Content uploaded by Daniel S Park
Author content
All content in this area was uploaded by Daniel S Park on Nov 27, 2018
Content may be subject to copyright.
rstb.royalsocietypublishing.org
Research
Cite this article: Park DS, Breckheimer I,
Williams AC, Law E, Ellison AM, Davis CC. 2018
Herbarium specimens reveal substantial and
unexpected variation in phenological sensitivity
across the eastern United States. Phil.
Trans. R. Soc. B 374: 20170394.
http://dx.doi.org/10.1098/rstb.2017.0394
Accepted: 15 October 2018
One contribution of 16 to a theme issue
‘Biological collections for understanding
biodiversity in the Anthropocene’.
Subject Areas:
ecology, evolution, taxonomy and systematics
Keywords:
citizen science, digitization, geographical
range, herbarium specimens, phenology,
phenological sensitivity
Authors for correspondence:
Daniel S. Park
e-mail: danielpark@fas.harvard.edu
Charles C. Davis
e-mail: cdavis@oeb.harvard.edu
†
These authors contributed equally to this
work.
Electronic supplementary material is available
online at https://dx.doi.org/10.6084/m9.
figshare.c.4274660.
Herbarium specimens reveal substantial
and unexpected variation in phenological
sensitivity across the eastern United States
Daniel S. Park1,†, Ian Breckheimer1,†, Alex C. Williams2, Edith Law2,
Aaron M. Ellison3and Charles C. Davis1
1
Department of Organismic and Evolutionary Biology and Harvard University Herbaria, Harvard University,
Cambridge, MA 02138, USA
2
David R. Cheriton School of Computer Science, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1
3
Harvard Forest, Harvard University, Petersham, MA 01366, USA
CCD, 0000-0001-8747-1101
Phenology is a key biological trait that can determine an organism’s survival
and provides one of the clearest indicators of the effects of recent climatic
change. Long time-series observations of plant phenology collected at conti-
nental scales could clarify latitudinal and regional patterns of plant
responses and illuminate drivers of that variation, but few such datasets
exist. Here, we use the web tool CrowdCurio to crowdsource phenological
data from over 7000 herbarium specimens representing 30 diverse flowering
plant species distributed across the eastern United States. Our results, span-
ning 120 years and generated from over 2000 crowdsourcers, illustrate
numerous aspects of continental-scale plant reproductive phenology. First,
they support prior studies that found plant reproductive phenology signifi-
cantly advances in response to warming, especially for early-flowering
species. Second, they reveal that fruiting in populations from warmer,
lower latitudes is significantly more phenologically sensitive to temperature
than that for populations from colder, higher-latitude regions. Last, we
found that variation in phenological sensitivities to climate within species
between regions was of similar magnitude to variation between species.
Overall, our results suggest that phenological responses to anthropogenic cli-
mate change will be heterogeneous within communities and across regions,
with large amounts of regional variability driven by local adaptation, pheno-
typic plasticity and differences in species assemblages. As millions of
imaged herbarium specimens become available online, they will play an
increasingly critical role in revealing large-scale patterns within assemblages
and across continents that ultimately can improve forecasts of the impacts of
climatic change on the structure and function of ecosystems.
This article is part of the theme issue ‘Biological collections for
understanding biodiversity in the Anthropocene’.
1. Introduction
Ecosystems on every continent have been affected by local, regional and global
changes in climate, especially increases in temperature [1]. Changes in phenol-
ogy—the timing of life-history events—are among the most conspicuous and
well-documented species responses to climatic change, especially for plants
[2–7]. Phenological disruption has already impacted species’ local persistence
and community diversity [8–10], which may have widespread consequences
for critical ecosystem processes, including carbon sequestration [11– 13],
ecosystem– atmosphere interactions [14] and trophic interactions [15–28].
Despite these trends, our knowledge of plant phenological responses to cli-
matic change remains inadequate. In particular, although phenological
responses may differ among species with different functional or life-history
traits and biogeographical origins [29– 32], long-term observational datasets
&2018 The Author(s) Published by the Royal Society. All rights reserved.
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
to assess such trends are limited in geographical, temporal
and taxonomic scope [33]. Many of these data track woody
plant species of the Northern Hemisphere (most commonly,
abundant tree species), and only for the last approximately
40 years (but see [34,35]). These biases limit our understand-
ing of variation in phenological responses across species and
biomes. Furthermore, although population-level variation in
phenology has been demonstrated for a few species [36 – 38],
there are very few studies that quantify both inter- and
intraspecific variation in phenological response [32,33].
Variability in species’ phenology is particularly relevant
because climatic change is not geographically uniform. For
example, high-latitude regions are warming faster than sub-
tropical and tropical ones [1,39]. Short growing seasons also
may cause high-latitude ecosystems to be especially sensitive
to temperature, leading to stronger selective pressures for
populations to initiate growth as soon as conditions become
favourable in early spring. Additionally, plants adapted to
highly variable climates may exhibit higher phenological
thresholds to temperature, as it provides a less reliable
signal [40]. Thus, the effects of climatic change on species’
phenology may differ across their ranges depending on vari-
ation in phenological sensitivity. Variability in phenological
responses to climatic change within species may also alter
patterns of gene flow, which could either promote or counter-
act adaptive evolution via the sharing of locally (mal)adapted
alleles [41–44]. Recent studies have shown that plant phenol-
ogy may be more responsive in more northern-ranging
populations where there are more variable and extreme cli-
mates, especially during the early part of the growing
season [44,45]. Although these studies are suggestive, they
are restricted in spatial scale and taxonomic scope, and
broad regional patterns of phenological response to climate
may differ from patterns at smaller, local scales.
Herbarium specimens represent snapshots of phenology
(i.e. flowering and fruiting) at a specific place and time, and
have shown tremendous promise to increase the spatial, tem-
poral and taxonomic resolution of phenological data [46– 48].
They provide rich historical depth, wide geographical scope
and taxonomic diversity, all of which allow researchers to
track long-term changes of vast numbers of species and com-
munities through space and time [4,46– 49]. Despite their
representation of phenological responses [48,50], herbarium
specimens have been used less frequently than other data
sources, such as field observations, to address phenological
change, in part because they have been inaccessible to
many researchers [46,51]. However, the widespread digitiz-
ation of herbarium collections [52] combined with new
approaches to collecting [46,53] and analysing [54,55] pheno-
logical data derived from herbarium specimens has the
power to transform our understanding of plant responses
to global climatic change.
Here, we applied a newly developed web-enabled crowd-
sourcing platform, CrowdCurio:Thoreau’s Field Notes (https://
www.crowdcurio.com/) [46], to examine more than 7000
specimens of 30 phylogenetically diverse flowering plant
species. The Thoreau’s Field Notes module facilitates the
rapid quantification of phenological traits via image annota-
tion and has been demonstrated to yield reliable data
regardless of the level of expertise among crowdsourcers
(i.e. expert versus non-expert scoring) [46]. We used these
crowdsourced data to infer the magnitude, direction, and
variability in reproductive phenological responses to spring
temperature across 238of latitude in the eastern United
States. We examined both native and introduced plant
species from northern coniferous forests, eastern deciduous
forests, subtropical evergreen forests, grasslands, wetlands,
alpine meadows and aquatic plant communities. Environ-
mental conditions in this region vary considerably across
species’ ranges, and populations may exhibit substantial
variation in phenological response across this latitudinal
gradient. Our overall goals with this study were: (i) to
demonstrate the power of characterizing phenology from
herbarium data using an efficient and rapid workflow that
leverages a nearly fully mobilized online flora of the eastern
United States [56,57]; (ii) to greatly increase the taxonomic
and ecological diversity of species sampled for this purpose
(from woody perennials to herbaceous annuals across a
range of biomes); and (iii) to sample species with broad lati-
tudinal ranges to assess regional and inter- and intra-species
variation in phenological responses.
2. Methods
(a) Specimen data collection
We examined phenological responses of species using digitized
specimens from two of the most comprehensive digitized regional
floras in the world, the Consortium of Northeastern Herbaria
(CNH; http://portal.neherbaria.org/portal/) [56] and Southeast
Regional Network of Expertise and Collections (SERNEC;
http://sernecportal.org/portal/index.php) [57]. These two
online portals include more than six million digitized herbarium
records, including specimen images. Our criteria for selecting
angiosperm species for analysis were that specimens: (i) included
at least county-level location data; (ii) included at least 50 unique
collections across space and time; (iii) were of species with rela-
tively easily identifiable and quantifiable reproductive structures;
and (iv) were from species with broad latitudinal ranges sufficient
to enable quantification of population-level variation.
