ArticlePDF Available

Herbarium records are reliable sources of phenological change driven by climate and provide novel insights into species' phenological cueing mechanisms


Abstract and Figures

Premise of the study: Climate change has resulted in major changes in the phenology of some species but not others. Long-term field observational records provide the best assessment of these changes, but geographic and taxonomic biases limit their utility. Plant specimens in herbaria have been hypothesized to provide a wealth of additional data for studying phenological responses to climatic change. However, no study to our knowledge has comprehensively addressed whether herbarium data are accurate measures of phenological response and thus applicable to addressing such questions. Methods: We compared flowering phenology determined from field observations (years 1852-1858, 1875, 1878-1908, 2003-2006, 2011-2013) and herbarium records (1852-2013) of 20 species from New England, United States. Key results: Earliest flowering date estimated from herbarium records faithfully reflected field observations of first flowering date and substantially increased the sampling range across climatic conditions. Additionally, although most species demonstrated a response to interannual temperature variation, long-term temporal changes in phenological response were not detectable. Conclusions: Our findings support the use of herbarium records for understanding plant phenological responses to changes in temperature, and also importantly establish a new use of herbarium collections: inferring primary phenological cueing mechanisms of individual species (e.g., temperature, winter chilling, photoperiod). These latter data are lacking from most investigations of phenological change, but are vital for understanding differential responses of individual species to ongoing climate change.
Content may be subject to copyright.
AMERICAN JOURNAL OF BOTANY 102 ( 10 ): 1 – 11 , 2015 ; © 2015 Botanical Society of America 1
e impacts of global climatic change on organisms have been
well documented in recent years ( Walther et al., 2002 ; Root et al.,
2003 ; Parmesan, 2006 ). Changes in phenological events—i.e., the
timing of speci c life-history events—are used widely to assess
responses of di erent organisms to climate change. Typically, re-
searchers have focused on relatively common phenological events
that are easily measured and have a history of regular observation
such as leaf-out,  owering, or fruiting in plants, or eclosion and mi-
gration in animals ( Sparks and Carey, 1995 ; Bradley et al., 1999 ;
Fitter and Fitter, 2002 ; Menzel, 2002 ; Parmesan and Yohe, 2003 ;
Root et al., 2003 ; Parmesan, 2007 ; Bertin, 2008 ; Miller-Rushing
et al., 2008 ; Miller-Rushing and Primack, 2008 ; Visser, 2008 ; Willis
et al., 2008 , 2010 ; Körner and Basler, 2010 ; Panchen et al., 2014 ).
However, the paucity of long-term data sets necessary to identify
the in uence of climatic change on phenological events remains
problematic, even for regions where associated climatic data are
available (e.g., long-term temperature trends). Moreover, most
data sets of this nature show a strong geographical and taxonomic
bias—they are largely from temperate regions, mostly include a
small subset of species or functional types within particular as-
semblages (e.g., dominant woody species), and do not sample the
variation in phenological response across the range of a species
( Wolkovich et al., 2014 ).
1 Manuscript received 22 May 2015; revision accepted 11 September 2015.
2 Department of Organismic and Evolutionary Biology and Harvard University Herbaria,
Harvard University, Cambridge, Massachusetts, USA;
3 Harvard University Center for the Environment, Harvard University, Cambridge,
Massachusetts, USA;
4 Massachusetts Natural Heritage and Endangered Species Program, Division of Fisheries &
Wildlife, Massachusetts, USA;
5 Department of Biology, Framingham State University, Framingham, Massachusetts, USA;
6 Harvard Forest, Harvard University, Petersham, Massachusetts, USA
7 Author for correspondence (e-mail:
Herbarium records are reliable sources of phenological
change driven by climate and provide novel insights
into species’ phenological cueing mechanisms
Charles C. Davis 2,7 , Charles G. Willis 2,3 , Bryan Connolly 4,5 , Courtland Kelly 2 , and Aaron M. Ellison 6
PREMISE OF THE STUDY: Climate change has resulted in major changes in the phenology of some species but not others. Long-term  eld observational
records provide the best assessment of these changes, but geographic and taxonomic biases limit their utility. Plant specimens in herbaria have been
hypothesized to provide a wealth of additional data for studying phenological responses to climatic change. However, no study to our knowledge has
comprehensively addressed whether herbarium data are accurate measures of phenological response and thus applicable to addressing such questions.
METHODS: We compared  owering phenology determined from  eld observations (years 1852–1858, 1875, 1878–1908, 2003–2006, 2011–2013) and her-
barium records (1852–2013) of 20 species from New England, United States.
KEY RESULTS: Earliest  owering date estimated from herbarium records faithfully re ected  eld observations of  rst  owering date and substantially in-
creased the sampling range across climatic conditions. Additionally, although most species demonstrated a response to interannual temperature varia-
tion, long-term temporal changes in phenological response were not detectable.
CONCLUSIONS: Our  ndings support the use of herbarium records for understanding plant phenological responses to changes in temperature, and also
importantly establish a new use of herbarium collections: inferring primary phenological cueing mechanisms of individual species (e.g., temperature,
winter chilling, photoperiod). These latter data are lacking from most investigations of phenological change, but are vital for understanding di erential
responses of individual species to ongoing climate change.
KEY WORDS climate change; climate variability; phenology; herbarium specimens; museum collections latest version is at
AJB Advance Article published on October 8, 2015, as 10.3732/ajb.1500237.
Copyright 2015 by the Botanical Society of America
e enormous collections of plants housed in herbaria around
the world provide a potential, largely untapped, alternative body of
data for studying long-term phenological responses to climatic
change ( Vellend et al., 2013 ; Kharouba and Vellend, 2015 ). Her-
barium specimens represent snapshots of phenological events (e.g.,
owering, fruiting) at a speci c place and time. Observations from
numerous specimens collected at multiple locations and times may
allow us to determine whether a given species has changed its phe-
nology in parallel with climate. Previous e orts have used herbar-
ium specimens in this manner ( Primack et al., 2004 ; Miller-Rushing
et al., 2006 ; Robbirt et al., 2011 ; Panchen et al., 2012 ), but only re-
cently has this e ort been scaled-up to investigate patterns of phe-
nology across large numbers of species and vast geographical areas.
For example, Calinger et al. (2013) combined data from herbarium
specimens of 141 species with climatic records to determine that
peak  owering has advanced 2.4 d/°C of warming over the last cen-
tury across ~116 000 km
2 in north-central North America.  ey
further identi ed di erences in phenological responses based on
whether a species was native, its pollination syndrome, growth
form, functional group, and flowering season. Similarly, Everill
et al. (2014) examined ~1600 herbarium records from 1834 to 2008
of 27 common New England tree species.  ey reported that spring
leaf-out dates were strongly associated with spring temperatures
and that tree species leafed out ~2.0 d/°C earlier now than in the
We applaud these e orts to leverage herbarium data to investi-
gate recent e ects of climate change. Despite the promise of these
studies, however, the e cacy and bias of herbarium records as ac-
curate measures of phenological response have seldom been as-
sessed ( Gardner et al., 2014 ; Moerman and Estabrook, 2006 ;
Loiselle et al., 2008 ; Sastre and Lobo, 2009 ). Such an assessment is
relevant because the purpose of herbarium collections, at least his-
torically, has not been to document phenological phenomena per
se, but rather to sample representative specimens of a species
throughout its geographic distribution. Thus, phenological data
collected from herbarium records are subject to numerous poten-
tial biases. For example, botanists might collect samples at the same
time every year out of habit or convenience, with little regard to
interannual climatic variation, occurrence of date of  rst  owering,
or time of spring leaf-out. Other sources of bias include misidenti-
cation of closely related species that vary in phenological response,
temporal gaps in collecting e ort that impede e orts to assess long-
term change (e.g., decline in collecting e orts since the mid-20th
century [ Gardner et al., 2014 ]), spatial gaps in collecting, and spa-
tial preferences when collecting samples (e.g., easily accessible ur-
ban areas, trails, roadsides).
At the same time, ecologists’ emphasis on reconstructing phe-
nology overlooks other important uses of herbarium data. With
rare exceptions [e.g., crops and model species like Arabidopsis
thaliana (L.) Heynh.], we know relatively little about environmen-
tal cues that regulate onset and duration of phenological events for
most plant species.  ese cues govern physiological mechanisms
that initiate phenological events associated with  tness traits (e.g.,
initiation of buds or  owering), and timing of these events ulti-
mately may determine how species will respond to future climatic
change. For example, species for which  owering is most sensitive
to temperature likely will be strongly a ected by changing tempera-
tures, especially when shi s in temperatures could create tempo-
ral mismatches with key pollinators that are more (or less) sensitive
to temperature cues ( Burkle et al., 2013 ). Fortunately, recent
advancements in process-based modeling and data-model fusion
have allowed researchers to distinguish among the relative impor-
tance of major environmental cues (e.g., temperature, winter chill-
ing, and photoperiod) ( Richardson et al., 2006 ; Morisette et al.,
2009 ; Migliavacca et al., 2012 ; Archetti et al., 2013 ; Siniscalco et al.,
2015 ). ese models, however, require extensive and temporally
dense (>10 years) collections of standardized observational data.
