Amer. J. Bot. 73(3): 353-363. 1986.

MORPHOLOGY AND DISPERSAL POTENTIAL OF WIND-DISPERSED

DIASPORES OF NEOTROPICAL TREESI

CAROL K. AUGSPURGER
Department of Plant Biology, University of lI1inois, 505 S. Goodwin, Urbana, lI1inois 6180 I

ABSTRACT
Morphological and aerodynamic traits affecting mean potential dispersal distance are quan-
tified for wind-dispersed diaspores of tree species on Barro Colorado Island, Panama. The
sample includes 34 species in 16 families and represents six aerodynamic groups. Mass and
area (maximum cross section) each vary over six orders of magnitude among the species. In
contrast, wing-loading, defined as weight divided by area, varies over only one order of mag-
nitude, as does the rate of descent. While the regression of rate of descent on the square root
of wing-loading is significant overall, the slopes vary significantly among five aerodynamic
groups. At comparable wing-loading values, diaspores of fluffy kapok fall faster than four other
aerodynamic groups and rolling autogyros fall faster than non-rolling autogyros. Assuming the
diaspores are released from their typical tree height and experience a mean windspeed of 1.75
m sec:", the expected mean dispersal distance varies among the 34 species from 22 to 194 m.
Rate of descent is weakly correlated with shade tolerance of seedlings for a subset of 18 species;
rate of descent is more strongly correlated with the log of dry mass of seed for all 34 species.
Given these wide differences in dispersal potential, any generalizations about tropical trees that
use wind dispersal are of dubious value.

MORPHOLOGICAL DESIGNS of wind-dispersed quences of this variation for rate of descent

diaspores appear to slow their rates of descent
and increase their chances of exposure to hor-
izontal or gusty winds. Rate of descent and
dispersal potential have been documented for
temperate herbaceous species (Sheldon and
Burrows, 1973; Werner and Platt, 1976; Platt
and Weis, 1977; Rabinowitz and Rapp, 1981)
and for temperate trees of North America (Sig-
gins, 1933; Green, 1980; Guries and Nord-
heim, 1984) and Eucalyptus trees of Australia
(Cremer, 1977). Values for rate of descent for
temperate tree seeds have been incorporated
into a model of wind dispersal that predicts
the seeds' spatial distribution under specified
wind conditions and crop size (Fields and
Sharpe, 1980; Sharpe and Fields, 1982). In the
species-rich tropical forests, there is much vari-
ation in morphology and size of wind-dis-
persed diaspores (Ridley, 1930). The conse-
I Received for publication 4 June 1985; revision ac-
and dispersal potential are the subject of this
study.
Horizontal winds spread diaspores over a
broad area and carry them away from the par-
ent. Hence, the potential dispersal distance and
area are inversely related to the rate ofdescent.
Because consistency ofwind direction will also
affect the total area over which diaspores are
dispersed, many studies of wind dispersal, in-
cluding the present one, focus on dispersal po-
tential. Dispersal potential here refers specif-
ically to potential mean dispersal distance,
although the area covered is implicitly in-
volved as it tends to increase with increasing
mean distance.
Green (1980) found for seven species oftem-
perate trees that a diaspore's rate of descent is
highly correlated with its (wing-loading) 1/2; here
wing-loading is defined as wt/area. His dia-
spores fall into two morphological groups; one
group simultaneously autorotates around the

cepted 27 September 1985.

