, such as the impacts derived
from excess nutrient inputs, which is a major driver of the proliferation of hypoxia in the
coastal ocean (Cloern 2001; Kemp et al. 2009). Dissolved oxygen is the property that has
changed more drastically in a shorter period of time in the marine environment (Diaz &
Rosenberg 1995; Diaz 2001). Oxygen deficiencies have increased in frequency, duration, and
severity in the world´s coastal areas during the last decades (Diaz & Rosenberg 2008). As a
consequence hypoxia is emerging as a major threat to marine coastal biodiversity (Vaquer-
Sunyer & Duarte 2008). Environmental factors, such as the presence of sulphide (Vaquer-
Sunyer & Duarte 2010); hypercapnia and low pH (Boleza et al. 2001; Rosa and Seibel 2008),
which are increasing with the increased pCO
2
and acidification of ocean waters (Caldeira &
Wickett 2003; Zeebe et al. 2008), have been shown to enhance the sensitivity of marine
organisms to hypoxia. Warming may also contribute to exacerbate hypoxia and its
consequences for marine life.
Temperature is a key factor controlling the extent of hypoxia (Conley et al. 2007),
acting through a multitude of interacting processes, including temperature effects on
increasing stratification and reducing ventilation of marine waters (Sarmiento et al. 1998).
The possibility of strengthened stratification alone, from increased surface water temperature,
is enough to worsen hypoxia where it presently exists and will trigger its occurrence in other
coastal areas (Rabalais et al. 2009). Stramma et al. (2008) documented the vertical expansion
of the intermediate-depth low oxygen zones in the eastern tropical Atlantic and the equatorial
Pacific during the past 50 years. The hypoxic boundary has shoaled up to 90 m in the
Californian Current System (Bograd et al. 2008). The oxygen content of the Oxygen
Minimum Zones (OMZ) has been documented to decline in the tropical Pacific, Atlantic and
Indian Oceans (Stramma et al. 2009; Stramma et al. 2010). The expansion and shoaling of
OMZ have been attributed to ocean warming (Keeling and Garcia 2002; Whitney et al.
2007).
The effects of climate change on water temperature are complex and modulated by
multiple factors such as changes in wind patterns with subsequent changes in surface
currents, circulation and mixing processes. Severe inner-shelf hypoxia off Oregon coast was
documented in 2002 (Service 2004). The formation of this new hypoxic area was attributed to
deviations in the circulation of the California Current System (Grantham et al. 2004) that
further reflect large-scale wind stress anomalies present over the northeast Pacific in 2002
(Murphree et al. 2003). Increasing temperature can produce intensification of coastal
upwelling (Bakun 1990; Bakun and Weeks 2004) with a subsequent oxygen decline in
bottom waters bellow the upwelling system when low grazing pressure allow the sinking and
subsequent respiratory decomposition of primary production (Bakun and Weeks 2004). In
these areas the upwelling of cold waters can lead to cooling, instead of warming, of seawater.
Despite this decrease of water temperature in regions with intensified upwelling, there is a
general trend toward a global warming of the upper ocean. Lyman et al. (2010) reported
significant warming of the upper ocean (from 0 to 700m depth) at global scales over the past
16 years. Recent observational surveys have shown significant warming of ocean bottom-
waters (Fukasawa et al. 2004; Masuda et al. 2010). The water temperature in shallow bays
from the Swedish West coast has also been reported to have increased (Cossellu and
Nordberg 2010). The seawater temperature at the West Mediterranean Sea has increased
between 0.12 and 0.5ºC from 1948 to 2005 in the upper layers (0 to 200 m), between 0.05
and 0.2 ºC from 1948 to 2000 in the mid layer (from 200 to 600 m depth) and between 0.03
and 0.1ºC from 1948 to 2005 in the deep water (from 1000 to 2000 m) (Bladé et al. 2010). A
coastal site (85 m depth) in the continental shelf of the Catalonian Sea (West Mediterranean)
sampled weekly exhibited intense warming trends at all depths ranging from 0.03 to
0.04ºC/year along the period from 1974 to 2001(Vargas-Yanez et al. 2005). Hence, despite
variable effects at specific locations, there is a well documented tendency for seawater
temperature to increase over the top 700 m, where most hypoxic events have been
documented, at the global scale.