As a result, the application of process-based models to studies of
phenological change has been limited to only a few dozen species,
o en from fairly restricted phylogenetic and life history groups.
To the extent that they prove to be reliable measures of interan-
nual phenological response, herbarium data o er the potential
for researchers to expand the temporal depth of phenological
data for large numbers of species and simultaneously propose
new hypotheses regarding physiological controls on phenological
In this paper, we address both the validity of herbarium records
for investigating phenological change and the use of herbarium re-
cords for identifying potential cueing mechanisms of phenological
events. Our focal region—New England in the United States—is
where researchers have suggested that many plants today  ower
much earlier than in the past because they are responsive to
changes in temperature, which has been rising rapidly in this re-
gion ( Miller-Rushing and Primack, 2008 ). We  rst combined direct
observational records and rich herbarium records collected during
the last 160 years from Concord, Massachusetts and adjacent coun-
ties to assess whether herbarium  owering data correspond with
rst  owering dates observed at irregular intervals in the  eld. We
then leveraged these results to identify likely physiological cueing
responses necessary for regulating  owering phenology. We fo-
cused on 20 species previously shown to exhibit dynamic responses
to climatic change ( Willis et al., 2008 ; Willis et al., 2010 ).  ese spe-
cies are abundant in New England herbaria because they are fre-
quently collected, conspicuous plants that produce large  owers,
which are easy to score from herbarium collections (e.g., lilies, or-
chids). Ultimately, results of our e orts will help to build a re ned
regional picture of how climatic change has a ected plant phenol-
ogy given the range of physiological mechanisms by which plants
are cued to  ower and fruit and how this is likely to shape plant
diversity in the near future.
Study site Concord (42 ° 27 38 N; 71 ° 20 54 W) is an approxi-
mately 67 km
2 town in Massachusetts (MA) with a wide range of
habitats, including peatlands, deciduous hardwood forests, and
prairies. Although Concord has undergone extensive development
since the 1850s, approximately 60% of the town’s land area remains
undeveloped or has been permanently protected from future devel-
opment ( Willis et al., 2008 ; Primack et al., 2009 ). During the mid-
19th and early 20th centuries, Henry David  oreau and Alfred
Hosmer, respectively, documented plant species occurrences and
rst- owering dates there ( Miller-Rushing and Primack, 2008 ;
Willis et al., 2008 ; Primack et al., 2009 ).  ese data, combined with
contemporary observations, have suggested that, depending on
spring temperatures, some species  ower as many as 8 d earlier
now than they did in the 1850s ( Miller-Rushing and Primack, 2008 ;
Willis et al., 2008 ; Primack et al., 2009 ). Moreover, Willis et al.
(2008 , 2009 ) identi ed phylogenetic e ects in these data on species
TABLE 1. Summar y of study species. Scienti c names, common names, native/introduced status, and growth habit from USDA PLANTS ( ).
Species Common Name Family Date range No. of years of data Native status
Daucus carota L. Queen Anne’s lace Apiaceae 1853–2008 35 Introduced
Aralia nudicalis L. wild sarsaparilla Araliaceae 1858–2012 47 Native
Barbarea vulgaris W.T.Aiton garden yellowrocket Brassicaceae 1877–2005 51 Introduced
Gaultheria procumbens L. eastern teaberry Ericaceae 1877–2011 32 Native
Gaylussacia baccata ( Wangenh.) K.Koch black huckleberry Ericaceae 1858–2011 36 Native
Vaccinium angustifolium Aiton lowbush blueberry Ericaceae 1878–2012 46 Native
Vicia cracca L. bird vetch Fabaceae 1877–2006 21 Introduced
Iris prismatica Pursh ex Ker Gawl. slender blue iris Iridaceae 1877–1934 24 Native
Arethusa bulbosa L. dragon’s mouth Orchidaceae 1861–1980 30 Native
Calopogon tuberosus (L.) Britton, Sterns &
tuberous grasspink Orchidaceae 1857–1984 19 Native
Corallorhiza maculata (Raf.) Raf. summer coralroot Orchidaceae 1854–1930 22 Native
Cypripedium acaule Aiton moccasin  ower Orchidaceae 1861–2012 51 Native
Platanthera grandi ora (Bigelow) Lindl. greater purple fringed orchid Orchidaceae 1861–1960 25 Native
Platanthera lacera (Michx.) G.Don green fringed orchid Orchidaceae 1854–1949 39 Native
Platanthera psycodes (L.) Lindl. lesser purple fringed orchid Orchidaceae 1854–1958 21 Native
Pogonia ophioglossoides (L.) Ker Gawl. snakemouth orchid Orchidaceae 1852–1962 44 Native
Chelidonium majus L. celandine Papaveraceae 1877–2011 21 Introduced
Aquilegia canadensis L. red columbine Ranunculaceae 1882–2012 39 Native
Ranunculus acris L. tall buttercup Ranunculaceae 1858–2011 34 Native/Introduced
diversity: clades whose  rst  owering time are less sensitive to tem-
perature and have shown little phenological change also have de-
clined signi cantly in abundance.
Study species We investigated 20 biennial or perennial species in
nine families ( Table 1 ), each of which met four criteria. (1)  e spe-
cies were represented in historical  eld observations by Hosmer
(1878–1903) from Concord ( Miller-Rushing and Primack, 2008 ).
ese data are unique in providing a reliable, uninterrupted 15-yr
period of phenological monitoring. (2)  e species had relatively
large  owers, which facilitated rapid and accurate assessment of
owering from herbarium specimens. (3)  e species were well
represented in herbarium collections from Massachusetts, includ-
ing Middlesex County (which includes Concord) and nearby coun-
ties. (4) Species with showy, ephemeral  owers and with relatively
short  owering time, such as orchids (Orchidaceae) and irises (Iri-
daceae), were preferred because they were more likely to have been
collected near to their  rst  owering date ( Robbirt et al., 2011 ). We
also included nonnative and invasive species (e.g., Barbarea vul-
garis [Brassicaceae], Chelidonium majus [Papaveraceae], respec-
tively) as well as species such as Vaccinium angustifolium , which
previously have been shown to be phenologically responsive to
warming ( Ellwood et al., 2013 ) and thus more likely to exhibit long-
term phenological shi s associated the secular trend of rising mean
temperatures. Additionally, taxon sampling is representative of the
breadth of seasonal  owering (e.g., spring ephemeral vs. summer
owering species).
Flowering time data Field observations of  rst owering date
for Concord were recorded by  oreau (1852–1858), Hosmer
(1875, 1878–1903), Miller-Rushing and Primack (2003–2006)
( Miller-Rushing and Primack, 2008 ), and this paper’s coauthors
Davis and Connolly (2011–2013:  eld data  rst reported and used
herein). Field observations by Davis and Connolly applied a simi-
lar method to the one outlined by Primack et al. (2009) : from
April to September, multiple sites throughout the Concord area
were visited 1–3 times weekly to systematically record  owering
dates. Similar to Primack et al., Davis and Connolly also consulted
local botanical experts about the location and  owering time of
certain species.
Estimates of earliest  owering dates for herbarium records were
based on data collected during visits to the Harvard University
Herbaria (HUH), New York Botanical Garden’s William and
Lynda Steere Herbarium (NY), Yale University Herbarium (YU),
and University of Connecticut’s George Sa ord Torrey Herbarium
(CONN) by coauthor Kelly.  ese herbaria collectively represent
the largest holdings of plants of the northeastern United States and
include both very old collections (HUH, YU) and more recent ones
(CONN). We  rst identi ed  owering specimens from MA for
each of our target species. Following Primack et al. (2004) , we re-
corded locality, collection date, accession number (when provided),
and collector for specimens with fully open  owers. When multiple
owers were present on a specimen, it was recorded as  owering if
≥75% of them were fully opened. Specimens that had a majority of
ower buds or fruit were ignored, as were those with insu cient or
illegible collection data.  e majority of herbarium specimens were
collected between the late 1800s and mid-1900s. When there were
multiple specimens for the same species in a given year (which oc-
curred only for <3% of the data collected), we used the earliest re-
cord for a species × county combination in a given year as our
estimate of earliest  owering date.
Finally, we emphasize that the earliest  owering date estimated
from herbarium specimens is di erent from  rst  owering date re-
corded by  eld observers. Observational records are the gold stan-
dard for phenological research and estimate  rst owering date
with a high degree of accuracy. In contrast, data from herbarium
specimens provides only an approximation of earliest  owering
date for those specimens, which may or may not be correlated with
rst  owering date observed in the  eld. However, our goal was to
assess changes in  owering as a function of both interannual tem-
perature variation and long-term changes in climate. Although we
expected di erences between the observed  rst day of  owering
and estimated date of  owering estimated from herbarium speci-
mens, the aim of this study was to assess whether these two data
types estimate similar responses to both short-term and long-term
climatic change. We hypothesized that estimated changes in earliest
owering date determined from herbarium specimens would be
correlated with observed changes in  rst  owering date as the cli-
mate has changed.
