El Nin?o and dry season rainfall influence hostplant
phenology and an annual butterfly migration from
Neotropical wet to dry forests
R O B E R T B . S R Y G L E Y *, R O B E R T D U D L E Y *w , E VA N D R O G . O L I V E I R A z, R A F A E L
A I Z P R U? A ? 1 , N I C O L E Z . P E L A E Z } 2 and A N D R E J . R I V E R O S k3
*Smithsonian Tropical Research Institute, Apdo. 0843-03092, Balboa, Republic of Panama, wDepartment of Integrative Biology,
University of California, Berkeley, CA 94720, USA, zFaculdade de Cie?ncias Biolo?gicas e Sau?de, Centro Universita?rio UNA ?
Campus Guajajaras, Rua Guajajaras 175, 30180-100 ? Belo Horizonte, MG, Brazil, ?Departamento de Bota?nica, Universidad de
Panama?, Repu?blica de Panama?, }Departamento de Ciencias Biolo?gicas, Universidad de Los Andes, Carrera 1A No. 18A-10, Bogota?,
Colombia, kDepartamento de Biolog??a, Universidad Nacional de Colo?mbia, Apdo. 14490, Bogota?, Colombia
Abstract
We censused butterflies flying across the Panama Canal at Barro Colorado Island (BCI)
for 16 years and butterfly hostplants for 8 years to address the question: What environ-
mental factors influence the timing and magnitude of migrating Aphrissa statira
butterflies? The peak migration date was earlier when the wet season began earlier
and when soil moisture content in the dry season preceding the migration was higher.
The peak migration date was also positively associated with peak leaf flushing of one
hostplant (Callichlamys latifolia) but not another (Xylophragma seemannianum). The
quantity of migrants was correlated with the El Nin?o Southern Oscillation, which
influenced April soil moisture on BCI and total rainfall in the dry season. Both hostplant
species responded to El Nin?o with greater leaf flushing, and the number of adults
deriving from or laying eggs on those new leaves was greatest during El Nin?o years. The
year 1993 was exceptional in that the number of butterflies migrating was lower than
predicted by the El Nin?o event, yet the dry season was unusually wet for an El Nin?o year
as well. Thus, dry season rainfall appears to be a primary driver of larval food production
and population outbreaks for A. statira. Understanding how global climate cycles and
local weather influence tropical insect migrations improves the predictability of ecolo-
gical effects of climate change.
Keywords: climate change, El Nin?o, ENSO, insect flight, insect migration, migratory behavior, resource
limitation, tropical rainforest
Received 12 March 2009 and accepted 24 April 2009
Introduction
Butterfly flight activity has provided important evidence
of global climate change (Parmesan, 2006). Climate is a
primary factor driving the advance of the first flight of
butterflies in central California (Forister & Shapiro, 2003)
and butterfly migrations to the British Isles (Sparks et al.,
2005). Climate change is likely to impact tropical biota, as
well; and yet, we know practically nothing about the
environmental factors driving the long-distance migra-
tion of tropical butterflies. This in part is due to the lack of
long-term data sets on migrating species in tropical
climates (e.g., Haber & Stevenson, 2004).
The roles of biotic and abiotic factors in regulating
insect populations are often nonadditively complex.
El Nin?o may be an exception. The global climate cycle
characterized as the El Nin?o Southern Oscillation
(ENSO) influences global patterns of primary productiv-
ity (Behrenfeld et al., 2001; Stenseth et al., 2002; Holmgren
et al., 2006; McPhaden et al., 2006). For example, a warm
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Correspondence: Present address: Robert B. Srygley, USDA-ARS-
NPARL, 1500 N. Central Ave., Sidney, MT 59270, USA, fax 1 1 406
433 5038, e-mail: robert.srygley@ars.usda.gov
1Present address: Flora Tropical, S.A., Apdo. 0819-07769, Repu?blica
de Panama?.
2Present address: Calle 84 # 7 ? 25 Int. 102, Bogota?, Colombia.
3Present address: Center for Insect Science and Division of Neu-
robiology, University of Arizona, Tucson, AZ 85721, USA.
Global Change Biology (2009), doi: 10.1111/j.1365-2486.2009.01986.x
r 2009 Blackwell Publishing Ltd 1
G C B 1 9 8 6 B Dispatch: 5.6.09 Journal: GCB CE:Journal Name Manuscript No. Author Received: No. of pages: 10 PE: xx
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
tropical Pacific Ocean surface is associated with
increased rainfall in deserts and tropical seasonally dry
forests worldwide, resulting in seed germination and
plant growth. The effects of increased primary produc-
tivity cascade upward into higher trophic levels resulting
in periodic outbreaks of herbivorous species (Jaksic,
2001; Holmgren et al., 2006). A number of researchers
have hypothesized a link between ENSO, primary pro-
ductivity, and insect migrations in deserts and tropical
seasonally dry forests (see Vandenbosch, 2003). Here, we
designate this link between ENSO and insect migrations:
the El Nin?o Migration Hypothesis.
In many tropical wet forests, reduced cloud cover and
rainfall and greater solar radiation and temperature are
associated with El Nin?o years (Kiladis & Diaz, 1989).