Increasing temperature diminishes oxygen solubility (Carpenter 1966; Garcia &
Gordon 1992), and increases the respiration rates of organisms (Jones 1977; Enquist et al.
2003), as temperature plays a fundamental role in regulating metabolic processes (Iriberri et
al. 1985; White et al. 1991).
Increased temperature will likely affect the responses of marine benthic organisms to
hypoxia because metabolic rates increase exponentially with temperature (Brown et al.
2004). Whereas both photosynthesis and respiration are enhanced with warming, within the
limits imposed by resources (light, CO
2
and nutrients, and oxygen concentration,
respectively), Metabolic Theory of Ecology (MTE, Brown et al., 2004) predicts that
respiration rates should increase faster with warming than photosynthetic rates as activation
energies for autotrophic processes are half of those for heterotrophic processes (Harris et al.
2006). On the basis of this differential response, Harris et al. (2006) predicted that an
hypothetically four degree increase in the summer water temperature of a north-eastern
Atlantic estuary will result in a 20% increase in net primary production and a 43% increase in
heterotrophic metabolism, resulting in a 16% decrease of the P:R ratios and an increasing
likelihood of system heterotrophy. These predictions, however, may be conservative, as they
refer to specific metabolic rates but do not consider possible effects of warming on
autotrophic and heterotrophic biomass. Müren et al. (2005) showed that the heterotrophic to
autotrophic biomass ratio increased 5 times and the production to respiration ratio decreased
six times when temperature was raised from 5 to 10ºC in experimental mesocosms. A
decrease in ecosystem P:R ratios with increasing temperature could result in a net decrease in
oxygen concentration, increasing the frequency and severity of hypoxic events. However,
increased temperature may also affect the vulnerability of organisms to low oxygen
concentration, as the increased organismal respiration rates increases their oxygen demand,
affecting the oxygen thresholds for hypoxia. Here we evaluate, on the basis of a meta-
analysis of available experimental results, the effects of temperature on the oxygen thresholds
for marine benthic macrofauna. Because the range of species where experimental assessment
of temperature effects on thresholds of hypoxia is limited, the generality of the conclusions
reached here must be tested further when data for species not included here become available.
Methods
We searched the Web of Science and Scholar Google for reports of hypoxia using the
keywords ‘hypoxia’, ‘marine’, ‘benthic’ and ‘sea’, and their combinations to guide the
search. This search delivered more than 6000 published reports of responses of benthic
marine organisms to hypoxia, which were then examined further for the availability of
experimental assessments of responses to reduced oxygen concentration that included
temperature and/or evaluated them at different experimental temperatures. We also searched
the list of papers cited in those retrieved by the search. This more restricted search delivered a
total of 363 experimental assessments examining the median lethal time (LT
50
), representing
the statistically derived time interval at which 50% of a given population dies after exposure
to low O
2
levels, involving 108 different species of marine benthos; and a total of 213
experimental assessments examining the median lethal concentration (LC
50
), representing the
statistically derived O
2
concentration at which 50% of the organisms in a given population
die, involving 39 different species of marine benthos. Data on the experimental water
temperature were derived from the paper, and where they were only reported in graphics
were extracted digitalizing data using Graph Click 2.9.2 software.
Quantile regression was used to assess changes in the probability distribution of
thresholds of hypoxia for marine benthic organisms with increasing temperature. The
response of the thresholds of hypoxia, as described by LC
50
(% sat. and mg O
2
L
-1
) and LT
50
(h), to temperature was described by fitting the relationship between the 95%, 50% (median)
and 5% quantiles for the distribution of these thresholds and water temperature. Quantile
regression estimates multiple rates of change (slopes), from the minimum to maximum
response, providing a more complete description of the relationships between variables
missed by other regression methods focused on prediction of the mean value (Cade & Noon
2003). Quantile regression can be considered as an extension of classical least squares
estimation of conditional mean models to the estimation of a compilation of models for
several conditional quantile functions, considering the median as the central parameter
(Koenker 2005). Statistical analyses were performed using JMP 7.02 for simple regression
analyses, ANOVA and ANCOVA, and R for quantile regression.