Temperature records Mean monthly temperatures (1885–present)
at Great Blue Hill, ~33 km southeast of Concord were obtained
from NOAA’s Global Historical Climatology Network (http:// We used the GHCNM v3 quality con-
trolled unadjusted data.  ese data are highly correlated ( r ≥ 0.995)
with available, but sparser, climatological data from Concord
( Miller-Rushing and Primack, 2008 ). Monthly temperature data
from 1831–1884 collected by the Blue Hill Meteorological Observa-
tory were provided to us by A. Miller-Rushing. Mean annual tem-
peratures ( Fig. 1A ) were calculated by averaging the mean monthly
temperatures for each year. Mean spring temperatures ( Fig. 1B )
were calculated as the average of each year’s February–May mean
monthly temperatures.  ese months were used because they had
been found previously to represent the months that are most pre-
dictive of flowering time ( Miller-Rushing and Primack, 2008 ;
Primack et al., 2009 ).
Statistical analysis We used linear mixed-e ects models in the nlme
library ( Pinheiro et al., 2015 ) of the R statistical software system
( R Core Team, 2014 ) version 3.1.0 to test for overall and species-
speci c relationships between spring temperature, calendar year,
and earliest  owering date. Earliest  owering date in a given year—
either from  eld or herbarium observations—was the response
variable in all models. In the “climate” model, mean spring tem-
perature was treated as a  xed predictor variable, whereas in the
“year” model, calendar year was the  xed predictor variable. In
both models, data type— eld observation or herbarium record—
was also treated as a  xed predictor variable. Species identity was
included as a random e ect.  e values of the random e ects (i.e.,
equivalent to the y intercept for each species) ordered the species
from earliest to latest  owering, so we also regressed (using a linear
regression model) the rate of change in earliest  owering date for
each species (i.e., the slope of the line relating earliest  owering date
to climate) against its random e ect term.  is latter analysis pro-
vided additional insights about potential species-speci c sensitivity
to spring temperature as a phenological cue.
To test the hypothesis that estimated changes in earliest  ower-
ing date determined from herbarium specimens were correlated
with observed changes in  rst owering date as the climate has
changed, we plotted the slopes of the lines  t to either the  eld ob-
servational data or the herbarium data in the “climate” model. We
tested the relationship between the slopes generated by these two
models in two ways. First, we did a simple paired t test on the slopes
(paired by species). Failure to reject the null hypothesis of no dif-
ference would suggest that the observed and herbarium data are
recording similar responses to climate. We also fit a Model II
regression to the paired slopes (Model II, or reduced major-axis
regression makes no assumption about the “independent” or “de-
pendent” variable ( Gotelli and Ellison, 2012 )).  e slope of this re-
gression tests whether the two sets of data vary in parallel, and the
intercept is an estimate of how the expected shi in  owering date
di ers between the two data sets.
In the main text, we report data only for the 600  eld observa-
tions from Concord combined with 297 herbarium records from
Middlesex County (where the town of Concord is located).  e re-
sults were qualitatively identical when we combined the Concord
observations with the 680 herbarium records from four nearby
counties (results in Appendix S1, see Supplemental Data with on-
line version of this paper). Raw data and model code are publicly
accessible from the Harvard Forest Data archive (http://harvardforest., data set HF-258.
Data density of  eld observations and herbarium records Field
observations (which we refer to henceforth as “observational data”)
of early  owering dates have been highly episodic ( Fig. 1C ).  o-
reau recorded dates of  rst owering in Concord annually from
1852 to 1858; Hosmer recorded  rst owering in 1875, and then
annually between 1878 and 1903; Primack and Miller-Rushing’s
data span 2003–2006, and our own observational data include
2011–2013. In contrast, we have 1108 herbarium records (which we
refer to henceforth as “herbarium data”) of  owering occurrences
in the state of Massachusetts collected between 1852 and 2012
( Fig. 1D ) with 297 records from the same county as Concord (Mid-
dlesex), and 680 remaining records from four nearby counties (on-
line Appendix S1 [containing Tables S1 and S2, and Figs. S1–S4];
see Fig. S1).
During the combined data interval (1852–2013), mean spring
temperatures varied widely, ranging from <1 ° to >8 ° C. Similarly,
mean annual temperatures ranged from <6 ° to >11 ° C. To charac-
terize this variation in temperature, we de ned the “climatic space”
of Concord since 1852 as the region encompassed by the range of
mean spring temperatures and mean annual temperatures ( Fig.
1E ). ree points are worth emphasizing about the sampling cover-
age of this climatic space. First, herbarium data covered a much
larger percentage of this climatic space than observational data
(91% vs. 76%, respectively). Second, observational data were nota-
bly lacking in years with unusually cool springs (i.e., those below
the regression line in Fig. 1E ).  ird, despite broad interannual
variability for both mean spring and mean annual temperatures
( Fig. 1A, 1B ), the historical and contemporary observational data
represent extreme endpoints in climatic space. Note that the his-
torical data amassed by  oreau and Hosmer were collected during
a relatively cold period, whereas the more contemporary data were
collected during a relatively warm period (i.e., Miller-Rushing and
Primack, Davis and Connolly) ( Fig. 1A, 1B ).  is sampling artifact
could bias inference about potential long-term secular trends on
phenology. However, the statistical bias of observational records
caused by this lack of overlap across the climate data are potentially
ameliorated by the herbarium data, which is distributed randomly
across the climatic space ( Fig. 1E ).
Herbarium data parallel  eld observations, but reduce long-term
estimates of phenological advancement attributed to climate
change — Overall, earliest recorded  owering dates in Middlesex
County were negatively associated with mean spring tempera-
tures for all species (i.e.,  owering was earlier; Fig. 2 ; overall slope =
–3.8 d/ ° C; F 1, 696 = 96.4, P < 0.001); results were similar for the four
nearby counties (Appendix S1: Fig. S2).  ere was no signi cant
interaction between mean spring temperature and observation type
on earliest  owering date ( F 1, 696 = 0.5, P = 0.47), suggesting that that
the overall relationship (i.e., slope) between mean spring tempera-
ture and earliest  owering date did not di er between  eld observa-
tions and herbarium data.
FIGURE 1 Climatic and phenological data. (A) Mean annual temperatures ( ° C) and (B) mean monthly temperatures recorded at Great Blue Hill, Massa-
chusetts (MA) (1885–present) and reconstructed by Miller-Rushing and Primack (1852–1884). (C) Observed  rst  owering dates in Concord, MA and
(D) earliest  owering dates on herbarium sheets from Middlesex County of the 20 species listed in Table 1 . (E) Coverage of the climatic space (1852–
2013; all boxes) by herbarium data (magenta boxes and magenta convex hull), Thoreau’s observations (orange dots and orange convex hull), Hosmer’s
observations (blue dots and blue convex hull), and contemporary observations (black dots and black convex hull). Unsampled points in the climate
space are represented by gray boxes without colored dots. Convex hulls encompass the outer boundaries of the climate space de ned by the most
extreme observations; they were  t using the “chull” function in R (base graphics). The gray line is the best- t regression line relating mean spring
temperature to mean annual temperature.
FIGURE 2 Relationship between mean spring temperature and earliest  owering date from  eld observations in Concord, Massachusetts (MA) or re-
corded on herbarium specimens from Middlesex County, MA. The blue points and lines are data and associated linear regressions for  eld observa-
tions; the red points and lines are data and associated linear regressions for herbarium specimens ( t using linear mixed e ect models); and the black
lines are common species-speci c robust linear regressions.
e paired t test comparing the slopes of the species-speci c re-
gression lines for observed and herbarium data shown in Fig. 2
found no significant differences ( t 18 = 0.45, P = 0.65). Although
the slopes of observed and herbarium data do not fall on a 1:1 line
( Fig. 3 ), the y -intercept of the plot, –2.8 d, suggests that observed
rst  owering dates are, on average, just under 3 d earlier than esti-
mated earliest  owering date of herbarium specimens.  us, we
were con dent (contra CaraDonna et al., 2014 ), that we could  t a
common climate model to these data as a whole, combining her-
barium data to  ll the gaps in the  eld observational data (black
lines in Fig. 2 ). To minimize e ects of outliers, however, we  t this
common slope using robust linear models ( Venables and Ripley,
2002 ).
Overall, our results support a previous study that has looked at
the fidelity of herbarium records with respect to field observa-
tions, but for a greatly reduced number of species and phylogenetic
diversity. Robbirt et al. (2011) compared an abundance of  eld and
herbarium data from across Europe for the single terrestrial orchid
Ophrys sphegodes and found no signi cant di erence between the
two data types for estimates of peak  owering time as a function of
spring temperature.