Applying these criteria yielded 30 species with varying life-
history traits, growth forms, native status and general reproduc-
tive seasonality (e.g. early- versus late-spring flowering). We
downloaded over 10 000 digital herbarium specimen images of
these species from CNH and SERNEC, removed duplicate, mis-
identified or sterile specimens, and those with notable insect
damage on reproductive structures, extensive physical damage
or poor preservation. We also removed all 30 specimens from
Florida, which were geographical and climatic outliers. Our
final dataset comprised 7722 specimens and spanned 120 years
across 512 United States counties (table 1). Species’ life-history
(annual versus perennial), growth form (woody versus herbac-
eous) and native status (native versus introduced) were
inferred from the United States Department of Agriculture
PLANTS Database (https://plants.usda.gov/). For individual
specimen metadata, see electronic supplementary material,
table S1.
(b) Crowdsourcing phenological data collection
The phenological state (phenophase) of a plant can be inferred
from the presence and quantity of relevant structures, such as
leaves, flowers or fruits [46,48]. Past researchers generally have
focused on the presence or the absence of a single structure or
trait (e.g. [58]) or applied majority estimates for scoring a
single phenophase (e.g. [49]). Here, we quantified data for two
reproductive phenophases, flowering and fruiting. Specimens
were scored as flowering if open flowers represented greater
than or equal to 50% of the total reproductive structure count
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
2
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
and scored as fruiting if they had less than 50% flowers and buds
and at least one fruit present. We used the Thoreau’s Field Notes
instance of CrowdCurio to crowdsource phenological data from
digitized herbarium specimens [46]. Citizen scientists hired
through Amazon’s Mechanical Turk service (MTurk; https://
www.mturk.com/) counted the number of buds, flowers and
fruits observed for a set of 10 specimen images. Participants
first watched a short (1 min) instructional video on how to
score phenological traits using CrowdCurio and then were
provided with three tutorial images of each reproductive struc-
ture for every species. The 2364 anonymous participants were
compensated at a rate of $0.12 per image.
To provide an estimate of measurement error, each 10-image
set scored by a single crowdsourcer included nine unique images
and a single duplicate image randomly selected from the other
nine [55]. We estimated the reliability score for each participant
based on the data for each 10-image set by dividing the absolute
difference in organ counts for each phenophase by the total
count of that specimen across the two duplicate specimens and
subtracting this value from 1 (1 – (jcount1 – count2j/(count1 þ
count2)) [55]. Reliability scores range from zero (unreliable/
inconsistent) to one (reliable/consistent). Participants who
reported no organs on one sheet and a non-zero number of the
same organ on the duplicate sheet were assigned a reliability
score of zero for that organ (i.e. the lowest reliability score).
We conservatively selected the lowest reliability score among
the three calculated for each organ per participant and
assigned it to each participant as their final score. That is, if a
participant got a reliability score equal to zero on one organ,
they would be assigned a reliability score of zero for all
organs. Specimen observations scored by crowdsourcers with a
reliability score of zero were excluded from the analysis. We
also spot checked for suspicious outliers manually and removed
such data.
Table 1. The number of specimens and different categorical traits of examined species. FL and FR refer to the number of specimens classified as flowering and
fruiting, respectively.
family species
time span
(years) FL FR lifespan growth form status
Ranunculaceae Anemone canadensis L. 116 41 59 perennial herbaceous native
Ranunculaceae Anemone hepatica L. 95 40 60 perennial herbaceous native
Ranunculaceae Aquilegia canadensis L. 119 251 238 perennial herbaceous native
Asteraceae Bidens vulgata Greene 119 83 136 annual herbaceous native
Celastraceae Celastrus orbiculatus Thunb. 103 151 220 perennial herbaceous introduced
Asteraceae Centaurea stoebe Tausch 111 93 161 perennial herbaceous introduced
Asteraceae Cirsium arvense (L.) Scop. 118 186 171 perennial herbaceous introduced
Asteraceae Cirsium discolor (Muhl. ex Willd.)
Spreng.
117 46 93 perennial herbaceous native
Geraniaceae Geranium maculatum L. 119 489 513 perennial herbaceous native
Geraniaceae Geranium robertianum L. 119 48 307 perennial herbaceous native
Xanthorrhoeaceae Hemerocallis fulva (L.) L. 115 144 45 perennial herbaceous introduced
Malvaceae Hibiscus moscheutos L. 119 105 100 perennial herbaceous native
Balsaminaceae Impatiens capensis Meerb. 120 153 501 annual herbaceous native
Iridaceae Iris pseudacorus L. 117 90 66 perennial herbaceous introduced
Iridaceae Iris versicolor L. 119 344 185 perennial herbaceous native
Liliaceae Lilium canadense L. 117 139 27 perennial herbaceous native
Caprifoliaceae Lonicerabella Zab. 107 37 62 perennial woody introduced
Caprifoliaceae Lonicera canadensis Bartram
ex Marshall
120 194 201 perennial woody native
Caprifoliaceae Lonicera japonica Thunb. 116 329 115 perennial woody introduced
Rosaceae Malus pumila Mill. 118 74 40 perennial woody introduced
Malvaceae Malva neglecta Wallr. 116 25 140 perennial herbaceous introduced
Onagraceae Oenothera perennis L. 120 194 214 perennial herbaceous native
Orobanchaceae Orobanche uniflora L. 118 213 105 annual herbaceous native
Rosaceae Rosa gallica L. 108 45 17 perennial woody introduced
Rosaceae Rubus odoratus L. 120 176 318 perennial woody native
Sarraceniaceae Sarracenia purpurea L. 119 234 75 perennial herbaceous native
Iridaceae Sisyrinchium mucronatum Michx. 117 86 157 perennial herbaceous native
Solanaceae Solanum rostratum Dunal 115 21 85 annual herbaceous native
Melanthiaceae Trillium grandiflorum (Michx.) Salisb. 119 129 40 perennial herbaceous native
Melanthiaceae Trillium undulatum Willd. 120 402 156 perennial herbaceous native
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
3
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
(c) Historical climate data
We used estimates of historic (1895– 2016) average monthly temp-
erature and precipitation data at 4 km resolution from PRISM
( product AN81 m; http://prism.oregonstate.edu/), which provide
high-resolution time-series estimates of climatic elements for the
contiguous UnitedStates. As accurate locality data are not available
for the majority of historic specimen records[59], we used county as
our geographical unit of analysis. The vast majority (79%) of speci-
mens used in this study were collected before 1980, and while 72%
of the specimens used in this study had associated coordinate data,
at least 91% of those coordinates had been georeferenced post hoc
(e.g. assigned county or township centroid coordinates), and thus
may not represent precise sampling locales. For each county and
year, we estimated the mean monthly temperature, precipitation
and elevation, and assigned these values to each specimen [59].
Although counties can vary in size and climate, counties in states
along the Atlantic coast of the United States are generally small in
size and geographically homogeneous. We estimated within-
county climatic heterogeneity as the standard deviation of
estimated monthly climatic values across each county and year
and included it in our initial analyses, but coefficients had Bayesian
credibility intervals that werenot credibly different from zero, sowe
dropped these terms from our final models.
(d) Statistical analyses
Phenological sensitivity to spring temperature—defined as the
mean of March, April and May temperatures [46]—was defined
as the slope of the linear relationship between the day of year
(DOY) of a phenophase and the spring temperature of the corre-
sponding location and year (shifts in days per degree Celsius
change: days/8C) [44,46]. These months have been used to
define spring across the east coast of the United States [46,60].
To calculate phenological sensitivity, we binned our specimen
data into both broad climatic zones [61,62] and finer-scale
National Ecological Observatory Network (NEON) domains.
Our data comprised two climatic zones (cold/very cold; mixed-
humid/hot-humid—hereafter referred to as cold and mixed-
warm) and five NEON ecoclimatic domains (NE, northeast;
MA, mid-Atlantic; AP, Appalachians & Cumberland Plateau;
OZ, Ozarks Complex; SE, southeast; electronic supplementary
material, table S2). We also estimated phenological sensitivity to
elevation as the slope of the linear relationship between the
DOY of a phenophase and metres above sea level (m a.s.l.).
We estimated the mean timing of flowering and fruiting phe-
nophases, and environmental influences on them, using Bayesian
hierarchical linear regression models [63]. In our models, species,
region, county and observer were considered random effects,
while spring temperature and county elevation were covariates.
The hierarchical nature of the model, in which the phenological
responses of individual species were assumed to be drawn
from statistical distributions instead of fixed estimates [64],
allowed us to better estimate their climatic sensitivities. These
models also more accurately quantified uncertainty in our esti-
mates and partitioned the variance in phenological timing and
phenological sensitivity within and between species and regions.
We fitted two models for each phenological phase. ‘Model 1’
estimated species-specific phenological sensitivities and parti-
tioned their variances. ‘Model 2’ provided a more powerful
comparison between phenological sensitivities found in the
warmer, lower latitudes of our study area and those in
the cooler, higher-latitude regions (figure 1).