For example on Barro Colorado Island (BCI) in central
Panama, Wright & Caldero?n (2006) found an increase in
solar radiation, higher temperatures and a reduction in
precipitation during El Nin?o events, and the opposite
effects during La Nin?a events. Plants appear to respond
favorably to increased sunlight and higher tempera-
tures in El Nin?o years. Although limited in scope,
annual tree ring data from central Panama indicate that
El Nin?o events may enhance tree growth (see discussion
in Wright & Caldero?n, 2006). Enhanced plant produc-
tivity with El Nin?o is also observed in measures of
reproduction: flower and seed production, although
extreme El Nin?o events may be detrimental to repro-
duction (Wright & Caldero?n, 2006). Given this response
in productivity, the El Nin?o Migration Hypothesis may
also apply to tropical wet forests.
For 16 years near BCI, we have censused migratory
butterflies flying over the Panama Canal at a point in a
migratory flyway from the Atlantic (Caribbean) coast to
the Pacific coast of the Republic of Panama (Srygley
et al., 1996). The butterfly migrations across the isthmus
differ from those frequently studied in which insects
move away from dry or cold environments (e.g., Larsen,
1976; Brower, 1995; Haber & Stevenson, 2004). Instead,
these Neotropical migrations occur near the beginning
of the wet season. Seasonal insect migrations are often
resource based with spatial and temporal distribution
of food supply being the main selective force governing
the migratory phenology, rate and direction (e.g., Rain-
ey, 1951; Southwood, 1977; Van Schaik et al., 1993; Lox-
dale & Lushai, 1999; Dingle & Drake, 2007). Tropical
butterfly migrations may follow a gradient of environ-
mental productivity, which in turn is governed by the
spatio-temporal distribution of rainfall. In this paper,
we address the question: What environmental factors
determine the timing and magnitude of migrants cross-
ing the Panama Canal? We evaluate the effect of the
ENSO climate cycle on hostplant phenology and a
butterfly migration in a Neotropical wet forest.
Materials and methods
Study organisms and study site
Annually from mid-May to mid-July, Aphrissa statira
Cramer (Pieridae) migrates directionally across the
Panama Canal near BCI (91100N, 791510W), a nature reserve
and biological field station administered by the Smith-
sonian Tropical Research Institute (STRI). The migra-
tory flyway extends from the Atlantic coastal wet forest
to the Pacific coastal dry forest, and the predominant
migratory direction at this time of year is south?south-
west (Srygley et al., 1996; Oliveira et al., 1998). A. statira
maintain their migratory direction with a time-compen-
sated sun compass and possess sophisticated mechan-
isms to maintain their course and reduce the energetic
cost while migrating (reviewed in Srygley & Dudley,
2008). As adults, A. statira feed on flower nectar, and
female A. statira lay eggs on lianas in the family Bigno-
niaceae, including Callichlamys latifolia (Rich.) K. Schum.
and Xylophragma seemannianum (Kuntze) Sandwith. The
latter is a new hostplant record (R. B. Srygley & R.
Aizpru?a, unpublished data). Both hostplant species
range from Mexico to Brazil (Flora of the Venezuelan
Guayana online at http://www.mobot.org).
The isthmus of Panama is characterized by two coastal
lowlands separated by a central mountain range stretch-
ing east?west along the length of the country. This range
falls to near sea level (ca. 200 m) where the migratory
flyway occurs in the region of the Panama Canal. Annual
rainfall declines from the Atlantic to Pacific coasts. On
average, Fort Sherman on the Atlantic coast receives
3020 mm of rain per year, BCI receives 2620 mm, Gamboa
in central Panama receives 2230 mm, and Panama City on
the Pacific coast receives 1850 mm. Across the isthmus, a
distinct dry season, during which rainfall is only one-tenth
the annual precipitation (300 mm on the Atlantic and
Gamboa, 285 mm on BCI, and 140 mm in Panama City),
occurs from mid-December through the end of April. The
rainfall gradient influences the proportion of evergreen vs.
deciduous trees in the forests across the isthmus.
Our working hypothesis is that the southwesterly
directional migration is a result of loss of larval re-
sources in the evergreen Atlantic coastal forest due to
overcast skies in the wet season (i.e., light limitation
hypothesis, Van Schaik et al., 1993). The deciduous
forest of the Pacific coast provides a suitable destination
because new growth begins with the onset of wet
season and continues to provide a predictable larval
resource due to the greater availability of sunlight
(Srygley & Oliveira, 2001). Interestingly, Aphrissa
butterflies migrate northeasterly in September?October
toward the evergreen forest before the end of the wet
season in December. This behavior may have evolved
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
2 R . B . S R Y G L E Y et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
due to a seasonal reduction in new leaf production in
the deciduous forest as dry season approaches and leaf-
flushing events in the wet forest during short dry
periods in September?October. Furthermore, the wet
forest provides a humid refuge to maintain the popula-
tion during the dry season (Srygley, 2001). Because of
logistic difficulties, we lack hostplant data for A. statira
at the origin and destination coastal sites, but we
present hostplant data from the isthmus to associate
leaf-flushing phenology with global and local climate.
Counts of migratory butterflies
During each migratory season, we censused the number
of butterflies that flew across two 300 m transects over
Lake Gatu?n (an artificial lake flooded in 1914 to form
a large part of the Panama Canal). We conducted
each census from a boat tied to a stump (9110.6220N,
79149.8730W) at the southwest apex of the two transects.