Although the response of earliest  owering date was similar
both for  eld and herbarium data, the intercepts di ered by 2.8 d.
is result should not be surprising because the observational data
that we used were collected with the explicit purpose of capturing
the earliest  owering day.  e natural historians and ecologists
who collected these data routinely sampled several, o en consecu-
tive days before  owering occurred and consulted local residents
and experts to increase the likelihood of identifying  rst  owering
events. In contrast, the collections represented by the herbarium
data that we used rarely were made expressly to capture  rst  ower-
ing events, but rather to document interesting, frequently abundant
plants in an area at multiple developmental stages (e.g.,  owering,
fruiting), usually meant for systematic and  oristic research.  ese
samples o en were obtained a er the time that the  rst  owers ap-
peared. Furthermore, our more conservative scoring of  owering
time for specimens (≥75% open  owers) potentially contributed to
the overall later date among herbarium records. Nonetheless, our
results provide the  rst broad validation, for a region in central
New England, that herbarium records can be used to address spatial
and temporal trends in phenology when and where  eld observa-
tional data are unavailable.  ese results underscore the enormous
promise of leveraging herbarium records for understanding the
impacts of climate change in New England, and perhaps more
We also demonstrated that the interannual variability in cli-
mate covered by the herbarium data fully encompassed and was
FIGURE 3 Relationship between phenological responses to climate esti-
mated from herbarium specimens and observed in the  eld. The values
on the x -axis are the slopes estimated for herbarium specimens (red lines
in Fig. 2 ), and the values on the y -axis are the slopes estimated for obser-
vational data (blue lines in Fig. 2 ). The dashed gray line is a 1:1 reference
substantially larger than the range of climate space encompassed by
observational data ( Fig. 1E ).  e coverage of the climate space was
large despite the potential biases observed in sampling temporal
variability, including episodic  eld observations, and herbarium
specimens collected predominantly before 1960.  is result dem-
onstrates for the  rst time to our knowledge that herbarium re-
cords represent key sources of data for  lling those parts of the
climatic space for which direct  eld observations are unavailable
and for determining how species dynamically adjust their  owering
time to interannual temperature variation. Analysis of our com-
bined  eld and herbarium data suggested an earlier  owering by
3.5 d/ ° C ( Fig. 2 ), similar to estimates from larger-scale studies that
have used herbarium records to assess phenological e ects of cli-
matic change. Calinger et al. (2013) , for example, reported an aver-
age change of 2.4 d/ ° C for  owering in 141 species in the midwestern
United States. Similarly, Everill et al. (2014) reported an advance-
ment of leaf out by 2 d/ ° C for 27 common deciduous woody species
in the northeastern United States.
Importantly, however, concluding that plants  ower earlier fol-
lowing warmer springs (or in warmer years) is not the same as say-
ing that these plants  ower earlier now than they did in the 1850s
or early 1900s (cf. Miller-Rushing and Primack, 2008 ; Ellwood
et al., 2013 ). When we regressed earliest  owering date on calen-
dar year, no signi cant e ect was observed ( F 1, 696 = 3.28, P = 0.07;
Fig. 4 ), nor was there an interaction between data type ( eld obser-
vation vs. herbarium) and calendar year ( F 1, 696 = 0.01, P = 0.75).
Even though both mean spring and mean annual temperatures are
clearly rising ( Fig. 1A, 1B ), interannual variation in both spring or
annual temperatures (>7 ° C) far exceed the long-term trend in tem-
perature (1.5 ° C/century): in fact, spring of 2012 was the warmest
(8.3 ° C) on record, but the spring of 2013 was nearly as cold (5.9 ° C)
as some of the warmest springs during Hosmer’s observations
more than a century ago (1898: 5.2 ° C; 1903: 6.7 ° C; Fig. 1B ). Pheno-
logical events in recent years illustrate this point remarkably well.
For example, in 2012, most species  owered early in the year (mean
observed  rst  owering date of all 20 species was 27 April), but in
2013, most species  owered much later (mean observed  rst  ow-
ering date was 23 May).  us, our  ndings indicate that researchers
should approach long-term phenological assessments using  eld or
observational data with caution given the high degree of interan-
nual variability in temperature.
ere are two likely explanations for the discrepancies in long-
term phenological trends we observed between our own results and
past studies for New England ( Primack et al., 2004 ; Miller-Rushing
et al., 2006 ; Willis et al., 2008 ; Panchen et al., 2012 ). First, historical
and contemporary observational data were collected in nonover-
lapping regions of climatic space ( Fig. 1E ). Despite pronounced
interannual variability in annual and spring temperatures ( Figs.
1A, 1B ), historical  eld observations were made during relatively
cool periods with late springs, while more recent observations have
been made during a record-setting warm period with early springs
( Fig. 1E ). Consequently, the use of  eld observational data alone is
biased toward  nding strong shi s in  owering over the last cen-
tury. As we have indicated already, herbarium data greatly help to
alleviate this sampling bias in climatic space. Second, because
spring  owering species are thought to be on average more respon-
sive to temperature, other studies of phenological advancement
have focused on these species with the premise that they would
likely exhibit the greatest long-term response, which indeed they
do ( Miller-Rushing and Primack, 2008 ). Our analyses, however,
demonstrate that the inclusion of later- owering species (summer
and early fall) results in this long-term trend being nonsigni cant
and, thus far, less dramatic when the seasonal variation of  owering
across the  ora is considered ( Fig. 5 ). We are not implying that
climatic change has not impacted or will not continue to impact
spring ephemeral communities. However, we caution against mak-
ing long-term phenological predictions based only on short-term
trends especially where interannual variability is high (regression
lines in Fig. 1A, 1B ).
We obtained qualitatively similar results when we included her-
barium data from the four adjacent counties in our analysis (Ap-
pendix S1: Fig. S3). However, with the inclusion of additional
herbarium data, the interaction term between calendar year and
observation type was signi cant (Appendix S1: Table S2). In other
words, not only the intercepts (as in the Middlesex County data
alone) but also the slopes of the regression lines relating  owering
date to calendar year di ered between observational and herbar-
ium data. Although the common slope  tted to each species was
FIGURE 4 Relationship between calendar year and earliest  owering date observed in the  eld or recorded on herbarium specimens from Middlesex
County in Massachusetts. The blue points and lines are data and associated linear regressions for  eld observations; red points and lines are data and
associated linear regressions for herbarium specimens; black lines are common species-speci c robust linear regressions.
FIGURE 5 Relationship between species-speci c random e ect size and the acceleration of  owering (i.e., the
values of the common slopes  t in Fig. 2 ).
essentially  at, the slopes  t to the observational data and the her-
barium data were not parallel to one another.  is result illustrates
that interannual temperature variability among locations is sub-
stantial and suggests potential limitations in using herbarium data
from areas that are not closely colocated with observational data.
Speci cally, the lack of colocated observational data may lead to
potentially spurious interpretation of phenological change across
larger areas where only herbarium records are available.
Previous studies of the Concord  ora have drawn a clear link
between short-term phenological sensitivity to temperature and
declining abundance ( Willis et al., 2008 ). Our revised estimates of
phenological sensitivity to interannual spring temperature raise
questions about one recently hypothesized mechanism driving this
decline—phenological mismatch. Bartomeus et al. (2011) found
that several common New England insect pollinators were sensitive
to spring temperature, advancing their  ight times by 3.6 d/ ° C,
remarkably similar to our own flowering phenology results of
3.5 d/ ° C. In contrast, Bartomeus et al. (2011) also found that these
same pollinators had advanced their phenology over the last
century by ~10 d, which is on par with previous studies of
plant phenology in New England
( Miller-Rushing and Primack,
2008 ). eir interpretation of these
results was that ecological mis-
matches in plant–pollinator mu-
tualisms were unlikely to explain
the decline among plants in the
region. Our results, however, dem-
onstrated no signi cant trend in
long-term  owering shi s among
New England plants ( Fig. 4 ; Ap-
pendix S1: Fig. S4, Table S2), in-
dicating that pollinators may be
emerging signi cantly earlier than
their plant hosts for all but the
most temperature-sensitive plant
species.  is reinterpretation of
the conclusions of Bartomeus et al.
(2011) reopens the question of
the importance of pollinator mis-
matches to the decline of those
less-temperature-sensitive species
in New England.
Using herbarium records to
assess phenological cueing
mechanisms — Finally, we sug-
gest that analysis of herbarium
data can be used to identify vari-
ability in physiological mecha-
nisms that cue phenological
events ( Fig. 5 ). It is clear that spe-
cies vary in their flowering re-
sponse to spring temperatures
( Figs. 2, 4 ). Including the ran-
dom e ects term (i.e., the species
e ect) in the model substantially
improved model  t (AIC full
model = 5584, AIC model with-
out species = 6902). Regression
of observed phenological advancement (days/ ° C) on the random
e ects term for each species (i.e., the change in y intercept relative
to a common model) revealed several interesting patterns ( Fig. 5 ).
First, spring-blooming species have much less variability in their
phenological responses to mean spring temperatures than do
summer- or fall-blooming species, suggesting that flowering in
spring-blooming species (i.e., those that bloom before early June) is
strongly controlled by temperature. In contrast, the large variability
in response of summer-, and fall-blooming species suggests that
owering in these species is controlled by a variety of di erent fac-
tors, including photoperiod and winter chilling ( Körner and Basler,
2010 ). When we included herbarium data from nearby counties in
this analysis (Appendix S1: Fig. S4), the variability in response of
later-blooming species was somewhat reduced, and the relation-
ship between the species-speci c random e ect size and advance-
ment of  owering time was more pronounced. Nonetheless, the
variability in response of later-blooming species still exceeded that
of spring-blooming species by more than 2-fold.
In either case, distinguishing more precisely between di erent
phenological cues, at least at broad scales, might now be possible
with the greatly expanded geographic and temporal sampling avail-
able from herbarium records.  is approach can further guide
more focused experiments to establish cues for di erent species,
but even the correlative associations we have identi ed between cli-
mate and phenology are valuable.  ese correlations are likely to
hold especially for species whose cueing mechanisms are simpler
and restricted primarily to a single variable. Such data are in great
demand, yet are seldom available for a large diversity of species
across a region. A recent review of  owering cues by Pau et al.