This research was supported by NSF grant BSR 8219856
and was made possible by the support facilities and per-
sonnel of the Smithsonian Tropical Research Institute,
especially D. Windsor and the Environmental Science Pro-
gram. I thank R. Beebe and his staff for use ofthe Krannert
Center for the Performing Arts, Univ. of 111., to determine
rates of descent, L. Siri Kimsey for the artwork, S. Franson
for statistical analyses, and B. Benner, S. Franson, K. Ho-
gan, C. Kelly, and H. Michaels for assistance in deter-
mining rates of descent. R. B. Foster collected the fruits
of Centrolobium sp. in Peru. K. Hogan and S. Franson
contributed constructive comments on earlier drafts of the
manuscript.
353
diaspore's longitudinal axis and autogyrates
around one end of the diaspore ("rolling
samaras"), while the other group only autogy-
rates ("non-rolling samaras"). For a given wing-
loading, the first group has a faster rate of de-
scent in still air than the second. The validity
of these relationships can be further tested us-
ing the wider variety of morphological groups
and sizes present among tropical species.
This paper presents data for 34 species of
tropical trees found on Barro Colorado Island,
Panama. I investigated the relationship of
354 AMERICAN JOURNAL OF BOTANY
(wing-loading)!" to the rate of descent for six
diverse morphological groups to determine
whether the groups vary in their dispersal po-
tentials. Using representative tree ht and wind
speeds, expected mean dispersal distances were
then calculated for each species. Finally, using
existing data for shade tolerance of seedlings
and dry wt of seeds of the species (Augspurger,
1984), I tested whether rates of descent are
related to the light requirements for seedlings
of these species. The probability of some di-
aspores landing in a light-gap increases as di-
aspores descend more slowly, thus traveling
greater distances and spreading over a larger
area. Hence, species requiring light-gaps for
early seedling survival may have slower rates
of descent of their diaspores.
STUDY SPECIES AND METHODS- The study in-
volves 34 tree species in 16 families and in-
cludes approximately 90% oftree species using
wind dispersal in the semi-deciduous forest on
Barro Colorado Island, Panama (Table 1, Fig.
1). Croat (1978) provides a detailed description
of each species. Foster and Brokaw (1982) de-
scribe the vegetation on the island. The trees
include emergent, canopy, and subcanopy trees.
Their crown ht range from 15-45 m, and the
ht of a tree relative to the surrounding vege-
tation is expected to affect wind dispersal as
well. The average canopy ht of the old forest
on Barro Colorado Island is 30-40 m; it is 10
m shorter in the young forest (Foster and Bro-
kaw, 1982).
Field collections of diaspores were made in
1982 and 1983. Because of the high species
diversity in this tropical forest, individual adults
of a particular species are sometimes widely
separated, and not all individuals of a species
fruit in the same year. Therefore it is possible
to locate isolated individual parents of these
species. Fifteen intact diaspores were collected
from the ground just outside the crown of a
specified parent tree for each species. This stan-
dard location was selected for all species to
allow interspecific comparisons. Previous
measurements of spatial distributions of dis-
persed diaspores indicated that this location
was typically where the peak density of seeds
occurred (c. Augspurger, unpubl. data). Al-
though these samples may be slightly biased
toward diaspores with low dispersal ability,
such a bias is consistent across species. Also
given the stochastic element of dispersal and
the fact that dispersal potential and distance
are only loosely coupled for a given diaspore,
even diaspores with high dispersal potential
may be included in the sample.
The mass and area for each diaspore were
[Vol. 73
determined shortly after the time of dispersal.
Diaspores with a mass ~ 100 mg were weighed
on a Cahn electrobalance, those with a mass
> 100 mg on a Mettler top-loading balance.
Area was defined as the maximum cross sec-
tion of the diaspore and was determined using
a Licor leaf area meter. The majority of the
diaspores were flat enough so they could be
passed directly through the meter. For dia-
spores with a pronounced third dimension, the
following modifications were required. For
species 4, 5, 8, 22, and 25 (Fig. 1), the radius
of the fluffy sphere was used to calculate the
maximum cross-sectional area of the sphere.
For species 21, 24, and 34 (Fig. 1), the shape
of the diaspores was traced onto paper, cut,
and passed through the meter. For species 3,
11, and 33 (Fig. 1), all wings were traced onto
paper, cut, and passed through the meter. For
species 6 (Fig. 1), all five wings were traced,
cut, and passed through the meter and mean
area per wing was determined; the area of this
diaspore producing lift at anyone time and
used to calculate its wing-loading was consid-
ered to equal the area of two wings. Finally,
for species 1, 9, and 20 (Fig. 1), the diaspores
were magnified using a photographic enlarger;
their magnified shadows were traced, cut, and
passed through the meter.
Wing-loading was defined as wt (mass x ac-
celeration) divided by area. This value was
used in subsequent analyses to determine
whether the rate of descent was proportional
to (wing-loading)':".
The rate ofdescent in still air was determined
for each diaspore shortly after field collection.
Each diaspore was dropped from the top of a
theater stage and the time required to fall
through 27 m of still air was determined. This
ht approximates the mean ht of reproductive
canopy trees on Barro Colorado Island. The
minute diaspores of species 1, 9, and 20 were
dropped from a ht of 4 m so that they could
be observed more readily. The rate of descent
(cm· sec:") is not precisely equivalent to the
terminal velocity because the diaspore requires
some time after release before it begins its aero-
dynamic movement and before it reaches its
terminal velocity. For most species this time
is a very minor portion of the total fall time;
the maximum time was approximately 1.5 sec
(8.4% of the total time) for those species with
large wing-loading values. This initial free fall
time is a small proportion of total time for
these tall trees, but can become important for
smaller trees and shrubs (see Guries and Nord-
heim, 1984).
The repeatability of the results for rate of
descent was determined for a subset of species.
March, 1986] AUGSPURGER-WIND-DISPERSED TROPICAL DIASPORES 355

FLOATER UNDULATOR

10 HELICOPTER

ROLLING AUTOGYRO ~05Cm~m

~C)~
II 33

\
2em 29
;;.~
30 31 ~
~ 23 3
~~
~ 13
17 ~ TUMBLER
~15~
O.5em lem
1--1 ~
--=-9
20

AUTOGYRO

~ 34 2em

~' NONCLASSI FIED

____--........24 ~7
28~
_~~~ <019 27~
--.:Jf7 21" 4:::> 18
2em
~

Fig. 1. Morphologies of the diaspores of 34 tree species using wind dispersal on Barro Colorado Island, Panama.
The number below a diaspore refers to its species number (see Table I for species identification). The label above a
cluster of diaspores refers to their aerodynamic classification (see Table 2 for a description of aerodynamic behaviors).
TABLE 1. Morphological characters oj wind-dispersed diaspores, rates of descent in still air, tree height, and expected mean dispersal distance at three wind speeds for 34 tree v.>
species on Barro Colorado Island, Panama. N = 15 for each species (except N = 7 for species 4, N = 3 Jar species 12, and N = 4 for species 17). Taxonomic nomenclature VI
0\
follows Groat (1978)

Rank
Expected mean of ex- Rank
dispersal distance (m) peeted of
(Wing- Rate of descent Tree _ _ _ _ _ _ _ _ disper- shade Wood
Type of Aerodynamic Mass (mg) Area (em') loadingj'v (em sec"} height 7 sal dis- toler- density
diaspore group x (SO) x (SO) x (SO) X (SD) (m) 1.75 3.5 m sec- 1b tance ances (g cm-1)d