Looking north for 1 min between stump and the western
promontory of Buena Vista Peninsula (9110.7090N,
79150.0010W), a person recorded the species of butterflies
that crossed the transect, the number of each species
flying west, and the number flying east. Looking east
for 1 min between stump and a southern promontory
of Buena Vista Peninsula (9110.7140N, 79150.7270W), a
person recorded the species of butterflies crossing the
transect, the number of each species flying south, and
the number flying north. Following these counts, we
collected environmental data, including air temperature,
wind speed and direction, the sun?s visibility, and the
proportion of sky covered by clouds. Then we counted
the number of butterflies a second time for each transect.
An entire census required approximately 5 min when two
persons were in the boat and a couple of minutes longer if
only one person collected data.
In this paper, we focus on the greatest migratory
activity of Aphrissa butterflies over the course of the
migratory months (Fig. 1, hereafter referred to as the
migratory peak). We pooled data for A. statira and
Aphrissa boisduvalii Felder because it is difficult to
distinguish these species when they are flying freely
across the transect lines. Capturing individual butter-
flies in 7 years between 1994 and 2005, a strong linear
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Fig. 1 Observed maximum number per minute of migrating Aphrissa butterflies flying south or west across either of two 300 m transects on
Lake Gatu?n, Panama during April?July of 1997?2006. For the migration phenology for 1991?1996, see fig. 1 in Oliveira et al. (1998).
E L N I N? O A N D A T R O P I C A L B U T T E R F LY M I G R A T I O N 3
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
relationship between the number of A. statira and the
total number captured each year (R25 0.997, Po0.0001)
permits us to reliably estimate that 89  2% of the
butterflies are A. statira. Daily migratory activity pat-
terns generally peak between 10:00 and 12:00 h (see fig.
3 in Oliveira et al., 1998). Because it was not always
practical to conduct a census each hour of each day
during the migratory season, we sometimes conducted
a single census, or more often, two censuses 1 h apart
near mid-morning (approximately 10:30?11:30 h) when
counts were typically greatest. From the results for a
season, we determined the date on which we saw the
maximum number of Aphrissa butterflies flying across
either one of the two transects in the prevailing migra-
tory direction (maximum number of butterflies per
minute per 300 m). We call this date the ?migration
peak? and we call the maximum number of butterflies
counted in this peak the ?maximum migratory rate?.
When the maximum migratory rate fell on more than
one date, we called the earlier one the migration peak.
Onset of wet season
We regressed the migration peak on the end of dry
season (here referred to as the onset of wet season), a
date designated each year by an algorithm created by
the Panama Canal Commission and subsequently used
by the Meteorological and Hydrological Branch of the
Autoridad del Canal de Panama (ACP). The algorithm
tracks 11 variables (see http://striweb.si.edu/esp/
physical_monitoring/summary_seasons.htm for more
information). There are no scientific publications justi-
fying its use; however, the need for operators of the
Panama Canal to be able to predict over the long term
the timing and length of the distinct dry season for a
transport system that requires fresh water to operate is
obvious. Thus, we assumed that the method employed
was a reasonable attempt by those that devised it.
Moreover, we benefited from an objective measure of
the onset of wet season rather than using our own
subjective measurement. The ACP also marks the date
that dry season begins each year from which we calcu-
lated the length of dry season in days.
Global climate
As an indicator of ENSO, we used the sea-surface
temperature (SST) anomaly from Nin?o region 3.4 pub-
lished by the Climate Prediction Center of the National
Oceanographic and Atmospheric Administration
(http://www.cpc.noaa.gov/data/indices). Region 3.4
is located between latitudes 51N?51S and longitudes
170?1201W. The Climate Prediction Center characterizes
the SST in this region as critical to characterizing ENSO.
The SST anomaly is a departure from the adjusted
optimum interpolation analysis (measured in 1C,
Reynolds et al., 2002). Information on the base periods
used to derive the anomaly may be found on the Climate
Prediction Center website (http://www.cpc.noaa.gov/
data/indices/Readme.index.shtml). We divided the year
into four calendar quarters. We used stepwise regression
analysis to evaluate relationships between the maximum
migratory rate and the mean SST anomalies for calendar
quarters 3 and 4 in the year before the migration rate and
calendar quarters 1 and 2 in the same year as the
migration rate.
Local climate
As an index of the water available for plant productivity,
we used April soil moisture at 30?40 cm depth on BCI.
We focused on the month of April because female
Aphrissa butterflies may be laying eggs on newly flushed
leaves. Their offspring will hatch and feed to grow into
the migratory generation. Expressed in percent soil wet
weight, deep soil moisture was measured by the Terres-
trial-Environmental Science Program of the STRI. It is
measured gravimetrically from 2.5-cm soil cores collected
every 2 weeks at 10 sites in the Lutz watershed. Further
details may be found on the STRI website (http://
striweb.si.edu/esp). We also totaled the rainfall in the
BCI laboratory clearing for January through April as a
measure of dry season rainfall. These data also came from
the Terrestrial-Environmental Science Program of STRI.
