Upload
victoria
View
1
Download
0
Embed Size (px)
Citation preview
DOI: 10.1111/j.1466-8238.2008.00408.x © 2008 The Authors
648
Journal compilation © 2008 Blackwell Publishing Ltd www.blackwellpublishing.com/geb
Global Ecology and Biogeography, (Global Ecol. Biogeogr.)
(2008)
17
, 648–657
RESEARCHPAPER
Blackwell Publishing Ltd
Global patterns in fruiting seasons
Steven Ting, Stephen Hartley and K. C. Burns*
ABSTRACT
Aim
To identify geographical and climatic correlates of the timing of fruit productionin fleshy fruited plant communities.
Location
Global.
Methods
We searched the literature for studies documenting monthly variation inthe number of fleshy fruited species bearing ripe fruits in plant communities (i.e.fruit phenologies). From these data, we used circular vector algebra to characterizeseasonal peaks in fruit production (mean date, as an angle) and the length of fruitingseasons (as a circular standard deviation). Generalized linear models and circularcorrelations were used to assess whether latitudinal patterns in fruit phenologiescould be explained by variation in temperature, precipitation and actual evapotran-spiration (AET).
Results
Dates of peak fruit production and the length of fruiting seasons showedconsistent differences with latitude. Annual peaks in fruit production occurred 1 to3 months after the summer solstice at high-latitude sites in both hemispheres.Fruiting seasonality increased with latitude, indicating that fruiting seasons werelonger in the tropics and shorter toward the poles. AET was the best climaticpredictor of fruit phenologies. Annual peaks in fruit production were positivelyassociated with annual peaks in AET and temperature, while fruiting seasons wereshorter in areas with pronounced annual variation in AET.
Main conclusions
Global patterns in fruiting seasons are associated with globalvariation in climate. Across the globe, fleshy fruits are produced after annual periodsof elevated water–energy availability. Fruiting seasonality is also more pronouncedin areas with strongly seasonal water–energy inputs. Therefore, the timing of repro-duction in fleshy fruited plant communities appears to be determined, at least inpart, by spatial and temporal variation in energy supplies needed to subsidise plantreproduction.
Keywords
Circular statistics, climate, evapotranspiration, frugivory, latitude, phenology,
seed dispersal.
*Correspondence: K. C. Burns, School of Biological Sciences, Victoria University of Wellington, PO Box 600, Wellington, New Zealand.E-mail: [email protected]
School of Biological Sciences, Victoria University
of Wellington, PO Box 600, Wellington, New
Zealand
INTRODUCTION
Latitudinal gradients in diversity are arguably the oldest and
most controversial patterns in ecology (Hawkins, 2001).
Previous work on latitudinal diversity gradients has typically
focused on taxonomic diversity, or species richness (see
Rosenzweig, 1995; Brown, 1999; Blackburn & Gaston, 2003).
Geographical clines in other measures of ecological diversity
remain poorly understood. For example, the number of species
bearing fruit varies through time in most plant communities and
seasonal changes in fruit availability have been documented in a
large number of plant communities inhabiting a wide range of
geographical locales. However, these data have yet to be synthe-
sized and global patterns in fruit phenologies are unresolved.
An understanding of global patterns in fruit phenologies is
important, given the link between plant phenology and global
climate change. Climate has long been recognized as a cue for
reproduction in fleshy fruited plants (Fisher, 1962). Global
Global patterns in fruit phenologies
© 2008 The Authors
Global Ecology and Biogeography
,
17
, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
649
temperatures have increased significantly over the last century
and the onset of flowering and fruiting in many plant species has
occurred progressively earlier during this time (Bradley
et al
.,
1999; McCarty, 2001; Walther
et al
., 2002; Chapman
et al
., 2005;
Franks
et al
., 2007; Sherry
et al
., 2007). Changes in the timing
and intensity of fruit set can have strong effects on ecosystem
structure and function. For example, in Barro Colorado Island,
droughts associated with El Niño–Southern Oscillation (ENSO)
events synchronize plant reproductive phenologies. Elevated
fruit production during ENSO events consumes plant energy
stores, which in turn generates future shortages in fruit supplies
and large-scale famine for fruit-eating animals in subsequent
years (Wright
et al
., 1999). Documenting global patterns in fruit
phenologies, and identifying the processes responsible for them,
is critical for an accurate understanding of how climate change
might affect ecosystem function.
Here, we test for global patterns in the phenology of fleshy
fruits. We collated studies from the literature that documented
annual cycles in the number of species producing fleshy fruits in
plant communities. These data were then used to test whether
peaks in fruit production and the length of fruiting seasons
vary consistently with latitude. We also evaluated the climatic
correlates of observed phenological patterns. We tested whether
annual peaks in fruit production are correlated with annual
peaks in precipitation, temperature and actual evapotranspiration
(AET, a proxy for productivity). We also tested whether the
length of fruiting seasons was associated with the annual dispersion
of precipitation, temperature and actual evapotranspiration.
