Ecology, 89(2), 2008, pp. 332–341
? 2008 by the Ecological Society of America
GLOBAL WARMING AND FLOWERING TIMES IN THOREAU’S CONCORD:
A COMMUNITY PERSPECTIVE
ABRAHAM J. MILLER-RUSHING1AND RICHARD B. PRIMACK
Department of Biology, Boston University, 5 Cummington Street, Boston, Massachusetts 02215 USA
than they did in the past. However, some species’ flowering times have changed much more
than others. Data at the community level can clarify the variation in flowering responses to
climate change. In order to determine how North American species’ flowering times respond
to climate, we analyzed a series of previously unstudied records of the dates of first flowering
for over 500 plant taxa in Concord, Massachusetts, USA. These records began with six years
of observations by the famous naturalist Henry David Thoreau from 1852 to 1858, continued
with 16 years of observations by the botanist Alfred Hosmer in 1878 and 1888–1902, and
concluded with our own observations in 2004, 2005, and 2006. From 1852 through 2006,
Concord warmed by 2.48C due to global climate change and urbanization. Using a subset of
43 common species, we determined that plants are now flowering seven days earlier on average
than they did in Thoreau’s times. Plant flowering times were most correlated with mean
temperatures in the one or two months just before flowering and were also correlated with
January temperatures. Summer-flowering species showed more interannual variation in
flowering time than did spring-flowering species, but the flowering times of spring-flowering
species correlated more strongly to mean monthly temperatures. In many cases, such as within
the genera Betula and Solidago, closely related, co-occurring species responded to climate very
differently from one another. The differences in flowering responses to warming could affect
relationships in plant communities as warming continues. Common St. John’s wort
(Hypericum perforatum) and highbush blueberry (Vaccinium corymbosum) are particularly
responsive to changes in climate, are common across much of the United States, and could
serve as indicators of biological responses to climate change. We discuss the need for
researchers to be aware, when using data sets involving multiple observers, of how varying
methodologies, sample sizes, and sampling intensities affect the results. Finally, we emphasize
the importance of using historical observations, like those of Thoreau and Hosmer, as sources
of long-term data and to increase public awareness of biological responses to climate change.
As a result of climate change, many plants are now flowering measurably earlier
climate change; Concord, Massachusetts; flowering times; global warming; Henry David
It is astonishing how soon and unexpectedly flowers
appear, when the fields are scarcely tinged with green.
Yesterday, for instance, you observed only the radical
leaves of some plants; to-day you pluck a flower.
—Henry David Thoreau (Thoreau 1962)
Climate change is already affecting biological systems
worldwide (Walther et al. 2002, Parmesan and Yohe
2003, Root et al. 2003). Several studies have detected
effects of climate change on changes in species distribu-
tions (Grabherr et al. 1994, Parmesan et al. 1999), rates
of extinctions (McLaughlin et al. 2002, Pounds et al.
2006), the storage of carbon in plants and soils (Shaver
et al. 2000), and the timing of life history or phenological
events (Menzel and Fabian 1999, Inouye et al. 2000,
2003, Primack et al. 2004). Of these biological responses
to climate change, changes in the timing of phenological
events are the most widely reported and probably the
most easily detectable (Parmesan and Yohe 2003, Root
et al. 2003). Climate-related changes in phenology, some
quite dramatic, have been observed on every continent
and in the oceans (e.g., Menzel and Fabian 1999, Inouye
et al. 2000, Schwartz and Chen 2002, Edwards and
Richardson 2004, Gordo et al. 2005, Barbraud and
Weimerskirch 2006, Beaumont et al. 2006).
In most instances, phenological events, such as
flowering, bird migration, and amphibian reproduction,
are now occurring earlier than in the past (Parmesan and
Yohe 2003, Root et al. 2003). However, it is clear that
species’ phenologies are changing at different rates. In
some cases, different phenological events are changing at
different rates even within a single species or individual
plant or animal (Post et al. 2008). These changes have
Manuscript received 12 January 2007; revised 20 April 2007;
accepted 1 May 2007. Corresponding Editor: S. Naeem. For
reprints of this Special Feature, see footnote 1, p. 319.
1Present address: Rocky Mountain Biological Laborato-
ry, P.O. Box 519, Crested Butte, Colorado 81224 USA and
Department of Biology, University of Maryland, College
Park, Maryland, USA. E-mail: email@example.com
the potential to alter relationships among many species
(Stenseth and Mysterud 2002, Visser and Both 2005)
and could alter species’ exposures to abiotic factors such
as frost (Inouye 2008). Some changes in intertrophic
interactions are already evident (Inouye et al. 2000,
Edwards and Richardson 2004, Both et al. 2006). For
plants, studies have shown significant variation in the
rates at which flowering times are changing across
species (Fitter et al. 1995, Bradley et al. 1999, Sparks et
al. 2000). For example, in England, some species are
flowering more than a month earlier than they did 50
years ago, while other species’ flowering times are not
changing (Fitter et al. 1995). From these findings arise
several questions: Why do species respond differently to
climate change? Can we better characterize these
differences? What species, or groups of species, are
most (or least) sensitive to changes in climate? More
specifically, does season of flowering or growth form
explain any of the variation we see in the responses of
individual species to variations in climate?
The answers to these questions could impact on
individual performance (Gross and Werner 1983, Parra-
Tabla and Vargas 2004) and population and community
dynamics (Inouye et al. 2000, Visser and Both 2005).
Rare and endangered species that do not adapt to these
changes could face extinction. Unfortunately, the
number of species included in most historical data sets
has limited previous studies. To our knowledge, only
one major study has examined the responses of enough
plant species (in this case, 243) to search for biological
patterns that might explain differences in species’
responses (Fitter et al. 1995, Fitter and Fitter 2002).