Results
We found a total of 363 published experiments involving 108 species pertaining to 10
different taxonomic groups of benthic macrofauna reporting the water temperature at which
the median lethal time (LT
50
, h) was assessed (Fig. 1a) and 213 experimental assessments
involving 39 species from 3 different taxonomic groups (mollusca, fishes and crustaceans) of
benthic marine fauna reporting the incubation temperature at which the median lethal
concentration (LC
50
, % saturation and mg O
2
L
-1
) was assessed (Fig 1b,c). The aim of this
ample comparison is to test for evidence of a temperature-dependence in the thresholds of
hypoxia for coastal benthic organisms. The experiments compiled include a diversity of
procedures and species, which may add variability to the analysis, contributing to the residual
variability.
Examination of the relationship between LT
50
and experimental temperature showed
that the range of LT
50
values observed declined with increasing temperature, with most
experiments conducted showing relatively low LT
50
values at high temperature (Fig. 1a). This
was confirmed using quantile regression fitted to the 95% and the 5% quantiles as well as the
50% quartile (median) of the change in LT
50
with increasing temperature (Fig 1a). The 95%
quantile regression, estimating the temperature dependence of the maximum LT
50
expected
for a given water temperature, indicated a decrease in the maximum LT
50
by 63.62 ± 19.10
hours per each degree of temperature increase, whereas the 5% quantile regression,
estimating the temperature dependence of the minimum LT
50
, showed only a decrease by
0.45 ± 0.31 hours (27 minutes) for each degree Celsius increase. The median LT
50
declined
![Figure 1. (a) The relationship between the median lethal time (LT50, h) and water temperature for the different experiments. The solid line represents the fitted regression for the median or the 50% quartile [LT50 (h) = 194.70 (±34.04)-3.95 (±1.67)* temperature; N=363, P<0.03]. The dashed lines represent the fitted regression for the 95% quantile [LT50 (h)=1956.25 (±430.54)-63.62 (±19.10) * temperature; N =363, P<0.001] and the 5% quantile [LT50 (h)=12.95 (± 6.63)-0.45 (±0.31) *temperature; N=363, P=0.15]. (b) The relationship between the median lethal concentration (LC50, %sat.) and the water temperature for the different experiments. The solid line represents the fitted regression for the median or the 50% quartile [LC50 (%sat.) =0.71 (±2.91)+1.02 (±0.15)* temperature; N=213, P<0.0001]. The dashed lines represent the fitted regression for the 95% quantile (LC50 (% sat.)=15.50 (±8.99)+2.75 (±0.47)*temperature; N=213, P<0.0001) and the 5% quantile (LC50 (% sat.)= -3.37 (±3.62) +0.50 (±0.15)* temperature; N=213, P<0.002). (c) The relationship between the median lethal concentration (LC50, mgO2 L-1) and the water temperature for the different experiments. The solid line represents the fitted regression for the median or the 50% quartile [LC50 (mgO2 L1)50.44 (±0.27)+0.06 (±0.01)* temperature; N=212, P<0.0001]. The dashed lines represent the fitted regression for the upper 95% quartile [LC50 (mgO2L-1) = 2.24 (±0.68)+0.15 (±0.04)*temperature; n=212, P<0.0001] and the lower 5% quartile [LC50 (mgO2L-1)=-0.10 (±0.32) +0.03 (±0.01)* temperature; N=212, P<0.02] (see Annex 7 at http://imedea.uib-](/figures/figure-1-a-the-relationship-between-the-median-lethal-time-2uzok6mq.webp)