METHODS
Fruit phenologies
Data used in the analyses were collated from previously pub-
lished studies (see Appendix S1 in Supplementary Material).
We searched the literature for studies that quantified annual
variation in the phenology of fleshy fruits. Studies were located
using web-based searches in Web of Science and Google Scholar
with the key words ‘fruit’, ‘fruiting’ or ‘phenology’, and from our
prior knowledge of the literature. We only included studies
reporting 12 or more consecutive months of phenological obser-
vations to promote unbiased comparisons between studies.
Those reporting less than one full year of observations, or
observations at intervals longer than 1 month, were excluded.
Monthly data from studies reporting more than 12 months of
observations were averaged between years prior to analyses.
We also limited our attention to studies reporting observations
on five or more fleshy fruited species.
Climatic data
We obtained historical monthly averages of three basic climate
variables: mean temperature (TMP), total precipitation (PPT)
and total AET. We chose these variables for several reasons.
First, temperature and precipitation have been suggested to
influence leaf, flower and fruit phenologies in regional studies
(van Schaik
et al
., 1993; Singh & Kushwaha, 2006). Second, AET
is a mechanistically based amalgamation of temperature, precip-
itation and solar radiation, which serves as a proxy for primary
productivity and the water–energy available for reproduction.
TMP and PPT were obtained from the International Panel for
Climate Change global data set, which contains climatic infor-
mation averaged over the years 1961–90 at a 0.5
°
geographical
resolution (New
et al
., 1999). We used the global data layer of
AET developed by Ahn & Tateishi (1994), downloaded from the
United Nations Environment Programme geodata repository
(http://www.grid.unep.ch/data/data.php, dataset gnv_183), also
presented at a 0.5
°
resolution.
Within-site description of annual fruiting and climatic cycles
For each study that fitted the criteria, we began with a simple
characterization of the within-site annual cycle of fruiting,
utilizing two metrics: (1) a measure of central tendency (mean
angle) which identifies the time of year when the greatest number
of species are expected to be in fruit and (2) a measure of disper-
sion (circular standard deviation) describing the degree to which
a variety of fruit species are available throughout the year. The
rationale behind these metrics, and their calculation, is described
in more detail below. The second stage (described later under Global-
scale Analyses) was to model the global trends and patterns of
these fruiting-cycle metrics with respect to latitude and climate.
The raw data consisted of monthly tallies of the number of
species bearing ripe fruits. According to the Gregorian calendar,
December is at the opposite end of the year from January. In
reality, however, there is no natural start or end point to the time
of year and thus quantities that vary on an annual cycle should be
analysed as ‘circular’ variables, where the time difference between
the eleventh and second month is recognized to be 3 months,
not 9 (Batschelet, 1981). With this in mind, the timing of annual
peaks in fruit production were quantified as mean angles (0–
360
°
, roughly equivalent to day of the year) and the dispersion
of fruiting (i.e. the lengths of the fruiting season) were quantified
as circular standard deviations.
Calculations followed Zar (1984) and proceeded as follows.
Each month
i
= (Jan, Feb, ... Dec) was associated with an angle
from the series
a
i
= (15
°
, 45
°
... 345
°
) and the corresponding
number of species bearing ripe fruits,
f
i
. The mean angle (date) of
fruit production,
µ
f
was calculated as:
(eqn 1)
where
µ
µ
f
f
Y
XX
Y
XX
arctan ( )
arctan ( )
=
>
= +
<
when
or
when
0
180 0
X
f a
n
ii
n
i
cos
= =∑
1
S. Ting
et al.
© 2008 The Authors
650
Global Ecology and Biogeography
,
17
, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
n
is the number of groups (i.e. 12 months) and
α
i
is the angle in
degrees from zero.
The second metric, the temporal dispersion of the fruiting
season at each site, was calculated as the circular standard
deviation of monthly variation in the number of species bearing
ripe fruits (
s
):
(eqn 2)
where
r
c
is the relative length of the mean vector, from 0
→
1, corrected
for grouped data.
The two metrics may be visualized as a ‘mean vector’ (Fig. 1)
with a specified length and direction, and where vector length is
an inverse function of standard deviation (equation 2). In the
same way, mean angles and circular standard deviations were
calculated for each of the environmental variables on a site-
by-site basis. Monthly values for temperature were translated
into kelvin to avoid complications caused by negative values.