(2011) underscores this demand.  eir meta-analysis summarized
115 studies from  eld observational data. From these studies, they
identi ed su cient data for only 325 species. While several tropical
and boreal species were included, the majority were from temper-
ate regions, primarily in the United States and western Europe
( Pau et al., 2011 ). Furthermore, all of these studies were restricted
to single sites and thus failed to capture the potential geographic
variation within species. These findings greatly emphasize the
limited taxonomic and geographic scope of field observational
data available for large-scale phenological research ( Wolkovich
et al., 2014 ).
In contrast, herbarium data hold great promise for overcoming
this impasse and improving assessments of how species will re-
spond to future climate change. In particular, our results could be
used in process-based models to distinguish the relative important
of temperature, chilling, and photoperiod across a wide diversity of
species ( Richardson et al., 2006 ; Morisette et al., 2009 ; Migliavacca
et al., 2012 ; Archetti et al., 2013 ; Siniscalco et al., 2015 ). Previous
studies have been limited to a few dozen species with su cient in-
terannual sampling, typically derived from a small number of well-
documented, long-term ecological study sites.  e reliability of
herbarium data, however, o ers the promise of greatly expanding
these studies to understand how species will respond to recent cli-
matic change and the potential to untangle the relative importance
of multiple cues (e.g., photoperiod, temperature) and how they vary
across space. Finally, the ability to study a broader diversity of species
could greatly expand our knowledge of deeper phylogenetic patterns
involving phenological response mechanisms ( Davies et al., 2013 ).
Our results indicate that herbarium data represent a valuable re-
source for studying both temporal trends and mechanisms of phe-
nological change.  e next challenge is to scale-up our assessments
of phenological responses and mechanisms to include the thou-
sands of species on the landscape that are also represented in her-
barium collections.  ese spatially and temporally explicit records
of biodiversity are increasingly becoming available digitally as a
result of investment in high-throughput digital imaging, GIS, and
rigorous spatial analyses.  e Harvard University Herbaria, along
with several collaborating institutions, are presently enhancing the
digital infrastructure for the  ora of New England by capturing
specimen-level metadata and images into digital form. Alongside
this e ort, coauthors Davis and Willis have created a crowdsourc-
ing platform ( Curio ) with Edith Law (University of Waterloo) to
engage volunteer botanists in detecting  owers, buds, and fruits on
herbarium records. We intend to use this platform to capture phe-
nological data from the ~1 million digitized specimens from New
England and use these data to understand how plants have re-
sponded, and will respond, to climatic change in this region. Future
studies focused on species that di er in their  owering season and
may respond di erently to climatic change (e.g., Vaccinium an-
gustifolium vs. Daucus carota ) but have large geographic ranges,
are well represented in herbaria, and can be identified easily by
amateur botanists, making them especially valuable for these ef-
forts. Moreover, by taking advantage of crowdsourcing, we will
be able to assess all of the relevant stages of plant phenology criti-
cal to climate change, including leaf-out, transitions from bud to
ower, peak  owering time, and transition to fruiting.
C.C.D. conceived the project. C.C.D. and C.K. developed the initial
research design. C.K. conducted herbarium sampling. C.C.D., B.C.,
and C.K. conducted  eldwork. A.M.E. analyzed and modeled the
data. Data were interpreted by C.C.D., C.G.W., and A.M.E. C.C.D.,
C.G.W., and A.M.E., largely did the writing, with input from B.C.
and C.K. Contributions of C.K. and C.C.D. were supported by a
Grants-in-Aid of Undergraduate Research from the Harvard
University Herbaria (to C.K.) and by grant EF 1208835 from the
U.S. National Science Foundation (to C.C.D.). A.M.E.’s work on this
project was supported by the Harvard Forest and by grant DEB
1237491 from the U.S. National Science Foundation.  is paper is a
contribution of the Harvard Forest LTER site.  e authors are
grateful to the following herbaria: Harvard University Herbaria,
New York Botanical Garden’s William and Lynda Steere Herbarium,
Yale University Herbarium, and University of Connecticuts George
Sa ord Torrey Herbarium.
Archetti , M. , A. D. Richardson , J. O’Keefe , and N. Delpierre . 2013 . Predicting
climate change impacts on the amount and duration of autumn colors in a
New England forest. PLoS One 8 : e57373 .
Bartomeus , I. , J. S. Ascher , D. Wagner , B. N. Danforth , S. Colla , S. Kornbluth ,
and R. Winfree . 2011 . Climate-associated phenological advances in bee
pollinators and bee-pollinated plants. Proceedings of the National Academy
of Sciences, USA 108 : 20645 – 20649 .
Bertin , R. I. 2008 . Plant phenology and distribution in relation to recent cli-
mate change. Journal of the Torrey Botanical Society 135 : 126 – 146 .
Bradley , N. L. , A. C. Leopold , J. Ross , and H. Wellington . 1999 . Phenological
changes re ect climate change in Wisconsin. Proceedings of the National
Academy of Sciences, USA 96 : 9701 – 9704 .
Burkle , L. A. , J. C. Marlin , and T. M. Knight . 2013 . Plant–pollinator interac-
tions over 120 years: Loss of species, co-occurrence, and function. Science
339 : 1611 – 1615 .
Calinger , K. M. , S. Queenborough , and P. S. Curtis . 2013 . Herbarium speci-
mens reveal the footprint of climate change on  owering trends across
north-central North America. Ecology Letters 16 : 1037 – 1044 .
CaraDonna , P. J. , A. M. Iler , and D. W. Inouye . 2014 . Shi s in  owering phe-
nology reshape a subalpine plant community. Proceedings of the National
Academy of Sciences, USA 111 : 4916 - 4921 .
Davies , T. J. , E. M. Wolkovich , N. J. Kra , N. Salamin , J. M. Allen , T. R. Ault ,
J. L. Betancourt , et al. 2013 . Phylogenetic conservatism in plant phenology.
Journal of Ecology 101 : 1520 – 1530 .
Ellwood , E. R. , S. A. Temple , R. B. Primack , N. L. Bradley , and C. C. Davis .
2013 . Record-breaking early owering in the eastern United States. PLoS
One 8 : e53788 .
Everill , P. H. , R. B. Primack , E. R. Ellwood , and E. K. Melaas . 2014 . Determining
past leaf-out times of New England’s deciduous forests from herbarium
specimens. American Journal of Botany 101 : 1293 – 1300 .
Fitter , A. H. , and R. S. R. Fitter . 2002 . Rapid changes in owering time in
British plants. Science 296 : 1689 – 1691 .
Gardner , J. L. , T. Amano , W. J. Sutherland , L. Joseph , and A. Peters . 2014 . Are
natural history collections coming to an end as time-series? Frontiers in
Ecology and the Environment 12 : 436 – 438 .
Gotelli , N. J. , and A. M. Ellison . 2012 . A primer of ecological statistics. 2nd ed.,
614. Sinauer, Sunderland, Massachusetts, USA.
Kharouba , H. M. , and M. Vellend . 2015 . Flowering time of butter y nectar
food plants is more sensitive to temperature than the timing of butter y
adult  ight. Journal of Animal Ecology 84 : 1311 – 1321 .
Körner , C. , and D. Basler . 2010 . Phenology under global warming. Science 327 :
1461 – 1462 .
Loiselle , B. A. , P. M. Jørgensen , T. Consiglio , I. Jiménez , J. G. Blake , L. G.
Lohmann , and O. M. Montiel . 2008 . Predicting species distributions from
herbarium collections: Does climate bias in collection sampling in uence
model outcomes? Journal of Biogeography 35 : 105 – 116 .
Menzel , A. 2002 . Phenology: Its importance to the global change community.
Climatic Change 54 : 379 – 385 .
Migliavacca , M. , O. Sonnentag , T. F. Keenan , A. Cescatti , J. O’Keefe , and A. D.
Richardson . 2012 . On the uncertainty of phenological responses to climate
change, and implications for a terrestrial biosphere model. Biogeosciences
9 : 2063 – 2083 .
Miller-Rushing , A. J. , T. L. Lloyd-Evans , R. B. Primack , and P. Satzinger . 2008 .
Bird migration times, climate change, and changing population sizes. Global
Change Biology 14 : 1959 – 1972 .
Miller-Rushing , A. J. , and R. B. Primack . 2008 . Global warming and ower-
ing times in  oreau’s Concord: A community perspective. Ecology 8 9 :
332 – 341 .
Miller-Rushing , A. J. , R. B. Primack , D. Primack , and S. Mukunda . 2006 .
Photographs and herbarium specimens as tools to document phenologi-
cal changes in response to global warming. American Journal of Botany 93 :
1667 – 1674 .
Moerman , D. E. , and G. F. Estabrook . 2006 . e botanist e ect: Counties with
maximal species richness tend to be home to universities and botanists.
Journal of Biogeography 33 : 1969 – 1974 .
Morisette , J. T. , A. D. Richardson , A. K. Knapp , J. I. Fisher , E. A. Graham , J.
Abatzoglou , B. E. Wilson , et al. 2009 . Tracking the rhythm of the seasons
in the face of global change: Phenological research in the 21st century.
