Download Rates of biotic interactions scale predictably with

Oikos 000: 001–008, 2014
doi: 10.1111/oik.01199
© 2014 The Authors. Oikos © 2014 Nordic Society Oikos
Subject Editor: Ulrich Brose. Accepted 1 April 2014
Rates of biotic interactions scale predictably with temperature
despite variation
William R. Burnside, Erik B. Erhardt, Sean T. Hammond and James H. Brown­
W. R. Burnside ([email protected]), S. T. Hammond and J. H. Brown, Dept of Biology, Univ. of New Mexico, Albuquerque,
NM 87131-0001, USA. Present address for WRB: National Socio-Environmental Synthesis Center (SESYNC), 1 Park Place, Suite 300,
Annapolis, MD 21401, USA. – E. B. Erhardt, Dept of Mathematics and Statistics, Univ. of New Mexico, Albuquerque, NM 87131-0001, USA.­
Most biological processes are temperature dependent. To quantify the temperature dependence of biotic interactions
and evaluate predictions of metabolic theory, we: 1) compiled a database of 81 studies that provided 112 measures of
rates of herbivory, predation, parasitism, parasitoidy, or competition between two species at two or more temperatures;
and 2) analyzed the temperature dependence of these rates in the framework of metabolic ecology to test our prediction
that the “activation energy,” E, centers around 0.65 eV. We focused on studies that assessed rates or associated times of
entire biotic interactions, such as time to consumption of all prey, rather than rates of components of these interactions,
such as prey encounter rate. Results were: 1) the frequency distribution of E for each interaction type was typically
peaked and right skewed; 2) the overall mean is E  0.96 eV and median E  0.78 eV; 3) there was significant variation
in E within but not across interaction types; but 4) average values of E were not significantly different from 0.65 eV by
interaction type and 5) studies with measurements at more temperatures were more consistent with E  0.65 eV. These
synthetic findings suggest that, despite the many complicating factors, the temperature-dependence of rates of biotic
interactions broadly reflect of rates of metabolism, a relationship with important implications for a warming world.
Understanding how temperature affects ecological relationships is increasingly urgent as our planet warms. In general,
most ecological processes go faster at higher temperatures,
reflecting the role of individual metabolism in ecology
and the role of temperature in metabolism (Brown et al.
2004). Metabolic rate increases approximately exponentially
with body temperature, at least up to some physiologically
stressful degree, reflecting the temperature dependence of
rates of biochemical reactions (Gillooly et al. 2001). The
simplest expectation is that rates of biotic interactions, such
as predation and competition, will reflect the temperaturedependence of metabolic rate, because the underlying
physiology and behavior are governed by metabolic processes (Peters 1983, Brown et al. 2004). However, biotic
interactions involve multiple individuals of often different
species. Slight differences in how quickly or how much
interacting organisms respond to temperature changes can
affect the outcome, so rates of biotic interactions may vary
widely. Current evidence on temperature dependence of
rates of biotic interactions is varied, incomplete, and inconsistent (Dell et al. 2011, Englund et al. 2011, Lemoine and
Burkepile 2012).
Recent efforts to build and test a metabolic theory of
ecology aim to conceptualize and quantify the mechanistic
role of metabolism in ecological patterns and processes.
Metabolic rate varies predictably with body size and
temperature. This relationship has been quantified in the
‘central equation’ of metabolic theory:
(1)
where B is mass-specific metabolic rate; B0 is a normalization
constant that typically varies with taxon, functional group,
and environmental setting; M is body mass; E is an
‘activation energy’ determined by the underlying biochemical reactions and physiological processes; k is Boltzmann’s
constant; and T is absolute temperature in Kelvin (Gillooly
et al. 2001, Brown et al. 2004, Brown and Sibly 2012).
The last term of Eq. 1 describes the temperature dependence
of metabolism in terms of the Arrhenius–Boltzmann factor,
e2E/kT. Activation energies for metabolic reactions vary
from approximately 0.2–1.2 eV. The middle of this range,
0.65eV, empirically characterizes many of these reactions,
including those that power aerobic respiration (Dell et al.
2011). This value corresponds to the physiologists’ Q10 of
approximately 2.5, meaning that metabolic rate increases
about 2.5 times with a 10°C increase in body temperature.