Global-scale analyses
Length of the fruiting season
First we examined the global relationship between latitude and
the length of fruiting season (measured as a circular standard
deviation), using a three-parameter Gaussian curve fit in Sigma-
Plot (2002). Next, a global-scale climatic prediction of the length
of fruiting seasons was determined using generalized least
squares (GLS) regression (Bailey & Gatrell, 1995; Pinheiro &
Bates, 2000). Sites were our sample unit, and the annual disper-
sion of fruiting measured as a circular standard deviation (
s
fru
)
was our response variable. The dispersion of TMP, PPT and AET,
also as circular standard deviations, formed our set of candidate
predictor variables. The circular standard deviations of TMP and
AET were ln-transformed to improve their normality and
correlations between variables were examined with Pearson’s
correlation test. Circular standard deviations vary from zero to
infinity and therefore can be used as continuous variables in
standard linear models.
GLS models are similar to general linear models, except that
they produce more realistic estimates of standard errors and
type-I errors in the presence of spatially correlated residuals
(Ver Hoef
et al
., 2001; Dormann
et al
., 2007). GLS models also
prevent clusters of sites from exerting undue (pseudo-replicated)
influence on estimates of beta coefficients, which may be an
important consideration if study sites are not uniformly distributed
in space (cf. Hawkins
et al
., 2007).
Y
f a
n
ii
n
i
sin
.= =∑
1
Figure 1 Graphical illustration of how estimates of central tendency (angle of the mean vector) and dispersion (circular standard deviation or vector length) relate to the monthly values of actual evapotranspiration (AET) (left) and the number of species bearing ripe fruits (right). Top, site 4 (Canada); bottom, site 44 (Uganda). In the text we refer to the mean angle as the ‘peak date of fruit production’ and the circular standard deviation as the ‘length of the fruiting season’.
s rc ln= −180
2π
rn
nX Yc
/
sin( / ) .= +
ππ
2 2
Global patterns in fruit phenologies
© 2008 The Authors
Global Ecology and Biogeography
,
17
, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
651
To apply a GLS, the spatial correlation of residuals should be
suitably described by a function fitted to an empirical variogram
(Bailey & Gatrell, 1995). We compared the fit of three common
variogram models: the ‘spherical’, ‘Gaussian’ and ‘exponential’
functions as part of a full GLS model containing all three envi-
ronmental predictors and selected the variogram function that
yielded the lowest Akaike information criterion (AIC) (Burnham
& Anderson, 2002). This function was used in all subsequent
model comparisons involving all subsets of one and two environ-
mental predictors. Because of problems of collinearity between
environmental predictors we felt it informative to examine the
predictive capability of each variable on its own, as well as in
combination with one or more additional variables. For compar-
ison, we also fitted versions of the model specifications without
the ‘correction’ for spatial autocorrelation (Hawkins
et al
., 2007).
Finally, we nominated the model with the lowest overall AIC as
the minimum adequate model.
GLS models were fit using the ‘gls’ command of the ‘nlme’
package in R (
R
Development Core Team, 2007). This function
only accepts Cartesian coordinates, therefore spatial separations
were calculated as the linear or chord distance between sites
located on a spherical model of the globe. Chord distances are a
monotonic function of great circle distances, so the main differ-
ence between the two distance descriptors is their effect on the
range parameter of the variogram model.
Timing of peak fruit production
Geographical patterns in annual peaks of fruit production were
assessed with polynomial regression. More specifically, we
evaluated the fit of a sine curve to the relationship between lati-
tude (a bounded ‘linear’ variable) and the mean angle of monthly
fruit production (a ‘circular’ variable) using SigmaPlot (2002).
To determine the best climatic correlate for the timing of
peak fruit production, we used circular–circular correlations
(Batschelet, 1981). Methods for handling more than one circular
‘predictor’ at a time are poorly developed; hence separate corre-
lations were conducted between the mean angle of fruit produc-
tion and mean angles of TMP, PPT and AET. Several alternative
correlation tests have been proposed for circular variables. We
chose the correlation test of Jupp & Mardia (1980, 1981) as it
does not require a uniform distribution of angles in the input
variables.
To obtain a form of residuals appropriate to circular data, we
first calculated the angular difference or ‘lag’ between the time of
peak fruiting and the time at which each of the environmental
predictors peaks, on a site by site basis: lag =
µ
fru
–
µ
env
.
pred
.
The lags were expressed in the range –180
→
+180
°
and the
difference between the lag at site
j
and the overall mean lag
was taken as the residual (residual
j
= mean lag – lag
j
). To our
knowledge, no methods have been developed to account for
spatial autocorrelation in circular–circular correlations.
Nonetheless, we were able to examine the spatial structure of the
‘residuals’ by plotting the
absolute
angular difference (0–180
°
)
between the two residuals of all pairwise combinations of sites
against their separation distance (in this case calculated as a great
circle distance): low angular differences indicating a pair of sites
with similar (and potentially correlated) residuals in their
relationship between time of peak fruiting and the environmental
predictor. The presence of positive spatial autocorrelation in the
residuals was assessed visually by fitting locally weighted scatter-
plot smoothing (LOWESS) trend lines to the plot (with short-
range and medium-range smoother spans) and examining the
lines for a strong upward trend. Any attempt to calculate Moran’s
I
or a GLS model from these data would be misleading because
two residuals of –150
°
and +160
°
would be treated as being 310
°
apart, instead of only 50
°
. As a final test for spatial structure, we
compared the distribution of lags between different tropical and
temperate zones using Watson’s
U
2
test (Batschelet, 1981). Data
and R programming code for the above analyses are available
in Appendices S2 and S3.
