ArticlePDF Available

Abstract and Figures

Climate change has been shown to induce shifts in the timing of life-history events. As a result, interactions between species can become disrupted, with potentially detrimental effects. Predicting these consequences has proven challenging. We apply structured population models to a well-characterised great tit-caterpillar model system and identify thresholds of temporal asynchrony, beyond which the predator population will rapidly go extinct. Our model suggests that phenotypic plasticity in predator breeding timing initially maintains temporal synchrony in the face of environmental change. However, under projections of climate change, predator plasticity was insufficient to keep pace with prey phenology. Directional evolution then accelerated, but could not prevent mismatch. Once predator phenology lagged behind prey by more than 24 days, rapid extinction was inevitable, despite previously stable population dynamics. Our projections suggest that current population stability could be masking a route to population collapse, if high greenhouse gas emissions continue.
This content is subject to copyright. Terms and conditions apply.
LETTER Phenological asynchrony: a ticking time-bomb for seemingly
stable populations?
Emily G. Simmonds,
Ella F. Cole,
Ben C. Sheldon
Tim Coulson
Climate change has been shown to induce shifts in the timing of life-history events. As a result,
interactions between species can become disrupted, with potentially detrimental effects. Predicting
these consequences has proven challenging. We apply structured population models to a well-
characterised great tit-caterpillar model system and identify thresholds of temporal asynchrony,
beyond which the predator population will rapidly go extinct. Our model suggests that phenotypic
plasticity in predator breeding timing initially maintains temporal synchrony in the face of envi-
ronmental change. However, under projections of climate change, predator plasticity was insuffi-
cient to keep pace with prey phenology. Directional evolution then accelerated, but could not
prevent mismatch. Once predator phenology lagged behind prey by more than 24 days, rapid
extinction was inevitable, despite previously stable population dynamics. Our projections suggest
that current population stability could be masking a route to population collapse, if high green-
house gas emissions continue.
Asynchrony, climate change, evolution, extinction, integral project model, mismatch, phenology,
plasticity, predictions.
Ecology Letters (2020) 23: 1766–1775
Rapid climate change is altering biological systems across the
globe (IPCC, 2007) with changes in phenology being the most
frequently documented response (Menzel et al., 2006; Cleland
et al., 2007; Parmesan, 2007; Thackeray et al., 2016). These tem-
poral shifts have been observed across systems, taxa and life-
history events (Menzel et al., 2006; Cleland et al., 2007; Lehikoi-
nen and Sparks, 2010; Plard et al., 2012; Gallinat et al., 2015).
However, the observed phenological changes are frequently
non-uniform, resulting in asynchrony between interacting spe-
cies. Such trophic asynchronies are hypothesised to cause fitness
reductions in one or both of the interacting species (Cushing,
1969), leading to population declines. Yet neither asynchrony
nor population declines are ubiquitous (Menzel et al., 2006;
Singer and Parmesan, 2010; Reed et al., 2013b, a; Johansson
et al., 2015) and our understanding of the mechanisms linking
asynchrony to their consequences is rudimentary (Samplonius
et al., 2020). Buffering from negative population consequences,
is suggested to occur through density-dependent feedbacks
(Grøtan et al., 2009; Schmidt et al., 2015) or positive effects of
other environmental drivers, such as winter conditions (Perrins,
1965; Van Balen, 1980; Reed et al., 2013a). However, an empiri-
cal understanding of exactly how this is occurring in the wild is
limited (Miller-Rushing et al., 2010; Bennett et al., 2015;
Johansson et al., 2015).
Previous work exploring the role of phenological asyn-
chrony on population dynamics has largely been theoretical
(Gienapp et al., 2014; Johansson et al., 2015) or considers the
influence of phenology in isolation of other drivers (Chevin
et al., 2010; Vedder et al., 2013). Combining detailed observa-
tional data with appropriate predictive models can improve
our understanding of the causes and consequences of pheno-
logical mismatch (Miller-Rushing et al., 2010), although this
approach cannot ultimately determine directions of causality.
To be useful, these models must simultaneously incorporate
the influence of determinants of both fitness and phenological
change on dynamics; few studies achieve this (Plard et al.,
2014; Childs et al., 2016). We aim to fill this gap by applying
a new predictive model structure (Coulson et al., 2017; Sim-
monds et al., 2019b) to an exemplary long-term data set of
great tit (Parus major) phenology and population size. We
explore the impacts of phenology on demography, while
simultaneously projecting and investigating the mechanisms
behind the phenological change. This is in contrast to studies
that have focussed on the demographic impacts of phenology
separately to the mechanisms behind phenotypic change (e.g.
Reed et al., 2013b).
Like many insectivorous birds in temperate zones, breeding
great tits rely on a short period of insect abundance to feed
their young. How well individuals match the peak energetic
demands of their offspring with the peak abundance of cater-
pillars, particularly that of the winter moth (Operophtera bru-
mata), strongly influences reproductive success (Simmonds
et al., 2017). We used hatch date rather than lay date because
the timing of hatching is more closely linked to reproductive
Department of Zoology, Edward Grey Institute, University of Oxford, Oxford
Department of Mathematical Sciences and Centre for Biodiversity Dynamics,
Norwegian University of Science and Technology (NTNU), Trondheim, Norway
*Correspondence and present address:
Emily G. Simmonds, Department of Mathematical Sciences, Norwegian Uni-
versity of Science and Technology (NTNU), Trondheim, Norway.
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use,
distribution and reproduction in any medium, provided the original work is properly cited.
Ecology Letters, (2020) 23: 1766–1775 doi: 10.1111/ele.13603
success than the onset of laying (Simmonds et al., 2017).
Shifts in nest attentiveness and incubation duration can occur
in response to temperature cues during the laying period (Sim-
monds et al., 2017), whereas the period from hatch to fledging
is unaffected by temperature (Buse et al., 1999). Hatch date is
therefore a more reliable indicator of phenological synchrony
than lay date. We project both great tit (predator) egg hatch
timing and the timing of the peak abundance of winter moth
caterpillars (prey). Consequently, we simultaneously consider
how future phenological change will impact trophic interac-
tions, and how these changes are predicted to impact popula-
tion dynamics.
We explore the conditions under which asynchrony arises
and its impact on population dynamics, using an evolutionar-
ily explicit integral projection model (IPM) (Coulson et al.,
2017; Simmonds et al., 2019b). Our stochastic, density-depen-
dent IPM combines demographic functions, phenotypic plas-
ticity (change in phenotype for the same genotype in response
to environmental change), and adaptive evolution to project
population dynamics and phenological change to the end of
this century. We use projections of the future weather from
the UK Climate Projection 2009 estimates (UKCP09) (Met
Office, 2009; Murphy et al., 2009) under three scenarios of
anthropogenic greenhouse gas emissions; low (+1.1 to 2.9 °C
by 2099), medium (+1.7 to 4.4 °C by 2099) and high (+2.4 to
6.4 °C by 2099) greenhouse gas emissions (Murphy et al.,
2009), which arise from different future anthropogenic beha-
viour (Nakicenovic and Swart, 2000). To tease apart the influ-
ence of evolution and plasticity we decompose the phenotype
(hatch date) into a bivariate distribution of breeding values
and environmental components of the phenotype.
Model overview
Here, we used an evolutionarily explicit IPM (Coulson et al.,
2017; Simmonds et al., 2019b) formed from four fundamental
functions, two of which represent key demographic rates (sur-
vival and recruitment) and two of which represent trait
change (development and inheritance). The survival, recruit-
ment, development and inheritance functions have logistic,
exponential and Gaussian forms respectively. The survival
and recruitment functions determined viability and fertility
selection, the drivers of adaptive evolution. Each of these
functions was parameterised using regression models (see,
Simmonds et al., 2019b for more details) that included the
effects of spring and winter weather conditions, and food sup-
ply (beech mast quantity and the timing of the peak abun-
dance of winter moth caterpillars). Beech mast is a key winter
food source for the great tits and has a strong impact on
annual survival (Perdeck et al., 2000). The development func-
tion kept an individual’s breeding value constant as they aged,
while capturing phenotypic plasticity by allowing the environ-
mental component of the phenotype to change as a function
of spring temperature. The inheritance function captured (1)
genetic inheritance by assuming genes are passed on assuming
the infinitesimal model of quantitative genetics (Lande, 1975),
and (2) non-genetic inheritance via an association between
parental and offspring environmental components of the phe-
notype that is determined by the environment in the breeding
This model was previously cross validated (Simmonds et al.,
2019b) and was shown to capture hatch date dynamics and
trends in population size, but under-predict absolute popula-
tion size, therefore we are confident that our model is robust
to initial conditions and can capture the trait dynamics of this
Simplifying assumptions
As with any model, our IPM contains several simplifying
Treatment of immigration we include immigration as a
fixed proportion of the predicted females born in Wytham
woods, assuming that the proportion of immigrants will
remain constant into the future. If the portion of immigrants
were to shift under novel environmental conditions, we might
expect higher or lower population sizes to occur than our
model projects, but we cannot be certain of the direction of
this effect.
Inclusion of the prey species while we do project the
timing of the peak abundance of the prey species, we do not
currently include evolution of the prey species phenology or
changes in abundance. While it would be desirable to include
a full two species model incorporating the demography of the
winter moth caterpillar, the data to parameterise such a model
is not currently available.