In Model 1, the dependent variable was the DOY for which a
given phenological phase (flowering or fruiting) was recorded
for the ith specimen. DOY
[i]
was assumed to be normally
distributed, with mean
m
[i]
and species-specific variance
s
[s]
.
DOY[i]N(
m
[i],
s
[s]):ð2:1Þ
The linear predictor
m
[i]
was estimated as a function of covariates,
including mean spring (March–May) air temperatures
(SpringT
[i]
) and the average elevation of the county in which
the specimen was recorded (Elev
[c]
). Additional intercept terms
(
a
1–
a
5) were added for each species (s), region (r), species
region combination (sr), county (c) and observer (o). The full
expression for estimating
m
[i]
was
m
[i]¼
a
1[s]þ
a
2[r]þ
a
3[sr]þ
a
4[c]þ
a
5[o]þ
b
1[s]SpringT[i]
þ
b
2[r]SpringT[i]þ
b
3[sr]SpringT[i]þ
b
4[s]Elev[c]:
ð2:2Þ
Species-specific slope and intercept terms (
a
1
[s]
,
b
1
[s]
and
b
4
[s]
in equation (2.2)) were drawn from normal distribu-
tions, with species assemblage means
m
a
1
,
m
b
1
and
m
b
4
, and
hypervariances
s
a
1
,
s
b
1
and
s
b
4
.
a
1[s]N(
m
a
1,
s
a
1), ð2:3Þ
b
1[s]N(
m
b
1,
s
b
1)ð2:4Þ
and
b
4[s]N(
m
b
4,
s
b
4):ð2:5Þ
Region and species region slopes (
b
2
[r]
,
b
3
[sr]
), and region,
species region, county and observer intercepts (
a
2
[r]
,
a
3
[sr]
,
a
4
[c]
,
a
5
[o]
) in equation (2.2) were drawn from zero-centred
normal distributions with hypervariances
s
b
2
,
s
b
3
,
s
a
2
,
s
a
3
,
s
a
4
and
s
a
5
, respectively. The species-specific sampling variation
(
s
[s]
) terms in equation (2.1) were estimated independently to
account for differences in the duration of flowering and fruiting
phases between species.
The three different groups of slopes estimated for spring
temperature decomposed variation in phenological sensitivity
into components representing between-species variability (
b
1
[s]
),
between-region variability (
b
2
[r]
) and within-species variability
across regions (
b
3
[sr]
). The accompanying hypervariances (
s
b
1
,
s
b
2
,
s
b
3
) directly represented these different sources of variability;
comparing their relative magnitudes quantified the contributions
of each source of variation to overall variation in phenological sen-
sitivity. This model structure also provided estimates of the
contributions of species turnover to differences in phenological
sensitivity between regions. We estimated these contributions
by analysing the output of Model 1, computing phenological sen-
sitivities for each observation for each iteration of our model. We
then used the mean and standard deviation of these estimates for
each region to create region-specific estimates of mean phenologi-
cal sensitivities and their variability. We assessed the contribution
of community turnover by comparing estimates that included all
three climate sensitivity terms (
b
1
[s]
þ
b
2
[r]
þ
b
3
[sr]
) with esti-
mates that included only the terms that represent species-level
variability in climate sensitivity (
b
1
[s]
). This strategy allowed us
to infer what the mean phenological sensitivities would be
across regions in the hypothetical case that they differed only in
species composition, and species responded identically to climate
across their ranges.
Model 2 differed from Model 1 in treating the region term
(
b
2
[r]
) as a two-level fixed effect representing the climatic
region from which the specimen was drawn.
m
[i]¼
a
2[r]þ
a
3[sr]þ
a
4[c]þ
a
5[o]þ
b
2[r]SpringT[i]
þ
b
3[sr]SpringT[i]þ
b
4[s]Elev[c]:
ð2:6Þ
Model 2 maximized statistical power to compare overall phe-
nological sensitivities of species between warmer, more southerly
parts of our study area and cooler, more northerly areas. Instead
of treating region-specific slopes and intercepts as normally dis-
tributed random grouping factors, we represented species
region, county and observer terms as zero-centred and normally
distributed. This structure allowed more direct inference about
overall differences in phenological sensitivity between cool and
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
4
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
mixed-warm climate regions (differences between estimates of
b
2
[r]
for the two regions). Unlike Model 1, however, Model 2
lacks species-specific parameters (
a
1
[s]
and
b
1
[s]
) and did not
provide species-specific estimates of phenological sensitivity or
permit comparison of variability in phenological sensitivity
within and between species.
We estimated all parameters of the two models using
Hamiltonian Monte Carlo (HMC) [65] implemented in Stan
(v. 2.17.3) [66] called from the rstan interface [67] in R [68].
HMC is a form of Markov chain Monte Carlo (MCMC) that effi-
ciently estimates hierarchical Bayesian regression models for
larger datasets like ours [69]. We used relatively uninformative
prior distributions: zero-centred normal priors for slopes and
intercepts, and truncated normal distributions for variances
and hypervariances. To account for sampling behaviour and sim-
plify prior choices, we scaled and centred the response variable
DOY
[i]
and all continuous predictors by subtracting the mean
and dividing by the standard deviation of each variable. For
each model run for each phenophase, we estimated parameters
using four MCMC chains of 4000 iterations each and discarded
the first 2000 iterations of each chain (as burn-in). We assessed
convergence of each model both visually and with the
Gelman–Rubin statistic (^
r,1:1 for all parameters). We also
assessed good model fit using visual posterior predictive
checks implemented in the bayesplot R package [70]. All
parameter estimates were based on at least 1000 effective pos-
terior samples. Estimates reported in the results were back-
transformed to the original data scale to facilitate illustration
and interpretation.
Code and data for reproducing these analyses are archived
by Harvard Forest [71].
3. Results
Our focal species spanned wide geographical and climatic
space (figure 1). They demonstrated diverse patterns of
phenology and significant variation in responses to climate
across species and geographies. Using Model 1, estimated
mean (non-leap-year) flower timing at 7.48C and 216 m a.s.l.
(mean collection conditions for the specimens) varied from
10 May (Day 130, Anemone hepatica) to 10 September (Day
253, Bidens vulgata) for flowering and 22 May (Day 142, A.
hepatica) to 14 September (Day 257, B. vulgata) for fruiting
(figure 2). The average lag time between flowering and fruit-
ing across all species was approximately 20 days. Most
species flowered and fruited earlier with warmer spring temp-
eratures (assemblage mean 22.56 days/8C, 95% CI 23.64 to
21.48, figure 2), and these responses were credibly different
from zero ( posterior probability .0.95) for 21 out of 30
species for flowering and 15 out of 30 species for fruiting
(electronic supplementary material, tables S3 and S4).
30
0
0 300 600
elevation a.s.l. (m)
900
5101520–85 –80 –75
longitude
longitude
spring temperature (ºC)
–70
–85 –80 –75 –70
NEON domain
climate zone
cold
mixed-warm
Appalachians (AP)
mid-Atlantic (MA)
northeast (NE)
Ozarks Complex (OC)
southeast (SE)
35
40
latitude
45
30
35
40
latitude
45
30
35
40
45
30
35
40
45
(c)(a)
(b)(d)
Figure 1. Distribution of herbarium specimens across geographical and environmental space, colour-coded by NEON domain (a,c) or climate region (b,d). Panels (a)
and (c) show specimens in flower, and panels (b) and (d) show specimens in fruit. Specimens are referenced to county centroids, and jitter has been added to
coordinates to reduce over-plotting. Average spring (March – May) air temperatures are strongly negatively correlated with latitude (Pearson correlation r¼20.92,
panel c), but elevation and latitude are largely independent (panel d).
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
5
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
For both flowering and fruiting, species with earlier
reproductive phenologies were substantially more sensitive
to spring temperature than species that flowered and fruited
later in the season. This sensitivity manifested in a strong
positive correlation between mean flowering and fruiting
date and spring temperature sensitivity, with a slope of
0.018 days/8C per day for flowering and 0.023 days/8C per
day for fruiting (figure 2a,b). These slopes were different
from zero with greater than 99% posterior probability. We
also found that, all other conditions being equal, flowering
and fruiting came earlier at higher elevations (community
means 21.71 to 20.11 days earlier per 100 m greater
elevation for flowering, and 22.07 to 0.14 days earlier for
fruiting, figure 2c,d). These effects were credibly different
from zero for 7 of 30 species for flowering and 8 of 30 species
for fruiting, but elevation influenced early-flowering and
late-flowering species approximately equally.