As a measure of sunlight available for butterfly
thermoregulation and plant productivity, we used total
solar radiation measured by a LiCor Pyranometer Q1, and
the photosynthetically active radiation (PAR) measured
by the Terrestrial-Environmental Science Program
above the forest canopy (at the top of Lutz Tower: 42?
48 m). We calculated the average daily solar radiation
from January through April of each year. January 1991
was excluded because both the pyranometer and the
PAR were not functional for the majority of the month,
and 1997 lacks PAR values because the data were not
recorded for most of February and all of April.
Hostplants
We monitored new leaf flushing by C. latifolia on BCI and
along Pipeline Road, approximately 4 km north of
Gamboa (for a map of the migratory flyway indicating
BCI and Gamboa, see Srygley et al., 1996). Most plants
extended up the forest edge to the canopy. With binocu-
lars, we observed 10 branches for each plant and counted
the number that were flushing new leaves. A. statira
butterflies are not active in the forest interior, and so the
proportion is a relative assessment of the availability of
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
4 R . B . S R Y G L E Y et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
new leaves for larval diet. Plants were added or sub-
tracted as they were discovered, died, or lost (on average,
4.6 plants were monitored at any one time). We observed
the plants approximately once each month from January
1995 to December 1997 (Fig. 2a). In 2002?2006, we mon-
itored the same plants every 2 weeks concentrating our
efforts on the migration period (early April to the end of
July, Fig. 2b). We averaged the plants? flushing propor-
tions on each date sampled. We then found the maximum
for each year and called that date the peak flushing date
and the quantity the peak proportion.
X. seemannianum was monitored for flushing of new
leaves along Pipeline Road. It too is a canopy liana, and
so we observed plants from the road edge. We discov-
ered A. statira?s use of Xylophragma in November 1994
and we began to monitor these hostplants at that time.
Data were collected in the same way as that for
C. latifolia. On average, 5.6 plants were monitored once
each month, year-round from January 1995 to Decem-
ber 1997 (Fig. 2c). In 2002?2006, we monitored five
plants every 2 weeks during the migration period
(Fig. 2d). Because none of the plants produced new
leaves during the migration period in 1996, we did not
assign a peak flushing date for Xylophragma. For that
year, the peak proportion was zero.
Results
Global and local climate
Over the past 100 years, the average date that wet
season began was May 4. Over the course of this study
(n5 16 years), the average onset of wet season was May
10 and ranged between April 24 and June 7. The date
that wet season began was significantly related with the
migration peak (Fig. 3a, F1, 145 5.56, y5 271 0.5x,
P5 0.034). The migration peak occurred approximately
4 weeks after the onset of wet season (mean
SD5 32.2  12.1 days). The migration peak was later
when dry season was longer (F1, 145 5.10, P5 0.040)
and sunnier (Pyranometer: F1, 145 7.05, P5 0.0188;
PAR: F1, 145 12.22, P5 0.0039). The migration peaked
earlier when deep soil on BCI was more moist in April
(Fig. 3b, F1, 145 5.43, P5 0.035).
The maximum migratory rate of Aphrissa butterflies
was not significantly autocorrelated from 1 year to the
next (r5 0.26). However, it tended to be negatively
correlated across a 2-year lag (partial correlation5
0.50). The median maximum migratory rate was 22
butterflies per minute per 300 m, yet in 3 years, 1992,
1997, and 2002, counts exceeded 90 per minute.
Log-transformed maximum migratory rate of Aphrissa
was best explained by the ENSO 3.4 SST for the second
quarter (Fig. 4, F1, 145 8.17, P5 0.013). The other quar-
ters were not significant (P40.91) following selection of
the ENSO 3.4 SST for the second quarter. However, the
ENSO 3.4 SST anomaly was highly autocorrelated
among calendar quarters, although only within half-
years across the duration of this study. The partial
correlation among contiguous quarters was 0.75
(Po0.0001) and it was 0.50 among half-year lag times
(Po0.0001), whereas three- to eight-quarter lags were
not significant (P40.15 in all cases). Thus El Nin?o
events, which often begin in October and end in the
first half of the following year, are independent across
years within the duration of this study. However, the
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Fig. 2 New leaf-flushing activity for (a, b) Callichlamys latifolia and (c, d) Xylophragma seemannianum (Bignoniaceae), larval hostplants for
Aphrissa statira. Hostplants were monitored all year for 3 years (a, c), and then during the migration season (April?July) for 5 years (b, d).
E L N I N? O A N D A T R O P I C A L B U T T E R F LY M I G R A T I O N 5
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
autocorrelations indicate that associating maximum
migratory rate of butterflies with the ENSO 3.4 SST in
the second quarter may also indicate a much more
general association with the ENSO event that spans
from the latter quarter of the previous calendar year to
the first or second quarter of the year of interest. Across
the course of this study, the mean ENSO 3.4 SST
anomaly in the second quarter of the year (April?June)
was 0.23. With warm SST anomalies (40.4 1C above
average), the years 1991, 1992, 1993, 1997, and 2002
are classified as El Nin?o events (1997 was a very strong
event). Defining La Nin?a events as a cold SST anomaly
(o0.4 1C below average), the year 1999 was a strong
event and 2000 a weak one.