That study showed that 16% of species flowered
significantly earlier in the 1990s than in previous
decades. Early-flowering species, annuals, and insect-
pollinated species showed the greatest sensitivity to
climate change (Fitter and Fitter 2002). Although their
findings are significant, the observed trends may be valid
only for Europe or central England, where the study
The purpose of our study was to characterize how
flowering times respond to variations in climate in North
America. To this end, we analyzed data from a
previously unstudied record of first flowering dates in
Concord, Massachusetts, USA. Our data set is unique in
several aspects. First, it spans an exceptionally long
period of time—155 years from 1852 to 2006—which we
accomplish by combining three individual sets of
observations. Second, our data set includes observations
on over 500 plant taxa, which allows us to identify
patterns that occur at the community level. Finally, our
data set begins with the observations of Henry David
Thoreau, the famous naturalist, philosopher, and author
of the widely read book Walden, which could make these
results particularly relevant to a nonscientist audience.
With this unique set of data, we investigated abiotic
factors contributing to variation in flowering responses
to climate change. We tested the hypothesis that climate
change has altered phenology, and we identified
potential mechanisms responsible for these phenological
Since the 1850s, several botanists have recorded
flowering times in Concord, Massachusetts, USA. These
records began with the work of Henry David Thoreau,
who observed the first flowering dates (FFD) of over 500
species of plants in Concord from 1852 to 1858
(Thoreau 1962; unpublished tables courtesy of B. P.
Dean). Alfred Hosmer, a shopkeeper and amateur
botanist, continued these observations of FFDs in
Concord for over 700 plant taxa in 1878 and 1888–
1902 (Hosmer 1878–1903). Thoreau’s and Hosmer’s
records included the flowering times of plants in all
habitat types. Later, from 1963 to 1993, Pennie
Logemann, a Concord landscape designer, maintained
records of flowering times for over 250 species of plants
that occurred on her property, which consisted primarily
of forest and wetland. Each of these botanists observed
new taxa in flower several days per week during the
flowering season. Thoreau intended to write a book
about phenology, but did not complete it before his
death (Thoreau 1993, 1999). We do not know why
Hosmer kept phenological records, as he never wrote
any papers based on his observations other than those
intended to update the flora of Concord (Hosmer
1899a, b). Logemann made phenological calendars as
an aid for designing gardens. We know that each of
these naturalists had a good working knowledge of the
flora of Concord, because of their abilities to distinguish
taxa that differ in subtle characteristics (Eaton 1974).
We made our own observations of flowering times in
Concord from 2003 to 2006. We purposefully used
methods similar to those of the previous naturalists,
particularly Thoreau and Hosmer. Two or three days a
week from March to October, we recorded plants in
flower across Concord. We observed over 500 species in
flower. For the analyses in this study, we did not use the
observations we made in 2003, because at that time we
were still learning the locations of the plants in Concord
and frequently missed the earliest flowering dates.
We analyzed in detail the FFDs for 43 common,
early-flowering species for which we had the most
flowering data. For these species, we included observa-
tions made by Thoreau (six years, 1852–1858), Hosmer
(16 years, 1878 and 1888–1902), and ourselves (three
years, 2004–2006), for a total of 25 years of observa-
tions. For each species, we had FFD data for at least 19
of the 25 years. Because we did not have observations
for each species in each year, we calculated the difference
between the FFD in each year and the FFD in the
benchmark year of 1893, a year for which we had
observations for all 43 species. This calculation mini-
mized biases caused by different species missing from
each year. We used regression analysis to determine the
relationship between FFD and mean monthly temper-
February 2008333 PHENOLOGY: AN INTEGRATIVE SCIENCE
atures. Logemann observed first flowering dates for
eight of these species in at least nine years during the
period 1963–1993. We included her observations in
regressions for these eight species.
For a broad survey of flowering phenology, we
analyzed the records of the 296 taxa (293 species, three
distinct subspecies) of flowering plants for which
Hosmer had made an observation in each of 15 years,
1888–1902. We compared FFD in each year with mean
monthly temperatures. For each taxon, we correlated
the FFD with the mean monthly temperatures of the
month of flowering and each of the 11 preceding
months. From those correlations, we found the months
for which the mean temperatures were best correlated
with FFD. We also correlated each FFD with the mean
temperature for January and the two months preceding
flowering, as temperatures in those months were often
significantly correlated with FFD. Following the exam-
ple of Fitter et al. (1995), we used standard deviations
about the mean FFD as a measure of interannual
variation in flowering time. We then used the regression
analyses and standard deviations to compare several
groups of taxa—e.g., plants that flower in different
months; native and nonnative taxa; and annuals,
perennial herbs, and woody plants—in order to find
patterns that might explain the overall variation in
response to year-to-year changes in climate.
For our analysis, we used dry bulb air temperatures
recorded in a standardized way at Blue Hill Meteoro-
logical Observatory in Milton, Massachusetts, USA (33
km southeast of Concord). Unfortunately, the weather
records for Concord were not complete for the time
period between 1888 and 1902. However, we correlated
the available Concord temperature records (1931–1949)
with those for Blue Hill Observatory and found that
mean monthly temperatures for each year had a
correlation coefficient of 0.995 or higher. Thus, we are
confident that the temperature in Concord was closely
related to that at Blue Hill Observatory.