Species in the warm and mixed-temperate climatic
regions showed greater mean sensitivities to spring tempera-
ture and also greater variability (standard deviation) in
climate sensitivity between species than in the cool-temperate
northeast and Appalachians (figure 3; electronic supplemen-
tary material, figures S1 and S2). Using Model 2, we
estimated that mean sensitivities in the mixed-warm region
(figure 1b) were 22.96 days/8C (95% CI 3.69 to 22.25) for
flowering and 23.37 days/8C (95% CI 24.12 to 22.60) for
fruiting, but were substantially closer to zero in the cool-
temperate region (22.51 days/8C, 95% CI 22.86 to 22.19
for flowering, and 22.09 days/8C, 95% CI 22.63 to 21.57
for fruiting, figure 3a). The mixed-warm climatic region also
had greater assemblage variability in phenological sensitivity
(figure 3b). All differences between cold and mixed-warm
climatic regions had a posterior probability greater than
0.95 except for mean differences in flowering, where differ-
ences had a posterior probability of 0.87. These qualitative
patterns were robust to an alternative spatial binning strategy
that used only latitude, and not climate, to differentiate more
northerly and southerly regions (electronic supplementary
material, figure S3).
Overall differences between cool, northern and warm,
southerly parts of the study area were accompanied by
large amounts of regional variation not explained by climate
or latitude (figure 4). For example, using Model 1 we esti-
mated that plants in the Ozark Complex and mid-Atlantic
NEON domains were substantially more phenologically
sensitive than those in the northeast and Appalachians
(figure 4a), while the northeast and Ozark Complex had
plant assemblages with greater variability in phenological
sensitivity (figure 4c). Differences in phenological sensitivity
between regions were not adequately explained by differ-
ences in species composition, as per-sample weighted
means computed using only species effects (
b
1
[s]
in equation
(2.2)), did not show strong regional differences (figure 4b,d).
Our hierarchical approach allowed us to compare within-
species, between-species and between-region sources of
variability for both mean flowering time and sensitivity to
spring temperature (figure 5). This analysis shows that
between-species variation dominated variability in mean
flowering time (figure 5a), but there was a similar amount
of variation in phenological sensitivity within species
between regions to that seen between species for both
flowering and fruiting (figure 5b).
1
23
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
−7
−6
−5
−4
−3
−2
−1
0
1
phenological sensitivity
(days/°C)
Anemone canadensis
Anemone hepatica
Aquilegia canadensis
Bidens vulgata
Celastrus orbiculatus
Centaurea stoebe
Cirsium arvense
Cirsium discolor
Geranium maculatum
Geranium robertianum
Hemerocallis fulva
Hibiscus moscheutos
Impatiens capensis
Iris pseudacorus
Iris versicolor
Lilium canadense
Lonicera bella
Lonicera canadensis
Lonicera japonica
Malus pumila
Malva neglecta
Oenothera perennis
Orobanche uniflora
Rosa gallica
Rubus odoratus
Sarracenia purpurea
Sisyrinchium mucronatum
Solanum rostratum
Trillium grandiflorum
Trillium undulatum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
species
community means
50% CI
1
2
3
4
56
78
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 21
22
23 24
25
26
27 28
29 30
100 150 200 250 100 150 200 250
−6
−5
−4
−3
−2
−1
0
1
2
phase mean at 7.4°C and 216 m a.s.l. (DOY)
phenological sensitivity
(days/100 m a.s.l.)
flowering fruiting
80% CI
(b)(a)
(c)(d)
Figure 2. Mean flowering and fruiting time compared with estimated phenological sensitivities to spring (March–May) temperatures (a,b), and collection elevation
(c,d) of 30 species estimated from herbarium specimens using a Bayesian hierarchical model (Model 1). Coloured and enumerated dots indicate species-specific
estimates, and white circles at panel margins represent estimated community means for each quantity. Thick and thin bars represent 50 and 80% posterior credible
intervals, respectively. Thick black lines represent credible linear relationships between quantities on x- and y-axes (posterior slope estimate different from zero
with greater than 90% probability). Dotted lines represent non-credible relationships (posterior slope estimate not different from zero with greater than
90% probability).
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
6
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
4. Discussion
Our analyses revealed that (i) plant species from the eastern
United States exhibit advanced timing of flowering and fruit-
ing in response to warmer spring temperatures, (ii) the
magnitude of these responses varies significantly between
and within species across their latitudinal ranges and (iii)
that phenological sensitivity to temperature tends to be
higher in the warmer, more stable climates of lower-latitude
regions.
(a) Differential responses to spring warming
across species
Consistent with previous field observations of community
phenology, we found that reproductive phenology of flower-
ing plants accelerated with warming spring temperatures
(e.g. [46,72,73]; but see [63]). The average number of days of
phenological advancement per degree increase in temperature
(22.56 days/8C) that we observed also fell within previous
estimates [46,74,75]. All else being equal, flowering and fruit-
ing tended to occur earlier at higher elevations. Higher
elevations tend to be relatively colder and have shorter
growing seasons, which exert pressure for species to initiate
growth as soon as conditions become favourable [44,76–78].
Despite these general trends, we observed significant
variation among species in their responses to warming. In
general, early-flowering and early-fruiting species were
more sensitive to spring temperatures than late-flowering/
fruiting species (figure 2), a pattern also observed at smaller
scales [75,79]. Warming-induced leaf budburst advancement
has been suggested to be less prominent in late-flushing
species compared with early-flushing ones owing to their
greater chilling requirements [80]. Similar mechanisms may
affect flowering and fruiting, where advances in the flowering
date of late-flowering species caused by spring warming
would be smaller than those of early-flowering species,
which would manifest as weaker phenological responses to
temperature in late-flowering species. Further, as flowering
and fruiting events later in the year are more separated from
spring climatic conditions, there is an extended window of
time in which other factors could affect or modify reproduc-
tive timing. For instance, late-flowering species may be more
sensitive to photoperiod or precipitation.
A large amount of variability in phenological sensitivity
across species suggests that phenological responses to cli-
matic change will be heterogeneous within communities.
This could cause temporal reorganization of the structure
and composition of plant communities, potentially altering
direct and indirect interactions among plant species and
between plant and animal species, and ecosystem services
[24,34,81–83].
(b) Phenological sensitivity to spring temperature tends
to decrease with latitude
The consequences of phenological shifts can be further com-
plicated by intraspecific variation in phenological sensitivity
to environmental cues [33,38]. For instance, using 20 years
of observational data, Preve
´yet al. [44] found that the pheno-
logical sensitivity to temperature of tundra plants at colder,
higher latitudes was greater than at warmer, lower latitudes.
However, contrary to such studies, we found that plants in
warmer, lower-latitude regions tended to be more pheno-
logically sensitive to temperature, especially for fruiting
(figure 3). We hypothesize that this is due to the lower and
less predictable winter and spring climates of the north-
eastern United States. In such environments, dynamic
phenological tracking of spring temperatures (i.e. high phe-
nological sensitivity to temperature) presents high risks to
reproductive success, because warm periods may often be fol-
lowed by periods of dramatic chilling [40]. At lower latitudes,
the advent and progression of spring is less variable and
average temperatures are higher; thus phenological tracking
of temperature is less risky (electronic supplementary
material, figures S1 and S2). Indeed, species exhibited a
larger amount of variability in their responses to temperature
in the warmer, lower-latitudinal parts of their ranges.
Climate and phenology might play different roles in filter-
ing species assemblages in regions with longer growing
seasons than in regions where the growing season is short
and reproductive phenologies are strongly constrained by
shorter freeze-free periods [35]. Indeed, studies synthesizing
plot-level observational data have suggested phenological
sensitivity of plant communities to warming may be
positively correlated with mean annual temperature, but
assemblage mean
assembla
g
e s.d.
cool warm-mixed
−4
−3
−2
−1
0
0
1
2
3
climate re
g
ion
phenological sensitivity
(days/°C)
phenological sensitivity
(days/°C)
flowering fruiting
**
*
**
(b)
(a)
Figure 3. Estimated community phenological sensitivities to spring (March –
May) air temperature in cold-temperate versus mixed-warm temperate climate
zones (depicted in figure 1) using a Bayesian hierarchical model (Model 2).
Panel (a) depicts species assemblage means, and panel (b) depicts assemblage
standard deviations. Black and grey represent flowering and fruiting stages, and
thick and thin bars represent 50 and 80% posterior credible intervals, respect-
ively. Comparisons with a posterior probability greater than or equal to 0.8 and
less than 0.95 are depicted with a ‘B’, comparisons with probabilities greater
than or equal to 0.95 and less than 0.99 are depicted with ‘*’, and comparisons
with probabilities greater than or equal to 0.99 are depicted with ‘**’.
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
7
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
negatively correlated with seasonal temperature range (i.e.
variability) in Europe [84] and China [85]. Alternatively, it
is possible that plants in warmer climates exist closer to
their response thresholds in terms of phenology, and thus
react more dramatically to small changes in temperature.