1. Alseis blackiana Seed Rolling 0.181 (0.050) 0.030 (0.006) 76.2 (10.4) 65.0 (13.5) 35 94 188 377 31 0.57
(Rubiaceae) autogyro
2. Aspidosperma cruenata Seed Undu1ator 871 (95) 64.5 (3.3) 115.7(7.1) 94.8 (8.7) 35 65 129 258 23 5 0.71
(Apocynaceae)
3. Astronium graveolens Fruits Helicopter 47.5 (5.5) 1.93 (0.21) 143.7 (11.2) 132.8 (6.3) 35 46 92 184 12 0.99
(Anacardiaceae)
4. Bombacopsis quinata Seed Floater 32.9 (5.2) 13.7 (3.7) 49.2 (4.0) 75.6 (15.0) 35 81 162 324 27 0.39
(Bombacaceae) ;I>
5. Bombacopsis sessilis Seed Floater 462 (65) 38.8 (7.2) 109.0 (11.2) 198.1 (24.4) 25 22 44 88 2 0.42 s::tr1
(Bombacaceae) ::0
6. Cavanillesia platanifolia Fruit Tumbler 3,055 (569) 155 (14) 137.9 (9.9) 140.4 (16.2) 45 56 112 224 18 16 0.12 n
(Bombacaceae) ~
7. Cedrela odorata Seed Autogyro 14.1 (2.1) 2.28 (0.31) 78.3 (6.9) 59.7(11.4) 35 103 205 410 32 (0.47) '-<
(Meliaceae) o
C
8. Ceiba pentandra Seed Floater 74.1 (13.4) 18.2 (2.3) 63.6 (6.0) 100.3 (17.7) 45 78 157 314 26 14 0.24 ::0
(Born bacaceae)
9. Cespedezia macrophylla Seed Rolling 0.132 (0.047) 0.068 (0.008) 43.1 (6.4) 31.6 (7.1) 35 194 388 775 34 0.64
~
r-
(Ochnaceae) autogyro o'r1
10. Cochlospermum vitifolium Seed Floater 24.8 (1.1) 5.79 (0.39) 64.8 (2.8) 109.1 (15.9) 15 24 48 96 2 17 0.30
I:l:I
(Coch1osperrnaceae) o...,
11. Cordia alliodora Fruit- Helicopter 6.25 (0.68) 0.50 (0.05) 110.7 (9.7) 124.4 (9.9) 25 35 70 140 6 10 0.44
(Boraginaceae) ~
12. Couratari panamensis Seed Rolling 236 (25) 14.1 (1.5) 128.4 (8.2) 107.0 (12.2) 35 57 114 229 20 0.49 -<
(Lecythidaceae) autogyro
13. Dalbergia retusa Fruit Rolling 248 (23) 10.7 (0.9) 150.6 (5.1) 149.8 (17.0) 25 29 58 117 3 7 0.95
(Papi1ionoideae) autogyro
14. Jacaranda copaia Seed Undu1ator 6.50 (1.52) 4.68 (0.26) 36.7 (4.3) 39.0 (6.2) 30 135 269 538 33 0.38
(Bignoniaceae)
15. Lafoensia punicifolia Seed Rolling 38.0 (2.4) 2.46 (0.25) 123.4 (3.9) 109.3 (11.4) 30 48 96 192 15 6 0.72
(Lythraceae) autogyro
16. Lonchocarpus pentaphyllus Fruit Rolling 143 (15) 8.33 (0.81) 129.9 (4.5) 132.1 (15.0) 35 46 93 185 13 8 (0.75)
(Papilionoideae) autogyro
17. Lonchocarpus velutinus Fruit Rolling 127 (54) 8.58 (0.75) 126.0 (19.8) 102.7 (13.7) 35 60 119 239 21 (0.75)
(Papi1ionoideae) autogyro
18. Luehea seemannii Seed Autogyro 2.10 (0.45) 0.26 (0.02) 88.7 (10.3) 70.0 (7.0) 35 88 175 350 30 13 (0.50)
(Tiliaceae) '<S
19. Luehea speciosa Seed Autogyro 4.95 (0.50) 0.30 (0.04) 128.1 (13.0) 96.7 (10.9) 25 45 90 181 11 (0.50)
(Tiliaceae) -...J
v.>
TABLE 1. Continued ~
II)
'"1
o
Rank
Expected mean of ex- Rank
r-.....
dispersal
distance (m) pected of \0
(Wing- Rate of descent Tree disper- shade Wood 00
Type of Aerodynamic Mass (mg} Area (em') Ioadingj's ' (em sec") height 7 sal dis- toler- density 0\
~
diaspore group x (SO) x (SO) x (SO) .e(SO) (m) 1.75 3.5 m sec-w tance ance- (g cm- 3) d