Inclusion of interspecific interactions - we include the
temporal change in the winter moth caterpillar lifecycle, but
other processes such as competition or diet switching (as the
winter moth is not the sole food source for great tit chicks
(Nour et al., 1998; Pagani-N´
nez et al., 2011)) could alter the
carrying capacity of the great tit population further under cli-
mate change.
Technical model summary
The model tracked a bivariate trait distribution of Gand E
(breeding values and environmental components of the pheno-
type respectively). The breeding values relate to hatch date
only. The focal trait was hatch date of great tits Z, this was
the date on which the first egg in a nest hatched, as deter-
mined in the field (see parameterisation section for more
detail) and was a trait related to the mother of that nest. We
assume Z¼GþEand Gis fixed for life.
The equation for the model is given in eqns 1 and 2. Here
ZG,EðÞ¼¼GþEis the hatch date at t, E0is the environmen-
tal component of the hatch date at tþ1, NG,E,tðÞis the
bivariate distribution of the trait at time t,N(G0,E0,t+1) is
the bivariate distribution of the trait time tþ1. SG,E,θ0,tðÞ,
RG,E,θ0,tðÞ,DðE0jE,θ,t,aÞand HðG0,E0jG,E,θ,tÞare the sur-
vival, recruitment, development, and inheritance conditional
on the hatch date and weather at time t(θ). Development was
conditional on age (a), making the model age- and stage-
structured (we only considered a distinction between first-year
birds and those older than 1 year). The population size at
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
Letter Phenological asynchrony a ticking time-bomb? 1767
tþ1 was calculated by summing the number of first-year birds
(Na1) and the number of birds older than one year (Na2).
The survival and recruitment functions were conditional on
both the environment at time tand t1(θ0), due to lagged
effects of spring conditions immediately before the census.
The census was immediately post-breeding.
The timing of peak abundance of the winter moth caterpil-
lars was projected at each time step using a linear equa-
tion (eqn 3), where CaterpillarTiming =the timing of peak
abundance of caterpillars, β0=the intercept of a linear model,
βj=the coefficients of the effect of environmental drivers on
timing, Xj=the values of the environmental drivers at time t.
CaterpillarTiming ¼β0þ
For greater detail on the individual functional forms, please
see Supplementary Information or (Simmonds et al., 2019b).
The IPM was parameterised for the Wytham Woods popula-
tion of great tits in Simmonds et al. (2019b). The data for
parameterisation spanned 1961 to 2010 (50 years). We used
four additional years of observed data (i.e. 20112014) from
this population as a start point for all simulations in this
The data were built up of population census data and phe-
nological records collected under a standardised protocol
(Perrins, 1979). A detailed data description is present in the
Supporting Information from Simmonds et al. (2019b, http:// The focal pheno-
logical trait in our model was hatch date. Observed hatch
dates were estimated in the field during routine nest box visits.
A hatch date was assigned to a nest either by observing hatch-
ing of the first egg or ageing-hatched chicks based on size/
weight and estimating the hatch date retrospectively. Timing
of the winter moth caterpillar peak abundance was estimated
from data on final instar larvae pupation dates. The median
date was assumed to indicate peak abundance. Asynchrony
was calculated as the difference between great tit and caterpil-
lar timing.
Our model is single sex, therefore assuming identical demog-
raphy for males and females. Each of the functions were
parameterised using various forms of linear model in R (R
version 3.5.1 ‘Feather Spray’).
Cues for great tits and caterpillars
While the model underlying this paper was of the same form
as in Simmonds et al. (2019b), we implemented some key
changes to create a more realistic representation of the phe-
nology of the study system. We used different phenological
cues for our two phenological traits of interest. The phenolog-
ical cues for each species were identified using an absolute
sliding time window approach implemented in the package
climwin’ (Bailey and van de Pol, 2016; van de Pol et al.,
2016) in R. This method would not identify a true cue, but a
proxy for predictive purposes; the true cues may be more
complex and could even change over time (Buse et al., 1999;
Simmonds et al., 2019a). But for the purposes of this study,
we made use of one of the best tools we have currently avail-
able (Simmonds et al., 2019a). An exhaustive search of win-
dows was performed with a reference date of 20th May for
both species, searching for a maximum length window of
365 days, with all variables standardised to z-scores. Several
aggregate statistics (mean, minimum, maximum or slope of
change across the window) and temperature measures (daily
mean, daily minimum and daily maximum temperature) were
included in the search. The optimal windows for the two spe-
cies had the temperature measure of mean of the maximum
daily temperatures across the window, but the duration and
exact position of the windows differed; 22nd February to 20th
May for caterpillars (R
=0.87) and 4th March to 10th May
for great tits (R
These identified cues were used within the development
function and the function to predict caterpillar peak abun-
dance. All other spring temperature measures in other func-
tions remained the same as in Simmonds et al., (2019b).
Weather projections
Projections from the UKCP09 (Met Office, 2009; Murphy
et al., 2009), grid square 4500210, were used as the basis for
our climate projections. Projections of future climate and
weather are made using our best knowledge of the climate
system in conjunction with anthropogenic forcing. We con-
sidered three levels of greenhouse gas emissions, to represent
uncertainty in human decision-making processes. They map
onto the IPCC’s Special Report on Emissions Scenarios
(SRES) scenarios as A1FI, A1B and B1 (Nakicenovic and
Swart, 2000). To move from yearly climate projections to
daily weather projections, we used the MET office Weather
Generator tool (Met Office, 2009) to generate 1000 equally
plausible daily mean temperature, daily maximum tempera-
ture, and daily precipitation projections from 2015 to 2100.
These daily environmental records were generated in three
sets using the UKCP09 projections for 2011 to 2039 as the
basis for our 2015 to 2040 daily projections, UKCP09 pro-
jections for 2040 to 2069 as the basis for our 2041 to 2070
daily projections, and UKCP09 projections for 2070 to 2099
as the basis for our 2071 to 2100 daily projections. The 90-
year period was split into three sections to match the time
periods projections exist for, rather than assuming the envi-
ronmental conditions for the first three decades will persist
for the remainder of the 21st century. While taking climate
predictions in 30-year blocks leads to a stepped pattern
appearing in some figures (Fig. S2 and Fig. S4), this is not
the case for individual time series (see Fig. S2 final column).
The steps appear visually as there are shared characteristics
across the different time series, not because all individual
time series follow jumps. Our conclusions here are not
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
1768 E. G. Simmonds et al. Letter
impacted by whether changes in weather are continuous or
in discrete steps.
The daily predictions for each period were combined to gen-
erate 1000 equally plausible daily environmental predictions
stretching continuously from 2015 to 2100. Observed values
of the environmental drivers from 2011 to 2014 were added to
the beginning of all climate predictions in order to begin each
model simulation on known values.
From these daily predictions we calculated annual values of
the environmental drivers used in model parameterisation:
Spring temperature (great tit) mean maximum temper-
ature for the period 4th March to 10th May. This window of
temperature was identified as the optimum critical window in
which great tits respond to temperature.
Spring temperature (caterpillars) mean maximum tem-
perature for the period 22nd February to 20th May. This win-
dow of temperature was identified as the optimum critical
window in which caterpillars respond to temperature.
Spring temperature (survival, recruitment, inheritance)
mean of daily mean temperature for the period 1st March to
9th May, following Simmonds et al. (2019b).
Spring precipitation total precipitation from 1st April
to 31st May to cover the egg-laying and incubation period
and when young chicks are in the nest.
Winter temperature mean temperature from the winter
following the breeding attempt from (1st December to 28th/
29th February).
Winter precipitation total precipitation from 1st
December to 28th/29th February.
The model was simulated 1000 times for each emission scenario,
using an equally likely projection of future climate for the period
of 2011 to 2100. At each time step of the simulation we calcu-
lated the population size (resident breeding females) and the G
and Edistribution. Values for continuous environmental drivers
(spring temperature, spring precipitation, winter temperature,
winter precipitation) were chosen from either observed values or
weather projections. Beech mast index could not be predicted as
the exact drivers of mast years have not yet been identified and
therefore was selected semi-stochastically at each time step based
on observed frequency from 1961 to 2010. We did not want to
make assumptions into the drivers of this process. Two high
mast years could not be allocated concurrently and if four con-
secutive non-mast years occur, then the next year is a high mast
year (Matthews, 1955). Beech mast values were different for each
of the 1000 simulations.
The timing of caterpillar peak abundance was also projected
at each time step during model simulations, following eqn 3.
Statistical analysis of simulation results
Perturbation analysis
To determine the key drivers of population dynamics, we
assessed the sensitivity of population size to perturbations in
different explanatory drivers. Values of continuous scaled
environmental drivers and synchrony were altered by 4
standard deviations. For the simulation, values of the focal
driver were held at the perturbed value for the duration of the
50 time steps, all other environmental drivers were held at
their mean. Beech mast index was held at full mast to avoid
extinction generated from too infrequent masting years. It is
unrealistic that beech mast would be this frequent however,
by fixing the beech mast index this allowed attribution of the
change in population growth rate to the variable of interest.
The population size after 50 years was calculated from each
perturbed model run.
Analysis of extinction probability
To explore the relationship between the probability of extinc-
tion and phenological synchrony, we used a generalised linear
model (GLM) with population extinction (0 or 1) as a
response and the maximum annual mean asynchrony experi-
enced during the 90-year simulation as an explanatory vari-
able. Only positive synchrony values, those where the
predator lags behind their prey, were included because this is
expected to be the more likely direction under climate change
(and the most common direction in our simulations). Exclud-
ing negative synchrony values allowed us to focus on the pre-
dominant pattern rather including a mixture of asynchrony
processes. The GLM had binomial errors and a logistic link.