The maximum migratory rate declined exponentially
with increased deep soil moisture on BCI in April (Fig. 5a,
F1, 145 6.95, P5 0.020). The maximum migratory rate also
declined exponentially with total dry season precipitation
(Fig. 5b, F1, 145 5.18, P5 0.039), which in turn declined
significantly with the ENSO 3.4 SST in the second quarter
(F1, 145 5.43, P5 0.035). The ENSO 3.4 SST was not sig-
nificantly related to solar radiation in the dry season
(Pyranometer: F1, 145 0.09, P5 0.765; PAR: F1, 135 0.50,
P5 0.494). Total solar radiation did not significantly effect
the maximum migratory rate (Pyranometer: F1, 145 0.58,
P5 0.459), but PAR did tend to have an effect on the
maximum migratory rate (F1, 135 4.04, P5 0.066).
Hostplant new leaf phenology
For X. seemannianum, the mean proportion of branches
producing new leaves declined significantly with rain-
fall in the dry season (measured locally in Gamboa,
Fig. 6a, F1, 65 7.70, P5 0.032) and increased signifi-
cantly with the ENSO index in the second quarter
(Fig. 6b, F1, 65 9.26, P5 0.023). It was not significantly
related to total solar radiation or PAR in the dry season
or in April (P40.34 for all four tests). For C. latifolia, the
mean proportion of branches producing new leaves
tended to decrease with rainfall on BCI (Fig. 6c,
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Fig. 3 The date that the migratory rate of Aphrissa butterflies peaked each year was positively associated with (a) the onset of wet season for
the same year (data courtesy of the Panama Canal Authority) and negatively associated with (b) deep soil moisture on Barro Colorado Island
(BCI) in April. Dates are graphed as days relative to the end of April, which is near to the 100-year average for the onset of wet season (May
4). Open triangles indicate El Nin?o years; closed triangles, La Nin?a years; open circles, neither El Nin?o nor La Nin?a years.
Fig. 4 The log-transformed maximum migratory rate for Aphrissa
butterflies relative to the average sea-surface temperature (SST)
anomaly measured in Nino region 3.4 in the second quarter of the
year (data courtesy of the National Oceanographic and Atmo-
spheric Administration). Symbols explained in Fig. 3 legend.
6 R . B . S R Y G L E Y et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
F1, 65 3.77, P5 0.100), but it was not significantly asso-
ciated with the ENSO 3.4 SST (Fig. 6d, F1, 652.02,
P50.205). It too was not significantly related to solar
radiation in the dry season or in April (P40.34 for all tests).
The maximum migration rate of Aphrissa was not
related to the peak amount of new leaf production in
Callichlamys (F1, 65 0.31, P5 0.60), but it tended to in-
crease with the peak amount of new leaf production in
Xylophragma (F1, 65 3.59, P5 0.107). The Aphrissa migra-
tion peaked approximately 22 days after Callichlamys
peaked flushing (F1, 65 6.65, y5 221 0.37x, P5 0.042),
whereas it was not related to the peak flushing of
Xylophragma (F1, 55 0.00, P5 0.99).
Discussion
The migration of Aphrissa butterflies across the Panama
Canal peaked approximately 4 weeks after the onset of
wet season. A. statira grew from egg to adult in approxi-
mately 22 days when reared in the laboratory (R. B.
Srygley, unpublished data). From these observations,
we hypothesize that in the first 4 or 5 days following the
initiation of wet season, females lay eggs on newly
flushed leaves. These eggs grow into a new generation
of adults that migrate from the wetter evergreen Atlan-
tic coastal forests to the drier deciduous forests on the
Pacific coast.
The maximum migratory rate was strongly influ-
enced by the ENSO. The maximum migratory rate
was greatest during drier El Nin?o events and least
during wetter La Nin?a events. The ENSO climate is
known to affect herbivore populations in subtropical
desert and seasonally dry grasslands and scrub forests.
Some of the best documented are the increases in rodent
populations as a result of increased rainfall during El
Nin?o years causing a pulse of plant productivity in
scrub and desert communities of western South Amer-
ica (Meserve et al., 2003) and western North America
(Polis et al., 1997; Holmgren et al., 2006). Painted-lady
butterflies Vanessa cardui also migrate in greater num-
bers following rainy El Nin?o events in the western
United States (Vandenbosch, 2003). Thus the typical
scenario in these dry regions is a pulse in plant pro-
ductivity positively affecting herbivore populations.
However in the case of the rodents, Meserve et al.
(2003) conclude that control shifts from bottom-up in
dry years when plant resources are limited to top-down
in wetter years marked by greater primary productivity.
Similarly, breeding success of rufous-crowned sparrows
in California scrub habitat was also subject to control
shifting from bottom-up during dry, resource-limited
La Nin?a to top-down during wet El Nin?o (Morrison &
Bolger, 2002). In the Mediterranean climate of Califor-
nia, increased rainfall suppresses herbivory by the
ghost moth Hepialis californicus because wet soils favor
the nematodes that regulate the moths (Preisser &
Strong, 2004). A variety of ecological effects may thus
derive from systematic changes in rainfall.
In addition to influencing the maximum migratory
rate, we assume that ENSO affects the population sizes
of migratory butterflies at the Atlantic coastal source and
the Pacific destination of this migration. Populations of
other lepidopteran species in Panama erupted during the
strong El Nin?o event of 1997?1998 (e.g., in the Pacific
coastal dry forest, Van Bael et al., 2004; and the Atlantic
coastal wet forest, J. Wright, unpublished data).