20. Macrocnemum glabrescens Seed Rolling 0.0194 (0.0056) 0.0029 (0.0004) 80.6 (9.8) 51.5 (9.0) 25 85 170 340 29 0.60
(Rubiaceae) autogyro
21. Myroxylon balsamum Fruit Autogyro 749 (138) 14.2 (2.1) 227.1 (15.0) 154.5 (22.4) 35 40 79 159 9 1 0.78
(Papilionoideae) e>
22. Ochroma pyramidale Seed Floater 14.8 (3.8) 7.31 (2.76) 45.9 (6.5) 61.3 (7.2) 20 57 114 228 19 18 0.13 0~
(Born bacaceae) 'tI
23. Platymiscium pinnatum Fruit Rolling 462 (52) 20.0 (2.7) 151.0 (11.9) 128.4 (13.3) 35 48 95 191 14 0.83
c
it!
(Papilionoideae) autogyro 0
rn
24. Platypodium elegans Fruit Autogyro 1,640 (218) 23.5 (2.5) 261.7 (8.8) 178.2 (12.6) 35 34 69 137 4 3 0.75 it!
(Papilionoideae) I
25. Pseudobombax septenatum Seed Floater 86.9 (9.6) 15.7 (4.4) 75.3 (7.9) 125.7 (11.8) 25 35 70 139 5 11 0.14 ~
(Bombacaceae) Z
26. Pterocarpus rohrii Fruit Undulator 337 (106) 35.8 (5.9) 95.0 (10.4) 91.4 (14.3) 35 67 134 268 24 0.57 0I
(Papilionoideae) 0
27. Tabebuia guayacan Seed Nonclassified 35.7 (4.3) 4.50 (0.72) 88.8 (9.0) 109.9(15.7) 35 56 111 223 17 0.85 Vi
'tI
(Bignoniaceae) tTl
it!
28. Tabebuia rosea Seed Nonclassified 35.3 (4.7) 4.17 (0.28) 91.0 (5.0) 85.5 (8.6) 30 61 123 246 22 15 0.52 tr:
tTl
(Bignoniaceae) 0
29. Tachigalia versicolor Fruit Rolling 1,421 (163) 44.2 (4.1) 178.2 (8.7) 173.2 (33.7) 35 35 71 141 7 (0.52) ...,
(Caesalpinoideae) autogyro it!
0
30. Terminalia amazonica Fruits Rolling 4.07 (1.07) 1.01 (0.10) 62.8 (9.0) 73.2 (7.8) 35 84 167 335 28 9 0.68 'tI
(Combretaceae) autogyro n
31. Terminalia oblonga Fruits Rolling 54.3 (12.5) 5.33 (1.03) 99.6 (2.9) 88.6 (12.1) 35 69 138 276 25 12 (0.70)
>r-
(Combretaceae) autogyro 0
32. Trichospermum mexicanum Seed Floater 2.96 (0.30) 0.48 (0.05) 77.8 (5.3) 118.9 (8.0) 25 37 74 147 8 0.25 ;;
~
(Tiliaceae) 'tI
33. Triplaris cumingiana Fruits Helicopter 76.7 (13.1) 7.88 (0.95) 97.7 (7.2) 83.0 (11.6) 25 53 105 211 16 4 0.56 0
it!
(Polygonaceae) tTl
tr:
34. Vatairea erythrocarpa Fruit Autogyro 1,181 (233) 23.3 (2.5) 222.4 (19.6) 138.9 (14.1) 35 44 88 176 10 (0.64)
(Papilionoideae)
aWing-loading = wt/area; weight is expressed in terms ofmillidynes (mgcmsec") and area is in cm-.
b1.75 m sec:" = (4 mph); mean velocity over 24 h for March-April in laboratory clearing on Barro Colorado Island. 3.5 m sec" = (8 mph); mean velocity at noon for
March-April in laboratory clearing in Barro Colorado Island. 7 m sec-' = (15 mph); velocity used by Green (1980) to estimate wind dispersal of some temperate trees.
C From Augspurger (1984).
d Oven dry wt/green volume; values primarily from Chudnoff (1979); values in parentheses are not species-specific, but refer to genus level values.
e Seed mass includes ovary tissues.
(,;J
VI
-...I
358 AMERICAN JOURNAL OF BOT ANY [Vol. 73

TABLE 2. A description of the aerodynamic behavior of morphological groups of wind-dispersed diaspores. See Fig. 1
for drawings ofthe various groups

Aerodynamic group Aerodynamic behavior in still air

Floater Floats downward in a vertical line.

Rolling autogyro = Rotates on two axes simultaneously:
Flettner-Rotating Autogyro I) around diaspore's longitudinal axis;
(Vogel 1981) = 2) around one end of the diaspore in a semi-tight spiral.
"Rolling Samara" (Green 1980)
Autogyro = Rotates tightly around the seed end of the diaspore.
Lift-producing Airfoil (Vogel 1981) =
"Non-rolling Samara" (Green 1980)
Undulator Glides and undulates, but not with cumulative forward motion (not
continuous glider or fugoid oscillator).
Helicopter Spins tightly around a vertical line; similar to autogyro with added
wings.
Tumbler Tumbles but not on a consistent axis; randomly oriented tumbles;
also rotates around a vertical line, but in a very large spiral.
Nonclassified Complex and variable behavior within and between diaspores; closer
to autogyro and undulator than rolling autogyro.