An important caveat is that Eq. 1, like all models, is a
deliberate simplification of a more complex reality. We
do not expect data to conform exactly to our expectations
for at least three reasons. First, data are always variable due
to intrinsic biological variation, imprecision in controlling
B  B0 M 3/4 e E /kT
EV-1
conditions, and measurement errors. Second, the Arrhenius–
Boltzmann expression itself, e2E/kT, is a simplification. The
full relationship between temperature and metabolic rate is
hump shaped: an approximately exponential curve rises to a
peak and then declines precipitously as the temperature
changes from optimal to stressful (Knies and Kingsolver
2010, Hoekman 2010). Although organisms encounter
stressfully high temperatures (Hochachka and Somero
1984, Englund et al. 2011), most of them operate most of
the time within a biologically relevant range of approximately 0–40° C, and many use behavior to seek out
optimal temperatures and avoid extremes (Martin and
Huey 2008, Stevenson 1985, and additional detail in the
Discussion). Within an organism’s optimal range, when it is
warm enough to be active but not past its peak performance,
an approximately exponential temperature-dependence
can be quantified as the slope of a linear regression in an
Arrhenius plot, where the logarithm of rate is plotted as a
function of 1/kT (Fig. 1). As illustrated in Fig. 1, such
Arrhenius plots often deviate from strict linearity, especially
at the highest temperatures. Third, the expected value of
E ≈ 0.65 eV is only an approximation based in part on the
balance of empirical evidence. Assuming that the overall rate
of aerobic respiration and metabolically dependent biological activities has approximately this temperature dependence
oversimplifies the complex biochemistry and kinetics of
metabolism (discussed in Ratkowsky et al. 2005 and Price
et al. 2012). Nevertheless, most biological processes governed by aerobic respiration have an E of approximately
0.65 eV, equivalent to a Q10 of approximately 2.5, although
Q10 itself is temperature dependent (Gillooly et al. 2001).
Biotic interactions are fundamentally metabolic because
they involve exchanges of energy and materials between
organisms and their environments. So the null expectation
Linear scale, exponential curve
(b)
2.5
Arrhenius−Boltzmann scale, linear fit
0.15
0.0
0.10
ln(Rate)
Rate (g leaf consumed / g caterpillar / hr)
(a)
of metabolic theory is that rates of most ecological processes
should exhibit similar temperature dependence as metabolic
rate (Eq 1; Brown et al. 2004). This prediction is supported
by empirical studies of processes at levels from individuals
to ecosystems (reviewed by Sibly et al. 2012), including
population growth (Savage et al. 2004). However,
community-level processes are inherently more complicated
because they involve biotic interactions among multiple
species – sometimes as different as animals, plants and
microbes – that may vary in the temperature-dependence of
their metabolic processes and behavior (discussed by Pörtner
and Farrell 2008, Kordas et al. 2011). The potential variety
of assymmetric responses to temperature, such as if the
escape velocity of prey rises faster than the pursuit velocity
of its predator and vice versa (Dell et al. 2014), suggests
there will be a great deal of variation in the temperaturedependence of biotic interaction rates.
Many empirical studies have measured the temperature
dependence of biotic interactions, especially predation.
Most of these have focused on just one of Holling’s (1959)
components of predation and have studied a single pair
of species in one specific environmental setting (Lemoine
and Burkepile 2012). Broad comparative studies have thus
far been limited to a handful of meta-analyses, notably:
1) Dell et al. (2011), which analyzed the temperaturedependence of some physiological and ecological traits,
including components of predation; 2) Englund et al.
(2011) and 3) Rall et al. (2012), both of which focused on
effects of temperature on parameters of the functional
response of predators; and 4) Rodríguez-Castañeda (2013),
which analyzed field studies on geographic variation in
the strength of top–down and bottom–up processes in
trophic cascades. These meta-analyses report varied, sometimes conflicting results. So a broader, more comprehensive
–2.5
0.05
0.00
–5.0
10
20
30
Temperature (deg C)
40
37
38
39
40
1/(k * Temperature (deg K))
41
Figure 1. Temperature dependence of the rate of an ecological interaction – here herbivory of caterpillars Pieris rapae on collard leaves
Brassica oleracea – plotted in two ways: (a) rate as a function of temperature (in°C) on linear axes, showing a typical exponential curve,
where the curve was estimated from the linear fit in (b); (b) as an Arrhenius plot, with the natural logarithm of rate plotted as a function of
inverse temperature, 1/kT, where k is Boltzmann’s constant and T is temperature in Kelvin, showing how the nearly linear relationship
can be fitted using weighted least squares (WLS), weighted by the number of observations contributing to each mean plotted, to obtain
the slope as a quantitative measure of temperature dependence, E. The data are from Kingsolver 2000.