For 43 common, spring-flowering species (33 native,
10 nonnative), we combined 25 years of observations by
three different observers (Thoreau, Hosmer, and our-
selves) that span the years 1852–2006. Over this time,
mean annual temperatures in Concord rose by 2.48C and
mean monthly temperatures in January, April, and May
rose by 2.38C, as determined by linear regression (Fig.
1). Our analysis of these observations showed that these
plants have flowered progressively earlier over the past
150 years (Fig. 2a). For the 43 species, the mean FFD
during Thoreau’s observations (1852–1858) was 14 May,
whereas the mean FFD for our observations (2004–
2006) was 7 May, seven days earlier. The mean FFD for
Hosmer (1878, 1888–1902) was 10 May, intermediate
between Thoreau’s and our own observations. The
differences in FFDs among the three time periods were
highly significant as determined by two-way ANOVA,
considering time period (Thoreau, Hosmer, and our-
selves) and species as factors (P , 0.001). The FFD for
some species changed dramatically from 1852 to 2006.
For example, highbush blueberry (Vaccinium corymbo-
sum), a native shrub, and yellow wood sorrel (Oxalis
europaea), a native herb, are now flowering 21 and 32
days earlier than they did 150 years ago, respectively.
to 2006. The upper line and open circles represent mean annual temperatures. The lower line and solid squares represent mean
monthly temperatures in January, April, and May, temperatures that were highly correlated with flowering times for many species.
Horizontal lines show long-term means for each (annual ¼ 8.38C; Jan, Apr, May ¼ 5.18C). Circles and squares show years with
Temperatures at Blue Hill Meteorological Observatory (33 km southeast of Concord, Massachusetts, USA) from 1852
ABRAHAM J. MILLER-RUSHING AND RICHARD B. PRIMACK 334 Ecology, Vol. 89, No. 2
The earlier flowering times were strongly correlated
with warming mean monthly temperatures in January,
April, and May over that time period. On average,
plants flowered 3.07 days earlier for each 18C increase in
mean monthly temperatures, as determined by linear
regression (43 species, R2¼ 0.609, P , 0.001; Fig. 2b).
The changes in FFD for native (33 species, 2.93 days
earlier per 18C, R2¼0.596, P , 0.001) and nonnative (10
species, 3.40 days earlier per 18C, R2¼0.428, P , 0.001)
were virtually identical. Average January, April, and
May temperatures were 4.38C during Thoreau’s obser-
vations, 5.08C during Hosmer’s observations, and 5.98C
during our own observations.
Of these 43 common species, Logemann observed
eight, all native, in at least nine years (1963–1993).
Inclusion of her observations improved the ability of
temperature to explain FFDs for three species—
shadbush (Amelanchier canadensis), bunchberry (Cornus
canadensis), and wild strawberry (Fragaria virginiana)—
as indicated by R2values from the flowering-tempera-
ture relationship; R2values increased when her obser-
vations were included. FFDs of three species were not
correlated with temperature, with or without Loge-
mann’s observations, and Logemann’s observations did
not improve the explanatory power for the remaining
For our broad survey of 296 species that Hosmer
observed from 1888–1902, mean FFD ranged from 4
March, for skunk cabbage (Symplocarpus foetidus), to
14 August, for swamp rose mallow (Hibiscus palustris).
Some species, such as witch hazel (Hamamelis virgin-
iana), flowered earlier or later, but Hosmer’s records for
those species were not complete. Most plant taxa
flowered in May (82), June (86), and July (76); fewer
taxa flowered in March (2), April (32), and August (18);
and the mean flowering date for all plants was 12 June.
On average, the Concord plants observed by Hosmer
responded to each 18C increase in mean January, April,
and May temperatures by flowering 3.28 days earlier
(296 taxa, R2¼ 0.84, P , 0.001). Of the 296 taxa we
examined, 279 (94%) flowered earlier in years with
warmer mean monthly temperatures in January and the
two months prior to flowering, as indicated by negative
correlations; 168 (57%) showed significant (P , 0.05)
correlations between FFD and mean monthly temper-
atures. No taxon showed a significant trend toward later
FFD with warmer mean monthly temperatures. For 20
taxa, mean monthly temperatures explained more than
60% of the variation in FFDs (R2. 0.60; Table 1). For
(FFD) for 33 native and 10 nonnative species (a)
over time and (b) in response to warming mean
monthly temperatures in January, April, and
May. (a) Symbols correspond to observers and
time periods: solid squares for observations by
Henry David Thoreau (1852–1858), open trian-
gles for observations by Alfred Hosmer (1878,
1888–1902), and solid circles for our observations
(2004–2006). Solid horizontal bars with standard
error bars represent the mean FFD for each
observer. Each point (other than solid bars) was
calculated by using the difference between when a
species flowered in a particular year and when it
flowered in the benchmark year of 1893, when all
species were observed. Then we averaged these
differences among species; each point represents
the mean difference in FFD from 1893 for all
species observed in a particular year. (b) Solid
diamonds and the solid line represent mean FFD
for 33 native species. Open circles and the dashed
line represent mean FFD for 10 nonnative
species. Means were calculated as described for
(a), as differences from FFD in 1893. Lines are
best-fit regressions. Natives flowered 2.93 days
earlier per 18C warming (R2¼ 0.609, P , 0.001).
Nonnatives flowered 3.40 days earlier per 18C
warming (R2¼ 0.428, P , 0.001).