However, Ko
¨rner & Basler [86] noted that cherry cultivars
from regions with less variable spring temperatures flowered
earlier in common gardens, suggesting phenological sensi-
tivity does vary with climate. Plants in regions with high
spring temperature variability also tend to be less pheno-
logically sensitive in terms of leaf out and bud break to
temperature than those in less variable climates [40]. Our
results demonstrating that phenological sensitivity to temp-
erature is higher in areas with low standard deviation of
intra-annual temperature and inter-annual variation in
spring temperature and high mean annual temperature
support these findings.
(c) Consequences of variation in phenological responses
across species ranges
Our results imply that with equal warming, individuals in
lower-latitude populations will advance their reproductive
phenology more dramatically than those at higher latitudes.
This observed variation in phenological response may reflect
adaptation to local climatic conditions, especially in annual
species. We found a large amount of regional variation in
phenological sensitivity that was not clearly linked to climate
or latitude. These regional differences were not explained
by species turnover, but rather suggest the presence of
inter-population variation driven by local adaptation or
phenological plasticity (figure 4). Further, within-species
variation in phenological sensitivity between regions was of
similar magnitude to differences in sensitivity between species
(figure 5). Other studies examining leaf out and senescence in
trees also have shown that individuals from geographically
and climatically separated populations differ in their
phenology even when grown in common gardens [40,87,88].
Because the eastern United States is experiencing geo-
graphically variable climatic change, the heterogeneity in
phenological responses to warming that we observed within
and among species may have important consequences for
plant communities in the near future. Colder, climatically
variable high-latitude regions are experiencing dispropor-
tionate warming and climatic homogenization (i.e. reduced
standard variation of intra-annual temperature), while
warmer, climatically less variable more southerly regions are
experiencing increases in intra-annual temperature variability
(electronic supplementary material, figure S1). These climatic
changes could alter patterns of overlap in reproductive timing
among species in a community and across their individual
ranges. Changes in phenological overlap across ranges could
have direct consequences for adaptive evolution and species
resilience to current and impending climatic changes, as
AP MA NE
assemblage meanassemblage s.d.
turnover +
local adaptation/plasticity
species
turnover only
phase
flowering
fruiting
SE OC AP MA NESE OC
1.0
1.5
2.0
2.5
–4.0
2.0
0
NEON domain NEON domain
phenological sensitivity (days/°C) phenological sensitivity (days/°C)
(c)
(a)
(b)(d)
Figure 4. Differences in sensitivity to spring temperatures between NEON domains are driven by regional variation due to local adaptation or plasticity, not com-
munity turnover between regions. This applies to both differences in species assemblage mean phenological sensitivities (a,b) and assemblage standard deviations
(c,d). Panels (a) and (b) represent best estimates of regional variations in phenological sensitivity incorporating species identity, NEON Domain and NEON Domain
species identity as random effects (
b
1
[s]
,
b
2
[r]
,
b
3
[sr]
) and panels (c) and (d) represent estimates incorporating only species-level effects (
b
1
[s]
). This means that in
panels (c) and (d), phenological sensitivity is assumed to be constant within a species across domains. Estimates for flowering are represented in black and
estimates for fruiting are represented in grey. Thick and thin bars represent 50 and 80% posterior credible intervals, respectively, from a hierarchical Bayesian
model (Model 1).
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
8
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
gene flow may increase among previously (temporally) iso-
lated populations of some species and decrease among
others [41–43,89–91]. Decreased phenological overlap could
genetically isolate fringe populations, potentially leading to
local extirpation and species range contractions. Moreover,
phenologically sensitive plants at lower latitudes may
especially be at risk due to increasingly variable temperatures
and increased probabilities of phenological mismatch with
mutualists [92]. However, we cannot fully rule out the
possibility that individuals of long-lived perennial species
may also be able to acclimate their phenological sensitivity
to changing climatic conditions over longer periods of time.
Some of the variation we observe in phenological sensi-
tivity to temperature across and within species may be due
to differences in microclimate. However, the lack of accurate
location data for most historic specimens limited our ability
to infer fine-scale climate, necessitating coarser, county-level
analyses. Also, we cannot ignore the possibility that the phe-
nological trends we observed are unique to the species that
we studied and/or reflect biases in herbarium collections
[93]. To minimize effects of spatial bias and uncertainty, we
studied common, well-collected species and accounted for
climatic heterogeneity present in each sampling locale in
our models. However, these taxa are not necessarily represen-
tative of species assemblages across regions, and our analyses
do not account explicitly for spatial and temporal sampling
biases; some regional differences could be due to differing
patterns of collection across space and time. Including
county-level random effects as we have done here minimizes
the impacts of these biases but does not eliminate them
altogether. It is possible that different patterns of pheno-
logical sensitivity may be observed across species ranges
depending on how climatic and/or geographical regions
are delimited, though testing an alternative threshold yielded
similar results, suggesting that the patterns we observe are
robust to spatial binning choices. Additionally, we addressed
crowdsourcing bias by including crowdsourcer random
effects and removing observations for crowdsourcers with
low reliability scores even though phenological data
collected by citizen scientists do not differ significantly in
quality from those collected by experts [46,55]. Lastly,
although spring temperature is a critical driver of flowering
phenology in temperate climates, we cannot fully exclude
the possibility that other variables correlated with latitude
or mean spring temperature may determine observed
variance in phenological sensitivity [46,73,75,94– 97]. For
example, spring temperature tends to be highly correlated
with mean annual and mean monthly temperatures in east-
ern North America (electronic supplementary material,
figure S4). Photoperiod or snow melt may further influence
and alter species phenological responses [98–103]. Future
research into how these environmental cues interact to trigger
phenological events is necessary and will greatly improve our
understanding of plant phenology.
5. Conclusion
Building on previous phenological research by scoring mul-
tiple phenological traits across over 8000 herbarium
specimens spanning 120 years, we have demonstrated that
phenological sensitivity can vary greatly across species’
ranges. This variance may be attributed to adaptation or
acclimation to local climates. The large amount of within-
species variation in phenological sensitivity that we observed
underlines the complex and contingent nature of phenological
sensitivities. Phenological responses of individual species to
climate are not stable phenotypic traits, but instead emerge
from a multitude of potentially reciprocal interactions between
organisms and their environment. Populations in different
regions could have differences in frequencies of genes that
control how climate affects the timing of development or
differences in microhabitat distributions between regions that
alter how populations experience local climate. The regions
themselves may have differences in unmeasured environ-
mental factors that interact with responses to temperature or
differences in species interactions that may alter phenological
signalling. The circumstances and extent to which these or
other factors explain regional variation in species responses
to climate is currently unknown. To untangle the roles of
ecological and evolutionary processes governing the hetero-
geneous phenological responses of plant species to warming,
researchers will have to take advantage of new techniques
and datasets. In addition to continued field observations and
laboratory analysis of mechanisms responsible for flowering
and fruiting, herbarium specimens can provide a comprehen-
sive, nuanced picture of phenological responses to ongoing
climatic change across many species. Our study further
demonstrates that we can now harness the treasure trove of
between
domain
(sb2)
between
species
(sb1)
within
species
(sb3)
0
10
20
30
40
0
1
2
3
variance component
variance (s)variance (s)
intercept (DOY)phenolo
g
ical sensitivit
y
(da
y
s/°C)
flowering fruiting
(b)
(a)
Figure 5. Variance components from Model 1, allowing comparisons of varia-
bility between NEON domains, between species and within species. Variance in
the intercept (a) represents variability in flowering (black) or fruiting times (grey)
at 7.48C spring temperature and 216 m elevation above sea level, median con-
ditions for the specimens. Variance in phenological sensitivity to spring air
temperature (days/8C) represents variability in the slope of the linear relationship
between spring temperature and flowering or fruiting times.
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
9
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
information in herbaria across the world to examine hundreds,
if not thousands of species across myriad plant lineages, habi-
tats and regions. Such efforts will be critical to enhance our
ability to forecast future changes in plant assemblages across
space and time in an era of accelerating climate change
[104,105].
Ethics. The use and collection of data by citizen scientists were
approved by an ethics review committee at the University of
Waterloo (ORE no. 21647).
Data accessibility. Data and R code are available from the Harvard Forest
Data Archive (http://harvardforest.fas.harvard.edu/dataarchive,
dataset HF309).
Authors’ contributions. C.C.D. conceived the study; D.S.P., C.C.D., A.C.W.
and E.L. designed the study; D.S.P. and A.C.W. collected data; D.S.P.,
A.M.E. and I.B. analysed the data; D.S.P. drafted the first version of
the manuscript; C.C.D. and D.S.P. made first substantial revisions
to this draft, and all authors contributed to subsequent revisions.
Competing interests. We declare we have no competing interests.
Funding. This study was funded as part of the New England Vascular
Plant Project to C.C.D. (NSF-DBI: EF1208835), NSF-DEB 1754584 to
C.C.D. and A.M.E., and an NSERC Discovery Grant to E.L.