EV-2
comparative study focused on the temperature dependence
of entire interactions is warranted.
Here we compile and analyze data from published
studies that measured herbivory, predation, parasitism,
parasitoidy, or competition between two species at two or
more different temperatures to ask: 1) do biotic interaction
rates vary systematically with temperature? And 2) is the
temperature dependence quantitatively similar to that of
metabolic rate, specifically the average rate of wholeorganism respiration, and hence consistent with predictions
of metabolic theory? Our treatment differs in several respects
from the meta-analyses discussed above. First, it is based
on a much larger sample of studies assessing interaction
rates per se, most of which were not included in the
previously published meta-analyses. Second, it includes data
on a wider range of interactions, including not only predation, herbivory, and parasitoidy, but also parasitism and
competition (Supplementary material Appendix 1). Third,
we restrict our database to studies that measured rates of
entire interactions, not just of components of interactions.
So, for example, we did not use studies of predation that
measured only search time, handling time, or consumption
rate. Fourth, perhaps because of the larger sample size and
greater scope, results of our study differ somewhat from
those of the previous meta-analyses. It will be important
to reconcile the differences to better understand inter­
specific interactions in the context of metabolic theory and
individual-level biological processes.
Methods
Study criteria and data sources
The present study is a meta-analysis in the sense that it is
based on a compilation and analysis of published data, but
we used methods from macroecology and metabolic ecology
rather than traditional meta-analytical procedures. We focus
on the quantitative form of temperature-rate relationships
(value of E, the negative slopes of Arrhenius plots) rather
than the extent of the temperature dependence (effect sizes).
We assembled a database of published studies that measured
interaction rate, such as the mean number of prey eaten
per day, or time to outcome (e.g. time to extinction of one
species, which we converted to a rate) at two or more
temperatures. We searched the literature for the
keywords ‘temperature’ and ‘rate’, with each of ‘herbivory’,
‘predation’, ‘parasitism’, ‘parasitic’, ‘parasitoid’, ‘parasitoidy’,
‘competition’ or ‘competitive’ using the ISI Web of Science
and Google Scholar databases. We found too few studies
on temperature dependence of mutualism to include in
our analysis. Our intent was to quantify and synthesize
data on temperature dependence of rates of representative
biotic interactions and to evaluate predictions of metabolic
theory rather than to perform an exhaustive literature survey.
Throughout the study we continued to find additional papers
with suitable data, so we have undoubtedly overlooked
some relevant studies, especially on predation. Nevertheless,
our sample of papers overlaps only slightly with those used
in recent, complementary meta-analyses (Dell et al. 2011,
Englund et al. 2011, Rodríguez-Castañeda 2013).
We included only studies that:
1. were published in the peer-reviewed literature,
2. used live organisms (e.g. no dead prey),
3. explicitly reported rates or times of biotic interactions
(e.g. time to competitive exclusion),
4. held everything except temperature constant within each
set of experiments or observations, so that the values of
E are comparable across studies,
5. provided data on at least two non-zero rates or times
measured at two or more different controlled or standardized temperatures within the normal thermal range
of the interacting species,
6. measured interactions between mobile organisms, except
for plants in cases of herbivory,
7. measured interactions directly; for example, we excluded
studies based on ontogenetic growth rates or parameters
of models fitted to data (e.g. capture rate, C, in Holling’s
(1959) disc equation), and
8. provided sufficient detail on the methods, including
measurements of rates or times, so as to ensure comparability for inclusion.
We used original data values of rates and temperatures
when reported and otherwise extracted these values from
published graphs using DataThief III (www.datathief.
org/). In the rare instances when either explicit values
were not reported or points on graphs were difficult to differentiate, we used the parameters calculated by the authors
directly from their data. We used the rate (or the inverse of
the time) of the entire biotic interaction rather than of
its subcomponents, such as search time or handling time,
measured separately (Supplementary Material Appendix A1,
columns G and I).
The units necessarily varied by study, but for predation
studies this typically meant the number of prey eaten
per predator per unit time at different temperatures, while
for competition this often meant the time required for
one species to drive another to extinction at different
temperatures. Our methodology is robust to these differences: if one experiment measured the biomass of prey
consumed at all test temperatures and another measured
the number of individual prey consumed at all test temperatures, the values of E should be comparable. In this
example, if the individual prey were all of about the
same body mass, the values of E should be similar and the
Arrhenius plots should be approximately parallel, offset by
a constant mass per individual. When studies repeated
experiments and reported values for the replicates, we
used the weighted mean rate value for each temperature.