Simulations with environmental drivers held at 1961 to 2010
To isolate the influence of each environmental driver, four
extra sets of 1000 simulations were run, with one environmen-
tal driver held at 1961 to 2010 levels. Only the high emission
scenario projections were used. The value of the ‘held’ envi-
ronmental driver at each time step was pulled from a random
normal distribution with a mean and standard deviation
matching observed data from 1961 to 2010.
Analysis of evolutionary and plastic change
To assess whether the rate of evolution changes over the
course of the simulations we tested whether a segmented
regression performed significantly better than a simple linear
regression at capturing the relationship between mean genetic
component of the phenotype and time, using the ‘chngpt’
package (Fong et al., 2017) in R. We tested a maximum of
two segments, a single change point, and whether this was a
statistically significant improvement over a single segment (a
simple linear regression) with mean of the genetic component
of the phenotype as a response and year through simulation
as an explanatory variable. This was done for the high emis-
sions scenario only from 2015 to 2100. The first 5 years, based
on observed environmental values, were excluded.
To explore whether changes in the plastic component of
the phenotype shifted during the same period as evolution,
we statistically tested changes in variance of the mean plastic
component. We would expect limits of plasticity to be shown
by variance reductions. We calculated the variance in the
mean plastic component of the phenotype for the period
prior to the change point identified for the genetic compo-
nent, and the period after the change point. We then tested
the difference in the two variance measures for the 1000 high
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
Letter Phenological asynchrony a ticking time-bomb? 1769
emission scenario simulations using a linear model with
group (pre or post change point) and a pairing variable (to
indicate which simulation run each variance measure is from)
as explanatory variables.
How will temporal asynchrony change over the 21st century?
Maximum reproductive success for individual great tits breed-
ing in Wytham Woods is reached when eggs hatch around
13 days prior to caterpillar peak abundance (Simmonds et al.,
2017); indeed, from 1961 to 2010 the mean hatching date for
great tits was 12.35 days (SE =0.06) prior to the caterpillar
peak. We consider an asynchrony value of 13 days between
hatching and caterpillar peak to represent optimal synchrony,
and all results are presented relative to this. Despite consider-
able interannual fluctuation, mean asynchrony under the low
emissions scenario increased only slightly in our simulations,
by at most five days for most of this century (Fig. 1), due to
both closely matched cues (see Table S3) and an advance in
the hatch dates of great tits (Fig. 2). In contrast, under the
high emissions scenario, our model forecasted that by the end
of the century great tit eggs would be hatching up to two
weeks after the peak caterpillar abundance, causing severe
asynchrony. Under this scenario, shared cues alone were not
sufficient to maintain synchrony between the interacting spe-
cies beyond the next 50 years. Medium emissions produced an
intermediate level of asynchrony in comparison to the low and
high emission scenarios (Fig. 1).
Figure 1 Projected asynchrony between predator and prey, 20112100. Heatmap intensity indicates data density; the more data points at a location the
greener it appears. Blue areas indicate no data. The heatmap is generated from all 1000 stochastic projections for each emission scenario. The 10 white
lines are single stochastic projections of asynchrony. Zero indicates optimal hatch timing (13 days prior to prey peak abundance). The first 5 years of each
simulation are generated from observed environmental data.
Figure 2 Projections of the phenotype, and its environmental, and genetic
components for the medium emission scenario. Heatmap intensity
indicates phenotype data density; the number of data points at a location
is indicated by the intensity of yellow, blue areas indicate no data. The
heatmap is generated from all 1000 stochastic projections for the high
emission scenario. The 10 black lines are single stochastic projections of
mean hatch date. Plasticity and evolution colours are scaled so that
stronger colours indicate more data points and both are mean centred.
Mean hatch date is in days since 31st March, i.e. 1 =1st April.
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
1770 E. G. Simmonds et al. Letter
Why does temporal synchrony degrade?
Across all emissions scenarios we projected an advance in
predator timing over the 21st century (Fig. 2). However, this
advance was slower than the predicted advance in timing of
prey abundance (Fig. 1).
Fig. 2 shows that phenotypic plasticity was the dominant
driver of interannual variation in phenology (the median trend
in plasticity over 90 years for all scenarios =0.12 days per
year, 1st quartile =0.16, 3rd quartile =0.08). The trend in
annual change due to each component of the phenotype was
quantified using linear models, because the direction of
changes fluctuates in both evolution and plasticity. The effect
of microevolution, changes in gene frequency in a population,
was an order of magnitude smaller than the effect of plasticity
(median trend in evolution over 90 years for all scenarios =
0.005, 1st quartile =0.013, 3rd quartile =0.0003). The
predominant direction of evolution in our simulations was
towards earlier hatching.
The contribution of phenotypic plasticity and evolution
were not static across our simulations: the rate of change
from each component shifted during the 90-year simulation.
During the initial 30 years of simulations, projected spring
temperature conditions for both predator and prey cues were
similar to the average conditions from the observed data (Fig.
S1). This led to asynchrony remaining at observed levels and
close to optimal during this period (Fig. 1) as a result of plas-
tic changes to the phenotype (Fig. 2). However, as the simula-
tions progressed, projected spring temperatures increased in
both the predator and prey cue windows (the increase in tem-
perature was almost identical for both species, Table S3). The
increase in temperatures led to a divergence in timing, as the
prey advanced more rapidly, causing phenological asynchrony
(Fig. 1). The increase in asynchrony coincided with an
increase in the rate of evolution (Fig. 2), due to negative
effects of asynchrony on survival and recruitment.
The increase in the rate of evolutionary change at approxi-
mately year 2072 (median, interquartile range 2065 to 2074)
was statistically significant in over 60% of high emissions simu-
lations (see Fig. S2a) and coincided with the most extreme pro-
jected weather changes (Fig. S1). At the same point, we saw a
statistically significant reduction in the variance of plasticity
(difference in mean variance =0.18, SE =0.006, Fig. S2b).
What are the population level consequences of phenological
Thresholds of extinction were identified in a perturbation
analysis (Fig. 3a). When prey timing was altered to be on
average, seven days earlier (positive asynchrony) or 11 days
later (negative asynchrony) than predator requirement, the
predator population went extinct. These thresholds exist due
to a quadratic effect of asynchrony on survival and reproduc-
tive success (shown in Fig. 3a), which was identified during
the parameterisation of the demographic functions (Simmonds
et al., 2019b).
The most extreme asynchrony values in our simulations
were associated with an elevated probability of extinction.
Probability of extinction, as quantified by a logistic regression,
showed a step change at maximum positive asynchrony values
of approximately 21 days from optimum (Fig. 3b) and all sim-
ulations that experienced an annual mean asynchrony value
of greater than 24 days, at any point in the 90-year simula-
tion, went extinct. 17% of populations went extinct in the
high emissions scenario simulations (0.5% under low emis-
sions and 3.8% under medium emissions).
Climate change drivers of population dynamics
Under all emissions scenarios, we projected an initial popula-
tion increase relative to the observed data (Fig. 4). This is a
result of improved winter conditions and spring precipitation
(Fig. S3), which both influence survival and recruitment (Sim-
monds et al., 2019b). The majority of populations remained
stable, fluctuating within the range of observed population
size, under moderate amounts of environmental change. But
the number of population extinctions increased rapidly as
greenhouse gas emissions increased. Fig. 4 shows a hotspot of
extinctions from 2075 onwards under all emissions scenarios,
but is most extreme in the high emissions scenario with a sub-
stantial number of simulations going extinct or close to
In this study, we show how phenological synchrony and pop-
ulation dynamics might change over time for a simple model
system of a predator (the great tit) and its primary prey dur-
ing the breeding season (the winter moth caterpillar). We
identify potential mechanisms underlying the lack of observed
population response to phenological asynchrony (Reed et al.,
2013b, a; Johansson et al., 2015; Samplonius et al., 2020), in
the form of thresholds of asynchrony, within which the popu-
lation is buffered from negative impacts of a loss of syn-
chrony, but beyond which the population declines rapidly.
Our results show several key patterns. The first is a mainte-
nance of temporal synchrony and stable population dynamics
under lower rates of climatic change (early in this century and
under lower emissions). During this period, the population
appears buffered from environmental change through a com-
bination behavioural changes in hatch timing (phenotypic
plasticity) and positive environmental effects on demography
(survival increases due to warmer winters and wetter springs,
while wetter winters and the wetter springs both increase
reproductive success). While some loss of synchrony occurs,
this is not sufficient to create a negative impact on population
dynamics. The second pattern is one of slowed mismatch
through a combination of plasticity and directional evolution-
ary change (see Fig. 1 and Fig. 2). Our simulations revealed
that phenotypic plasticity is the dominant driver of interan-
nual variation in phenology supporting previous work in wild
great tit populations (Gienapp et al., 2006; Caro et al., 2013;
Charmantier and Gienapp, 2014). However, evolution had a
consistent directional contribution to the phenotype (Gienapp
et al., 2006). While microevolution does help slow the occur-
rence of asynchrony, it is not sufficient to maintain synchrony
across the 90-year simulation (Ramakers et al., 2018). Addi-
tionally, plasticity becomes dampened during the very end of
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
Letter Phenological asynchrony a ticking time-bomb? 1771
the century (Fig. 2 and Fig. S2), suggesting that current levels
of plasticity may not be sufficient to keep pace with consistent
directional change (Visser et al., 2004; Visser, 2008; Thackeray
et al., 2010; Ramakers et al., 2018).