Frontiers in Ecology and the Environment 7 : 253 – 260 .
Panchen , Z. A. , R. B. Primack , T. Anisko , and R. E. Lyons . 2012 . Herbarium
specimens, photographs, and  eld observations show Philadelphia area
plants are responding to climate change. American Journal of Botany 99 :
751 – 756 .
Panchen , Z. A. , R. B. Primack , B. Nordt , E. R. Ellwood , A.-D. Stevens , S. S.
Renner , C. G. Willis , et al. 2014 . Leaf out times of temperate woody plants
are related to phylogeny, deciduousness, growth habit and wood anatomy.
New Phytologist 203 : 1208 – 1219 .
Parmesan , C. 2006 . Ecological and evolutionary responses to recent climate
change. Annual Review of Ecology Evolution and Systematics 37 : 637 – 669 .
Parmesan , C. 2007 . In uences of species, latitudes and methodologies on es-
timates of phenological response to global warming. Global Change Biology
13 : 1860 – 1872 .
Parmesan , C. , and G. Yohe . 2003 . A globally coherent ngerprint of climate
change impacts across natural systems. Nature 421 : 37 – 42 .
Pau , S. , E. M. Wolkovich , B. I. Cook , T. J. Davies , N. J. B. Kra , K. Bolmgren ,
J. L. Betancourt , and E. E. Cleland . 2011 . Predicting phenology by inte-
grating ecology, evolution and climate science. Global Change Biology 17 :
3633 – 3643 .
Pinheiro , J. , D. Bates . S. DebRoy , D. Sarkar , and R. Core Team . 2015 . nlme:
Linear and Nonlinear Mixed E ects Models. R package version 3.1-120,
available at .
Primack , D. , C. Imbres , R. B. Primack , A. J. Miller-Rushing , and P. Del Tredici .
2004 . Herbarium specimens demonstrate earlier  owering times in re-
sponse to warming in Boston. American Journal of Botany 91 : 1260 – 1264 .
Primack , R. B. , A. J. Miller-Rushing , and K. Dharaneeswaran . 2009 . Changes in
the  ora of  oreau’s Concord. Biological Conservation 142 : 500 – 508 .
R Core Team . 2014 . R: A language and environment for statistical comput-
ing. R Foundation for Statistical Computing, Vienna. Available at http:// .
Richardson , A. D. , A. S. Bailey , E. G. Denny , C. W. Martin , and J. O’Keefe . 2006 .
Phenology of a northern hardwood forest canopy. Global Change Biology
12 : 1174 – 1188 .
Robbirt , K. M. , A. J. Davy , M. J. Hutchings , and D. L. Roberts . 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. Journal of
Ecology 99 : 235 – 241 .
Root , T. L. , J. T. Price , K. R. Hall , S. H. Schneider , C. Rosenzweig , and J. A.
Pounds . 2003 . Fingerprints of global warming on wild animals and plants.
Nature 421 : 57 – 60 .
Sastre , P. , and J. M. Lobo . 2009 . Taxonomist survey biases and the unveiling of
biodiversity patterns. Biological Conservation 142 : 462 – 467 .
Siniscalco , C. , R. Caramiello , M. Migliavacca , L. Busetto , L. Mercalli , R.
Colombo , and A. Richardson . 2015 . Models to predict the start of
the airborne pollen season . International Journal of Biometeorology 5 9 :
837 – 848 .
Sparks , T. H. , and P. D. Carey . 1995 . e responses of species to climate over
two centuries: An analysis of the Marsham phenological record, 1736–1947.
Journal of Ecology 83 : 321 – 329 .
Vellend , M. , C. D. Brown , H. M. Kharouba , J. L. McCune , and I. H. Myers-
Smith . 2013 . Historical ecology: Using unconventional data sources to test
for e ects of global environmental change. American Journal of Botany 100 :
1294 – 1305 .
Venables , W. , and B. Ripley . 2002 . Modern applied statistics with S, 4th ed.
Springer-Verlag, Berlin, Germany.
Visser , M. E. 2008 . Keeping up with a warming world; assessing the rate of
adaptation to climate change. Proceedings of the Royal Society of London, B,
Biological Sciences 275 : 649 – 659 .
Walther , G. R. , E. Post , P. Convey , A. Menzel , C. Parmesan , T. J. C. Beebee , J.
M. Fromentin , et al. 2002 . Ecological responses to recent climate change.
Nature 416 : 389 – 395 .
Willis , C. G. , B. Ruhfel , R. B. Primack , A. J. Miller-Rushing , and C. C. Davis .
2008 . Phylogenetic patterns of species loss in  oreau’s woods are driven
by climate change. Proceedings of the National Academy of Sciences, USA
105 : 17029 – 17033 .
Willis , C. G. , B. Ruhfel , R. B. Primack , A. J. Miller-Rushing , and C. C. Davis .
2009 . Reply to McDonald et al.: Climate change, not deer herbivory,
has shaped species decline in Concord, Massachusetts. Proceedings of the
National Academy of Sciences, USA 106 : E29 .
Willis , C. G. , B. R. Ruhfel , R. B. Primack , A. J. Miller-Rushing , J. B. Losos , and
C. C. Davis . 2010 . Favorable climate change response explains non-native
species’ success in  oreau’s woods. PLOS One 5 : e8878 .
Wolkovich , E. M. , B. I. Cook , and T. J. Davies . 2014 . Progress towards an inter-
disciplinary science of plant phenology: Building predictions across space,
time and species diversity. New Phytologist 201 : 1156 – 1162 .
... These various phenology data sources come along with advantages and limitations for climate change research (Davis et al., 2015;Willis et al. 2017;Taylor et al. 2019;Funk 2018;Belitz et al. 2020). Most notably, herbarium specimens and citizen science networks include biases in sampling locations, taxonomic groups, and gaps in sampling years over time (Delisle et al. 2003;Daru et al. 2018;Panchen et al. 2019). ...
... The digitization process itself may introduce further error or biases for phenology or other data derived from digitized herbarium specimens (Groom et al. 2019). Due to these limitations, supplementing phenology data derived from herbarium specimens with other types of phenology data is proving to be potentially valuable approach (Davis et al. 2015;Spellman and Mulder 2016;Jones and Daehler 2018). ...
... The trends we obtained from herbarium data are comparable to findings of other phenology studies based on long-term field-collected data in the UK and elsewhere in Europe for the same species, with flowering advancing 6-7 days per increase in degrees Celsius (Sparks et al. 2000;Menzel 2003). The importance of long time series for studying the phenological response to climate has long been recognized, and numerous studies have now demonstrated the value of herbarium specimens in recreating long-term phenology data and advancing the study of global change (Primack et al. 2004;Calinger et al. 2013;Davis et al. 2015;Lang et al. 2019). Citizen science data of equal sample size to our herbarium data from a directed phenology monitoring program detected no trends between flowering DOY and climate. ...
Full-text available
Phenology has emerged as a key metric to measure how species respond to changes in climate. Innovative means have been developed to extend the temporal and spatial range of phenological data by obtaining data from herbarium specimens, citizen science programs, and biodiversity data repositories. These different data types have seldom been compared for their effectiveness in detecting environmental impacts on phenology. To address this, we compare three separate phenology datasets from Denmark: (i) herbarium specimen data spanning 145 years, (ii) data collected from a citizen science phenology program over a single year observing first flowering, and (iii) data derived from incidental biodiversity observations in iNaturalist over a single year. Each dataset includes flowering day of year observed for three common spring-flowering plant species: Allium ursinum (ramsons), Aesculus hippocastanum (horse chestnut), and Sambucus nigra (black elderberry). The incidental iNaturalist dataset provided the most extensive geographic coverage across Denmark and the largest sample size and recorded peak flowering in a way comparable to herbarium specimens. The directed citizen science dataset recorded much earlier flowering dates because the program objective was to report the first flowering, and so was less compared to the other two datasets. Herbarium data demonstrated the strongest effect of spring temperature on flowering in Denmark, possibly because it was the only dataset measuring temporal variation in phenology, while the other datasets measured spatial variation. Herbarium data predicted the mean flowering day of year recorded in our iNaturalist dataset for all three species. Combining herbarium data with iNaturalist data provides an even more effective method for detecting climatic effects on phenology. Phenology observations from directed and incidental citizen science initiatives will increase in value for climate change research in the coming years with the addition of data capturing the inter-annual variation in phenology.
... With many herbaria dating back to some 200 yr, and hundreds of millions of specimens worldwide, herbaria are a tremendous treasure for studying phenology changes both long term and large scale. Previous studies have indeed found strong patterns of long-term phenology changes in herbarium data (Primack et al., 2004;Miller-Rushing et al., 2006;Davis et al., 2015;Willis et al., 2017;Lang et al., 2019;Park et al., 2019;reviewed by Jones & Daehler, 2018), and they have also demonstrated that phenology trends estimated from herbarium data are comparable to those from field observations (Davis et al., 2015;Jones & Daehler, 2018;Miller et al., 2021). However, almost all previous studies have been carried out in the USA, and there has been little work so far on herbaria and plant phenology in Europe (but please refer to Robbirt et al., 2011;Diskin et al., 2012;Molnar et al., 2012). ...