To avoid satiation of predators in studies reporting functional responses, we used rates measured at intermediate
prey densities. We detail this, for each study, in the associated Notes column (G) of the database.
Database
Our database, compiled using the above criteria, consisted
of 81 studies on a total of 112 different interactions providing estimates of E (Table 1). For each study, we used the
rate or 1/time reported at each different temperature. For
studies with data for three or more temperatures we used
EV-3
Table 1. Statistics for the temperature dependence of biotic inter­actions analyzed here. Temperature dependence was measured as the
value of E, the negative slope of the data in Arrhenius plot form for each interaction. There was considerable variation in the values of E for
each interaction type but no significant difference in the overall distribution centers among interaction types.
Interaction
NumStudies
Mean
Median
SD
SE
Skewness
Kurtosis
20
15
16
21
40
112
1.2
0.93
1.4
0.95
0.73
0.97
0.87
0.7
0.76
0.87
0.75
0.79
1.2
0.71
1.4
0.5
0.66
0.92
0.26
0.18
0.36
0.11
0.11
0.087
1.6
1.7
0.82
0.78
20.45
1.6
2.2
2.1
20.79
20.027
3.7
4.5
Competition
Herbivory
Parasitism
Parasitoidy
Predation
All
only the values along the rising portion of the thermal performance curve (Fig. 1a); when there were only two temperatures, we used both values. We included studies with
only two data points because our effort to find suitable
studies strongly suggests there are relatively few that satisfied our criteria and because comparison with studies with
more temperatures suggested no underlying biases, just a
lack of resolution (see Fig. 4 and Results for additional
detail). Excluding those with only two or three points
would have restricted the size of this sample much further.
We calculated the natural logarithm of the rate, converted
the temperature from Celsius to Kelvin, constructed an
Arrhenius plot, and estimated E as the negative slope using
weighted least squares (WLS) regression. The complete
database, which includes detailed information on each study
and interaction we included, are in the Supplementary
material Appendix A1. Columns “Rate definition” and
“Rate terms” of the Supplementary material Appendix 1
includes the units of the rate or time assessed, which varied
by study.
Statistical analyses
We calculated the temperature dependence of the inter­
action as E, the negative slope of the WLS regression in
the Arrhenius plot (Fig. 1b). WLS, a variation on OLS, is
appropriate because interaction rate clearly depends on
temperature, and because temperature was usually closely
controlled and hence measured with less error than
interaction rate (Smith 2009, White et al. 2012). We
compiled summary statistics for each interaction type
separately (Table 1, Fig. 2, 3) and for all types combined
(Table 1).
Results
Results are presented in Fig. 2–4, Table 1 and the
Supplemental material (Appendix 1 contains all the data).
Figure 2 shows all of the data in Arrhenius format, with
each of the 112 interactions plotted separately and
color-coded by interaction type. The vertical displacement
(intercept) of the fitted regression lines is uninformative
because the reported rates depend on the units used to
measure them, which varied by study. The negative slope of
the regression gives an estimate of E, which we took as
our quantitative measure of temperature dependence. The
negative slope for E  0.65 eV, the approximate predicted
value, is plotted for reference.
EV-4
The distributions of activation energy, E, by interaction
type with mean and 95% bootstrapped confidence interval
are plotted in Fig. 3, with dashed reference line at 0.65.
There is a central tendency slightly greater than the predicted value of 0.65 eV. Figure 3 conveys three additional
messages as well: 1) there is wide variation in the estimated
values of E for each type of interaction; 2) the overall frequency distribution of E is strongly peaked and modestly
right-skewed, and 3) the frequency distributions of E for
different interaction types overlap broadly.
As shown in Table 1, there is insufficient evidence to
reject the null hypothesis that the mean slopes (negative
activation energy E) between interaction types are
equal (ANOVA p-value  0.14, Kruskal–Wallis p  0.79,
permutation test p  0.14). Individually, parasitoidy has
a significantly different mean from 0.65 (t-test p  0.013,
Wilcoxon p  0.013, Sign test p  0.027) while the other
four groups do not (minimum t-test p  0.059, Wilcoxon
p  0.20, Sign test p  0.43). However after correcting for
family-wise type I error using Bonferroni, the parasitoidy
p-value also exceeds p  0.05.
To maximize the number and variety of studies, we
included all studies that met our a priori selection criteria.