The third pattern is a pronounced loss of synchrony when
climate change increases during the second half of the 21st
century under a high emissions scenario. The predicted loss of
synchrony arises from differences in the response of predator
and prey species to temperature changes (Gienapp et al.,
2014). Despite the temperature cues for both species being
composed of the same aggregate statistic, covering a similar
temporal window, and being projected to change at a similar
Figure 3 Perturbation analysis and extinction probability against maximum annual asynchrony. (a) Population size after 50 years of simulation against the
amount of change in the explanatory driver that was perturbed. All environmental drivers are scaled to be z-scores, (b) Extinction probability is calculated
based on a logistic regression (see Table S4 for full results). The black line indicates the fitted logistic regression line. Points indicate individual simulations
across all emission scenarios (n=3000). Zero indicates optimal hatch timing (13 days prior to prey peak abundance), negative changes to synchrony
indicate prey abundance is earlier than expected.
Figure 4 Projected population size of great tits from 2011 to 2100, for the low and high emission scenarios. Heatmap intensity indicates data density, the more
data points the higher the intensity of red. Black areas indicate no data. The heatmap is generated from all 1000 single stochastic projections for each emission
scenario. The 10 white lines are single stochastic projections of future population size. All simulations begin by projecting population size from 2011 to 2015
using observed environmental conditions. The red lines at the top of the plot indicate extinction of a population. Horizontal solid and dashed white lines
indicate the mean, the minimum and the maximum population size of females born in Wytham woods, respectively, for the 1961 to 2010 period.
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
1772 E. G. Simmonds et al. Letter
rate (see Table S3) asynchrony still occurred. This pattern
arises because a change of 1 °C to the cue for each species
produces a different amount of phenological change. We esti-
mated that great tits advance 4.35 days for every 1 °C change
to their cue (SE =0.5, on original scale), whereas caterpillars
advance 6.38 days for every 1 °C change to their cue (SE =
0.003, on original scale). Even though the two cues are not
directly comparable, if they are predicted to change similarly
under climate change, then the prey species will advance more
rapidly and asynchrony will result. The divergence predicted
from our model supports previous theoretical work suggesting
resource reaction norms have evolved to be steeper than con-
sumers (Gienapp et al., 2014). Spring temperatures as a whole
had the largest effect on population dynamics and hatch date
(Fig. S1).
The overall consequence of the increased asynchrony
towards the end of this century is a rapid increase in extinc-
tion probability of the simulated predator population. This
is most extreme under the high emissions scenario. However,
extinctions still occurred but at a low level under both low
and medium emissions. These population declines occurred
very rapidly with asynchrony acting as a threshold. Any
population that experienced positive asynchrony of greater
than 24 days went extinct. This is driven again by the quad-
ratic effect of asynchrony on the key demographic rates in
our model. As asynchrony increases, so does its effect on
survival and recruitment, eventually leading to population
Buffering from positive environmental effects on vital rates
and an initial lack of impact from asynchrony can mask
increasing divergence between predator and prey timing, cre-
ating the appearance of a stable population. Despite apparent
stability, populations may undergo rapid declines when differ-
ences in environmental sensitivity between predator and prey
lead to predator plasticity no longer being adaptive. These
findings have wide reaching consequences. It is likely that
many wild populations have not yet reached these limits to
adaptive plasticity or asynchrony extreme enough to produce
a population level signal. These thresholds suggest that cur-
rent population stability due to buffering should not be
assumed to continue. If the thresholds of asynchrony are
exceeded as a consequence of future climate change, rapid
population declines and extinctions could occur, even in seem-
ingly stable populations.
Our identified thresholds suggest caution, but it is not cer-
tain that these negative outcomes will be realised for this
population, or others. During cross validation, our model
was shown to under predict population size (Simmonds
et al., 2019b). Therefore, we might expect it to take longer
before the negative effects projected here are seen. However,
it should also be noted that the projected population sizes in
these simulations fluctuate largely within the observed range
of resident breeding females, suggesting that systematic
under prediction is not as common as in the previous study.
There are also several compensatory factors that could not
be included in our current model that could mitigate the
worst case scenario, such as increases in immigration, switch-
ing of prey species, changes to winter food availability (such
as beech mast (Bogdziewicz et al., 2020)), reduced
competition, or evolution of the reaction norm to better
keep pace with the prey (although this latter possibility
seems unlikely (Ramakers et al., 2018)). However, the exact
form any of these processes might take is not yet known,
they could either buffer the local population from declines
or exacerbate them. For example, more complex systems
with many species within each trophic level may be more
resilient. However, it is likely that the same patterns we
show here could still emerge, regardless of system complex-
ity, if lower trophic levels still respond more rapidly than
higher trophic levels (Gienapp et al., 2014). The abundance
of the prey species could also be altered by environmental
change and can itself lead to phenological change (R´
et al., 2003). If the peak or timespan of abundance of a prey
species increased, this could lead to positive consequences
for the predator reproductive success (Reneerkens et al.,
2016; Seress et al., 2018). Experimental studies would ulti-
mately be required to determine any causal links. Our cur-
rent model gives a first indication of how the Wytham
Woods great tits might respond in isolation, with a fixed
proportion of immigrants to the population, and a single
preferred food source. This model framework could easily be
extended to include the demography of multiple species or
incorporate immigration. Including the demography, abun-
dance, and evolution of the prey species would be an excit-
ing future direction for this framework in systems where
individual level data on prey survival and reproductive suc-
cess are available. The biggest mitigation factor to prevent
population declines would be to limit the amount of climate
change experienced. Under projections of low and medium
greenhouse gas emissions, population stability is maintained
for the majority of simulated populations. Therefore, if emis-
sions can be kept lower, the chances of survival for the pop-
ulation and the chance of micro-evolutionary rescue are
greatly increased.
We are grateful to all of the Wytham fieldworkers who col-
lected population census data on the Wytham Woods great
tits over the years and to all others involved in this project.
This work was supported by NERC grant NE/K006274/1to
Ben C. Sheldon.
EGS led the design, analyses and writing of this manuscript,
with substantial contribution from TC. EGS, EFC, and BCS
have all contributed to data collection and management. EFC
and BCS provided feedback on the methodological
approaches, results of the analysis and writing. All authors
gave final approval for publication.
The code used to parameterise the model used in this manu-
script is available on the Github repository (https://github.
com/emilygsimmonds/Evolutionarily_Explicit_IPM) and Zen-
odo (10.5281/zenodo.4016389).
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
Letter Phenological asynchrony a ticking time-bomb? 1773
The peer review history for this article is available at https://
The data used to parameterise the model used in this manuscript
are available on the Github repository (
lygsimmonds/Evolutionarily_Explicit_IPM) and Zenodo (10.
5281/zenodo.4016389). The climate data and projections are
used free of charge under the following licence: ©Crown Copy-
right 2009. The UK Climate Projections (UKCP09) have been
made available by the Department for Environment, Food and
Rural Affairs (Defra) and the Department of Climate Change
(DECC) under licence from the Met Office, UK Climate
Impacts Programme, British Atmospheric Data Centre, New-
castle University, University of East Anglia, Environment
Agency, Tyndall Centre and Proudman Oceanographic Labora-
tory. These organisations give no warranties, express or implied,
as to the accuracy of the UKCP09 and do not accept any liabil-
ity for loss or damage, which may arise from reliance upon the
UKCP09 and any use of the UKCP09 is undertaken entirely at
the user’s risk.
Bailey, L.D. & van de Pol, M. (2016). Climwin: An R toolbox for climate
window analysis. PLoS One, 11, e0167980.
Van Balen, J.H. (1980). Population fluctuations of the Great Tit and
feeding conditions in winter. Ardea, 68, 143164.
Bennett, N.L., Severns, P.M., Parmesan, C. & Singer, M.C. (2015).
Geographic mosaics of phenology, host preference, adult size and
microhabitat choice predict butterfly resilience to climate warming.
Oikos, 124, 4153.
Bogdziewicz, M., Kelly, D., Thomas, P.A., Lageard, J.G.A. & Hacket-
Pain, A. (2020). Climate warming disrupts mast seeding and its fitness
benefits in European beech. Nat. Plants,6,8894.
Buse, A., Dury, S.J., Woodburn, R.J.W., Perrins, C.M. & Good, J.E.G.
(1999). Effects of elevated temperature on multi-species interactions: The
case of pedunculate oak, winter moth and tits. Funct. Ecol., 13, 7482.
Caro, S.P., Schaper, S.V., Dawson, A., Sharp, P.J., Gienapp, P. & Visser,
M.E. (2013). Is microevolution the only emergency exit in a warming
world? Temperature influences egg laying but not its underlying
mechanisms in great tits. Gen. Comp. Endocrinol., 190, 164169.
Charmantier, A. & Gienapp, P. (2014). Climate change and timing of
avian breeding and migration: Evolutionary versus plastic changes.
Evol. Appl.,7,1528.
Chevin, L.-M., Lande, R. & Mace, G.M. (2010). Adaptation, plasticity,
and extinction in a changing environment: Towards a predictive theory.
PLoS Biol., 8, e1000357.
Childs, D.Z., Sheldon, B.C. & Rees, M. (2016). The evolution of labile
traits in sex- and age-structured populations. J. Anim. Ecol., 85, 329342.
Cleland, E.E., Chuine, I., Menzel, A., Mooney, H.A. & Schwartz, M.D.