Change in mean first flowering dates
February 2008335PHENOLOGY: AN INTEGRATIVE SCIENCE
example, 73% of the variation in the FFD of common
St. John’s wort (Hypericum perforatum) was explained
by changes in mean monthly temperatures in January,
April, and May. Two other species sensitive to mean
monthly temperatures, privet (Ligustrum vulgare) and
robin’s plantain (Erigeron pulchellus), each responded to
each 18C increase in temperatures by flowering about six
Because taxa respond more to temperatures in
individual months rather than annual temperatures
(Fitter et al. 1995, Sparks and Carey 1995), we examined
the relative importance of each month’s mean temper-
ature in predicting changes in mean FFD. Of the 221
taxa with FFDs that were significantly correlated with
the mean temperature of at least one month (P , 0.05),
116 (52%) were correlated with mean May temperatures,
while 100 (45%) were correlated with mean January
temperatures (Fig. 3). The FFDs of 162 (73%) taxa were
correlated with the mean temperatures either in the
month of flowering or in one of the two months prior to
months prior to flowering, out of a sample of 296 taxa.
The 20 taxa with first flowering dates (FFD) best predicted by mean monthly temperatures in January and the two
Species Common nameR2
common St. John’s wort
downy yellow violet
late low blueberry
larger blue flag
marsh blue violet
Notes: Change is given as days/8C. P , 0.001 for all taxa shown. Negative change indicates a change toward earlier FFD in
? Nonnative species.
of a taxon. Only significant correlations are shown (P , 0.05). The pattern is consistent whether the number of taxa or the
percentage of possible occurrences is considered. The FFDs for a total of 221 taxa were significantly correlated with mean
temperatures in at least one month. We tested correlations between FFD and temperatures in the month of flowering and in the 11
preceding months. Months are shown as occurring during the year of flowering or during the previous calendar year.
Frequency with which each month’s mean temperature was significantly correlated with the first flowering date (FFD)
ABRAHAM J. MILLER-RUSHING AND RICHARD B. PRIMACK336Ecology, Vol. 89, No. 2
flowering. Fig. 3 shows three peaks where months were
relatively important in predicting FFD compared to
nearby months: large peaks in April, May, and January
of the flowering year. A small number of species had
FFDs correlated with temperatures in August, Septem-
ber, and October of the year prior to flowering. This
pattern, which is consistent whether absolute number of
occurrences or percent of possible occurrences are
considered (data not shown), suggests that many species
are particularly sensitive to cold January temperatures
and to warming temperatures early in the spring or just
prior to flowering.
The peak in January is particularly striking, because
FFDs were often significantly correlated with mean
January temperature, but very rarely correlated with the
adjacent months, i.e., December of the previous year
and February of the flowering year. Mean January
temperatures were significantly correlated with FFDs
for equal percentages of all growth forms (approximate-
ly 33% each of annuals, perennial herbs, and woody
shrubs). It is possible that these correlations were due to
severely cold temperatures in January. January was the
coldest month in eight out of the 15 years that Hosmer
kept records (1888–1902).
Due to the large number of taxa included in our
analysis, we were also able to examine the effects of
season of flowering, growth form, nativeness, and
habitat on responses to climate change. We found that
FFDs for early-flowering (March, April, May, June)
taxa were more correlated with mean monthly temper-
atures than were FFDs for late-flowering (July, August)
taxa as shown by regression analysis (296 taxa, P ,
0.001; Fig. 4). In other words, the mean FFDs of early-
flowering taxa were better predicted by and more
responsive to mean monthly temperatures than were
late-flowering taxa. Even though FFDs of early-flower-
ing species were more correlated with temperature, we
found that late-flowering taxa had greater standard
deviations about their mean flowering dates than did
early-flowering taxa, as determined by regression
analysis (P ¼ 0.016).
When we analyzed the same relationship according to
growth form, we found that growth forms differed
significantly in their patterns. The standard deviations of
annuals were not significantly affected by season of
flowering (18 taxa, P ¼ 0.120); late-flowering perennial
herbs had greater standard deviations than early-
flowering taxa (194 taxa, P , 0.001). Woody plants,
however, displayed an opposite trend: early-flowering
taxa had greater standard deviations than late-flowering
ones (66 taxa, P¼0.032). Because the majority of taxa in
the data set were perennial herbs (65%), it is likely that
the relationship between standard deviation and mean
FFD for perennial herbs drove the trend seen when all
taxa were considered together. In addition, on average,
annuals showed a marginally significantly greater
standard deviation about their mean FFDs than did
perennial herbs (11.2 compared to 8.7 days, t ¼ 1.92,
two-tailed P ¼ 0.07), which in turn showed a signifi-
cantly greater standard deviation than did woody plants
(8.7 compared to 7.4 days, t¼?2.77, two-tailed P¼006).
That is, life-form explained, in part, why some taxa had
more year-to-year variation in flowering compared to
We found that nonnative taxa did not differ from
native taxa in their flowering responses to temperature.
Both native (239 taxa) and nonnative (54 taxa) taxa
showed a great deal of variation of response, but neither
standard deviations about mean FFDs (8.6 vs. 9.2 days,
t ¼ ?0.971, two-tailed P ¼ 0.33) nor correlations with
mean monthly temperatures (correlation coefficients of
?0.490 vs.?0.523, t¼0.863, two-tailed P¼0.39) differed
significantly between the two groups. Similarly, habitat
(aquatic, forest, grassland, roadside, wetland) did not
explain any of the variation in flowering responses to
The FFDs of many closely related (i.e., within the
same genus) and co-occurring species responded to
changes in temperature at very different rates. For
example, black birch (Betula lenta) and gray birch
(Betula populifolia), which occur in many of the same
habitats in Concord, show very different responses to
temperature (Miller-Rushing and Primack, in press).