Figure 2, 3 and 4 show all the data. We did not want to
exclude studies with limited resolution because they
measured rates at only a few temperatures. Therefore
the ‘regression lines’ plotted for each study in Fig. 2 and 3
include many cases with only two data points. We expected,
however, that increased sample sizes of test temperatures
would give more accurate estimates of E, and that these
would converge more closely around the expected value
of approximately 0.65 eV. This is supported by the
solid points in Fig. 3 and by plotting values of E as a function
of sample size in Fig. 4. In general, the temperaturedependence of interactions were slightly greater than
E ≈ 0.65 and were qualitatively closer to 0.65 as measurements at more temperatures provided greater resolution.
Discussion
Overall the results support our expectation that rates of
biotic interactions will reflect the temperature-dependence
of metabolic rate generally, in particular the mean rate of
whole-organism respiration. Rates of the biotic interactions
we assessed center near the expected value of E ≈ 0.65 eV,
equivalent to Q10 ≈ 2.5. For each interaction type, the
mean and median is slightly greater than but not significantly different from 0.65 eV (Fig. 3). As expected, there is
competition
herbivory
parasitoidy
predation
parasitism
10
0
ln(Rate)
–10
10
0
–10
37
38
39
40
41
42
37
38
39
40
1/kT
41
42
Figure 2. Arrhenius plots of the temperature–rate relationships for all data in the analysis. Each point is the temperature dependence of a
single mean rate from a single study, and the colored lines connecting them are WLS regressions fitted to the mean points for different
temperatures in that study. Different kinds of interactions are color-coded. The black dashed lines running diagonally have the predicted
slope of 20.65 and are for reference. The Y-axis units are arbitrary because the reported rates depend on the units of measurement in a
given study. While some slopes vary substantially, most are parallel to each other and to the predicted slope.
(a)
(b)
Interaction
competition
herbivory
4
parasitism
parasitoidy
15
2
Count
Estimated activation energy, E (eV)
20
predation
10
5
0
0
competition herbivory
parasitism parasitoidy predation
Interaction type
0
2
4
Estimated activation energy, E (eV)
Figure 3. (a) Estimated activation energy, E, by biotic interaction type for each study with mean and bootstrapped 95% confidence interval.
Temperature dependence is quantified as the negative slope of a WLS regression in an Arrhenius plot (Fig. 1, 2). Open circles are estimates
based on two or three temperatures, while solid circles have at least four temperatures. Circles are slightly offset from their associated lines
to avoid overlap. (b) Histogram of the frequencies of activation energies for all interactions plotted together.
EV-5
Interaction
competition
herbivory
4
parasitism
parasitoidy
Estimated activation energy, E (eV)
predation
2
0
2
4
Number of data points
6
8
Figure 4. Plot of the estimated magnitude of temperature dependence, measured as the value of E from the negative slopes of Arrhenius
plots, as a function of the number of temperatures at which rates were measured in each study. In general, the temperature-dependence of
interactions converged toward E  0.65 (dashed line) as measurements at more temperatures, which often covered a wider temperature
span, provided greater resolution. The two low values with four data points correspond to predation rates of carabid beetles on fruit flies
(Kruse et al. 2008), a case, discussed more below, in which the predators took advantage of their prey’s inability to fly, a better means of
escape, at lower temperatures.
considerable variation around the center. Because the data
are somewhat right-skewed, the overall mean is somewhat
higher (0.96 eV) than the overall median (0.79 eV), which
coincides with the results of Dell et al. (2011). Without
observation-level data, we have insufficient information to
assess the fit of the Boltzmann–Arrhenius model, but more
than 70% of studies with at least three temperatures had
BIC values favoring the fit of regressions on the Arrhenius–
Boltzmann scale compared to regressions on the original,
linear scale (Fig. 1b versus 1a). We also did not find
significant differences in means or medians of E among the
different interaction types (Table 1, Fig. 3a).
Like all meta-analyses based on published studies, this
one has potential sources of error and bias. First, although
we controlled for differences among studies as much as
possible by using strict criteria for inclusion and consistent
methods to quantify rates, some of the variation we found
may be due to differences in methodology. We used the
units and rates reported by the study authors, which
were consistent within studies. However, so long as study
authors did not systematically round in one direction, and
we found no evidence they did, such variation will not affect
the mean value of E for a reasonably-sized sample of studies.
EV-6
Furthermore, data points within studies were typically mean,
often per capita rates from a number of replicates, each with
a number of individual antagonist organisms, all of which
would increase the precision of reported rates that much
more. Differences between studies in rate units, such as mass
of prey in mg versus number of gram-sized prey consumed,
analogous to variation from rounding to different levels of
precision, can cause modest variation in calculated activation
energies. Furthermore, systematic rounding only up or
only down to the nearest rate unit can cause minor slope
bias on the log(rate) scale since a unit-sized change in a small
log-transformed rate (the rate at the lowest temperature)
has a greater effect than the same size change in a larger
log-transformed rate (the rate at the highest temperature).