Using a paired t-test it was determined that no
significant differences occurred between the first
and second releases of a given diaspore.
Based upon observations of their aerody-
namic behavior in still air and their mor-
phology, each species was placed into one of
six categories (Table 2); the motion of species
27 and 28 were too inconsistent to classify.
The expected mean dispersal distances were
estimated using Cremer's (1977) formula: D =
Vw(H/Vr) where D equals the horizontal dis-
tance to which wind with a velocity of Vwwill
carry a diaspore falling at a mean velocity of
Vrfrom a height H; here Vwis the average wind
velocity between the release point and the
ground. Estimates were made using three dif-
ferent wind velocities: I. 75 m- sec", the mean
velocity over 24 h in March-April in the lab-
oratory clearing on Barro Colorado Island; 3.5
m .sec-I, the mean velocity at noon for March-
April in the laboratory clearing on Barro Col-
orado Island, and 7 m- sec:", the velocity used
by Green (1980) to estimate dispersal distances
by diaspores of some temperate trees. Based
on observations ofdistances ofdiaspores from
known parent trees (C. Augspurger, unpubl.
data), it appears that mean wind speeds be-
tween I. 75 and 3.5 m sec:" are closest to the
mean speeds dispersing diaspores on Barro
Colorado Island during the dry season when
the majority of these species disperse their
seeds. Distances estimated using a mean wind
speed of 7 m- sec- 1 appear to more closely ap-
proximate the observed maxima, not the mean
dispersal distance. Typical release height ofdi-
aspores was estimated for each species based
on general observations of these tree species
over many years.
RESULTS - Both mass and area, two variables
expected to strongly affect the diaspore's rate
ofdescent, vary widely among the species; each
variable covers six orders of magnitude. Mass
varies from 0.0194-3055 mg; area varies from
0.0029-155 em- (Table 1). Mass and area do
not differ proportionately and, consequently,
the species have different values for (wing-
loading)!"; they range from 36.7-261.7 (mil-
lidynescm <)!" (Table 1). The rate of descent
in still air ranges from 31. 6-19 8.1 em sec- 1
among the species (Table 1).
A regression analysis of rate of descent on

TABLE 3. Regression analyses of rate of descent in still air on (wing-loading)'!' for five aerodynamic groups of wind-
dispersed diaspores. Each regression is significant (P < 0.00001). Significant differences exist among the slopes of
the five groups (F•.4U = 77.891, P < 0.00001;from Zar, 1978, pp. 230-233). Multiple comparisons of slopes are
indicated by underlinings, where separate lines indicate significant differences (P < 0.05)

Aerodynamic group Floaters Rolling autogyros Helicopters Undulators Autogyros

Sample size 97 157 45 45 90

r2 0.848 0.837 0.511 0.848 0.913
Intercept -22.61 -8.72 21.61 14.10 15.94
Slope 1.950 0.989 0.787 0.741 0.599
Multiple comparison
result
March, 1986] AUGSPURGER-WINO-DISPERSED TROPICAL DIASPORES 359

250
,
FLOATER
,,
5 '

o FLOATER
• ROLLING AUTOGYRO
o AUTOGYRO
• UNDULATOR
6. HELICOPTER
• TUMBLER
<> NONCLASSIFIED

28 56 84 112 140 168 196 224 252 280

(WI NG-LOADI NG)112
Fig. 2. The regression of rate of descent on (wing-loading)"? for individual diaspores of five aerodynamic groups
from Table 3. Dotted lines extend beyond the range of validity for each regression line for visual clarity only. The
symbols are plotted at the mean values for each species; see Table I for species' names.