RESULTS
Forty-eight studies met the inclusion criteria (see Appendix S1).
Figure 2 graphically illustrates several important components
of our data set, which covered a large geographical extent
(17 counties) and a wide range of climatic conditions.
Arrows represent (a) fruit phenologies, (b) AET and (c) the time
lag between the two. Long arrows in Fig. 2(a) represent the
concentration of fruiting into a short and well-defined period
of time, thus arrow length (
r
) is inversely related to the length
of the fruiting season (circular standard deviation,
s
; equation 2).
Arrow directionality (angle) illustrates the date of peak fruit
production.
Length of fruiting season
Latitudinal changes in the circular standard deviations of annual
fruiting cycles followed a Gaussian distribution (
R
2
= 0.491,
P
< 0.001). Fruiting seasons are shorter at high latitudes and
become progressively longer towards the equator (Fig. 3).
The length of fruiting seasons was also correlated with climatic
variables. When tested separately as single variable models, all
predictors were significantly associated with the circular standard
deviation of fruit production. The best single climatic predictor
was AET (
t
= 6.748,
P
< 0.001,
n
= 48), followed by temperature
(
t
= 6.523,
P
< 0.001,
n
= 48) and then precipitation (
t
= 3.100,
P
= 0.002,
n
= 48). AET and TMP were both strongly correlated
with absolute latitude, and with each other (all
r
> 0.87, all
P
<< 0.001,
n
= 48), while precipitation was not significantly
correlated with latitude or either of the other two environmental
predictors. The minimum adequate model retained terms for
AET and temperature, although only the term for AET was
significant (Table 1). The GLS models described above included
an exponential function to describe the spatial autocorrelation of
residuals (Appendix S4). The most parsimonious aspatial model
explained 55% of the variation in the length of fruiting seasons
using the two variables AET and TMP; it gave similar results
to the GLS though with lower coefficients, a slightly higher AIC
and reduced significance for the environmental predictors
(Appendix S4).
S. Ting
et al.
© 2008 The Authors
652
Global Ecology and Biogeography
,
17
, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
Figure 2 Map illustrating the location of study sites and a summary of (a) each site’s community-level fruiting phenology, (b) seasonality of actual evapotranspiration (AET) and (c) the lag between the two (mean angle of fruiting–mean angle of AET). Numbers correspond to study sites listed in Appendix S1. Vectors (arrows) point to annual peaks in fruit production (or AET). Vector length (rc) is inversely related to the length of fruiting seasons (or availability of AET) and inversely related to circular standard deviation (s). The map projection is Plate Carée, lines of latitude indicate the Tropic of Cancer and Tropic of Capricorn.
Global patterns in fruit phenologies
© 2008 The Authors
Global Ecology and Biogeography
,
17
, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
653
Timing of peak fruit production
Latitudinal changes in the mean angle of monthly fruit produc-
tion followed a sine curve (
R
2
= 0.410,
P
< 0.001). Annual peaks
in fruit production occurred approximately 1–3 months after the
summer solstice at high-latitude sites; between September and
October in the Northern Hemisphere and between March and
April in the Southern Hemisphere (Fig. 4). Annual peaks in fruit
production were also correlated with climatic variables. At a
global scale, the timing of peak fruiting showed similar strength
correlations with each of the environmental predictors (PPT,
TMP and AET) (Table 2a). As expected, the correlations were
weak (and non-significant) in tropical regions, but were stronger
in the temperate zones of both hemispheres (Table 2b–d). In
temperate zones, peak fruiting occurred 1–2 months after peaks
in temperature and 2–4 months after peaks in AET. Correlations
between fruiting and precipitation were not as strong and the
relationship between fruiting and precipitation was particularly
weak in the Northern Hemisphere (Table 2b,c). The timing of the
lags did not differ significantly between the three geographically
defined subregions (Watson
U
2
< 0.10,
P
> 0.10). However,
fruiting appeared to peak only 6 weeks after the summer solstice
in the Southern Hemisphere (
c
. 1 February), whereas in the
Northern Hemisphere the peak was approximately 10 weeks
after the summer solstice (
c
. 10 September) (Watson
U
2
= 0.154,
0.05 <
P <
0.10). We note that at a global scale, the seasonal
timing of peaks in AET, TMP and PPT were all highly correlated
with each other, particularly AET and temperature (PPT:TMP
r
2
= 0.333,
P
= 0.003; PPT:AET
r
2
= 0.488,
P
= 0.0001; TMP:AET
r
2
= 0.871,
P
<< 0.0001, all
n
= 48; circular correlation test of
Jupp & Mardia, 1980, 1981). Some evidence for spatial auto-
correlation was found in the angular residuals of the fruiting:
AET correlation, among temperate sites separated by less than
5000 km (Fig. S3 in Appendix S4), suggesting that the signifi-
cance values of the associated correlation test may need to be
viewed with caution.