Black birch flowered 2.83 days earlier for each 18C
increase in January, March, and April temperatures (R2
¼0.376, P¼0.015), whereas gray birch FFDs showed no
relationship with temperature (P ¼ 0.535). In an even
more dramatic example, rough-stemmed goldenrod
(Solidago rugosa) flowered 11.17 days earlier for each
18C increase in January, May, and June temperatures
(R2¼ 0.554, P ¼ 0.001), whereas the FFDs of lance-
leaved goldenrod (Solidago graminifolia) and most other
goldenrods showed no relationship with temperature (P
¼ 0.535). Among the 52 genera for which Hosmer
(FFD) from 1888 to 1902 and the correlation between FFD and
mean monthly temperatures in January and the two months
prior to flowering for 296 plant taxa. Each point represents one
taxon. Negative correlation coefficients indicate earlier flower-
ing in warmer years. Slope ¼ 0.003, R2¼ 0.141, P , 0.001.
The relationship between mean first flowering date
February 2008337PHENOLOGY: AN INTEGRATIVE SCIENCE
observed more than one species, 31 (60%) contained at
least one species with FFDs significantly correlated to
mean monthly temperatures in January and the two
months prior to flowering and at least one species with
FFDs that were not correlated with temperature. Of the
25 genera for which Hosmer observed just two species,
10 (40%) contained one species with FFDs significantly
correlated to mean monthly temperatures and one
species with FFDs not correlated to temperature.
On average, plants in Concord appear to flower now
seven days earlier than they did when Thoreau made his
observations (1852–1858). Most of this change in
flowering time is probably due to rising winter and
spring temperatures. Temperatures in eastern Massa-
chusetts have increased more rapidly than in many other
areas of the world due to the combination of global
warming and the urban heat island effect (New England
Regional Assessment Group 2001). The rate at which
Concord plants responded to warming—3.3 days earlier
flowering for each 18C increase in mean monthly
temperatures in January, April, and May—fits well with
findings in Europe (Sparks and Carey 1995, Chmielew-
ski and Ro ¨ tzer 2001, Fitter and Fitter 2002) and North
America (Schwartz and Reiter 2000, Cayan et al. 2001).
We also found that mean monthly temperatures in
January and the two months immediately preceding
flowering were significantly correlated with the FFDs
for many species. It is known that plants respond to
temperatures from the previous fall (Fitter et al. 1995).
Cooling temperatures in the fall and winter often
contribute to the vernalization process, in which colder
temperatures lead to increased competence and earlier
flowering (Chuine 2000, Sung and Amasino 2004).
However, we found that colder January temperatures
were correlated to later flowering times. One reason for
the discrepancy could be the difference between climate
patterns in the northeastern United States and Western
Europe, the site of many previous studies of plant
phenology (e.g., Fitter et al. 1995, Sparks et al. 2000).
Winter temperatures in Western Europe’s maritime
climate tend to be significantly milder than those in
the continental climate of the northeastern United States
(Hartmann 1994, Seager et al. 2002). For an average of
12 days in each January (1963–2006), there is no snow
cover in Concord to insulate plants from the extreme
cold. Thus, overwintering plants and seeds in the
northeastern United States may be much more suscep-
tible to damage from January’s extreme cold tempera-
tures than are plants in Western Europe. It is also
possible that the phenologies of species in colder
climates may simply be particularly sensitive to climate
(Tho ´ rhallsdo ´ ttir 1998).
In another surprising finding, early-flowering peren-
nial herbs had FFDs that displayed less interannual
variability than did those of late-flowering taxa,
although the variation in early-flowering taxa was more
closely linked to mean monthly temperatures (Fig. 2).
Previous studies have found individual examples of
highly variable flowering times in late-flowering peren-
nial herbs, such as bird’s foot trefoil (Lotus corniculatus;
Ollerton and Lack 1998), but we do not know of
previous evidence suggesting that it may be a widespread
pattern. Woody species showed the more usual pattern
of greater interannual variation in flowering times for
early-flowering species rather than late-flowering species
(Fitter et al. 1995, Post and Stenseth 1999). It seems that
the flowering times of many late-flowering perennial
herbs may have been linked to an indicator, or set of
indicators, that were more variable than mean monthly
temperatures, or that late-flowering perennial herbs have
inherently variable flowering times. Possibilities of non-
temperature indicators for flowering times include
phenomena such as rainfall, shading, and land use. It
is also possible that monthly temperatures were too
coarse to have a detectable effect on flowering times for
these species, and that daily temperatures may be more
appropriate. Further study is clearly necessary to isolate
the factors responsible for the high variation in the
flowering dates of late-flowering perennial herbs. In-
triguingly, Rich et al. (2008) found additional differenc-
es between woody and herbaceous species in a pin ˜ on–
juniper woodland suggesting that herbaceous species are
more responsive to environmental variation than are
The flowering times of several species appear sensitive
enough to changes in temperature that they could serve
as indicator species and be used to measure biological
responses to changes in climate over time. Among the
species in our study, two particularly common species—
common St. John’s wort (Hypericum perforatum) and
chicory (Cichorium intybus)—both had high correlations
with mean monthly temperatures (R2. 0.60) and had
mean FFDs that advanced more than three days per 18C
increase in temperature (Table 1). Although these
species are nonnative, they are both common in urban
and rural areas across the United States and easy to
identify. Common and widespread native species, such
as highbush blueberry (Vaccinium corymbosum), Canada
mayflower (Maianthemum canadense), and larger blue
flag (Iris versicolor; see Plate 1) could also serve as
indicator species (Table 1). Before these species are
utilized as indicator species across their ranges, however,
we suggest that studies determine if their sensitivity to
changes in temperatures is consistent across their ranges.
Their usefulness as indicators for interacting species
should also be tested, as Both et al. (2006) have shown
that phenological changes may differ among different
parts of a food chain.