(2007). Shifting plant phenology in response to global change. Trends
Ecol. Evol., 22, 357365.
Coulson, T., Kendall, B.E., Barthold, J., Plard, F., Schindler, S., Ozgul,
A. et al. (2017). Modeling adaptive and non-adaptive responses of
populations to environmental change. Am. Nat., 190, 313336.
Cushing, D.H. (1969). The regularity of the spawning season of some
fishes. J. du Cons. Int. pour l’Exploration la Mer, 33, 8197.
Fong, Y., Huang, Y., Gilbert, P.B. & Permar, S.R. (2017). chngpt:
Threshold regression model estimation and inference. BMC
Bioinformatics, 18, 17.
Gallinat, A.S., Primack, R.B. & Wagner, D.L. (2015). Autumn, the neglected
season in climate change research. Trends Ecol. Evol., 30, 169176.
Gienapp, P., Postma, E. & Visser, M.E. (2006). Why breeding time has
not responded to selection for earlier breeding in a songbird
population. Evolution (N., Y), 60, 23812388.
Gienapp, P., Reed, T. & Visser, M.E. (2014). Why climate change will
invariably alter selection pressures on phenology. Proc. R. Soc. B Biol.
Sci., 281, 18.
Grøtan, V., Saether, B.-E., Engen, S., Van Balen, J.H., Perdeck, A.C. &
Visser, M.E. (2009). Spatial and temporal variation in the relative
contribution of density dependence, climate variation and migration to
fluctuations in the size of great tit populations. J. Anim. Ecol., 78, 447459.
IPCC (2007). The Physical Science Basis: Contribution of Working Group
1 to the Fourth Assessment Report of the IPCC. Cambridge University
Johansson, J., Kristensen, N.P., Nilsson, J.-
A. & Jonz´
en, N. (2015). The
eco-evolutionary consequences of interspecific phenological asynchrony
- a theoretical perspective. Oikos, 124, 102112.
Lande, R. (1975). The maintenance of genetic variability by mutation in a
polygenic character with linked loci. Genet. Res. (Camb), 26, 221235.
Lehikoinen, E. & Sparks, T.H. (2010). Changes in migration. In: Effects
of Climate Change on Birds (eds Møller, A.P., Fiedler, W. & Berthold,
P.). Oxford University Press, Oxford, pp. 89112.
Matthews, J.D. (1955). The influence of weather on the frequency of
beech mast years in England. Forestry, 28, 107116.
Menzel, A., Sparks, T.H., Estrella, N., Koch, E., Aasa, A., Ahas, R.
et al. (2006). European phenological response to climate change
matches the warming pattern. Glob. Chang. Biol., 12, 19691976.
Met Office (2009). UKCP09 gridded observation datasets. Available at: Last accessed.
Miller-Rushing, A.J., Hoye, T.T., Inouye, D.W., Post, E., Høye, T.T.,
Inouye, D.W. et al. (2010). The effects of phenological mismatches on
demography. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 365, 31773186.
Murphy, J., Sexton, D., Jenkins, G., Boorman, P., Booth, B., Brown, K.
et al. (2009). UK Climate Projections Science Report: Climate change
Nakicenovic, N. & Swart, R. (2000). Emissions scenarios - special report
of the Intergovernmental Panel on Climate Change. United Kingdom.
Nour, N., Currie, D., Matthysen, E., Van Damme, R. & Dhont, A.A. (1998).
Effects of habitat fragmentation on provisioning rates, diet and breeding
success in two species of tit (great tit and blue tit). Oecologia, 114, 522530.
nez, E., Ruiz, I., Quesada, J., Senar, J.C. & Negro, J.J. (2011). The
diet of Great Tit Parus major nestlings in a Mediterranean Iberian forest:
The important role of spiders. Anim. Biodivers. Conserv., 34, 355361.
Parmesan, C. (2007). Influences of species, latitudes and methodologies on
estimates of phenological response to global warming. Glob. Chang.
Biol., 13, 18601872.
Perdeck, A.C., Visser, M.E. & Van Balen, J.H. (2000). Great tit Parus
major survival and the beech crop cycle. Ardea, 88, 99108.
Perrins, C.M. (1965). Population fluctuations and clutch-size in the great
tit. J. Anim. Ecol., 34, 601647.
Perrins, C.M. (1979). British tits. Collins, London.
Plard, F., Bonenfant, C., Delorme, D. & Gaillard, J.-M. (2012). Modeling
reproductive trajectories of roe deer females: Fixed or dynamic
heterogeneity? Theor. Popul. Biol., 82, 317328.
Plard, F., Gaillard, J.-M., Coulson, T., Hewison, A.J.M., Delorme, D.,
Warnant, C. et al. (2014). Mismatch between birth dates and vegetation
phenology slows the demography of roe deer. PLoS Biol., 12, 18.
van de Pol, M., Bailey, L.D., McLean, N., Rijsdijk, L., Lawson, C.R. &
Brouwer, L. (2016). Identifying the best climatic predictors in ecology
and evolution. Methods Ecol. Evol., 7, 12461257.
Ramakers, J.J.C., Gienapp, P. & Visser, M.E. (2018). Phenological
mismatch drives selection on elevation, but not on slope, of breeding
time plasticity in a wild songbird. Evolution (N. Y), 73, 175187.
eale, D., McAdam, A.G., Boutin, S. & Berteaux, D. (2003). Genetic and
plastic responses of a northern mammal to climate change. Proc. Biol.
Sci., 270, 591596.
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
1774 E. G. Simmonds et al. Letter
Reed, T., Grøtan, V., Jenouvrier, S., Sæther, B.-E. & Visser, M.E.
(2013a). Population growth in a wild bird is buffered against
phenological mismatch. Science, 80-. ), 340, 488491.
Reed, T., Jenouvrier, S. & Visser, M.E. (2013b). Phenological mismatch
strongly affects individual fitness but not population demography in a
woodland passerine. J. Anim. Ecol., 82, 131144.
Reneerkens, J., Schmidt, N.M., Gilg, O., Hansen, J., Hansen, L.H.,
Moreau, J. et al. (2016). Effects of food abundance and early clutch
predation on reproductive timing in a high Arctic shorebird exposed
to advancements in arthropod abundance. Ecol. Evol., 6, 73757386.
Samplonius, J.M., Atkinson, A., Hassall, C., Keogan, K., Thackeray,
S.J., Assmann, J.J. et al (2020). Systematic weaknesses in the evidence
base for temperature-mediated phenological mismatch. Nat. Ecol. Evol.
Schmidt, K.A., Johansson, J., Kristensen, N.P., Massol, F. & Jonz´
en, N.
(2015). Consequences of information use in breeding habitat selection
on the evolution of settlement time. Oikos, 124, 6980.
Seress, G., Hammer, T., B´
okony, V., Vincze, E., Preiszner, B., Pipoly, I.
et al. (2018). Impact of urbanization on abundance and phenology of
caterpillars and consequences for breeding in an insectivorous bird.
Ecol. Appl., 28, 11431156.
Simmonds, E.G., Cole, E.F. & Sheldon, B.C. (2019a). Cue identification
in phenology: A case study of the predictive performance of current
statistical tools. J. Anim. Ecol., 88, 14281440.
Simmonds, E.G., Cole, E.F., Sheldon, B.C. & Coulson, T. (2019b).
Testing the effect of quantitative genetic inheritance in structured
models on projections of population dynamics. Oikos, 129(4),
Simmonds, E.G., Sheldon, B.C.C., Coulson, T. & Cole, E.F. (2017).
Incubation behaviour adjustments, driven by ambient temperature
variation, improve synchrony between hatch dates and caterpillar peak
in a wild bird population. Ecol. Evol., 7, 94159425.
Singer, M.C. & Parmesan, C. (2010). Phenological asynchrony between
herbivorous insects and their hosts: signal of climate change or pre-
existing adaptive strategy? Philos. Trans. R. Soc. B Biol. Sci., 365,
Thackeray, S.J., Henrys, P.A., Hemming, D., Bell, J.R., Botham, M.S.,
Burthe, S. et al. (2016). Phenological sensitivity to climate across taxa
and trophic levels. Nature, 535, 241245.
Thackeray, S.J., Sparks, T.H., Frederiksen, M., Burthe, S., Bacon, P.J.,
Bell, J.R. et al. (2010). Trophic level asynchrony in rates of
phenological change for marine, freshwater and terrestrial
environments. Glob. Chang. Biol., 16, 33043313.
Vedder, O., Bouwhuis, S. & Sheldon, B.C. (2013). Quantitative
assessment of the importance of phenotypic plasticity in adaptation to
climate change in wild bird populations. PLoS Biol., 11, e1001605.
Visser, M.E. (2008). Keeping up with a warming world; assessing the rate of
adaptation to climate change. Proc. R. Soc. B Biol. Sci., 275, 649659.
Visser, M.E., Both, C. & Lambrechts, M. (2004). Global climate change
leads to mistimed avian reproduction. Adv. Ecol. Res., 2504, 89110.
Additional supporting information may be found online in
the Supporting Information section at the end of the article.
Editor, Elsa Cleland
Manuscript received 11 May 2020
First decision made 29 June 2020
Manuscript accepted 11 August 2020
©2020 The Authors. Ecology Letters published by John Wiley & Sons Ltd.