... With many herbaria dating back to some 200 yr, and hundreds of millions of specimens worldwide, herbaria are a tremendous treasure for studying phenology changes both long term and large scale. Previous studies have indeed found strong patterns of long-term phenology changes in herbarium data (Primack et al., 2004;Miller-Rushing et al., 2006;Davis et al., 2015;Willis et al., 2017;Lang et al., 2019;Park et al., 2019;reviewed by Jones & Daehler, 2018), and they have also demonstrated that phenology trends estimated from herbarium data are comparable to those from field observations (Davis et al., 2015;Jones & Daehler, 2018;Miller et al., 2021). However, almost all previous studies have been carried out in the USA, and there has been little work so far on herbaria and plant phenology in Europe (but please refer to Robbirt et al., 2011;Diskin et al., 2012;Molnar et al., 2012). ...
... In Europe, land temperatures have increased c. 1.5°C since 1900 (Luterbacher et al., 2004;Harris et al., 2014;European Environmental Agency, 2020), so the magnitude of overall phenological changes we observed is similar to what would be expected based on climate change and the observed temperature sensitivities (1.5°C × 3.6 d/°C = 5.4 d, vs our observed average shift of c. 6 d). Previous herbarium studies from the temperate zone estimated similar flowering-time advancements of −2.4 to −6.3 d per 1°C temperature increase (Primack et al., 2004;Miller-Rushing et al., 2006;Panchen et al., 2012;Calinger et al., 2013;Hart et al., 2014;Bertin, 2015;Davis et al., 2015;Bertin et al., 2017). Again, most of these studies were from the northeastern USA, and they were often geographically very restricted. ...
Today plants often flower earlier due to climate warming. Herbarium specimens are excellent witnesses of such long‐term changes. However, the magnitude of phenological shifts may vary geographically, and the data are often clustered. Therefore, large‐scale analyses of herbarium data are prone to pseudoreplication and geographical biases. We studied over 6000 herbarium specimens of 20 spring‐flowering forest understory herbs from Europe to understand how their phenology had changed during the last century. We estimated phenology trends with or without taking spatial autocorrelation into account. On average plants now flowered over 6 d earlier than at the beginning of the last century. These changes were strongly associated with warmer spring temperatures. Flowering time advanced 3.6 d per 1°C warming. Spatial modelling showed that, in some parts of Europe, plants flowered earlier or later than expected. Without accounting for this, the estimates of phenological shifts were biased and model fits were poor. Our study indicates that forest wildflowers in Europe strongly advanced their phenology in response to climate change. However, these phenological shifts differ geographically. This shows that it is crucial to combine the analysis of herbarium data with spatial modelling when testing for long‐term phenology trends across large spatial scales.
... Rather, we find that timing of both spring-occurring CL and CH flowers has advanced across the past century ( Figure 5A, B). These findings add to a growing body of literature on how climate change drives phenological advancement (e.g., Menzel et al., 2006;Miller-Rushing and Primack, 2008;Calinger et al., 2013;Davis et al., 2015;Jones and Daehler, 2018) and, to the best of our knowledge, provides the first documentation of obligately selfing CL flowers having advanced across the 20th century. Interestingly, while we find that both flower types have progressively occurred earlier across this time period, climate was only associated with phenology shifts of CH flowers; CL phenology shifts were not associated with either mean annual temperature or total annual precipitation. ...
Full-text available
Premise: Although the balance between cross‐and self‐fertilization is driven by the environment, no long‐term study has documented whether anthropogenic climate change is affecting reproductive strategy allocation in species with mixed mating systems. Here, we test whether the common blue violet (Viola sororia; Violaceae) has altered relative allocation to the production of potentially outcrossing flowers as the climate has changed throughout the 20th century. Methods: Using herbarium records spanning from 1875 to 2015 from the central United States, we quantified production of obligately selfing cleistogamous (CL) flowers and potentially outcrossing chasmogamous (CH) flowers by V. sororia, coupled these records with historic temperature and precipitation data, and tested whether changes to the proportion of CL flowers correlate with temporal climate trends. Results: We find that V. sororia progressively produced lower proportions of CL flowers across the past century and in environments with lower mean annual temperature and higher total annual precipitation. We also find that both CL and CH flower phenology has advanced across this time period. Conclusions: Our results suggest that V. sororia has responded to lower temperatures and greater water availability by shifting reproductive strategy allocation away from selfing and toward potential outcrossing. This provides the first long‐term study of how climate change may affect relative allocation to potential outcrossing in species with mixed mating systems. By revealing that CL flowering is associated with low water availability and high temperature, our results suggest the production of obligately selfing flowers is favored in water limited environments.
... Alternative tools that use images of the specimen on the herbarium mounting sheet would greatly help botanists and taxonomists working in herbaria. There is increasing interest in the digitization of herbarium specimens (Daru et al., 2018;Davis et al., 2015;Espinosa and Pinedo Castro, 2018;Lang et al., 2019;Pearson, 2019;Wang, 2018). This is mainly due to current initiatives to digitize biodiversity data and make them accessible to a larger audience worldwide (Hedrick et al., 2020). ...
Herbaria contain the treasure of millions of specimens that have been preserved for several years for scientific studies. To increase the rate of scientific discoveries, digitization of these specimens is currently ongoing to facilitate the easy access and sharing of data to a wider scientific community. Online digital repositories such as Integrated Digitized Biocollection and the Global Biodiversity Information Facility have already accumulated millions of specimen images yet to be explored. This presents the perfect time to take advantage of the opportunity to automate the identification process and increase the rate of novel discoveries using computer vision (CV) and machine learning (ML) techniques. In this study, a systematic literature review of more than 70 peer-reviewed publications was conducted focusing on the application of computer vision and machine learning techniques to digitized herbarium specimens. The study categorizes the different techniques and applications that are commonly used for digitized herbarium specimens and highlights existing challenges together with their potential solutions. We hope this study will serve as a firm foundation for new researchers in the relevant disciplines and will also be enlightening to both computer science and ecology experts.
... One question that has attracted renewed interest in the context of global change biology is related to the phenology of symbiot ic associations (Rafferty, CaraDonna, & Bronstein, 2015). The utilization of herbarium specimens to investigate plant phenology has emerged as a leading research area in recent decades (Davis, Willis, Connolly, Kelly, & Ellison, 2015). Because herbarium labels contain collection dates and specimens capture key phenological events (e.g., flowering, fruiting, leaf-out) they represent an underutilized, yet powerful record of phenology across time, space, and phylogeny. ...
1. The widespread digitization of natural history collections, combined with novel tools and approaches is revolutionizing biodiversity science. The “extended specimen” concept advocates a more holistic approach in which a specimen is framed as a diverse stream of interconnected data. Herbarium specimens that by their very nature capture multi‐species relationships, such as certain parasites, fungi, and lichens, hold great potential to provide a broader and more integrative view of the ecology and evolution of symbiotic interactions. This particularly applies to parasite‐host associations, which owing to their interconnectedness are especially vulnerable to global environmental change. 2. Here, we present an overview of how parasitic flowering plants is represented in herbarium collections. We then discuss the variety of data that can be gathered from parasitic plant specimens, and how they can be used to understand global change impacts at multiple scales. Finally, we review best practices for sampling parasitic plants in the field, and subsequently preparing and digitizing these specimens. 3. Plant parasitism has evolved 12 times within angiosperms, and similar to other plant taxa, herbarium collections represent the foundation for analyzing key aspects of their ecology and evolution. Yet these collections hold far greater potential. Data and metadata obtained from parasitic plant specimens can inform analyses of co‐distribution patterns, changes in eco‐physiology and species plasticity spanning temporal and spatial scales, chemical ecology of tripartite interactions (e.g., host‐parasite‐herbivore), and molecular data critical for species conservation. Moreover, owing to the historic nature and sheer size of global herbarium collections, these data provide the spatiotemporal breadth essential for investigating organismal response to global change. 4. Parasitic plant specimens are primed to serve as ideal examples of extended specimen concept and help motivate the next generation of creative and impactful collection‐based science. Continued digitization efforts and improved curatorial practices will contribute to opening these specimens to a broader audience, allowing integrative research spanning multiple domains and offering novel opportunities for education.
... Pressed, preserved plant specimens collected over centuries and around the globe can serve as snapshots of plant phenology contributing to a wealth of research relating to the timing of phenological events (Jones and Daehler, 2018). Although most herbarium specimens were not collected to detect phenological stages, phenological data from specimens have proven reliable and irreplaceable for understanding plant phenology (Davis et al., 2015). The collection of phenological data from herbarium specimens is fundamentally based on the presence and absence of key reproductive or vegetative traits. ...
Full-text available
Climate change represents one of the most critical threats to biodiversity with far-reaching consequences for species interactions, the functioning of ecosystems, or the assembly of biotic communities. Plant phenology research has gained increasing attention as the timing of periodic events in plants is strongly affected by seasonal and interannual climate variation. Recent technological development allowed us to gather invaluable data at a variety of spatial and ecological scales. The feasibility of phenological monitoring today and in the future depends heavily on developing tools capable of efficiently analyzing these enormous amounts of data. Deep Neural Networks learn representations from data with impressive accuracy and lead to significant breakthroughs in, e.g., image processing. This article is the first systematic literature review aiming to thoroughly analyze all primary studies on deep learning approaches in plant phenology research. In a multi-stage process, we selected 24 peer-reviewed studies published in the last five years (2016–2021). After carefully analyzing these studies, we describe the applied methods categorized according to the studied phenological stages, vegetation type, spatial scale, data acquisition- and deep learning methods. Furthermore, we identify and discuss research trends and highlight promising future directions. We present a systematic overview of previously applied methods on different tasks that can guide this emerging complex research field.