There are a number of mechanisms that could lead to
an outbreak in the Aphrissa populations in El Nin?o years
and a decline in La Nin?a years. El Nin?o years are
associated with increased temperatures, increased solar
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Fig. 5 (a) The maximum migratory rate for Aphrissa butterflies declined logarithmically with deep soil moisture on Barro Colorado
Island (BCI) in April, and (b) the total rainfall during the dry season. Symbols explained in Fig. 3 legend.
E L N I N? O A N D A T R O P I C A L B U T T E R F LY M I G R A T I O N 7
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
radiation, and decreased rainfall on BCI (Wright &
Caldero?n, 2006). X. seemannianum, the bignoniaceous
liana that is a host for A. statira responded to the
decreased rainfall during the dry season and dry April
soils associated with El Nin?o with an increase in leaf
flushing. C. latifolia, the other known host for A. statira,
had its greatest leaf production during intermediate
years of moderate dry season rainfall. Measuring seed
production, Wright & Caldero?n (2006) found that Calli-
chlamys and other lianas responded favorably to El Nin?o
years as long as they did not result in extreme droughts.
Given the general effect of water stress on plant pro-
ductivity in tropical wet forest (e.g., Engelbrecht &
Kursar, 2003), it is surprising that the lianas respond
so favorably to dry soils.
We propose two hypotheses to explain the increase in
plant productivity with El Nin?o years in this tropical
wet forest. First, the host lianas may produce new
leaves more synchronously at the onset of wet seasons
that end droughts associated with El Nin?o years. Our
observations do not support this hypothesis. In 3 out of
7 years, Xylophragma peaked before the wet season
began. Factors timing leaf production in these host
lianas are not known, and threshold levels of drought
can synchronize flowering and enhance seed set in
some tropical forest plants (Alvim, 1960).
Second, greater availability of light for photosynth-
esis during El Nin?o relative to La Nin?a years may
promote the production of new leaves. This hypothesis
is supported by observations that solar radiation limits
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Fig. 6 The maximum proportion of Xylophragma branches flushing new leaves (a) declined with total local rainfall during the dry
season and (b) increased with the sea-surface temperature (SST) anomaly for region 3.4 in the second quarter. The maximum proportion
of Callichlamys branches flushing new leaves tended to have the same relationships as Xylophragma with (c) dry season rainfall and (d) the
SST anomaly. Symbols explained in Fig. 3 legend.
8 R . B . S R Y G L E Y et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
photosynthetic activity in Neotropical wet forest
(reviewed in Van Schaik et al., 1993), and experimental
addition of radiation to a Pacific coastal forest during
the wet season resulted in increased tree growth and
reproduction (Graham et al., 2003). Moreover, Amazo-
nian rainforests produce and drop leaves in anticipation
of seasonal changes in solar radiation (Myneni et al.,
2006).Q2 Although dry season sunlight measured above
the canopy on BCI during the period of our study was
unaffected by El Nin?o (but see Wright & Caldero?n,
2006), drought may cause big trees to lose leaves and
the thinned canopy may allow more sunlight onto the
host lianas. On BCI, understory Connarus turczaninowii
lianas also responded positively to the duration of the
dry season (Aide & Zimmerman, 1990), which the
authors postulate is due to greater canopy tree leaf loss
during El Nin?o years reducing competition for water
and light in the understory.
For this study, we did not measure hostplant quality
or predation pressure, both of which may contribute to
population outbreaks. As a result of poor defense in
drought years, hostplants might also suffer greater
herbivore pressure and the insects might survive better
during development. Because of increased solar radia-
tion, El Nin?o years might be favorable for herbivore
development, or adults might be more active to find
food plants or oviposit on larval resources. El Nin?o
might also result in a decrease in the natural enemies of
herbivores, resulting in an ecological release of the
lepidopteran populations (e.g., Preisser & Strong,
2004). Thus our demonstration that outbreaks in migra-
tory populations of Aphrissa butterflies is associated
with increased plant productivity in Panama does not
preclude potential roles of plant defenses and top-down
pressure from parasitoids and predators in regulating
migratory populations (Stireman et al., 2005).
In Panama, an El Nin?o year is typically characterized
by drier than normal conditions, and the Aphrissa
butterflies respond with an outbreak in the migratory
population. The El Nin?o year of 1993 was a notable
exception. The rate of butterflies migrating was much
lower than predicted by its ENSO index (Fig. 4). How-
ever, in Panama the dry season of 1993 was unusually
wet for an El Nin?o year. When deep soil moisture in
April is the predictor, the maximum migratory rate for
1993 falls in line (Fig. 5a). Thus, both global and local
environmental factors influenced the number of
migrants traversing the Panama Canal.
In the Meteorological Research Institute model to
predict climate change, ENSO becomes more extreme
in its peaks with an overall warming of the mean SST;
however, this is not true for the Geophysical Fluid
Dynamical Laboratory model (Yeh & Kirtman, 2007).