(RGPIN-2015-04543). A.M.E.’s participation in this project was sup-
ported by Harvard Forest. D.S.P.’s contribution was supported by
the Harvard University Herbaria and NSF-DEB 1754584. I.B.’s contri-
bution was supported by the National Science Foundation’s
Postdoctoral Research Fellowship in Biology (NSF-DBI-1711936).
Acknowledgements. The authors thank X. Feng, S. Worthington, and
members of the Davis lab for their invaluable insights and comments
on the project and manuscript.
References
1. Field CB et al. 2014 Climate change 2014: impacts,
adaptation, and vulnerability. Part A: global and
sectoral aspects. In Contribution of Working Group II
to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change. Geneva,
Switzerland: IPCC.
2. Root TL, Price JT, Hall KR, Schneider SH, Rosenzweig
C, Pounds JA. 2003 Fingerprints of global warming
on wild animals and plants. Nature 421, 57– 60.
(doi:10.1038/nature01333)
3. Parmesan C, Yohe G. 2003 A globally coherent
fingerprint of climate change impacts across natural
systems. Nature 421, 37–42. (doi:10.1038/
nature01286)
4. Miller-Rushing AJ, Primack RB, Primack D, Mukunda
S. 2006 Photographs and herbarium specimens as
tools to document phenological changes in
response to global warming. Am. J. Bot. 93,
1667–1674. (doi:10.3732/ajb.93.11.1667)
5. Menzel A, Sparks TH, Estrella N, Roy DB. 2006
Altered geographic and temporal variability in
phenology in response to climate change. Glob.
Ecol. Biogeogr. 15, 498–504. (doi:10.1111/
j.1466-822X.2006.00247.x)
6. Chambers LE et al. 2013 Phenological changes in
the southern hemisphere. PLoS ONE 8, e75514.
(doi:10.1371/journal.pone.0075514)
7. Ovaskainen O, Skorokhodova S, Yakovleva M,
Sukhov A, Kutenkov A, Kutenkova N, Shcherbakov A,
Meyke E, del Delgado M.2013 Community-level
phenological response to climate change. Proc. Natl
Acad. Sci. USA 110, 13 434 –13 439. (doi:10.1073/
pnas.1305533110)
8. Willis CG, Ruhfel B, Primack RB, Miller-Rushing AJ,
Davis CC. 2008 Phylogenetic patterns of species loss
in Thoreau’s woods are driven by climate change.
Proc. Natl Acad. Sci. USA 105, 17 029–17 033.
(doi:10.1073/pnas.0806446105)
9. Inouye D. 2008 Effects of climate change on
phenology, frost damage, and floral abundance of
montane wildflowers. Ecology 89, 353– 362.
(doi:10.1890/06-2128.1)
10. Wheeler HC, Høye TT, Schmidt NM, Svenning JC,
Forchhammer MC. 2015 Phenological mismatch
with abiotic conditions—implications for flowering
in Arctic plants. Ecology 96, 775–787. (doi:10.
1890/14-0338.1)
11. White MA, Nemani RR. 2003 Canopy duration has
little influence on annual carbon storage in the
deciduous broad leaf forest. Glob. Change Biol. 9,
967–972. (doi:10.1046/j.1365-2486.2003.00585.x)
12. Richardson AD, Hollinger DY, Dail DB, Lee JT,
Munger JW, O’keefe J. 2009 Influence of spring
phenology on seasonal and annual carbon balance
in two contrasting New England forests. Tree
Physiol. 29, 321–331. (doi:10.1093/treephys/
tpn040)
13. Barford CC et al. 2001 Factors controlling long- and
short-term sequestration of atmospheric CO
2
in a
mid-latitude forest. Science 294, 1688– 1691.
(doi:10.1126/science.1062962)
14. Fitzjarrald DR, Acevedo OC, Moore KE. 2001 Climatic
consequences of leaf presence in the eastern United
States. J. Clim. 14, 598–614. (doi:10.1175/1520-
0442(2001)014,0598:CCOLPI.2.0.CO;2)
15. Inouye DW, Barr B, Armitage KB, Inouye BD. 2000
Climate change is affecting altitudinal migrants and
hibernating species. Proc. Natl Acad. Sci. USA 97,
1630–1633. (doi:10.1073/pnas.97.4.1630)
16. Visser ME, Both C. 2005 Shifts in phenology due to
global climate change: the need for a yardstick.
Proc. R. Soc. B 272, 2561–2569. (doi:10.1098/rspb.
2005.3356)
17. Doiron M, Gauthier G, Le
´vesque E. 2015 Trophic
mismatch and its effects on the growth of young in
an Arctic herbivore. Glob. Change Biol. 21,
4364–4376. (doi:10.1111/gcb.13057)
18. Post E, Forchhammer MC. 2008 Climate change
reduces reproductive success of an Arctic herbivore
through trophic mismatch. Phil. Trans. R. Soc. B
363, 2367–2373. (doi:10.1098/rstb.2007.2207)
19. Post E, Pedersen C, Wilmers CC, Forchhammer MC.
2008 Warming, plant phenology and the spatial
dimension of trophic mismatch for large herbivores.
Proc. R. Soc. B 275, 2005–2013. (doi:10.1098/rspb.
2008.0463)
20. Plard F, Gaillard J-M, Coulson T, Hewison AJM,
Delorme D, Warnant C, Bonenfant C. 2014
Mismatch between birth date and vegetation
phenology slows the demography of roe deer.
PLoS Biol. 12, e1001828. (doi:10.1371/journal.pbio.
1001828)
21. Kharouba HM, Ehrle
´n J, Gelman A, Bolmgren K,
Allen JM, Travers SE, Wolkovich EM. 2018 Global
shifts in the phenological synchrony of species
interactions over recent decades. Proc. Natl Acad.
Sci. USA 115, 5211–5216. (doi:10.1073/pnas.
1714511115)
22. Navarro-Cano JA, Karlsson B, Posledovich D,
Toftegaard T, Wiklund C, Ehrle
´n J, Gotthard K. 2015
Climate change, phenology, and butterfly host plant
utilization. Ambio 44, 78– 88. (doi:10.1007/s13280-
014-0602-z)
23. Posledovich D, Toftegaard T, Wiklund C, Ehrle
´nJ,
Gotthard K. 2018 Phenological synchrony
between a butterfly and its host plants:
experimental test of effects of spring temperature.
J. Anim. Ecol. 87, 150–161. (doi:10.1111/1365-
2656.12770)
24. Burkle LA, Marlin JC, Knight TM. 2013 Plant-
pollinator interactions over 120 years: loss of
species, co-occurrence, and function. Science 339,
1611–1615. (doi:10.1126/science.1232728)
25. Kudo G, Ida TY. 2013 Early onset of spring increases
the phenological mismatch between plants and
pollinators. Ecology 94, 2311– 2320. (doi:10.1890/
12-2003.1)
26. Iler AM, Inouye DW, Høye TT, Miller-Rushing AJ,
Burkle LA, Johnston EB. 2013 Maintenance of
temporal synchrony between syrphid flies and floral
resources despite differential phenological responses
to climate. Glob. Change Biol. 19, 2348–2359.
(doi:10.1111/gcb.12246)
27. Visser ME, van Noordwijk AJ, Tinbergen JM, Lessells
CM. 1998 Warmer springs lead to mistimed
reproduction in great tits (Parus major). Proc. R. Soc.
Lond. B 265, 1867–1870. (doi:10.1098/rspb.1998.
0514)
28. Mayor SJ et al. 2017 Increasing phenological
asynchrony between spring green-up and arrival of
migratory birds. Sci. Rep. 7, 1902. (doi:10.1038/
s41598-017-02045-z)
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
10
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
29. Wolkovich EM, Cleland EE. 2011 The phenology of
plant invasions: a community ecology perspective.
Front. Ecol. Environ. 9, 287– 294. (doi:10.1890/
100033)
30. Craine JM, Wolkovich EM, Towne EG, Kembel SW.
2012 Flowering phenology as a functional trait in a
tallgrass prairie. New Phytol. 193, 673– 682.
(doi:10.1111/j.1469-8137.2011.03953.x)
31. Colautti RI, A
˚
gren J, Anderson JT, Colautti RI, Jon A,
Colautti RI. 2017 Phenological shifts of native and
invasive species under climate change: insights from
the Boechera– Lythrum model. Phil. Trans. R. Soc. B
372, 20160032. (doi:10.1098/rstb.2016.0032)
32. Pau S, Wolkovich EM, Cook BI, Davies TJ, Kraft NJB,
Bolmgren K, Betancourt JL, Cleland EE. 2011
Predicting phenology by integrating ecology,
evolution and climate science. Glob. Change Biol.
17, 3633–3643. (doi:10.1111/j.1365-2486.2011.
02515.x)
33. Wolkovich EM, Cook BI, Davies TJ. 2014 Progress
towards an interdisciplinary science of plant
phenology: building predictions across space, time
and species diversity. New Phytol. 201, 1156– 1162.