Second, because of the time, effort, and equipment required,
careful studies of interactions have typically been conducted
at only a small number of controlled temperatures. As
mentioned above, such small sample sizes inherently limit
statistical inference and the precision of estimates of E.
The greatest variation occurred when there were only two or
three measurements encompassing only a limited range of
temperatures (Fig. 4). With such limited data, it is difficult
to assess whether the measurements were taken within the
normal temperature range or included stressful conditions
(Izem and Kingsolver 2005, Angilletta et al. 2010, Englund
et al. 2011). Third, there are issues in using the slopes of
Arrhenius plots to quantify temperature dependence. The
Arrhenius–Boltzmann term in Eq. 1 is meant to apply only
to the approximately exponential, ascending part of the
thermal performance curve in Fig. 1a. To obtain an accurate
estimate of E it is necessary to exclude values from the
descending part of the curve, where temperatures are
generally assumed to be outside the normal activity range.
Fourth, our database is highly biased toward interactions
involving mobile animals, especially predators, reflecting the
current availability of studies.
We intentionally did not include studies on mutualism,
competition in plants, or interactions involving microbes
for two reasons. First, the relevant rate-limiting metabolic
processes in these groups – photosynthesis in plants and
the variety of energy-transforming biochemical pathways
in microbes – might be expected a priori to have different
‘activation energies’ (Anderson-Teixeira et al. 2008, Okie
2012), complicating quantitative comparisons. Second,
the different traditions of plant ecology and microbiology,
as opposed to animal community ecology, have resulted in
investigators asking different questions and using different
methodologies, again making comparisons across groups
difficult to interpret. Although there is a rich tradition
of experimentation in plant ecology, most studies of
interactions have not manipulated temperature. Work in
microbial ecology has thus far focused largely on nonexperimental studies to document taxonomic composition
and functional properties.
Despite these issues of methodology and potential
bias, however, much of the observed variation in temperature dependence is undoubtedly real and warrants further
study. Some of this variation in the estimated value of E is
likely artifactual or methodological. There are formidable
challenges in accurately measuring the overall interaction
rates in some standardized way that can be compared
across studies. Furthermore, not all organisms in all environments are equally subject to the “tyranny of Boltzmann”
(Clarke and Fraser 2004). There is an extensive literature on
thermal acclimation, acclimatization, and adaptation showing how physiology and behavior may be altered to confer
short-term phenotypic plasticity or longer-term evolutionary adaptation to changing environmental conditions
(Bartholomew 1964, Ayala 1966, Brett 1971, Stevenson
1985, Huey and Kingsolver 1989, Dunson and Travis 1991,
Deutsch et al. 2008, Martin and Huey 2008, Kearney et al.
2009, Somero 2010, Sunday et al. 2011, Dell et al. 2014).
Another complication comes from characterizing the
temperature dependence with just a single value of E,
because the two interacting species almost certainly
differ somewhat in the temperature dependence of their
ecological processes. All types of interactions studied here
are fundamentally antagonistic. In the cases of predation,
parasitoidy, parasitism and herbivory one species benefits by
consuming biomass of the other, presumably negatively
affecting its population’s growth and fitness. In the case of
interspecific competition, the species negatively affect each
other by consuming shared resources or by such direct interference. These interactions often lead to ‘coevolutionary
arms races’ in which species sequentially evolve traits in
response to one another. Previous authors have pointed out
that such coevolutionary adaptations may include shifts in
thermal performance. One consequence may be asymmetries in the temperature dependent responses of predators
and prey (Dell et al. 2011, 2014, Lemoine and Burkepile
2012). In Fig. 3, the two data points with exceptionally low
values of E are from a study of predation (Kruse et al. 2008)
in which carabid beetles take advantage of a mismatch
between the temperature dependence of their activity and
that of their fruit fly prey. Although both the flies and
the beetles were more active and moved faster at higher
temperatures, the beetles caught more flies at lower temperatures because the flies were even more sluggish and
could only escape by walking rather than flying, presumably
because flying requires more energy.
Our results, based on the 81 studies of 112 interactions
analyzed here, are generally consistent with the findings of
Dell et al. (2011), who analyzed data for a wider variety
of ecological traits and included some studies of predation.