Our study suggests that flowering times are changing
at different rates for several closely related, co-occurring
species, such as those within the genera Betula and
Solidago. As the timing of flowering and other
correlated life history traits change for these species,
interactions among the species will also change. Un-
ABRAHAM J. MILLER-RUSHING AND RICHARD B. PRIMACK 338Ecology, Vol. 89, No. 2
doubtedly, these changes will be to the advantage of
some species and disadvantage of others, although it is
difficult to predict the winners and losers. It is also clear
that the net effect of phenological changes on the fitness
of individuals or species will depend on complex, timing-
based interactions, sometimes spanning multiple trophic
levels (Stenseth and Mysterud 2002, Visser and Both
2005). For example, plants with particularly rapid
responses to changes in temperature could flower before
the emergence of their pollinators, thus decreasing their
chances of reproductive success (Kudo et al. 2008). In
other cases, plants may become more susceptible to frost
events or benefit from the lengthening of the growing
season (Inouye 2008, Kudo et al. 2008). As studies like
ours identify the species most sensitive to changes in
climate, researchers can specifically include these species
in their examinations of the ecological and evolutionary
impacts of non-synchronous shifts in flowering times.
In many instances, the best long-term phenological
data may contain observations made by several observ-
ers over long periods of time, as in our study. In these
cases, researchers must be mindful of the different time
periods and the methods that various observers might
use, including sampling effort (days/week, hours/day,
total area examined) and definitions of what constitutes
an open flower. For example, the statistical power of our
analysis was limited because of the heterogeneity of our
data, with only three years of recent observations. In
such a circumstance, one anomalous year could alter
results. In our case, mean January, April, and May
temperatures in 2004 and 2005 were colder than most
years since 1990 (Fig. 1). Thus, our estimates of changes
in flowering times are probably quite conservative. In
addition, we, Thoreau, and Hosmer observed flowering
times throughout Concord, while Logemann observed
flowering times only on her property in Concord.
Because Logemann observed a smaller area and fewer
plants, the first flowering dates she observed for many
species were later than they were for the other observers
(data not shown). Many phenological records document
changes in first observations, such as first flowering
dates rather than mean flowering dates. Changes in
population size or sampling effort can affect these first
observations independently of changes in the changes in
the population mean (Tryjanowski and Sparks 2001). If
populations decline over time or if sampling intensity
declines, first observations can occur later even when the
population mean does not change. Similarly, if popula-
tions increase over time or if sampling intensity
increases, first observations can occur earlier even when
the population mean does not change. Based on
Hosmer’s descriptions of species abundance, population
sizes in Concord remained fairly constant over the last
century for 32 of 43 of the main study species.
Population sizes declined over time for the remaining
11 species, meaning that estimates of changes in
Concord, Massachusetts (USA). The flowering date of this species may provide a good indicator of biotic responses to climate
change. Photo credit: A. J. Miller-Rushing and R. B. Primack.
The first flowering dates of larger blue flag (Iris versicolor) were highly correlated with changes in temperature in
February 2008339 PHENOLOGY: AN INTEGRATIVE SCIENCE
flowering times are probably overly conservative for
these 11 species. Clearly, it is critical that researchers
combine phenological data with descriptions of obser-
vation methods and changes in population sizes. Given
these caveats, however, observations made by different
individuals, or even using different methods, can yield
surprisingly high quality, consistent results (Miller-
Rushing et al. 2006).
Because of the clear ecological and evolutionary
importance of phenological responses to climate change,
we suggest that researchers increase efforts to collect
long-term phenological data via new projects and
searches of historical records. Many Long-term Ecolog-
ical Research (LTER) sites, as well as other research
sites, already collect such data. In addition, phenological
data sets already exist in many libraries, herbaria,
museums, and private homes (Ledneva et al. 2004,
Primack et al. 2004, Miller-Rushing et al. 2006).
Thoreau’s and Hosmer’s records were freely available
at various libraries, but had never been previously
analyzed. Logemann quite willingly shared her note-
books and charts, which she kept in her home. By using
such pre-existing records and adding new sites for
phenological studies, researchers could greatly enhance
our understanding of how phenological changes vary
according to location and species and how they might
affect other aspects of ecology and evolution (Betan-
court and Schwartz 2005).
In addition, evidence of phenological changes can
improve public awareness of the effects that climate
change is already having on biological systems. People
can see changes in phenology in their immediate
environment: plants flowering in gardens, fruits ripen-
ing, and birds arriving at bird feeders. We believe that
building on the observations of a well known figure such
as Thoreau can show that plants are responding to
climate change and increase the potential for public
outreach. Other studies of changes in phenology made
by famous individuals such as Aldo Leopold (Bradley et
al. 1999), or in well-known locations such as Wash-
ington, D.C. (Abu-Asab et al. 2001) and Boston
(Primack et al. 2004, Miller-Rushing et al. 2006)
generate similar public interest. Thoreau was keenly
aware of the importance of educating people about
environmental issues. He helped his townsmen to
appreciate wild nature, and he encouraged them to
protect it. He wrote, ‘‘I think that each town should have
a park, or rather a primitive forest of five hundred or a
thousand acres, either in one body or several, where a
stick should never be cut for fuel, nor for the navy, nor
to make wagons, but stand and decay for higher uses—a
common possession forever, for instruction and recre-
ation.’’ Residents of Concord and the government have
followed this advice; about 40% of Concord’s land is
preserved in parks and protected areas. With the help of
these protected areas, we have been able to continue the
same observations of flowering times made by Thoreau
at the same localities in Concord. We now hope that
Thoreau’s observations and our own work will promote
broad discussion of the effects of climate change on
biological systems. Only with an understanding of the
changes taking place can people make informed
decisions regarding climate change.