(doi:10.1111/nph.12599)
34. CaraDonna PJ, Iler AM, Inouye DW. 2014 Shifts in
flowering phenology reshape a subalpine plant
community. Proc. Natl Acad. Sci. USA 111,
4916–4921. (doi:10.1073/pnas.1323073111)
35. Diez JM, Iba
´n
˜ez I, Miller-Rushing AJ, Mazer SJ,
Crimmins TM, Crimmins MA, Bertelsen CD, Inouye
DW. 2012 Forecasting phenology: from species
variability to community patterns. Ecol. Lett. 15,
545–553. (doi:10.1111/j.1461-0248.2012.01765.x)
36. Banta JA, Ehrenreich IM, Gerard S, Chou L, Wilczek
A, Schmitt J, Kover PX, Purugganan MD. 2012
Climate envelope modelling reveals intraspecific
relationships among flowering phenology, niche
breadth and potential range size in Arabidopsis
thaliana.Ecol. Lett. 15, 769–777. (doi:10.1111/
j.1461-0248.2012.01796.x)
37. Wilczek AM et al. 2009 Effects of genetic
perturbation on seasonal life history plasticity.
Science 323, 930–934. (doi:10.1126/science.
1165826)
38. Matthews ER, Mazer SJ. 2016 Historical changes in
flowering phenology are governed by
temperature precipitation interactions in a
widespread perennial herb in western North
America. New Phytol. 210, 157–167. (doi:10.1111/
nph.13751)
39. Barros VR et al. 2014 Climate change 2014: impacts,
adaptation, and vulnerability. Part B: regional
aspects.. In Contribution of Working Group II to the
Fifth Assessment Report of the Intergovernmental
Panel on Climate Change. Geneva, Switzerland: IPCC.
40. Zohner CM, Renner SS. 2014 Common garden
comparison of the leaf-out phenology of woody
species from different native climates, combined
with herbarium records, forecasts long-term change.
Ecol. Lett. 17, 1016–1025. (doi:10.1111/ele.12308)
41. Lenormand T. 2002 Gene flow and the limits to
natural selection. Trends Ecol. Evol. 17,183 – 189.
(doi:10.1016/S0169-5347(02)02497-7)
42. Alleaume-Benharira M, Pen IR, Ronce O. 2006
Geographical patterns of adaptation within a
species’ range: interactions between drift and gene
flow. J. Evol. Biol. 19, 203– 215. (doi:10.1111/
j.1420-9101.2005.00976.x)
43. Sexton JP, Strauss SY, Rice KJ. 2011 Gene flow
increases fitness at the warm edge of a species’
range. Proc. Natl Acad. Sci. USA 108,
11 704–11 709. (doi:10.1073/pnas.1100404108)
44. Preve
´yJet al. 2017 Greater temperature sensitivity
of plant phenology at colder sites: implications for
convergence across northern latitudes. Glob. Change
Biol. 23, 2660–2671. (doi:10.1111/gcb.13619)
45. Riihima
¨ki M, Savolainen O. 2004 Environmental and
genetic effects on flowering differences between
northern and southern populations of Arabidopsis
lyrata (Brassicaceae). Am. J. Bot. 91, 1036– 1045.
(doi:10.3732/ajb.91.7.1036)
46. Willis CG et al. 2017 CrowdCurio: an online
crowdsourcing platform to facilitate climate change
studies using herbarium specimens. New Phytol.
215, 479–488. (doi:10.1111/nph.14535)
47. Willis CG et al. 2017 Old plants, new tricks:
phenological research using herbarium specimens.
Trends Ecol. Evol. 32, 531– 546. (doi:10.1016/j.tree.
2017.03.015)
48. Davis CC, Willis CG, Connolly B, Kelly C, Ellison AM.
2015 Herbarium records are reliable sources of
phenological change driven by climate and provide
novel insights into species’ phenological cueing
mechanisms. Am. J. Bot. 102, 1599–1609. (doi:10.
3732/ajb.1500237)
49. Primack D, Imbres C, Primack RB, Miller-Rushing AJ,
Del Tredici P. 2004 Herbarium specimens
demonstrate earlier flowering times in response to
warming in Boston. Am. J. Bot. 91, 1260– 1264.
(doi:10.3732/ajb.91.8.1260)
50. Robbirt KM, Davy AJ, Hutchings MJ, Roberts DL.
2011 Validation of biological collections as a source
of phenological data for use in climate change
studies: a case study with the orchid Ophrys
sphegodes.J. Ecol. 99, 235 – 241. (doi:10.1111/
j.1365-2745.2010.01727.x)
51. Vellend M, Brown CD, Kharouba HM, McCune JL,
Myers-Smith IH. 2013 Historical ecology: using
unconventional data sources to test for effects of
global environmental change. Am. J. Bot. 100,
1294–1305. (doi:10.3732/ajb.1200503)
52. Sweeney PW, Starly B, Morris PJ, Xu Y, Jones A,
Radhakrishnan S, Grassa CJ, Davis CC. 2018
Large-scale digitization of herbarium specimens:
development and usage of an automated, high-
throughput conveyor system. Taxon 67, 165– 178.
(doi:10.12705/671.9)
53. Yost JM et al. 2018 Digitization protocol for scoring
reproductive phenology from herbarium specimens
of seed plants. Appl. Plant Sci. 6,e1022. (doi:10.
1002/aps3.1022)
54. Pearse WD, Davis CC, Inouye DW, Primack RB,
Davies TJ. 2017 A statistical estimator for
determining the limits of contemporary and historic
phenology. Nat. Ecol. Evol. 1, 1876. (doi:10.1038/
s41559-017-0350-0)
55. Williams AC, Goh J, Willis CG, Ellison AM, Brusuelas
JH, Davis CC, Law E. 2017 De
´ja vu: characterizing
worker reliability using task consistency. In Proc. 5th
AAAI Conf. Human Computation and Crowdsourcing,
HCOMP 2017, 23– 26 October 2017, Que
´bec City,
Que
´bec, Canada (eds S Dow, AT Kalai),
pp. 197–205. Palo Alto, CA: AAAI Press.
56. Schorn C, Weber E, Bernardos R, Hopkins C, Davis C.
2016 The New England Vascular Plants Project:
295 000 specimens and counting. Rhodora 118,
324. (doi:10.3119/15-34)
57. SouthEast Regional Network of Expertise and
Collections. 2017 SERNEC Data Portal. See http://
sernecportal.org/portal/index.php (accessed 28 June
2018).
58. Hart R, Salick J, Ranjitkar S, Xu J. 2014 Herbarium
specimens show contrasting phenological responses
to Himalayan climate. Proc. Natl Acad. Sci. USA 111,
10 615–10 619. (doi:10.1073/pnas.1403376111)
59. Park DS, Davis CC. 2017 Implications and
alternatives of assigning climate data to
geographical centroids. J. Biogeogr. 44,
2188–2198. (doi:10.1111/jbi.13029)
60. Pen
˜A JE, Duncan R, Browning H. 1996 Seasonal
abundance of Phyllocnistis citrella (Lepidoptera:
Gracillariidae) and its parasitoids South Florida
citrus. Environ. Entomol. 25, 698–702. (doi:10.
1093/ee/25.3.698)
61. Baechler MC, Gilbride TL, Cole PC, Hefty MG, Ruiz K.
2010 Guide to determining climate regions by
county. Build. Am. Best Pract. Ser. 7, 1– 45.
(doi:10.2172/969181)
62. Zhang Y, Bielory L, Georgopoulos PG. 2014 Climate
change effect on Betula (birch) and Quercus (oak)
pollen seasons in the United States.
Int. J. Biometeorol. 58, 909–919. (doi:10.1007/
s00484-013-0674-7)
63. Hobbs NT, Hooten MB. 2015 Bayesian models: a
statistical primer for ecologists. Princeton, NJ:
Princeton University Press.
64. Gelman A, Hill J. 2006 Data analysis using regression
and multilevel/hierarchical models. Cambridge, UK:
Cambridge University Press.
65. Hoffman MD, Gelman A. 2014 The No-U-turn
sampler: adaptively setting path lengths in
Hamiltonian Monte Carlo. J. Mach. Learn. Res. 15,
1593–1623.
66. Carpenter B et al. 2017 Stan: a probabilistic
programming language. J. Stat. Softw. 76, 1– 32.
(doi:10.18637/jss.v076.i01)
67. Stan Development Team. 2018 RStan: the R
interface to Stan. R package version 2.18.1. See
http://mc-stan.org/.
68. R Core Team. 2018 R: a language and environment
for statistical computing. Vienna, Austria: R
Foundation for Statistical Computing. See http://
www.R-project.org.