Letter Phenological asynchrony a ticking time-bomb? 1775
... Using this approach, we identified heretofore undetected individual and population level fitness effects of mismatches in the Alaskan breeding population of Hudsonian godwits (see Senner et al. 2017). Our study joins the growing literature suggesting that quantifying mismatches in terms of resource availability and consumer energetic demands can provide important nuance about the risks mismatches pose to consumer populations (Simmonds et al. 2020;Takimoto and Sato 2020). ...
... The need to accurately identify mismatches is made most clear by the accumulating evidence for variable and non-linear responses by consumer populations to mismatches Both 2005, Phillimore et al. 2016). So called 'tipping points' -thresholds past which an effect abruptly changes -buffer consumer populations from the negative impacts of moderate mismatches and may contribute to the lack of consistent responses to mismatches across consumer populations (Simmonds et al. 2020). In this population of godwits, we found that greater population level mismatches consistently drove poorer fledging success, but that there may be thresholds past which the effects are most severe. ...
Climate change has caused shifts in seasonally recurring biological events leading to the temporal decoupling of consumer‐resource pairs – i.e., phenological mismatching. Although mismatches often affect individual fitness, they do not invariably scale up to affect populations, making it difficult to assess the risk they pose. Individual variation may contribute to this inconsistency, with changes in resource availability and consumer needs leading mismatches to have different outcomes over time. Nevertheless, most models estimate a consumer’s match from a single timepoint, potentially obscuring when mismatches matter to populations. We analyzed how the effects of mismatches varied over time by studying precocial Hudsonian godwit (Limosa haemastica) chicks and their invertebrate prey from 2009 to 2019. We developed individual and population level models to determine how age‐specific variation affect the relationship between godwits and resource availability. We found that periods with abundant resources led to higher growth and survival of godwit chicks, but also that chick survival was increasingly related to the availability of larger prey as chicks aged. At the population level, estimates of mismatches using age‐structured consumer demand explained more variation in annual godwit fledging rates than more commonly used alternatives. Our study suggests that modeling the effects of mismatches as the disrupted interaction between dynamic consumer needs and resource availability clarifies when mismatches matter to both individuals and populations.
... Great Tits, Parus major, at Whytham Woods; Simmonds et al., 2020), anticipated population declines from such phenological mismatches have not been documented, although models suggest this may be possible in the future (Simmonds et al., 2020). For migratory birds, phenological mismatches can be compounded by the need for food to fuel migration along the migratory route. ...
... Great Tits, Parus major, at Whytham Woods; Simmonds et al., 2020), anticipated population declines from such phenological mismatches have not been documented, although models suggest this may be possible in the future (Simmonds et al., 2020). For migratory birds, phenological mismatches can be compounded by the need for food to fuel migration along the migratory route. ...
1. Global climate change impacts species and ecosystems in potentially harmful ways. For migratory bird species, earlier spring warm-up could lead to a mis-match between nesting activities and food availability. CO 2 provides a useful proxy for temperature and an environmental indicator of climate change when temperature data are not available for an entire time series. 2. Our objectives were to (a) examine nesting phenology over time; (b) determine how nesting phenology relates to changes in atmospheric CO 2 concentration; and (c) demonstrate the usefulness of historical museum collections combined with modern observations for trend analyses. 3. We assessed changes in nesting dates of 72 bird species in the Upper Midwest of the United States by comparing contemporary lay dates with those obtained from archived, historical museum nest records over a 143-year period (1872-2015). 4. Species-specific changes in lay date per one unit change in the CO 2 residual ranged from −0.75 (95% CI: −1.57 to −0.10) to 0.45 (95% CI: −0.29 to 1.43). Overall, lay dates advanced ~10 days over the 143-year period. Resident, short-distance migrants and long-distance migrants lay dates advanced by ~15, 18 and 16 days on average respectively. Twenty-four species (33.3%) significantly advanced , one (1.4%) significantly delayed and we failed to detect an advance or delay in lay date for 47 species (65.3%). Overall mean advance in first lay date (for the species that have significantly advanced laying date) was 25.1 days (min: 10.7, max: 49.9). 5. Our study highlights the scientific importance of both data gathering and ar-chiving through time to understand phenological change. The detailed archived information reported by egg collectors provide the early data of our study. As with studies of eggshell thinning and pesticide exposure, our use of these data illustrates another scientific utility of egg collections that these pioneer naturalists never imagined. As museums archive historical data, these locations are also ideal candidates to store contemporary field data as it is collected. Together, such information will provide the ability to track, understand and perhaps predict responses to human-driven environmental change.
... The alignment and/or mismatch of phenological adaptation among interacting species indeed affect population dynamics (Hunter et al., 1997;McLaughlin et al., 2002;Hellmann et al., 2003;Deacy et al., 2017;Simmonds et al., 2020). Because of the species-specific phenological response to environmental changes, the interspecific variation in phenological shifts may disrupt species interactions, particularly when these interactions strongly depend on each other (Visser and Both, 2005;Donnelly et al., 2011;Forrest, 2015). ...
Full-text available
In recent years, phenology – the seasonal timing of biological life cycles – has received increasing attention as climate change threatens to shift phenology. Phenology is crucial to the life cycle of organisms and their interactions with intimate partner species; hence, phenology has important fitness consequences suggesting that phenology can change through adaptive processes caused by species interaction. However, to date, there is limited understanding of how phenological adaptation occurs among interacting species and consequently affects ecological population dynamics. In this study, a phenological predator–prey co-adaptation model was evaluated to determine how adaptive phenological changes occur in prey and predator and how phenological coadaptation affects their coexistence. Population fluctuations tend to decrease and become stabilized when adaptation occurs rapidly. Furthermore, when adaptation is slow, predator–prey dynamics can be stabilized or destabilized depending on the initial difference in phenological timing between species. These results suggest that phenology shaped by slow coevolution can shift with changes in activity timing caused by environmental changes and simultaneously alter the stability of predator–prey dynamics. In contrast, phenology caused by rapid adaptation, such as phenotypic plasticity, may be robust to environmental change and maintain the stability of predator–prey dynamics. Understanding the types of adaptative processes that shape species phenologies may be crucial for predicting the ecological effects of climate change.
... We found mean timing was most sensitive to temperatures from early March to mid-April (ordinal dates 65-106, 5 th March -15 th April in non-leap year, Fig. 3a), height was most sensitive to temperatures later in the spring (100-141, 9 th April -20th May, Fig. 3a) and width to temperatures that spanned the spring (58-155, 27 th February -3rd June, Fig. 3a). For mean timing this period is similar to that identified as important in other European studies 26,36,37 , whereas for width our time window is broader than identified in Visser et al 26 , though we note a high degree of uncertainty in the position of the window for this phenological parameter (Fig. S1). We used the most sensitive windows (i.e. ...
Advances in spring phenology are among the clearest biological responses to climate warming. In the ephemeral temperate deciduous forest food webs, at the vanguard of research on temperature’s effect on trophic interactions, most work has focused on the average timing of phenological events. In comparison, effects of temperature on the abundance of individuals and their seasonal spread is understudied, despite the potential for profound impacts on trophic interactions. Here we use a new method to show that for the guild of forest caterpillars, warmer spring conditions not only advance the timing of the phenological distribution of abundance by -4.96 days oC-1, but also increase its height by 34% oC-1. This increase in the maximum density of caterpillars with rising temperatures is likely to have major implications for both herbivory pressure and the resources available to secondary consumers.
... The population-and community-level consequences of changes in mismatch depend on the specific consumerresource system or food web. For example, an increase of the phenological mismatch may limit the consumer's food intake, which affects its biomass, its life cycle and potentially even its persistence (Simmonds et al. 2020). The consumer may also switch resources (Stålhandske et al. 2016). ...
Full-text available
Anthropogenic climate change increasingly affects species phenology. Because trophic interactions often occur at specific phenological stages, changes in one species' phenology may affect others through phenological mismatch. When a consumer and a resource both exhibit a seasonal resting period, the synchrony of the end of their respective resting periods is fundamental for the persistence of their interaction. Since the consumer and its resource may react differently to changes in temperature regime, the synchrony between them could be altered. We investigate potential effects of climate change on species' synchrony. We propose a general model that determines the duration of the resting period according to temperature, and its effects on the mismatch between phenological stages of two interacting species. We illustrate our approach using the spruce budworm–balsam fir system in eastern Canada. We find that an increase in temperature advances the end of the resting period. However, the effects of a warm or cold spell during the resting period strongly vary according to the timing and the duration of the spell. Depending on how a consumer and its resource react to the same temperature shift, the mismatch between them may increase or decrease. The spruce budworm–balsam fir model predicts that an increase in temperature may increase the mismatch between the insect and the tree in southern sites, but may increase the synchrony in northern sites. This modelling approach is of prime importance to investigate potential effects of climate change on consumer–resource systems.
... For example, since 1896, spring temperatures have increased by about 1.12°C in the contiguous United States of America, and snowmelt has advanced by 14.6 days in parts of Greenland over ten years prior to 2007 (Høye et al., 2007;NCEI, 2016). Furthermore, differential responses across species within the same community indicate that the temporal coordination of species interactions is changing, which raises concerns that interacting species are becoming temporally "mismatched" (Both et al., 2009;Farzan & Yang, 2018;Kharouba et al., 2018;Renner & Zohner, 2018;Simmonds et al., 2020;Visser & Both, 2005). ...