We appreciate the assistance of numerous Boston University
undergraduates, especially Kiruba Dharanaswareem, Sharda
Mukunda, Dan Primack, and Ellen Reed, who helped with data
collection and entry. Ray Angelo, Brad Dean, Mary Walker,
and Leslie Wilson helped us to locate and understand the flora
of Concord and the data sets of Thoreau and Hosmer. Brad
Dean, in particular, gave us copies of Thoreau’s unpublished
tables of flowering times, which Thoreau had extracted from his
journals. Pennie Logemann generously shared her observations
with us. We thank Helge Bruelheide, Phil Cafaro, Brad Dean,
Alistair Fitter, David Foster, David Inouye, Claude Lavoie,
Annette Menzel, Jeff Ollerton, Nathan Philips, Elizabeth Platt,
Eric Post, Gian-Reto Walther, and one anonymous reviewer
whose comments improved drafts of the manuscript. The
National Science Foundation, the Arnold Arboretum of
Harvard University, and Boston University provided funds
for this project.
Abu-Asab, M. S., P. M. Peterson, S. G. Shetler, and S. S. Orli.
2001. Earlier plant flowering in spring as a response to global
warming in the Washington, DC, area. Biodiversity and
Barbraud, C., and H. Weimerskirch. 2006. Antarctic birds
breed later in response to climate change. Proceedings of the
National Academy of Sciences (USA) 103:6248–6251.
Beaumont, L. J., I. A. W. McAllan, and L. Hughes. 2006. A
matter of timing: changes in the first date of arrival and last
date of departure of Australian migratory birds. Global
Change Biology 12:1339–1354.
Betancourt, J. L., and M. D. Schwartz. 2005. Implementing a
U.S. national phenology network. Eos 86:539–542.
Both, C., S. Bouwhuis, C. M. Lessells, and M. E. Visser. 2006.
Climate change and population declines in a long-distance
migratory bird. Nature 441:81–83.
Bradley, N. L., A. C. Leopold, J. Ross, and W. Huffaker. 1999.
Phenological changes reflect climate change in Wisconsin.
Proceedings of the National Academy of Sciences (USA) 96:
Cayan, D. R., S. A. Kammerdiener, M. D. Dettinger, J. M.
Caprio, and D. H. Peterson. 2001. Changes in the onset of
spring in the western United States. Bulletin of the American
Meteorological Society 82:399–415.
Chmielewski, F. M., and T. Ro ¨ tzer. 2001. Response of tree
phenology to climate change across Europe. Agricultural and
Forest Meteorology 108:101–112.
Chuine, I. 2000. A unified model for budburst of trees. Journal
of Theoretical Biology 207:337–347.
Eaton, R. J. 1974. A flora of Concord. Museum of Compar-
ative Zoology, Harvard University, Cambridge, Massachu-
Edwards, M., and A. J. Richardson. 2004. Impact of climate on
marine pelagic phenology and trophic mismatch. Nature 430:
Fitter, A. H., and R. S. R. Fitter. 2002. Rapid changes in
flowering time in British plants. Science 296:1689–1691.
Fitter, A. H., R. S. R. Fitter, I. T. B. Harris, and M. H.
Williamson. 1995. Relationships between first flowering date
and temperature in the flora of a locality in central England.
Functional Ecology 9:55–60.
Gordo, O., L. Brotons, X. Ferrer, and P. Comas. 2005. Do
changes in climate patterns in wintering areas affect the
ABRAHAM J. MILLER-RUSHING AND RICHARD B. PRIMACK340 Ecology, Vol. 89, No. 2
timing of the spring arrival of trans-Saharan migrant birds?
Global Change Biology 11:12–21.
Grabherr, G., M. Gottfried, and H. Pauli. 1994. Climate effects
on mountain plants. Nature 369:448.
Gross, R. S., and P. A. Werner. 1983. Relationships among
flowering phenology, insect visitors, and seed-set of individ-
uals: experimental studies on four co-occurring species of
goldenrod (Solidago: Compositae). Ecological Monographs
Hartmann, D. L. 1994. Global physical climatology. Academic
Press, San Diego, California, USA.
Hosmer, A. W. 1978–1903. Alfred W. Hosmer Botanical
Manuscripts, 1878–1903. William Munroe Special Collec-
tions. Concord Free Public Library, Concord, Massachu-
Hosmer, A. W. 1899a. On the plants introduced by Minot Pratt
at Concord, Massachusetts. Rhodora 1:170–172.
Hosmer, A. W. 1899b. Further additions to the flora of
Middlesex County, Mass. Rhodora 1:223–224.
Inouye, D. W. 2008. Effects of climate change on phenology,
frost damage, and floral abundance of montane wildflowers.
Inouye, D. W., B. Barr, K. B. Armitage, and B. D. Inouye.
2000. Climate change is affecting altitudinal migrants and
hibernating species. Proceedings of the National Academy of
Sciences (USA) 97:1630–1633.
Inouye, D. W., F. Saavedra, and W. Lee-Yang. 2003.
Environmental influences on the phenology and abundance
of flowering by Androsace septentrionalis (Primulaceae).
American Journal of Botany 90:905–910.
Kudo, G., T. Y. Ida, and T. Tani. 2008. Linkages between
phenology, pollination, photosynthesis, and reproduction in
deciduous forest understory plants. Ecology 89:321–331.
Ledneva, A., A. J. Miller-Rushing, R. B. Primack, and C.