(wing-loading)"? for all diaspores, irrespective
of species or aerodynamic group, is significant
(rate of descent = 0.559 [wing-loading]"? +
44.62;N=479,r 2 = 0.503,P < 0.00001). The
coefficient of determination (r 2) increases
greatly when regression analyses are done on
each of five aerodynamic groups separately
(Table 3). The slopes of the regression of rate
ofdescent on (wing-loading) 1/2 are not the same
for the five groups (Table 3). At a given value
for (wing-loading)!", diaspores of floaters fall
significantly more rapidly in still air than each
of the other four groups (Table 3, Fig. 2). In
addition, the rolling autogyros fall more rap-
idly than the non-rolling autogyros (Table 3;
Fig. 2). The sample sizes ofthe remaining two
groups, tumblers and nonclassifieds, are in-
sufficient for similar analysis.
Rankings among the 34 species for their ex-
pected mean dispersal distances are provided
in Table I. The expected mean dispersal dis-
tances vary nearly nine-fold among the 34
species; using a windspeed of 1.75 m .sec' the
values range from 22-194 m (Table 1).
Analyses were performed to determine
whether rate ofdescent is correlated with shade
tolerance of seedlings and dry mass of seeds.
A parent tree with diaspores that have a slow
rate ofdescent spreads its diaspores over a wide
area and thus increases the probability that
some of its diaspores land in light-gaps where
survival may be enhanced particularly for
shade-intolerant species. Shade-tolerant species
are expected, in general, to have heavier seeds
than shade-intolerant species (Grime and Jef-
fery, 1965; Ng, 1978). Earlier studies of 18 of
these wind-dispersed species determined the
survival rate (slope of the regression of log
number seedlings surviving on time) of seed-
lings under sun and shaded forest conditions
(Augspurger, 1984). The survival rate in the
shade is used as the index to rank the shade
tolerance of the species (Table 1). The species
represent a continuum of shade tolerance and
do not fit a dichotomy of shade-tolerant vs.
shade-intolerant species. Wood density ofadult
trees ofthese 18 species (Table 1) is positively
correlated with their shade tolerance as seed-
lings; however, shade tolerance and dry mass
of seed are not significantly correlated (Aug-
spurger, 1984).
Rate ofdescent shows a weak, but significant,
rank correlation with the shade tolerance of
the 18 species (Table 4). The diaspores of the
floater group also show a significant rank cor-
relation between rate of descent and shade tol-
erance (Table 4). Rate of descent shows a sig-
nificant correlation (r = 0.70, P < 0.00 I) with
360 AMERICAN JOURNAL OF BOTANY
TABLE 4. Values of Kendall's Tau for rank correlations
of mean rate of descent with shade tolerance. Rank
correlations were performed first for all species for
which data were available and second for the three
aerodynamic groups with the most representatives
Shade tolerance
n dynamics certainly varies among the diverse
All species 18 0.372 a
Floaters 5 0.800a
Rolling autogyros 5 00400
Autogyros 5 0.333
a = P < 0.05.
the log of dry mass of seeds of the 34 species.
The correlation is lower (r = 0.54, P < 0.001)
when values for dry mass rather than log dry
mass are used. This result suggests that the
relation between rate of descent and dry mass
of seeds is curvilinear; however, a description
of that relationship is impossible because of
the limited number of species with high mass.
DISCUSSION - This study illustrates the wide
array of morphological design found among
species with wind-dispersed diaspores within
one tropical forest. It further documents the
wide range in dispersal potential that occurs
among them. The most abundant morpholog-
ical group represented is the rolling autogyro.
Evolutionary convergence onto that morphol-
ogy has occurred in six families. Evolutionary
divergence is illustrated by the Bombacaceae
with five species of floaters and one very dif-
ferent tumbler, and the Leguminosae repre-
sented by three autogyros, five rolling auto-
gyros, and one undulator. One aerodynamic
group conspicuously missing from the trees on
Barro Colorado Island is the true glider capable
of self-propelled forward momentum in still
air. Such a morphology occurs among the lia-
nas on the island. McCutcheon (1977) sug-
gested that the overall rarity of gliders can be
explained because of their relative inferiority
to autogyros and the rolling autogyros in tur-
bulent air. He predicted such unstable gliders
would be found in the still air of the tropical
rainforest. The three undulators of this study
occasionally glide forward for a portion oftheir
descent in still air, but more commonly they
undulate back and forth in short arcs and ac-
complish little, if any, forward motion. Bur-
rows (1975) pointed out that gliders ought to
undergo "fugoid oscillations" resulting in a
flight path in which the diaspore sweeps out
elongated L-shaped arcs during its descent.
These arcs are necessarily undirectional with
no back and forth motion. The three undula-
tors in this study may therefore represent a
new aerodynamic category.
[Vol. 73
Given the extreme morphological diversity
among the six groups, it is difficult to know
what part of the diaspore to consistently mea-
sure for area to determine (wing-loading) 1/2.
How the cross-sectional area measured reflects
the actual forces affecting air columns and aero-
groups. Further studies are needed to deter-
mine the aerodynamics of the different groups
and to ascertain which aspects of morphology
best predict the rate of descent.
The values for the wt/area ratio and the rate
of descent for diaspores of these 34 tropical
species do not deviate strongly from those of
earlier studies of temperate species. Despite a
range of six orders of magnitude for mass and
area, the 34 species converge on a relatively
restricted range of values for (wing-loading)"?
(37 to 262 millidynes-cm q-", The seven
r
species of temperate trees of Green's (1980)
study range from 70to 230 (millidynes-cm ")!",
although their values for mass and area cover
a smaller range than those in this study. The
rates of descent of the 34 species range from
39 to 198 em-sec:". These values are within
the same order of magnitude of other studies
of trees (Green, 1980; Guries and Nordheim,
1984). Slightly lower values occur in a few
species of Compositae with a pappus (Sheldon
and Burrows, 1973) and in some species of
Apocynaceae/Asclepidaceae with a parachute
to lower their rate of descent (Platt and Weis,
1977). The maximum rate of descent reported
among north temperate trees is 293 cm'sec l
for Chamaecyparis thyoides (Siggins, 1933). All
15 species of Eucalyptus studied by Cremer
(1977) exceed the values in this study; they
range from 210 to 554 em ·sec- I . None ofthose
species have any wings or dispersal appendages