Thus, extreme El Nin?o and La Nin?a years may become
more frequent. Other tropical rainforest Lepidoptera
appear to have a similar response to rainfall as the
Aphrissa butterflies in Panama (e.g., Van Bael et al.,
2004; Kunte, 2005). If this phenomenon extends beyond
Lepidoptera, then insect outbreaks during El Nin?o and
population crashes during La Nin?a may become more
frequent. Over the past century, the numbers of Lepi-
doptera migrating to Britain were associated with high-
er temperatures in the migratory route in phase with the
North Atlantic Oscillation (Sparks et al., 2005). Similarly,
a long-term increase in the number of migrants may be
linked to global warming (Sparks et al., 2007).
The El Nin?o migration syndrome may apply to
desert, savannah, and tropical wet forest (e.g., Vanden-
bosch, 2003), but the mechanisms may differ. In El Nin?o
years, rainfall in deserts and savannah tends to increase,
while it declines in tropical wet forests (see fig. 1 in
Holmgren et al., 2006). However in all three ecotypes,
hostplants respond by producing more leaves, enhan-
cing migratory butterfly populations.
Acknowledgements
A Senior Postdoctoral Fellowship to R. B. S. from the Smith-
sonian Institution and grants from the National Geographic
Society Committee for Research and Exploration supported the
research. The Autoridad Nacional del Ambiente (ANAM)
granted permission to conduct the research in Panama. Hydro-
logical and radiation data were provided by the Terrestrial-
Environmental Science Program of the STRI. Season begin and
end dates were provided by the Meteorological and Hydrologi-
cal Branch of the Panama Canal Authority (ACP). For presenta-
tional purposes, data were modified from their original form. We
thank the meteorologists and their parent institutions for their
long-term commitment to these essential programs. We also
thank S. Paton for numerous discussions and assistance with
BCI climate and its effects, and Joe Wright and Egbert Leigh for
commenting on the manuscript.
References
Aide TM, Zimmerman JK (1990) Patterns of insect herbivory,
growth, and survivorship in juveniles of a Neotropical liana.
Ecology, 71, 1412?1421.
Alvim PdeT (1960) Moisture stress as a requirement for flower-
ing of coffee Q3. Science, 132, 354.
Behrenfeld MJ, Randerson JT, McClain CR et al. (2001) Biospheric
primary production during an ENSO transition Q4. Science, 291,
2594?2597.
Brower LP (1995) Understanding and misunderstanding the
migration of the monarch butterfly (Nymphalidae) in North
America. Journal of the Lepidopterists? Society, 49, 304?385.
Dingle H, Drake VA (2007) What is migration? Bioscience, 57,
113?121.
Engelbrecht BMJ, Kursar TA (2003) Comparative drought-resis-
tance of seedlings of 28 species of co-occurring tropical woody
plants. Oecologia, 136, 383?393.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
E L N I N? O A N D A T R O P I C A L B U T T E R F LY M I G R A T I O N 9
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
Forister ML, Shapiro AM (2003) Climatic trends and advancing
spring flight of butterflies in lowland California. Global Change
Biology, 9, 1130?1135.
Graham EA, Mulkey SS, Kitajima K, Phillips NG, Wright SJ
(2003) Cloud cover limits net CO2 uptake and growth of a
rainforest tree during tropical rainy seasons. Proceedings of the
National Academy of Sciences, 100, 572?576.
Haber WA, Stevenson RD (2004) Diversity, migration, and con-
servation of butterflies in northern Costa Rica. In: Biodiversity
Conservation in Costa Rica: Learning the Lessons in a Seasonal Dry
Forest (eds Frankie G, Mata A, Vinson S), pp. 99?115.
University of California Press, Los Angeles, CA.
Holmgren M, Stapp P, Dickman CR et al. (2006) Extreme climatic
events shape arid and semiarid ecosystems. Frontiers in Ecology
and the Environment, 4, 87?95.
Jaksic FM (2001) Ecological effects of El Nin?o in terrestrial
ecosystems of western South America. Ecography, 24,
241?250.
Kiladis GN, Diaz HF (1989) Global climatic anomalies associated
with extremes in the Southern Oscillation. Journal of Climate, 2,
1069?1090.
Kunte K (2005) Species compositions, sex ratios, and move-
ment patterns in danaine butterfly migrations in southern
India. Journal of the Bombay Natural History Society, 102,
280?286.
Larsen TB (1976) The importance of migration to the butterfly
faunae of Lebanon, East Jordan, and Egypt. Notulae Entomolo-
gicae, 56, 73?83.
Loxdale HD, Lushai G (1999) Slaves of the environment: the
movement of herbivorous insects in relation to their ecology
and genotype. Philosophical Transactions of the Royal Society of
London B, 354, 1479?1495.
McPhaden MJ, Zebiak SE, Glantz MH (2006) ENSO as an
integrating concept in Earth science. Science, 314, 1740?1745.
Meserve PL, Kelt DA, Milstead WB, Gutie?rrez JR (2003) Thirteen
years of shifting top-down and bottom-up control. Bioscience,
533, 633?646.
Morrison SA, Bolger DT (2002) Variation in a sparrow?s repro-
ductive success with rainfall: food and predator-mediated
processes. Oecologia, 133, 315?324.
Myneni RB, Yang W, Nemani RR et al. (2006) Large seasonal
swings in leaf area of Amazon rainforests. Proceedings of the
National Academy of Sciences, 104, 4820?4823.