Methodological differences likely account for the differences
between our results and those of Englund et al. (2011),
who compiled and analyzed data on components of the
functional response in predator–prey interactions. They
questioned the utility of Arrhenius plots and single values
of E to characterize the temperature dependence, but
their meta-analysis included data from stressfully high temperatures on the downward slope of thermal performance
curves.
The bottom line is that despite the complications and
resulting variation, the majority of biotic interactions exhibit
temperature dependence within the range expected on
the basis of metabolic theory (Brown et al. 2004, Sibly
et al. 2012). This finding extends to biotic interactions a
growing list of physiological, behavioral, ecological, and
evolutionary rate process that scale with body size and temperature as predicted by metabolic theory, with E ≈ 0.65 eV,
or Q10 ≈ 2.5. This should not be surprising. Common
metabolic mechanisms underlie most aspects of ecological
performance. Predators, herbivores and parasites must consume food at rates that meet their metabolic needs. The
dynamics of interactions also depend on the scaling of life
history attributes and population growth rates, which in
turn scale with metabolic rate.
Finally, the temperature dependence of biotic inter­
actions has important implications for understanding
the ecological consequences of climate change. Higher
temperatures are already affecting individual organisms,
populations, communities, and ecosystems (Parmesan and
Yohe 2003). There are too many species and environmental
settings to tackle individually, so some general framework
is needed to interpret these changes (Hoekman 2010,
Kordas et al. 2011, Vucic-Pestic et al. 2011). Communities
and ecosystems are complex systems, so there will undoubtedly be many uncertainties, contingencies and unexpected
consequences. Nevertheless, metabolic theory and the
empirical results reported here suggest how rates of biotic
interactions will increase as the climate warms. Being
able to predict the first-order effects of temperature on
rates of interactions will be an important tool in dealing
with the increasing magnitude of climate change.
EV-7
Acknowledgements – We thank Eva Robinson at the Univ. of New
Mexico, for assistance with initial coding and figures, members of
the Brown lab for useful feedback, and NSF Macrosystems Biology
Grant EF 1065836 for financial support.
­­­­References
Anderson-Teixeira, K. J. et al. 2008. Amplified temperature
dependence in ecosystems developing on the lava flows
of Mauna Loa, Hawai’i. – Proc. Natl Acad. Sci. USA 105:
228–233.
Angilletta, M. et al. 2010. Thermodynamic effects on organismal
performance: is hotter better? – Phys. Biochem. Zool. 83:
197–206.
Ayala, F. J. 1966. Reversal of dominance in competing species
of Drosophila. – Am. Nat. 100: 81–83.
Bartholomew, G. A. 1964. The roles of physiology and
behaviour in the maintenance of homeostasis in the desert
environment. – In: Hughes, G. M. (ed.), Homeostasis and
feedback mechanisms. Cambridge Univ. Press, pp. 7–29.
Brett, J. R. 1971. Energetic responses of salmon to temperature. A
study of some thermal relations in the physiology and
freshwater ecology of sockeye salmon (Oncorhynchus
nerkd). – Am. Zool. 11: 99–113.
Brown, J. H., and Sibly, R. M. 2012. The metabolic theory of
ecology and its central equation. – In: Sibly, R. M. et al. (eds),
Metabolic ecology: a scaling approach. Wiley, pp. 21–33.
Brown, J. H. et al. 2004. Toward a metabolic theory of ecology.
– Ecology 85: 1771–1789.
Clarke, A. and Fraser, K. P. P. 2004. Why does metabolism scale
with temperature? – Funct. Ecol. 18: 243–251.
Dell, A. I. et al. 2011. Systematic variation in the temperature
dependence of physiological and ecological traits. – Proc.
Natl Acad. Sci. USA 108: 10591–10596.
Dell, A. I. et al. 2014. Temperature dependence of trophic
interactions are driven by asymmetry of species responses and
foraging strategy. – J. Anim. Ecol. 82: doi: 10.1111/13652656.12065.
Deutsch, C. A. et al. 2008. Impacts of climate warming on
terrestrial ectotherms across latitude. – Proc. Natl Acad. Sci.
USA 105: 6668–6672.
Dunson, W. A. and Travis, J. 1991. The role of abiotic factors
in community organization. – Am. Nat. 138: 1067–1091.
Englund, G. et al. 2011. Temperature dependence of the functional
response. – Ecol. Lett. 14: 914–921.
Gillooly, J. F. et al. 2001. Effects of size and temperature on
metabolic rate. – Science 293: 2248–2251.