... The recent and massive digitization of natural history collections, particularly of herbaria, offers new material available for many research activities [16][17][18] such as the analysis of plant species morphology and diversity [19,20]. Because most of the collections are dated and geolocated, they constitute a valuable source of information for (i) determining the proven or potential distribution areas of species [21][22][23], whether native or exotic (dynamics of biological invasions), with direct applications in conservation biology [24], and for (ii) determining the reproductive phenological patterns of species (e.g., date and duration of flowering and fruiting periods) [25][26][27][28]. Research in these fields has been particularly stimulated by questions related to climate change and its effect on the range of species distribution [29,30] or their biological rhythms [31][32][33][34][35][36][37][38][39]. ...
Full-text available
A better knowledge of tree vegetative growth phenology and its relationship to environmental variables is crucial to understanding forest growth dynamics and how climate change may affect it. Less studied than reproductive structures, vegetative growth phenology focuses primarily on the analysis of growing shoots, from buds to leaf fall. In temperate regions, low winter temperatures impose a cessation of vegetative growth shoots and lead to a well-known annual growth cycle pattern for most species. The humid tropics, on the other hand, have less seasonality and contain many more tree species, leading to a diversity of patterns that is still poorly known and understood. The work in this study aims to advance knowledge in this area, focusing specifically on herbarium scans, as herbariums offer the promise of tracking phenology over long periods of time. However, such a study requires a large number of shoots to be able to draw statistically relevant conclusions. We propose to investigate the extent to which the use of deep learning can help detect and type-classify these relatively rare vegetative structures in herbarium collections. Our results demonstrate the relevance of using herbarium data in vegetative phenology research as well as the potential of deep learning approaches for growing shoot detection.
... Much work has been published in recent years (Davis et al. 2015;Park and Schwartz 2015;Willis et al. 2017a;Park et al. 2018) that has confirmed the value of herbaria in characterizing phenological changes. A synthesis of methods and applications of these has been published recently. ...
Significance Adaptive evolution can help species to persist and spread in new environments, but it is unclear how the rate and duration of adaptive evolution vary throughout species ranges and on the decadal timescales most relevant to managing biodiversity for the 21st century. Using herbarium records, we reconstruct 150 y of evolution in an invasive plant as it spread across North America. Flowering phenology evolves to adapt to local growing seasons throughout the range but stalls after about a century. This punctuated, convergent evolution recapitulates long-term dynamics in the fossil record, implicating limits to evolutionary rates that are not evident for the first century of spread.
Changes in climate can alter the phenology of organisms, potentially decoupling partners within mutualisms. Previous studies have shown that plant and pollinator phenologies are shifting over time, but these shifts have primarily been documented for generalists and within small geographic regions, and the specific climatic cues regulating these shifts are not well‐understood. We examined phenological shifts in a specialist pollinator and its host plant species over a 117‐year study period using a digitized dataset of over 4000 unique collection records. We assess how climatic cues regulate these organisms' phenologies using PRISM weather data associated with each record. We tested the hypothesis that rates of phenological change would be greater at northern latitudes. We found that the phenology of the specialist bee pollinator Habropoda laboriosa is changing over time, but at different rates across its range. Specifically, phenology is advancing to a greater degree in more northern populations, with increasing phenological advances of 0.04 days/year with each degree of latitude, and with a delay in phenology in more southern populations. In contrast, only one species in the host plant genus Vaccinium is experiencing phenological change over time. For this plant, rates of change are also variable across latitudes, but in a pattern opposite that of the bee; while phenology is advancing across its range, rates of advance are highest in more southern populations, with decreasing phenological advances of 0.01 days/year with each degree of latitude. The phenologies of both the bee and three of four Vaccinium spp. were regulated primarily by spring temperature, with phenologies overall advancing with increasing temperature, and with the strongest responses shown by the bee in northern populations. Our study provides partial support for the hypothesis that phenologies advance most at northern latitudes, but demonstrates that pollinators and plants do not adhere similarly to this prediction. Additionally, we illustrate the potential for phenological mismatch between a specialist pollinator and its host plants by showing that plants and pollinators are advancing their phenologies at different rates across space and time and with differing responses to changing climatic cues.
Full-text available
Aerobiological data can be used as indirect but reliable measures of flowering phenology to analyze the response of plant species to ongoing climate changes. The aims of this study are to evaluate the performance of several phenological models for predicting the pollen start of season (PSS) in seven spring-flowering trees (Alnus glutinosa, Acer negundo, Carpinus betulus, Platanus occidentalis, Juglans nigra, Alnus viridis, and Castanea sativa) and in two summer-flowering herbaceous species (Artemisia vulgaris and Ambrosia artemisiifolia) by using a 26-year aerobiological data set collected in Turin (Northern Italy). Data showed a reduced interannual variability of the PSS in the summer-flowering species compared to the spring-flowering ones. Spring warming models with photoperiod limitation performed best for the greater majority of the studied species, while chilling class models were selected only for the early spring flowering species. For Ambrosia and Artemisia, spring warming models were also selected as the best models, indicating that temperature sums are positively related to flowering. However, the poor variance explained by the models suggests that further analyses have to be carried out in order to develop better models for predicting the PSS in these two species. Modeling the pollen season start on a very wide data set provided a new opportunity to highlight the limits of models in elucidating the environmental factors driving the pollen season start when some factors are always fulfilled, as chilling or photoperiod or when the variance is very poor and is not explained by the models.
Full-text available
Unlabelled: • Premise of the study: There is great interest in studying leaf-out times of temperate forests because of the importance of leaf-out in controlling ecosystem processes, especially in the face of a changing climate. Remote sensing and modeling, combined with weather records and field observations, are increasing our knowledge of factors affecting variation in leaf-out times. Herbarium specimens represent a potential new source of information to determine whether the variation in leaf-out times observed in recent decades is comparable to longer time frames over past centuries.• Methods: Here we introduce the use of herbarium specimens as a method for studying long-term changes in leaf-out times of deciduous trees. We collected historical leaf-out data for the years 1834-2008 from common deciduous trees in New England using 1599 dated herbarium specimens with young leaves.• Key results: We found that leaf-out dates are strongly affected by spring temperature, with trees leafing out 2.70 d earlier for each degree C increase in mean April temperature. For each degree C increase in local temperature, trees leafed out 2.06 d earlier. Additionally, the mean response of leaf-out dates across all species and sites over time was 0.4 d earlier per decade. Our results are of comparable magnitude to results from studies using remote sensing and direct field observations.• Conclusions: Across New England, mean leaf-out dates varied geographically in close correspondence with those observed in studies using satellite data. This study demonstrates that herbarium specimens can be a valuable source of data on past leaf-out times of deciduous trees.
Variation among species in their phenological responses to temperature change suggests that shifts in the relative timing of key life cycle events between interacting species are likely to occur under climate warming. However, it remains difficult to predict the prevalence and magnitude of these shifts given that there have been few comparisons of phenological sensitivities to temperature across interacting species. Here, we used a broad-scale approach utilizing collection records to compare the temperature sensitivity of the timing of adult flight in butterflies vs. flowering of their potential nectar food plants (days/°C) across space and time in British Columbia, Canada. On average, the phenology of both butterflies and plants advanced in response to warmer temperatures. However, the two taxa were differentially sensitive to temperature across space vs. across time, indicating the additional importance of non-temperature cues and/or local adaptation for many species. Across butterfly-plant associations, flowering time was significantly more sensitive to temperature than the timing of butterfly flight and these sensitivities were not correlated. Our results indicate that warming-driven shifts in the relative timing of life cycle events between butterflies and plants are likely to be prevalent, but that predicting the magnitude and direction of such changes in particular cases is going to require detailed, fine-scale data. This article is protected by copyright. All rights reserved. This article is protected by copyright. All rights reserved.
Leaf out phenology affects a wide variety of ecosystem processes and ecological interactions and will take on added significance as leaf out times increasingly shift in response to warming temperatures associated with climate change. There is, however, relatively little information available on the factors affecting species differences in leaf out phenology. An international team of researchers from eight Northern Hemisphere temperate botanical gardens recorded leaf out dates of c. 1600 woody species in 2011 and 2012. Leaf out dates in woody species differed by as much as 3 months at a single site and exhibited strong phylogenetic and anatomical relationships. On average, angiosperms leafed out earlier than gymnosperms, deciduous species earlier than evergreen species, shrubs earlier than trees, diffuse and semi-ring porous species earlier than ring porous species, and species with smaller diameter xylem vessels earlier than species with larger diameter vessels. The order of species leaf out was generally consistent between years and among sites. As species distribution and abundance shift due to climate change, interspecific differences in leaf out phenology may affect ecosystem processes such as carbon, water, and nutrient cycling. Our open access leaf out data provide a critical framework for monitoring and modelling such changes going forward.