attached to the small seed.
On a world-wide basis the wind-dispersed
diaspore with one of the most extreme values
appears to be the huge autogyro of Centrolo-
bium sp., a canopy tree growing in the Ama-
zonian lowland forests. Mean values for a sam-
ple from lowland Peru were: mass = 17.84 g,
area = 109.4 cm-, (wing-loading)"? = 407.1
(millidynes-cm ")!", and rate of descent =
367.6 em-sec>' (N = 14) (c. Augspurger, un-
pub. data). These values greatly exceed those
for any species in this study although the re-
lationship of rate of descent to wing-loading
for Centrolobium sp. is within the predicted
range based on regression analysis of the 34
species (Fig. 2).
The specific values for rate of descent re-
ported here are unlikely to be repeatable when
trees in other locations are studied. Siggins
(1933) found considerable intraspecific vari-
ation in rate ofdescent over a broad geographic
March, 1986] AUGSPURGER-WINO-DISPERSED TROPICAL DlASPORES
range of one species of Pinus. More locally on
Barro Colorado Island, significant variation in
wt and area ofsamaras and their rate ofdescent
occurs among trees ofone species in this study,
Platypodium elegans. Among 20 trees of this
species the mean values for (wing-loading)!"
range from 244 to 319 (millidynes- cm- 2) 112 and
mean values for rate ofdescent range from 178
to 236 em- sec:" (c. Augspurger, unpubl. data).
Therefore, species' values reported here pro-
vide only relative data for making very general
comparisons among species.
Knowledge of mass and area of diaspores
allows predictions of their rate of descent. The
correlation between (wing-loading)"? and rate
of descent is significant in this study when all
individual diaspores ofall aerodynamic groups
are combined. The correlation improves mark-
edly when the analysis is restricted to any given
aerodynamic group, indicating that the specific
relationship varies among aerodynamic groups.
Except for the helicopter group (r? = 0.51), the
coefficient of determination is high for each
group (r? ranges from 0.84 to 0.91). The im-
portance of identifying these strong relation-
ships lies principally in their predictive power.
Rate of descent in still air can be accurately
predicted without experimentation for other
species, provided their wt, area, and mor-
phology are known.
The regression analyses of rate of descent on
(wing-loading)"? indicate strong differences
among morphological groups in the slope de-
scribing that relationship (Fig. 2). The floaters
have the steepest slope (1.95). This group is
dominated by species of Bombacaceae; their
diaspore consists of a seed surrounded by a
sphere of silky fibers (kapok) that provides a
drag, but no lift, component to its aerodynamic
motion (M. Denny, pers. comm.). All other
groups have both lift and drag components.
The slope for the autogyros (0.60) differs sig-
nificantly from that of the rolling autogyros
(0.99). The rolling autogyros fall faster than
autogyros, but at progressively smaller amounts
as the (wing-loading)"? declines. These results
differ from those found for temperate trees
(Green, 1980). He found that rolling autogyros
and autogyros have almost identical slopes
(0.52 and 0.53, respectively), but different in-
tercepts, such that at a given (wing-loading)!",
a rolling autogyro falls faster than an auto-
gyro.
Similar evolutionary shifts in wing-loading
among morphological groups should have quite
different consequences for dispersal. Any com-
parable change in wing-loading resulting from
altering seed weight or diaspore area will have
a greater effect on rate of descent and dispersal
distance for groups with steeper slopes than for
361
groups with shallow slopes. Hence the potential
risk of losing dispersal distance while modi-
fying wing-loading is greatest for floaters;
alternatively, with the same reduction in wing-
loading floaters can gain more dispersal dis-
tance than other groups, although they still fall
faster and therefore disperse less far than other
groups.
Any increase in seed wt or in the size of
dispersal appendages that alters rate of descent
represents a cost to the parent. Some morpho-
logical designs are undoubtedly more costly to
build or alter than others. For example, pro-
duction ofmore fluffy kapok ofthe floaters may
require more or fewer resources than increasing
the woody wing of an auto-rotating samara,
depending on the wing's photosynthetic con-
tribution. In addition, for two diaspores with
comparable wing-loadings, but different areas,
a similar decrease in wing-loading achieved by
increasing area is more costly on an absolute
basis for the diaspore with the larger initial
area.
The narrow range in wing-loading values
among the 34 species, relative to the wide ranges
in wt and area, indicates a constraint on se-
lection acting on the relationship between wt
and area. Selection for increased seed wt ap-
parently has been accompanied by selection
for increased area of dispersal appendages in
order to diminish any loss of dispersal poten-
tial. If a plant is resource limited, selection
acting to change dispersal potential is inextri-
cably linked to the number of diaspores that
can be made. No data are available for inter-
specific comparisons in the size of fruit crops
of the study species. Species with larger dia-
spore areas and/or heavier seeds are predicted
to have smaller crop sizes.
Knowledge of rate of descent, ht of release,
and mean windspeed allows predictions of ex-
pected mean dispersal distance for each species.
Despite the narrow range among species in
wing-loading, relative to their range in wt and
in area, the species still have a nine-fold range
in predicted mean dispersal distance; they range
from 22 to 194 m at a mean windspeed of 1.75
m- sec-I. Actual dispersal distance achieved in
the field will be affected by a diversity of fac-
tors, e.g., windspeed required for abscission,
turbulence and variability in windspeed during
descent, and the surrounding vegetation. The
expected mean distances are not likely to be
accurate predictions. However, such values are
useful in providing a general idea of how the
various species rank in their dispersal potential
and in illustrating the tremendous range in dis-
persal potential that exists within one tropical
forest. Ultimately values for rate of descent
can be incorporated into a model that predicts
362 AMERICAN JOURNAL OF BOT ANY
the spatial distribution ofwind-dispersed seeds,
such as that developed by Fields and Sharpe
(1980). Before such a model can be applied to
these tropical trees, additional measurements
will be required of the variability in rate of
descent within a species (and within individual
trees) and ofthe wind environment of tropical
forests.
The predicted mean dispersal distance rep-
resents a one-dimensional value, but seeds are
dispersed over a two-dimensional area. Be-