Hochachka, P. W. and Somero, G. N. 1984. Biochemical
adaptation: mechanism and process in physiological evolution.
– Princeton Univ. Press.
Hoekman, D. 2010. Turning up the heat: temperature influences
the relative importance of top–down and bottom–up
effects. – Ecology 91: 2819–2825.
Holling, C. S. 1959. Some characteristics of simple types of
predation and parasitism. – Can. Entomol. 91: 385–398.
Huey, R. B. and Kingsolver, J. G. 1989. Evolution of thermal
sensitivity of ectotherm performance. – Trends Ecol. Evol.
4: 131–135.
Izem, R. and Kingsolver, J. G. 2005. Variation in continuous
reaction norms: quantifying directions of biological interest.
– Am. Nat. 166: 277–289.
Supplementary material (available as Appendix oik.01199 at
www.oikosjournal.org/readers/appendix). Appendix 1.
EV-8
Kearney, M. et al. 2009. The potential for behavioral thermo­
regulation to buffer ‘cold-blooded’ animals against climate
warming. – Proc. Natl Acad. Sci. USA 106: 3835–3840.
Knies, J. L. and Kingsolver, J. G. 2010. Erroneous Arrhenius:
modified Arrhenius model best explains the temperature
dependence of ectotherm fitness. – Am. Nat. 176:
227–233.
Kordas R. L. et al. 2011. Community ecology in a warming
world: the influence of temperature on interspecific
interactions in marine systems. – J. Exp. Mar. Biol. Ecol. 400:
218–226.
Kruse, P. D. et al. 2008. Temperature and prey capture: opposite
relationships in two predator taxa. – Ecol. Entomol. 33:
305–312.
Lemoine, N. P. and Burkepile, D. E. 2012. Temperature-induced
mismatches between consumption and metabolism reduce
consumer fitness. – Ecology 93: 2483–2489.
Martin, T. L. and Huey, R. B. 2008. Why ‘suboptimal’ is
optimal: Jensen’s inequality and ectotherm thermal preferences.
– Am. Nat. 171: E102–E118.
Okie, J. G. 2012. Microorganisms. – In: Sibly, R. M. et al. (eds),
Metabolic ecology: a scaling approach. Wiley-Blackwell,
pp. 135–153.
Parmesan, C. and Yohe, G. 2003. A globally coherent fingerprint
of climate change impacts across natural systems. – Nature
421: 37–42.
Peters, R. H. 1983. The ecological implications of body size.
– Cambridge Univ. Press.
Pörtner, H. O. and Farrell, A. P. 2008. Physiology and climate
change. – Science 322: 690–692.
Price, C. A. et al. 2012. Testing the metabolic theory of ecology.
– Ecol. Lett.15: 1465–1474.
Rall, B. C. et al. 2012. Universal temperature and bodymass scaling of feeding rates. – Phil. Trans. R. Soc. B 367:
2923–2934.
Ratkowsky, D. A. et al. 2005. Unifying temperature effects on
the growth rate of bacteria and the stability of globular
proteins. – J. Theor. Biol. 233: 351–362.
Rodríguez-Castañeda, G. 2013. The world and its shades of
green: a meta-analysis on trophic cascades across temperature
and precipitation gradients. – Global Ecol. Biogeogr. 22:
118–130.
Savage, V. M. et al. 2004. Effects of body size and temperature
on population growth. – Am. Nat. 163: 429–441.
Sibly, R. M. et al. (eds) 2012. Metabolic ecology: a scaling approach.
– Wiley.
Smith, R. J. 2009. Use and misuse of the reduced major axis
for line fitting. – Am. J. Phys. Anthropol. 140: 476–486.
Somero, G. N. 2010. The physiology of climate change: how
potentials for acclimatization and genetic adaptation
will determine ‘winners’ and ‘losers’. – J. Exp. Biol. 213:
912–920.
Stevenson, R. D. 1985. The relative importance of behavioral and
physiological adjustments controlling body temperature in
terrestrial ectotherms. – Am. Nat. 126: 362–386.
Sunday, J. M. et al. 2011. Global analysis of thermal tolerance and
latitude in ectotherms. – Proc. R. Soc. B 278: 1823–1830.
Vucic-Pestic, O. et al. 2011. Warming up the system: higher
predator feeding rates but lower energetic efficiencies. – Global
Change Biol. 17: 1301–1310.
White, E. P. et al. 2012. Methodological tools. – In: Sibly, R. M.
et al. (eds), Metabolic ecology: a scaling approach. WileyBlackwell, pp. 9–20.