DISCUSSION
The phenology of fleshy fruits shows consistent geographical
variation across the globe. Equatorial regions have prolonged
fruiting seasons and peak dates of fruit production that occur
throughout the year. Temperate regions have short fruiting
seasons that generally occur in late summer or autumn, 1–3
months after the summer solstice. As a result, annual variation in
fruit phenologies is more uniform in tropical regions, whereas
temperate regions are characterized by pronounced annual peaks
in production followed by periods of fruit scarcity.
Figure 3 Latitudinal variation in the length of fruiting seasons. The circular standard deviation of monthly variation in fruit production is shown on the x-axis. Degree latitude (Southern Hemisphere sites are negative) is shown on the y-axis. A Gaussian curve is fitted to the corresponding relationship (R2 = 0.491, P < 0.001, n = 48).
Table 1 Minimum adequate model (using spatial generalized least squares) for explaining the length of fruiting seasons (i.e. circular standard deviations in monthly distributions of the number of species bearing ripe fruits). Competing models are presented in Appendix S4.
Coefficient Std. Error t P-value
Intercept –593.27 264.29 –2.245 0.030
log(TMP) 94.45 63.98 1.476 0.147
log(AET) 43.43 19.11 2.273 0.028
Akaike information criterion (AIC) = 406.7, estimated using restricted
maximum likelihood (REML), total d.f. = 48, residual d.f. = 44. Spatial
correlation of residuals described using an exponential model with
range = 359.8 km and nugget = 0.00088. TMP, mean temperature;
AET, actual evapotranspiration.
Figure 4 Latitudinal variation in annual peaks in fruit production. The mean angle of monthly variation in fruit production is shown on the x-axis latitude. Months are used as x-axis labels instead of angles (i.e. Jan = 15°, Feb = 45°, ... Dec = 345°). Degree latitude (Southern Hemisphere sites are negative) is shown on the y-axis. A sine curve is fitted to the corresponding relationship (R2 = 0.410, P < 0.001, n = 48).
S. Ting et al.
© 2008 The Authors654 Global Ecology and Biogeography, 17, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
Variable patterns in dates of peak fruit production in the
tropics are quantitatively linked to the lack of fruiting seasonality
in the tropics. Estimates of a central tendency in circular distri-
butions are more difficult to establish when the underlying data
are uniformly distributed (i.e. platykurtic; see Batschelet, 1981).
Therefore, greater variability of peak dates of fruit production in
the tropics is a reflection of more uniform monthly distributions
in the number of species bearing ripe fruits. Conversely, latitudinal
patterns in peak dates of fruit production in temperate regions
are facilitated by strong seasonality in fruit production. Although
we analysed peak dates of fruit production and the length of
fruiting seasons separately, the two are best interpreted jointly,
given the quantitative connection between them.
Geographical patterns in fruit phenologies were associated
with geographical variation in climate. In general, AET was a better
predictor than temperature, which was a better predictor than
precipitation for both the timing and length of fruiting seasons.
This indicates that geographical locales with longer fruiting
seasons had more uniform monthly AET distributions, while
sites with less uniform AET distributions usually peaked in fruit
production 2–4 months after annual peaks in water–energy
inputs.
Although the mechanism underpinning the time lag between
annual peaks in AET and annual peaks in fruit production is
unclear, it could result from time lags in the process of plant
reproduction. Plant reproduction is a sequential process. Flower
development, pollen dispersal and fruit maturation do not occur
simultaneously, and each requires a substantial period of time to
take place. Time lags in the development of fleshy fruits resulting
from the earlier processes of plant reproduction can have
substantial influence on broad-scale patterns in plant–animal
interactions (Burns, 2002). If the process of plant reproduction is
initiated by annual increases in water–energy inputs, a time lag
between peak AET and fruit maturation would be generated
by the developmental time required for flower and fruit
development. van Schaik et al.’s (1993) global analysis of flower
phenologies supports this interpretation. They found that peak
dates in flower production occurred concomitantly with annual
periods of increased solar radiation.