Climate change is altering the relative timing of species interactions by shifting when species first appear in communities and modifying the duration organisms spend in each developmental stage. However, community contexts, such as intraspecific competition and alternative resource species, can prolong shortened windows of availability and may mitigate the effects of phenological shifts on species interactions. Using a combination of laboratory experiments and dynamic simulations, we quantified how the effects of phenological shifts in Drosophila‐parasitoid interactions differed with concurrent changes in temperature, intraspecific competition, and the presence of alternative host species. Our study confirmed that warming shortens the window of host susceptibility. However, the presence of alternative host species sustained interaction persistence across a broader range of phenological shifts than pairwise interactions by increasing the degree of temporal overlap with suitable development stages between hosts and parasitoids. Irrespective of phenological shifts, parasitism rates declined under warming due to reduced parasitoid performance, which limited the ability of community context to manage temporally mismatched interactions. These results demonstrate that the ongoing decline in insect diversity may exacerbate the effects of phenological shifts in ecological communities under future global warming temperatures.
... These IPMs were constructed using either climate-only models of vital rates (hypothesis C1) or vital rate models that included climate and competition as predictors (hypothesis C2), both with interactions between the climate variables. Other examples of environmentally sensitive IPMs include Coulson et al. (2011) and Simmonds et al. (2020). ...
Ecologists have long sought to understand the controls of species’ geographic distributions. Two important hypotheses have been that range limits are determined 1) predominantly by climate or 2) by competition in addition to climate, with competitive interactions dominating where climate is benign. If the first hypothesis is correct, the effect of changing climate on species’ distributions should be adequately predicted by climate-only models. Here, we used demographic range models (DRMs) to evaluate the influence of climate and competition on the geographic distribution of Pinus edulis, a desert pine in the southwestern U.S. We parameterized DRMs with data on 23,426 trees in 1,941 forest inventory plots. Vital rate responses were consistent with the predictions of the second hypothesis: negative effects of warm-dry conditions at low-elevation sites and competition at cool-humid, high-elevation sites. Further, including both climate and competition in the DRM yielded a better fit of DRM-predicted population growth rate to the observed distribution than a climate-only DRM. However, at cool-humid sites, negative effects of competition were too weak to offset positive climate effects, resulting in a mismatch between predicted population growth rate and observed distribution. This result leads us to formulate a new hypothesis: climate has complex influences on species’ distributions through its effects on range-limiting processes at a variety of scales, for example, biotic and disturbance feedbacks at community, ecosystem, landscape, and macrosystem scales. The implications of this complex systems hypothesis are that range dynamics may be poorly predictable from static, climate-only models, and may be characterized by sudden changes in abundance, even distribution collapse, with climate change.
... The seasonal timing of important life history decisions, such as when to breed, hibernate, or migrate, are typically under natural selection toward some optimum set by ecological circumstances (Visser et al., 2010;McNamara et al., 2011). This optimum can vary through time or across space (Phillimore et al., 2010(Phillimore et al., , 2016Tansey et al., 2017;Gamelon et al., 2018;de Villemereuil et al., 2020), particularly under climate change or other anthropogenic stressors, with possible demographic consequences if populations cannot keep pace evolutionarily (Visser, 2008;Gienapp et al., 2013;Vedder et al., 2013;Simmonds et al., 2020). Breeding time in birds, in particular, has received a lot of attention (Perrins, 1970;Verhulst & Nilsson, 2008), given the relative ease with which data can be collected on egg laying/hatching dates and fitness components such as number of fledglings, recruits, or adult survival (Newton et al., 1989;Clutton-Brock & Sheldon, 2010). ...
Full-text available
The mismatch between when individuals breed and when we think they should breed has been a long‐standing problem in evolutionary ecology. Price et al. is a classic theory paper in this field and is mainly cited for its most obvious result: if individuals with high nutritional condition breed early, then the advantage of breeding early may be overestimated when information on nutritional condition is absent. Price at al.'s less obvious result is that individuals, on average, are expected to breed later than the optimum. Here, we provide an explanation of their non‐intuitive result in terms of hard selection, and go on to show that neither of their results are expected to hold if the relationship between breeding date and nutrition is allowed to evolve. By introducing the assumption that the advantage of breeding early is greater for individuals in high nutritional condition, we show that their most cited result can be salvaged. However, individuals, on average, are expected to breed earlier than the optimum, not later. More generally, we also show that the hard selection mechanisms that underpin these results have major implications for the evolution of plasticity: when environmental heterogeneity becomes too great, plasticity is selected against, prohibiting the evolution of generalists.
Full-text available
Discrete time structured population projection models are an important tool for studying population dynamics. Within this field, Integral Projection Models (IPMs) have become a popular method for studying populations structured by continuously distributed traits (e.g. height, weight). Databases of discrete time, discrete state structured population models, for example DATLife (life tables) and COMPADRE & COMADRE (matrix population models), have made quantitative syntheses straightforward to implement. These efforts allow researchers to address questions in both basic and applied ecology and evolutionary biology. Since their introduction in 2000, over 300 works containing IPMs have been published, offering opportunities for ecological synthesis too. We describe a novel framework to quickly reconstruct these models for subsequent analyses using Rpadrino R package, which serves as an interface to PADRINO, a new database of IPMs. We introduce an R package, Rpadrino, which enables users to download, subset, reconstruct, and extend published IPMs. Rpadrino makes use of recently created software, ipmr, to provide an engine to reconstruct a wide array of IPMs from their symbolic representations and conduct subsequent analyses. Rpadrino and ipmr are extensively documented to help users learn their usage. Rpadrino currently enables users to reconstruct 280 IPMs from 40 publications that describe the demography of 14 animal and 26 plant species. All of these IPMs are tested to ensure they reproduce published estimates. Rpadrino provides an interface to augment PADRINO with external data and modify parameter values, creating a platform to extend models beyond their original purpose while retaining full reproducibility. PADRINO and Rpadrino provide a toolbox for asking new questions and conducting syntheses with peer‐reviewed published IPMs. Rpadrino provides a user‐friendly interface so researchers do not need to worry about the database structure or syntax, and can focus on their research questions and analyses. Additionally, Rpadrino is thoroughly documented, and provides numerous examples of how to perform analyses which are not included in the package’s functionality.
The ability to track variation in climate is important for species to persist in a given environment. Lack of responses to both long-term changes and inter-annual variation in climate parameters can result in reduced fitness and population decline. Furthermore, migration strategy can influence the ability to track climatic variation due to the potential to use reliable environmental cues. Here, we studied the temporal relationship between birch leafing and onset of breeding for three bird species with contrasting migration strategies over a 20-year period in a subalpine habitat in Central Norway. We found no temporal change in birch leafing date or breeding onset for the three bird species over the study period. However, we found a statistically significant difference in the ability to track inter-annual variation in birch leafing date between the resident and two long-distance migratory species. The resident great tit Parus major was more capable of initiating egg laying in closer association to variation in birch leafing in early springs, than the long-distance migratory European pied flycatcher Ficedula hypoleuca and common redstart Phoenicurus phoenicurus. Long-distance migrants seem to have been constrained by arrival date or time from arrival to entering the breeding areas, in contrast to resident birds, which might be better able track early initiation of spring in breeding areas by adjusting egg laying date. Our findings highlight the importance of not solely studying directional long-term climatic change, but also pay attention to inter-annual variation.
Full-text available
Many plants benefit from synchronous year-to-year variation in seed production, called masting. Masting benefits plants because it increases the efficiency of pollination and satiates predators, which reduces seed loss. Here, using a 39-year-long dataset, we show that climate warming over recent decades has increased seed production of European beech but decreased the year-to-year variability of seed production and the reproductive synchrony among individuals. Consequently, the benefit that the plants gained from masting has declined. While climate warming was associated with increased reproductive effort, we demonstrate that less effective pollination and greater losses of seeds to predators offset any benefits to the plants. This shows that an apparently simple benefit of climate warming unravels because of complex ecological interactions. Our results indicate that in masting systems, the main beneficiaries of climate-driven increases in seed production are seed predators, not plants.
Full-text available
Global climate change is altering the timing of life history events for species living in seasonal environments. These shifts in phenology can lead to the disruption of interspecific relationships with implications for individual fitness. Predicting phenological change and its population level consequences can provide insights into population persistence. Achieving this is challenging for labile traits as current structured population models do not explicitly distinguish between the roles of phenotypic plasticity and micro‐evolution, hindering realistic predictions of trait change. In this study we present the first empirical test of a new integral projection model (IPM) framework, which allows phenotypic plasticity and micro‐evolution to be teased apart by incorporating a quantitative genetic inheritance function. We parameterise this model for a population of wild great tits Parus major and test its predictive capabilities through K‐fold cross validation. We test the predictive accuracy of the quantitative genetic IPM in comparison to the standard IPM. We demonstrate that adding genetic inheritance rules maintains high accuracy of projections of phenological change, relative to the standard IPM. In addition, we find almost identical projections of population dynamics in this population for both IPMs, demonstrating that this model formulation allows researchers to investigate the contributions of phenotypic plasticity and micro‐evolution to trait change, without sacrificing predictive accuracy. Modelling in this way reveals that, under directional environmental change, both micro‐evolution and plasticity contribute to an advance of phenology, although the effect of plasticity is an order of magnitude higher than evolution. Despite this, synchrony between great tits and their caterpillar prey was reduced and population declines occurred. Our approach demonstrates that this model framework provides a promising avenue through which to explore the roles of phenotypic plasticity and evolution in trait changes and population dynamics. This article is protected by copyright. All rights reserved.