Imbres. 2004. Climate change as reflected in a naturalist’s
diary, Middleborough, Massachusetts. Wilson Bulletin 116:
McLaughlin, J. F., J. J. Hellmann, C. L. Boggs, and P. R.
Ehrlich. 2002. Climate change hastens population extinc-
tions. Proceedings of the National Academy of Sciences
Menzel, A., and P. Fabian. 1999. Growing season extended in
Europe. Nature 397:659.
Miller-Rushing, A. J., and R. B. Primack. In press. A
comparison of the impacts of winter temperatures on two
birch (Betula) species. Tree Physiology.
Miller-Rushing, A. J., R. B. Primack, D. Primack, and S.
Mukunda. 2006. Photographs and herbarium specimens as
tools to document phenological changes in response to global
warming. American Journal of Botany 93:1667–1674.
New England Regional Assessment Group. 2001. New England
regional assessment. University of New Hampshire, Institute
for the Study of Earth, Oceans, and Space, Durham, New
Ollerton, J., and A. Lack. 1998. Relationships between
flowering phenology, plant size and reproductive success in
Lotus corniculatus (Fabaceae). Plant Ecology 139:35–47.
Parmesan, C., et al. 1999. Poleward shifts in geographical
ranges of butterfly species associated with regional warming.
Parmesan, C., and G. Yohe. 2003. A globally coherent
fingerprint of climate change impacts across natural systems.
Parra-Tabla, V., and C. F. Vargas. 2004. Phenology and
phenotypic natural selection on the flowering time of a deceit-
pollinated tropical orchid, Myrmecophila christinae. Annals
of Botany 94:243–250.
Post, E. S., C. Pedersen, C. C. Wilmers, and M. C.
Forchhammer. 2008. Phenological sequences reveal aggre-
gate life history response to climatic warming. Ecology 89:
Post, E., and N. C. Stenseth. 1999. Climatic variability, plant
phenology, and northern ungulates. Ecology 80:1322–1339.
Pounds, J. A., et al. 2006. Widespread amphibian extinctions
from epidemic disease driven by global warming. Nature 439:
Primack, D., C. Imbres, R. B. Primack, A. J. Miller-Rushing,
and P. Del Tredici. 2004. Herbarium specimens demonstrate
earlier flowering times in response to warming in Boston.
American Journal of Botany 91:1260–1264.
Rich, P. M., D. D. Breshears, and A. B. White. 2008.
Phenology of mixed woody–herbaceous ecosystems following
extreme events: net and differential responses. Ecology 89:
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.
Schwartz, M. D., and X. Chen. 2002. Examining the onset of
spring in China. Climate Research 21:157–164.
Schwartz, M. D., and B. E. Reiter. 2000. Changes in North
American spring. International Journal of Climatology 20:
Seager, R., D. S. Battisti, J. Yin, N. Gordon, N. Naik, A. C.
Clement, and M. A. Cane. 2002. Is the Gulf Stream
responsible for Europe’s mild winters? Quarterly Journal of
the Royal Meteorological Society 128:2563–2586.
Shaver, G. R., J. Canadell, F. S. Chapin, J. Gurevitch, J. Harte,
G. Henry, P. Ineson, S. Jonasson, J. Mellilo, L. Pitelka, and
L. Rustad. 2000. Global warming and terrestrial ecosystems:
a conceptual framework for analysis. BioScience 50:871–882.
Sparks, T. H., and P. D. Carey. 1995. The responses of species
to climate over two centuries: an analysis of the Marsham
phenological record, 1736–1947. Journal of Ecology 83:321–
Sparks, T. H., E. P. Jeffree, and C. E. Jeffree. 2000. An
examination of the relationship between flowering times and
temperature at the national scale using long-term phenolog-
ical records from the UK. International Journal of Biome-
Stenseth, N. C., and A. Mysterud. 2002. Climate, changing
phenology, and other life history traits: nonlinearity and
match-mismatch to the environment. Proceedings of the
National Academy of Sciences (USA) 99:13379–13381.
Sung, S., and R. M. Amasino. 2004. Vernalization and
epigentics: how plants remember winter. Current Opinion
in Plant Biology 7:4–10.
Thoreau, H. D. 1962. The journal of Henry D. Thoreau. B.
Torrey and F. H. Allen, editors. Dover Publications, New
York, New York, USA.
Thoreau, H. D. 1993. Faith in a seed: the dispersion of seeds
and other late natural history writings. B. P. Dean, editor.
Island Press, Washington, D.C., USA.
Thoreau, H. D. 1999. Wild fruits: Thoreau’s rediscovered last
manuscript. B. P. Dean, editor. W. W. Norton and
Company, New York, New York, USA.
Tho ´ rhallsdo ´ ttir, T. E. 1998. Flowering phenology in the central
highland of Iceland and implications for climatic warming in
the Arctic. Oecologia 114:43–49.
Tryjanowski, P., and T. H. Sparks. 2001. Is the detection of the
first arrival date of migrating birds influenced by population
size? A case study of the Red-backed Shrike Lanius collurio.
International Journal of Biometeorology 45:217–219.
Visser, M. E., and C. Both. 2005. Shifts in phenology due to
global climate change: the need for a yardstick. Proceedings
of the Royal Society of London, B 272:2561–2569.
Walther, G. R., E. Post, P. Convey, A. Menzel, C. Parmesan,
T. J. C. Beebee, J. M. Fromentin, O. Hoegh-Guldberg, and
F. Bairlein. 2002. Ecological responses to recent climate
change. Nature 416:389–395.
February 2008341PHENOLOGY: AN INTEGRATIVE SCIENCE