Previous studies on smaller spatial scales have linked
community-level phenological events to precipitation (Singh &
Kushwaha, 2005; Boulter et al., 2006) and/or temperature
(Williams-Linera, 1997; Berlin et al., 2000). The discrepancy
between these results and those reported here could be explained
by several factors. First, most proxies of productivity (including
AET) are derived from both temperature and precipitation
measurements; hot, wet locales are more productive than cooler,
drier locales. Previous studies on the relationship between plant
reproductive phenologies and climate have often focused on
tropical regions that experience annual periods of drought. In
these locales, the strongest limitation on productivity is likely to
be water stress. Therefore, any relationship between phenologies
and precipitation in tropical dry forests is likely to reflect a
similar relationship with productivity. A second explanation
might be that the processes occurring on smaller, regional scales
may be different from the processes operating at the global scale
(see Blackburn & Gaston, 2002; Aukema, 2004; Burns, 2004;
García & Ortiz-Pulido, 2004).
Several aspects of our study warrant a cautious approach
to interpreting the results. Spatial autocorrelation was not
Table 2 Correlation between the peak date of fruit production (mean lag) and the peak concentration of three environmental variables. P-value determined from nr2 distributed as χ2 with 4 d.f.; r 2 may vary from 0 to 2 (Jupp & Mardia, 1980, 1981).
Environmental predictor r 2 P Timing of fruit production (mean lags)
(a) Global data set, n = 48
Precipitation 0.298** 0.006 4 months before peak PPT
Temperature 0.300** 0.006 2 months after peak TMP
AET 0.330** 0.003 3.3 months after peak AET
(b) Northern temperate, n = 12
Precipitation 0.099 NS 5 months before peak PPT
Temperature 0.639 0.104 2 months after peak TMP
AET 0.757+ 0.059 4 months after peak AET
Calendar date 10 September: 2.6 months after summer solstice
(c) Southern temperate, n = 15
Precipitation 0.772* 0.021 3.3 months before peak PPT
Temperature 0.949** 0.007 1.3 months after peak TMP
AET 0.917** 0.008 2 months after peak AET
Calendar date 1 February: 1.3 months after summer solstice
(d) Tropical, n = 21
Precipitation 0.264 NS 4.7 months before peak PPT
Temperature 0.144 NS 1.5 months after peak TMP
AET 0.051 NS 3.3 months after peak AET
‘NS’ = P > 0.20, +P < 0.1, *P < 0.05, **P < 0.01.AET, actual evapotranspiration; PPT, total precipitation; TMP, mean temperature.
Global patterns in fruit phenologies
© 2008 The Authors Global Ecology and Biogeography, 17, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd 655
accounted for in analyses of circular variables. There was also a
difference in the scale at which phenological and climatic data
were gathered. Phenological data were collected at particular
points in space that were usually less than several square kilometres
in area, whereas climatic data were estimated for 0.5° grid cells.
We only assessed relationships between phenological parameters,
latitude and climate. However, many other variables, including
soil conditions, elevation and anthropogenic disturbance, may
also be important. The potential effects of plant species richness
were also not assessed. Global patterns in plant species richness
are correlated with the same climatic variables that were used in
our analyses, complicating tests for causal relationships between
variables. Identifying causal relationships between plant species
richness and fruit phenologies represents a major challenge to
future work.
Although our analyses suggest climate is a major driver of
global patterns in fruit phenologies, biotic interactions could
also be important. Animal fruit consumers are a critical compo-
nent of successful reproduction in fleshy fruited plant species
(Herrera, 2002). Therefore, the actions of seed dispersers could
profoundly influence the timing of fruit production, both
ecologically (habitat shaping) or evolutionarily (natural
selection; see Burns, 2004). Tropical regions typically house large
assemblages of fruit-eating animals, many of which are non-
migratory. Protracted fruiting seasons in the tropics might
therefore facilitate seed dispersal by sedentary frugivores.
Conversely, many temperate regions receive massive annual
influxes of migratory fruit-eating birds (Herrera, 1998; Teller’a &
Pérez-Tris, 2003, 2007). Shorter fruiting seasons, which coincide
with autumn migrations of avian seed dispersers that winter
in the tropics, may facilitate the spread of seeds over larger
distances (Thompson & Willson, 1979). Our analyses did not
address this possibility. Identifying the relative influence of
biotic and abiotic processes affecting global patterns in the
phenology of fleshy fruits presents another interesting challenge
to future research.
In conclusion, our results showed consistent geographical
patterns in the phenology of fleshy fruits in plant communities.
High-latitude sites have strong annual peaks in fruit production,
typically in late summer or autumn, in both hemispheres. Tropical
sites have less pronounced annual peaks in fruit production that
occur throughout the year. Fruit phenologies were more strongly
associated with AET than other climatic variables, suggesting
that the timing of reproduction in most plant communities is
initiated by the combination of elevated periods of solar energy
and water availability. A better understanding of global patterns
in plant reproduction hinges on future research into how
pollinators and seed dispersers might interact with climate to
determine global phenological cycles.
ACKNOWLEDGEMENTS
We would like to thank Daniel Burket, Angela Moles, Shirley
Pledger, Hannah Sutton, David Ting, Sue Ting and Gemma
Bowker-Wright for helpful advice and Victoria University of
Wellington for funding.