Full-text available
Phenotypic plasticity is an important mechanism for populations to respond to fluctuating environments, yet may be insufficient to adapt to a directionally changing environment. To study whether plasticity can evolve under current climate change, we quantified selection and genetic variation in both the elevation (RNE) and slope (RNS) of the breeding time reaction norm in a long‐term (1973–2016) study population of great tits (Parus major). The optimal RNE (the caterpillar biomass peak date regressed against the temperature used as cue by great tits) changed over time, whereas the optimal RNS did not. Concordantly, we found strong directional selection on RNE, but not RNS, of egg‐laying date in the second third of the study period; this selection subsequently waned, potentially due to increased between‐year variability in optimal laying dates. We found individual and additive genetic variation in RNE but, contrary to previous studies on our population, not in RNS. The predicted and observed evolutionary change in RNE were, however, marginal, due to low heritability and the sex limitation of laying date. We conclude that adaptation to climate change can only occur via micro‐evolution of RNE, but this will necessarily be slow and potentially hampered by increased variability in phenotypic optima. This article is protected by copyright. All rights reserved
Full-text available
Urbanization can have marked effects on plant and animal populations’ phenology, population size, predator‐prey interactions and reproductive success. These aspects are rarely studied simultaneously in a single system, and some are rarely investigated, e.g. how insect phenology responds to urban development. Here, we study a tri‐trophic system of trees – phytophagous insects (caterpillars) – insectivorous birds (great tits) to assess how urbanization influences i) the phenology of each component of this system, ii) insect abundance and iii) avian reproductive success. We use data from two urban and two forest sites in Hungary, central Europe, collected over four consecutive years. Despite a trend of earlier leaf emergence in urban sites there is no evidence for an earlier peak in caterpillar abundance. Thus, contrary to the frequently stated prediction in the literature, the earlier breeding of urban bird populations is not associated with an earlier peak in caterpillar availability. Despite this the seasonal dynamics of caterpillar biomass exhibited striking differences between habitat types with a single clear peak in forests, and several much smaller peaks in urban sites. Caterpillar biomass was higher in forests than urban areas across the entire sampling period, and between 8.5 and 24 times higher during the first brood's chick‐rearing period. This higher biomass was not associated with taller trees in forest sites, or with tree species identity, and occurred despite most of our focal trees being native to the study area. Urban great tits laid smaller clutches, experienced more frequent nestling mortality from starvation, reared fewer offspring to fledging age, and their fledglings had lower body mass. Our study strongly indicates that food limitation is responsible for lower avian reproductive success in cities, which is driven by reduced availability of the preferred nestling diet, i.e. caterpillars, rather than phenological shifts in the timing of peak food availability. This article is protected by copyright. All rights reserved.
Full-text available
Background Threshold regression models are a diverse set of non-regular regression models that all depend on change points or thresholds. They provide a simple but elegant and interpretable way to model certain kinds of nonlinear relationships between the outcome and a predictor. Results The R package chngpt provides both estimation and hypothesis testing functionalities for four common variants of threshold regression models. All allow for adjustment of additional covariates not subjected to thresholding. We demonstrate the consistency of the estimating procedures and the type 1 error rates of the testing procedures by Monte Carlo studies, and illustrate their practical uses using an example from the study of immune response biomarkers in the context of Mother-To-Child-Transmission of HIV-1 viruses. Conclusion chngpt makes several unique contributions to the software for threshold regression models and will make these models more accessible to practitioners interested in modeling threshold effects. Electronic supplementary material The online version of this article (doi:10.1186/s12859-017-1863-x) contains supplementary material, which is available to authorized users.
Full-text available
For organisms living in seasonal environments, synchronizing the peak energetic demands of reproduction with peak food availability is a key challenge. Understanding the extent to which animals can adjust behavior to optimize reproductive timing, and the cues they use to do this, is essential for predicting how they will respond to future climate change. In birds, the timing of peak energetic demand is largely determined by the timing of clutch initiation; however, considerable alterations can still occur once egg laying has begun. Here, we use a wild population of great tits (Parus major) to quantify individual variation in different aspects of incubation behavior (onset, duration, and daily intensity) and conduct a comprehensive assessment of the causes and consequences of this variation. Using a 54-year dataset, we demonstrate that timing of hatching relative to peak prey abundance (synchrony) is a better predictor of reproductive success than clutch initiation or clutch completion timing, suggesting adjustments to reproductive timing via incubation are adaptive in this species. Using detailed in-nest temperature recordings, we found that postlaying, birds improved their synchrony with the food peak primarily by varying the onset of incubation, with duration changes playing a lesser role. We then used a sliding time window approach to explore which spring temperature cues best predict variance in each aspect of incubation behavior. Variation in the onset of incubation correlated with mean temperatures just prior to laying; however, incubation duration could not be explained by any of our temperature variables. Daily incubation intensity varied in response to daily maximum temperatures throughout incubation, suggesting female great tits respond to temperature cues even in late stages of incubation. Our results suggest that multiple aspects of the breeding cycle influence the final timing of peak energetic demand. Such adjustments could compensate, in part, for poor initial timing, which has significant fitness impacts.
Full-text available
Understanding how the natural world will be impacted by environmental change over the coming decades is one of the most pressing challenges facing humanity. Addressing this challenge is difficult because environmental change can generate both population-level plastic and evolutionary responses, with plastic responses being either adaptive or nonadaptive. We develop an approach that links quantitative genetic theory with data-driven structured models to allow prediction of population responses to environmental change via plasticity and adaptive evolution. After introducing general new theory, we construct a number of example models to demonstrate that evolutionary responses to environmental change over the short-term will be considerably slower than plastic responses and that the rate of adaptive evolution to a new environment depends on whether plastic responses are adaptive or nonadaptive. Parameterization of the models we develop requires information on genetic and phenotypic variation and demography that will not always be available, meaning that simpler models will often be required to predict responses to environmental change. We consequently develop a method to examine whether the full machinery of the evolutionarily explicit models we develop will be needed to predict responses to environmental change or whether simpler nonevolutionary models that are now widely constructed may be sufficient.
Full-text available
When studying the impacts of climate change, there is a tendency to select climate data from a small set of arbitrary time periods or climate windows (e.g., spring temperature). However, these arbitrary windows may not encompass the strongest periods of climatic sensitivity and may lead to erroneous biological interpretations. Therefore, there is a need to consider a wider range of climate windows to better predict the impacts of future climate change. We introduce the R package climwin that provides a number of methods to test the effect of different climate windows on a chosen response variable and compare these windows to identify potential climate signals. climwin extracts the relevant data for each possible climate window and uses this data to fit a statistical model, the structure of which is chosen by the user. Models are then compared using an information criteria approach. This allows users to determine how well each window explains variation in the response variable and compare model support between windows. climwin also contains methods to detect type I and II errors, which are often a problem with this type of exploratory analysis. This article presents the statistical framework and technical details behind the climwin package and demonstrates the applicability of the method with a number of worked examples.
Full-text available
Climate change may influence the phenology of organisms unequally across trophic levels and thus lead to phenological mismatches between predators and prey. In cases where prey availability peaks before reproducing predators reach maximal prey demand, any negative fitness consequences would selectively favor resynchronization by earlier starts of the reproductive activities of the predators. At a study site in northeast Greenland, over a period of 17 years, the median emergence of the invertebrate prey of Sanderling Calidris alba advanced with 1.27 days per year. Yet, over the same period Sanderling did not advance hatching date. Thus, Sanderlings increasingly hatched after their prey was maximally abundant. Surprisingly, the phenological mismatches did not affect chick growth, but the interaction of the annual width and height of the peak in food abundance did. Chicks grew especially better in years when the food peak was broad. Sanderling clutches were most likely to be depredated early in the season , which should delay reproduction. We propose that high early clutch pre-dation may favor a later reproductive timing. Additionally, our data suggest that in most years food was still abundant after the median date of emergence, which may explain why Sanderlings did not advance breeding along with the advances in arthropod phenology.
1.Changes in the timing of life history events (phenology) are a widespread consequence of climate change. Predicting population resilience requires knowledge of how phenology is likely to change over time, which can be gained by identifying the specific environmental cues that drive phenological events. Cue identification is often achieved with statistical testing of candidate cues. As the number of methods used to generate predictions increases, assessing the predictive accuracy of different approaches has become necessary. 2.This study aims to (i) provide an empirical illustration of the predictive ability of five commonly‐applied statistical methods for cue identification (absolute and relative sliding time window analyses, penalised signal regression, climate sensitivity profiles, and a growing degree‐day model), and (ii) discuss approaches for implementing cue identification methods in different systems. 3.Using a dataset of mean clutch initiation timing in wild great tits (Parus major), we explored how the days of the year identified as most important, and the aggregate statistic identified as a cue, differed between statistical methods and with respect to the timespan of data used. Each method's predictive capacity was tested using cross validation and assessed for robustness to varying sample size. 4.We show that the identified critical time window of cue sensitivity was consistent across four of the five methods. The accuracy and precision of predictions differed by method with penalised signal regression resulting in the most accurate and most precise predictions in our case. Accuracy was maximal for near future predictions and showed a relationship with time. The difference between predictions and observations systematically shifted across the study from preceding observations to lagging. 5.This temporal trend in prediction error suggests that the current statistical tools either fail to capture a key component of the cue‐phenology relationship, or the relationship itself is changing through time in our system. These two influences need to be teased apart if we are to generate realistic predictions of phenological change. We recommend future phenological studies to challenge the idea of a static cue‐phenology relationship and should cross‐validate results across multiple time periods. This article is protected by copyright. All rights reserved.