69. Betancourt M, Girolami M. 2015 Hamiltonian Monte
Carlo for hierarchical models. Curr. Trends Bayesian
Methodol. Appl. 79, 30. (doi:10.1201/b18502-5)
70. Gabry J, Simpson D, Vehtari A, Betancourt M,
Gelman A. 2017 Visualization in Bayesian workflow.
See http://arxiv.org/abs/1709.01449.
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
11
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
71. Park DS, Breckheimer I, Williams AC, Law E, Ellison
AM, Davis CC. 2018 Assessing plant phenological
patterns in the eastern United States over the last
120 years. Harvard Forest Data Archive, HF309.
(doi:10.6073/pasta/
f6afa728bb5edfd79f458d7d5e23f559)
72. Roberts AMI, Tansey C, Smithers RJ, Phillimore AB.
2015 Predicting a change in the order of spring
phenology in temperate forests. Glob. Change Biol.
21, 2603–2611. (doi:10.1111/gcb.12896)
73. Yu H, Luedeling E, Xu J. 2010 Winter and
spring warming result in delayed spring phenology
on the Tibetan Plateau. Proc. Natl Acad. Sci. USA
107, 22 151–22 156. (doi:10.1073/pnas.
1012490107)
74. Wolfe DW, Schwartz MD, Lakso AN, Otsuki Y,
Pool RM, Shaulis NJ. 2005 Climate change and
shifts in spring phenology of three horticultural
woody perennials in northeastern USA.
Int. J. Biometeorol. 49, 303–309. (doi:10.1007/
s00484-004-0248-9)
75. Miller-Rushing AJ, Primack RB. 2008 Global
warming and flowering times in Thoreau’s Concord:
a community perspective. Ecology 89, 332 – 341.
(doi:10.1890/07-0068.1)
76. Bliss LC. 1962 Adaptations of arctic and alpine
plants to environmental conditions. Arctic 15,
117–144. (doi:10.14430/arctic3564)
77. Billings WD, Mooney HA. 1968 The ecology of arctic
and alpine plants. Biol. Rev. 43, 481–529. (doi:10.
1111/j.1469-185X.1968.tb00968.x)
78. Oberbauer SF et al. 2013 Phenological response of
tundra plants to background climate variation
tested using the International Tundra Experiment.
Phil. Trans. R. Soc. B 368, 20120481. (doi:10.1098/
rstb.2012.0481)
79. Fitter AH, Fitter RSR. 2002 Rapid changes in
flowering time in British plants. Science 296,
1689–1691. (doi:10.1126/science.1071617)
80. Murray MB, Cannell MGR, Smith RI. 1989 Date of
budburst of fifteen tree species in Britain following
climatic warming. J. Appl. Ecol. 26, 693– 700.
(doi:10.2307/2404093)
81. Theobald EJ, Breckheimer I, HilleRisLambers J. 2017
Climate drives phenological reassembly of a
mountain wildflower meadow community. Ecology
98, 2799–2812. (doi:10.1002/ecy.1996)
82. Miller-Rushing AJ, Inouye DW. 2009 Variation in the
impact of climate change on flowering phenology
and abundance: an examination of two pairs of
closely related wildflower species. Am. J. Bot. 96,
1821–1829. (doi:10.3732/ajb.0800411)
83. Memmott J, Craze PG, Waser NM, Price MV.
2007 Global warming and the disruption of
plant–pollinator interactions. Ecol. Lett.
10, 710–717. (doi:10.1111/j.1461-0248.
2007.01061.x)
84. Lapenis A, Henry H, Vuille M, Mower J. 2014
Climatic factors controlling plant sensitivity to
warming. Clim. Change 122, 723–734. (doi:10.
1007/s10584-013-1010-2)
85. Zhang H, Yuan W, Liu S, Dong W, Fu Y. 2015
Sensitivity of flowering phenology to changing
temperature in China. J. Geophys. Res. Biogeosci.
120, 1658–1665. (doi:10.1002/2015JG003112)
86. Ko
¨rner C, Basler D. 2010 Phenology under global
warming. Science 327,1461– 1462. (doi:10.1126/
science.1186473)
87. Zohner CM, Benito BM, Fridley JD, Svenning J,
Renner SS. 2017 Spring predictability explains
different leaf-out strategies in the woody floras of
North America, Europe and East Asia. Ecol. Lett. 20,
452–460. (doi:10.1111/ele.12746)
88. Zohner CM, Renner SS. 2017 Innately shorter
vegetation periods in North American species
explain native– non-native phenological
asymmetries. Nat. Ecol. Evol. 1, 1655. (doi:10.1038/
s41559-017-0307-3)
89. Savolainen O, Pyha
¨ja
¨rvi T, Knu¨rr T. 2007 Gene flow
and local adaptation in trees. Annu. Rev. Ecol. Evol.
Syst. 38, 595– 619. (doi:10.1146/annurev.ecolsys.
38.091206.095646)
90. Nagy ES, Rice KJ. 1997 Local adaptation in two
subspecies of an annual plant: implications for
migration and gene flow. Evolution 51,
1079–1089. (doi:10.1111/j.1558-5646.1997.
tb03955.x)
91. Fox GA. 2003 Assortative mating and plant
phenology: evolutionary and practical consequences.
Evol. Ecol. Res. 5, 1– 18.
92. Thomson JD. 2010 Flowering phenology, fruiting
success and progressive deterioration of pollination
in an early-flowering geophyte. Phil. Trans. R. Soc. B
365, 3187–3199. (doi:10.1098/rstb.2010.0115)
93. Daru BH et al. 2018 Widespread sampling biases
in herbaria revealed from large-scale digitization.
New Phytol. 217, 939–955. (doi:10.1111/nph.
14855)
94. Richardson AD, Bailey AS, Denny EG, Martin CW,
O’Keefe J. 2006 Phenology of a northern hardwood
forest canopy. Glob. Change Biol. 12, 1174– 1188.
(doi:10.1111/j.1365-2486.2006.01164.x)
95. Ellwood ER, Temple SA, Primack RB, Bradley NL,
Davis CC. 2013 Record-breaking early flowering in
the eastern United States. PLoS ONE 8,1–9.
(doi:10.1371/journal.pone.0053788)
96. Augspurger CK. 2013 Reconstructing patterns of
temperature, phenology, and frost damage over 124
years: spring damage risk is increasing. Ecology 94,
41–50. (doi:10.1890/12-0200.1)
97. Gu¨sewell S, Furrer R, Gehrig R, Pietragalla B. 2017
Changes in temperature sensitivity of spring phenology
with recent climatewarming in Switzerland are related
to shifts of the preseason. Glob. Change Biol. 23,
5189– 5202. (doi:10.1111/gcb.13781)
98. Bjorkman AD, Elmendorf SC, Beamish AL, Vellend
M, Henry GHR. 2015 Contrasting effects of warming
and increased snowfall on Arctic tundra plant
phenology over the past two decades. Glob. Change
Biol. 21, 4651–4661. (doi:10.1111/gcb.13051)
99. Way DA, Montgomery RA. 2015 Photoperiod
constraints on tree phenology, performance and
migration in a warming world. Plant. Cell Environ.
38, 1725–1736. (doi:10.1111/pce.12431)
100. Nakagawa H, Yamagishi J, Miyamoto N, Motoyama
M, Yano M, Nemoto K. 2005 Flowering response of
rice to photoperiod and temperature: a QTL analysis
using a phenological model. Theor. Appl. Genet.
110, 778–786. (doi:10.1007/s00122-004-1905-4)
101. Dunne JA, Harte J, Taylor KJ. 2003 Subalpine
meadow flowering phenology responses to climate
change: integrating experimental and gradient
methods. Ecol. Monogr. 73, 69–86. (doi:10.1890/
0012-9615(2003)073[0069:SMFPRT]2.0.CO;2)
102. Wolkovich EM, Davies TJ, Schaefer H, Cleland EE,
Cook BI, Travers SE, Willis CG, Davis CC. 2013
Temperature-dependent shifts in phenology
contribute to the success of exotic species with
climate change. Am. J. Bot. 100, 1407–1421.
(doi:10.3732/ajb.1200478)
103. Park IW, Mazer SJ. In press. Overlooked climate
parameters best predict flowering onset: assessing
phenological models using the elastic net. Glob.
Change Biol. (doi:10.1111/gcb.14447)
104. Chuine I, Beaubien EG. 2001 Phenology is a major
determinant of tree species range. Ecol. Lett. 4,
500–510. (doi:10.1046/j.1461-0248.2001.00261.x)
105. Chuine I. 2010 Why does phenology drive species
distribution? Phil. Trans. R. Soc. B 365, 3149– 3160.
(doi:10.1098/rstb.2010.0142)
rstb.royalsocietypublishing.org Phil. Trans. R. Soc. B 374: 20170394
12
on November 27, 2018http://rstb.royalsocietypublishing.org/Downloaded from
A preview of this full-text is provided by The Royal Society.
Content available from Philosophical Transactions B
This content is subject to copyright.