REFERENCES
Ahn, C.H. & Tateishi, R. (1994) Development of global land
surface evapotranspiration and water balance data sets. Journal
of the Japan Society Photogrammetry and Remote Sensing, 33,
48–61.
Aukema, J.E. (2004) Distribution and dispersal of desert mistletoe
is scale-dependent, hierarchically nested. Ecography, 27, 137–144.
Bailey, T.C. & Gatrell, A.C. (1995) Interactive spatial data analysis.
Longman Scientific and Technical, Harlow, UK.
Batschelet, E. (1981) Circular statistics in biology. Academic Press,
New York.
Berlin, K.E., Pratt, T.K., Simon, J.C., Kowalsky, J.R. & Hatfield,
J.S. (2000) Plant phenology in a cloud forest on the island of
Maui, Hawaii. Biotropica, 32, 90–99.
Blackburn, T.M. & Gaston, K.J. (2002) Scale in macroecology.
Global Ecology and Biogeography, 11, 185–189.
Blackburn, T.M. & Gaston, K.J. (eds) (2003) Macroecology:
concepts and consequences. Blackwell Publishing, Oxford, UK.
Boulter, S.L., Kitching, R.L. & Howlett, B.G. (2006) Family,
visitors and the weather: patterns of flowering in tropical
rain forests of northern Australia. Journal of Ecology, 94, 369–
382.
Bradley, N.L., Leopold, A.C., Ross, J. & Huffaker, W. (1999)
Phenological changes reflect climate change in Wisconsin.
Proceedings of the National Academy of Sciences USA, 96, 9701–
9704.
Brown, J.H. (1999) Macroecology: progress and prospects.
Oikos, 87, 3–14.
Burnham, K.P. & Anderson, D.R. (2002) Model selection and
multimodel inference: a practical information-theoretic approach.
Springer-Verlag, New York, USA.
Burns, K.C. (2002) Seed dispersal facilitation and geographic
consistency in bird–fruit abundance patterns. Global Ecology
and Biogeography, 11, 253–259.
Burns, K.C. (2004) Scale and macroecological patterns in seed
dispersal mutualisms. Global Ecology and Biogeography, 13,
289–293.
Chapman, C.A., Chapman, L.J., Struhsaker, T.T., Zanne, A.E.,
Clark, C.J. & Poulsen, J.R. (2005) A long-term evaluation of
fruiting phenology: importance of climate change. Journal of
Tropical Ecology, 21, 31–45.
Dormann, C.F., McPherson, J.M., Araújo, M.B., Bivand, R.,
Bolliger, J., Carl, G., Davies, R.G., Hirzel, A., Jetz, W., Kissling,
W.D., Kühn, I., Ohlemüller, R., Peres-Neto, P.R., Reineking, B.,
Schröder, B., Schurr, F.M. & Wilson, R. (2007) Methods to
account for spatial autocorrelation in the analysis of species
distributional data: a review. Ecography, 30, 609–628.
Fisher, D.V. (1962) Heat units and number of days required to
mature some pome and stone fruits in various areas of North
America. Proceedings of the American Society for Horticultural
Science, 80, 114–124.
Franks, S.J., Sim, S. & Weis, A.E. (2007) Rapid evolution of
flowering time by an annual plant in response to a climate
fluctuation. Proceedings of the National Academy of Sciences
USA, 104, 1278–1282.
S. Ting et al.
© 2008 The Authors656 Global Ecology and Biogeography, 17, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd
García, D. & Ortiz-Pulido, R. (2004) Patterns of resource
tracking by avian frugivores at multiple spatial scales: two case
studies on discordance among scales. Ecography, 27, 187–196.
Hawkins, B.A. (2001) Ecology’s oldest pattern? Trends in Ecology
and Evolution, 16, 470.
Hawkins, B.A., Diniz-Filho, J.A.F., Bini, L.M., De Marco, P. &
Blackburn, T.M. (2007) Red herrings revisited: spatial auto-
correlation and parameter estimation in geographical ecology.
Ecography, 30, 375–384.
Herrera, C.M. (1998) Long-term dynamics of Mediterranean
frugivorous birds and fleshy-fruits: a 12-year study. Ecological
Monographs, 68, 511–538.
Herrera, C.M. (2002) Seed dispersal by vertebrates. Plant–animal
interactions: an evolutionary approach (ed. by C.M. Herrera
and O. Pellmyr), pp. 185–208. Blackwell Science, Oxford.
Jupp, P.E. & Mardia, K.V. (1980) A general correlation coefficient
for directional data and related regression problems. Biometrika,
67, 163–173.
Jupp, P.E. & Mardia, K.V. (1981) Amendments and corrections: a
general correlation coefficient for directional data and related
regression problems. Biometrika, 68, 738.