Oliveira EG, Srygley RB, Dudley R (1998) Do Neotropical
migrant butterflies navigate using a solar compass? Journal of
Experimental Biology, 201, 3317?3331.
Parmesan C (2006) Ecological and evolutionary responses to
recent climate change. Annual Review of Ecology and Systema-
tics, 37, 637?669.
Polis GA, Hurd SD, Jackson CT, Sa?nchez Pin?ero F (1997) El Nin?o
effects on the dynamics and control of an island ecosystem in
the Gulf of California. Ecology, 78, 1884?1897.
Preisser EL, Strong DR (2004) Climate affects predator control of
an herbivore outbreak. American Naturalist, 163, 754?762.
Rainey RC (1951) Weather and the movements of locusts
swarms: a new hypothesis. Nature, 168, 1057?1060.
Reynolds RW, Rayner NA, Smith TM, Stokes DC, Wang W (2002)
An improved in situ and satellite SST analysis for climate.
Journal of Climate, 15, 1609?1625.
Sparks TH, Dennis RLH, Croxton PJ, Cade M (2007) Increased
migration of Lepidoptera linked to climate change. European
Journal of Entomology, 104, 139?143.
Sparks TH, Roy DB, Dennis RLH (2005) The influence of tem-
perature on the migration of Lepidoptera into Britain. Global
Change Biology, 11, 507?514.
Southwood TRE (1977) Habitat: the templet for ecological stra-
tegies? Journal of Animal Ecology, 46, 337?365.
Srygley RB (2001) Compensation for fluctuations in crosswind
drift without stationary landmarks in butterflies migrating
over seas. Animal Behaviour, 61, 191?203.
Srygley RB, Dudley R (2008) Optimal strategies for insects
migrating in the flight boundary layer: mechanisms and con-
sequences. Integrative and Comparative Biology, 48, 119?133.
Srygley RB, Dudley R, Oliveira EG (1996) Wind drift compensa-
tion, flyways, and conservation of diurnal, migrant Neotropi-
cal Lepidoptera. Proceedings of the Royal Society London B, 263,
1351?1357.
Srygley RB, Oliveira EG (2001) Sun compass and wind drift
compensation in migrating butterflies. Journal of Navigation, 54,
405?417.
Stenseth NC, Mysterud A, Ottersen G, Hurrell JW, Chan K-S,
Lima M (2002) Ecological effects of climate fluctuations.
Science, 297, 1292?1296.
Stireman JO, Dyer LA, Janzen DH et al. (2005) Climatic unpre-
dictability and parasitism of caterpillars: implications of global
warming. Proceedings of the National Academy of Sciences, 102,
17384?17387.
Van Bael SA, Aiello A, Valderrama A, Medianero E, Samaniego
M, Wright SJ (2004) General herbivore outbreak following an
El Nin?o-related drought in a lowland Panamanian forest.
Journal of Tropical Ecology, 20, 625?633.
Vandenbosch R (2003) Fluctuations of Vanessa cardui butterfly
abundance with El Nin?o and Pacific Decadal Oscillation cli-
matic variables. Global Change Biology, 9, 785?790.
Van Schaik CP, Terborgh JW, Wright SJ (1993) The phenology of
tropical forests: adaptive significance and consequences for
primary consumers. Annual Review of Ecology and Systematics,
24, 353?377.
Wright SJ, Caldero?n O (2006) Seasonal, El Nin?o and longer term
changes in flower and seed production in a moist tropical
forest. Ecological Letters, 9, 35?44.
Yeh SW, Kirtman BP (2007) ENSO amplitude changes due to
climate change projections in different coupled models. Journal
of Climate, 20, 203?217.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
10 R . B . S R Y G L E Y et al.
r 2009 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2009.01986.x
GCB 1986(BW
UK
 G
CB
 19
86
.PD
F 0
5-J
un
-09
 20
:2 
41
46
83
 B
yte
s 1
0 P
AG
ES
 n 
op
era
tor
=jn
m.
Ch
ris
tin
a)
Author Query Form
_______________________________________________________
_______________________________________________________
Dear Author,
During the copy-editing of your paper, the following queries arose. Please respond to these by marking up your proofs with the necessary 
changes/additions. Please write your answers clearly on the query sheet if there is insufficient space on the page proofs. If returning the 
proof by fax do not write too close to the paper's edge. Please remember that illegible mark-ups may delay publication. 
Journal?? ? ? ? ? GCB
Article??? ? ? ? 1986 
Query No. Description Author Response 
Q1 ? Author: Please give manufacturer information for LiCor Pyranometer: company name, town, state (if USA), and country.
? 
?
Q2 ? Author: The publication year of 2007 has been changed to 2006 match the publication year given in the reference list for the reference Myneni et al. (2006). Please confirm that this is correct.
? 
?
Q3 ? Author: If this is not a one-page article please supply the first and last pages for the reference Alvim (1960).? ?
Q4 ? Author: If there are 8 or more authors, please list the first three authors, then et al. If there are 7 authors or fewer, please list all authors for references with three authors et al.
? 
?
?
? 
?
?
? 
?
?
? 
?
?
? 
?
?
? 
?
?
? 
?