McCarty, J.P. (2001) Ecological consequences of recent climate
change. Conservation Biology, 15, 320–331.
New, M., Hulme, M. & Jones, P. (1999) Representing twentieth-
century space–time climate variability. Part I: development of
a 1961–90 mean monthly terrestrial climatology. Journal of
Climate, 12, 829–856.
Pinheiro, J.C. & Bates, D.M. (2000) Mixed effects models in S and
S-Plus. Springer, New York.
Rosenzweig, M.L. (1995) Species diversity in space and time.
Cambridge University Press, Cambridge, UK.
R Development Core Team (2007). R: a language and environ-
ment for statistical computing. R Foundation for Statistical
Computing, Vienna, Austria. http://www.R-project.org.
van Schaik, C.P., Terborgh, J.W. & Wright, S.J. (1993) The
phenology of tropical forests: adaptive significance and
consequences for primary consumers. Annual Review of
Ecology and Systematics, 24, 353–377.
Sherry, R.A., Zhou, X., Gu, S., Arnone, J.A., Schimel, D.S.,
Verburg, P.S., Wallace, L.L. & Luo, Y. (2007) Divergence of
reproductive phenology under climate warming. Proceedings
of the National Academy of Sciences USA, 104, 198–202.
SigmaPlot (2002) Release 8.02 for Windows. SPSS Inc., Chicago,
USA.
Singh, K.P. & Kushwaha, C.P. (2006) Diversity of flowering and
fruiting phenology of trees in a tropical deciduous forest in
India. Annals of Botany, 97, 265–276.
Tellería, J.L. & Pérez-Tris, J. (2003) Seasonal distribution of a
migratory bird: effects of local and regional resource tracking.
Journal of Biogeography, 30, 1583–1591.
Tellería, J.L. & Pérez-Tris, J. (2007) Habitat effects on resource
tracking ability: do wintering blackcaps Sylvia atricapilla track
fruit availability? Ibis, 149, 18–25.
Thompson, J.N. & Willson, M.F. (1979) Evolution of temperate
fruit/bird interactions: phenological strategies. Evolution, 33,
973–982.
Ver Hoef, J.M., Cressie, N., Fisher, R.N. & Case, T.J. (2001)
Uncertainty and spatial linear models for ecological data.
Spatial uncertainty in ecology: implications for remote sensing
and GIS applications (ed. by C.T. Hunsaker, M.F. Goodchild,
M.A. Friedl and T.J. Case), pp. 214–237. Springer-Verlag, New
York.
Walther, G.R., Post, E., Convey, P., Menzel, A., Parmesank, C.,
Beebee, T.J.C., Fromentin, J.M., Hoegh-Guldberg, O. &
Bairlein, F. (2002) Ecological responses to recent climate
change. Nature, 416, 389–395.
Williams-Linera, G. (1997) Phenology of deciduous and
broadleaved-evergreen tree species in a Mexican tropical lower
montane forest. Global Ecology and Biogeography Letters, 6,
115–127.
Wright, S.J., Carrasco, C., Calderón, O. & Paton, S. (1999) The El
Niño Southern Oscillation, variable fruit production, and
famine in a tropical forest. Ecology, 80, 1632–1647.
Zar, J.H. (1984) Biostatistical analysis. Prentice Hall, Englewood
Cliffs, NJ.
SUPPLEMENTARY MATERIAL
The following supplementary material is available for this article:
Appendix S1 Published studies used in analyses, their biblio-
graphic references and our circular description of phenologies.
Appendix S2 Monthly fruiting data and climatic conditions.
Appendix S3 Programming code for the statistical analyses con-
ducted in R.
Appendix S4 Semivariogram analysis of model residuals, and
non-spatial equivalents to the GLS models.
This material is available as part of the online article from:
http://www.blackwell-synergy.com/doi/abs/10.1111/
j.1466-8238.2008.00408.x
(This link will take you to the article abstract).
Please note: Blackwell Publishing is not responsible for the con-
tent or functionality of any supplementary materials supplied by
the authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.
Global patterns in fruit phenologies
© 2008 The Authors Global Ecology and Biogeography, 17, 648–657, Journal compilation © 2008 Blackwell Publishing Ltd 657
Editor: José Alexandre F. Diniz-Filho
BIOSKETCHES
Steven Ting is an honours student at Victoria University of Wellington who is interested in macroecological patterns in plant–
animal interactions.
Stephen Hartley is an assistant professor at Victoria University of Wellington. He is interested in population and community
ecology, conservation biology, invasive species and bioclimatic modelling, with a focus on spatial analysis. Current research topics
include the spread of Argentine ants, patterns of rarity in plants, insect forging behaviour and coastal sand dune communities.
K. C. Burns is an assistant professor at Victoria University of Wellington. His research interests include bird behaviour,
community assembly rules, plant–animal interactions and plant morphology.