cause the area ofa circle increases by the square
of its radius, any difference among species in
mean distance will result in even greater dif-
ferences among species in their area over which
seeds are dispersed. Selection on morpholog-
ical traits that results in only a small increase
in mean dispersal distance should have a sub-
stantially greater effect on the area ofa parent's
seed distribution. This study places emphasis
on mean dispersal distances. Anecdotal ob-
servations suggest extreme values in distance
are sometimes achieved by wind-dispersed di-
aspores (Whitmore, 1984; C. Augspurger, pers.
obs.). What is unknown is the relative impor-
tance of mean versus extreme values of dis-
persal distance for successful seedling estab-
lishment.
Given these background data, the next step
is to determine whether these rates of descent
and expected dispersal distances are realized
under field conditions. Previous studies con-
firm the general predictability of dispersal dis-
tance under field conditions from lab results
of rate of descent in still air. A field study of
Lonchocarpus pentaphyllus, one of the species
in the present study, demonstrated that fruits
with 0, 1, 2, and 3 seeds have increasing values
for (wing-loading) 1/2 and rate of descent in still
air. Actual dispersal distance declined among
the four fruit types with an increase in rate of
descent (Augspurger and Hogan, 1983). For
herbs Platt and Weis (1977) found a negative
correlation between rate of descent in still air
and dispersal distance under field conditions.
In contrast, Rabinowitz and Rapp (1981) found
no such relationship for grass species and
stressed the importance of specific morpho-
logical features in affecting dispersal. For seeds
of Melaleuca trees, Woodall (1982) found the
dispersal distance he predicted to be shorter
than actual distance by a factor of two. By
adding a correction factor for wind turbulence,
he achieved a reasonable prediction of actual
distance based on knowledge of rate ofdescent
and wind conditions. Studies on Barro Colo-
rado Island are currently underway to docu-
ment the seed distributions of many of the
[Vol. 73
study species to test the validity of the predic-
tions from this study.
The dispersal potential of a species has im-
portant implications for other aspects of its
biology, e.g., size of fruit crop, traits affecting
seedling survival and establishment, the spatial
relationship of juveniles and adults, and the
genetic structure ofthe population due to gene
flow via seed dispersal. Given the differences
among the species in their dispersal potential,
they are expected to also differ widely in ad-
ditional aspects affected by their dispersal bi-
ology.
Species with a more restricted dispersal dis-
tance and area must have other life history
characters that allow seedling establishment
nearer the parent. Factors affecting such es-
tablishment include shade tolerance and resis-
tance to mortality factors that act in a density-
and/or distance-dependent manner, e.g., seed/
seedling pathogens, seed predators, seedling
herbivores, and seedling competitors. This
study shows that rate of descent correlates
weakly with two traits affecting seedling estab-
lishment, viz. log ofdry mass ofseed and shade
tolerance of seedlings. It appears that species
requiring light-gaps as seedlings have lighter
seed mass and/or larger area of their diaspores
so as to lower their rate of descent. Airborne
for a longer time, they potentially disperse over
a broader area and increase the chance that
some of parent's diaspores land in light-gaps.
Dispersal potential also has implications for
the genetic structure and speed of movement
ofnew alleles through a population. With great-
er dispersal potential, offspring of one parent
are more likely to encounter those of other
parents. This overlapping of seed distributions
enhances the probability of outbreeding and
increases the movement of alleles in the pop-
ulation.
In summary, considerable morphological
variation exists among the tropical tree species
with wind-dispersed diaspores on Barro Col-
orado Island, Panama. Knowledge of their
mass, area, and wing-loading values allows ac-
curate predictions oftheir rates ofdescent, par-
ticularly within a given morphological group.
The species differ greatly in their dispersal po-
tential. These differences are likely to have im-
portant implications for their fruit crop size,
population demography and spatial patterns of
surviving seedlings, and the overall genetic
structure ofthe population. The study indicates
the danger inherent in assuming that knowl-
edge of the general dispersal mode ofa species,
e.g., wind dispersal, is sufficient to predict other
aspects of its biology. Generalizations about
March, 1986] AUGSPURGER - WIND-DISPERSED TROPICAL DIASPORES
species that use wind dispersal in tropical for-
ests are unlikely to be valid.
LITERATURE CITED
AUGSPURGER, C. K. 1984. Light requirements of neo-
tropical tree seedlings: a comparative study ofgrowth
and survival. J. Ecol. 72: 777-795.
- - - , AND K. P. HOGAN. 1983. Wind dispersal offruits
with variable seed number in a tropical tree (Lon-
chocarpus pentaphyllus: Leguminosae). Amer. J. Bot.
70: 1031-1037.
BURROWS, F. M. 1975. Wind-borne seed and fruit move-
ment. New Phytol. 75: 405-418.
CHUDNOFF, M. 1979. Tropical timbers ofthe world. U.S.
Forest Products Lab., U.S. Dept. ofAgriculture, Mad-
ison, Wise.
CREMER, K. W. 1977. Distance of seed dispersal in eu-
calypts estimated from seed weights. Austral. For.
Res. 7: 225-228.
CROAT, T. B. 1978. Flora of Barro Colorado Island. Stan-
ford Univ. Press, Stanford, Calif.
FIELDS, D. E., AND D. M. SHARPE. 1980. SEDFAL: a
model of dispersal of tree seeds by wind. 0 RNL Re-
port EDFB/IBP-78/2. Oak Ridge Natl. Laboratory,
Oak Ridge, TN.
FOSTER, R. B., AND N. V. L. BROKAW. 1982. Structure
and history of the vegetation on Barro Colorado Is-
land. In E. G. Leigh, Jr., A. S. Rand, and D. M.
Windsor [eds.], The ecology of a tropical forest, pp.
67-82. Smithsonian Institution Press, Washington,
D.C.
GREEN, D. S. 1980. The terminal velocity and dispersal
of spinning samaras. Amer. J. Bot. 67: 1218-1224.
GRIME, J. P., AND D. W. JEFFREY. 1965. Seedling estab-
lishment in vertical gradients of sunlight. J. Ecol. 53:
621-642.
GURIES, R. P., AND E. V. NORDHEIM. 1984. Flight char-
363
acteristics and dispersal potential of maple samaras.
Forest Sci. 30: 434-440.
MCCUTCHEN, C. W. 1977. The spinning rotation of ash
and tulip tree samaras. Science 197: 691-692.
NG, F. S. P. 1978. Strategies ofestablishment in Malayan
forest trees. In P. B. Tomlinson and M. H. Zimmer-
mann [eds.], Tropical trees as living systems, pp. 129-
163. Cambridge Univ. Press, Cambridge.
PLATT, W. J., AND J. M. WEIS. 1977. Resource parti-
tioning and competition within a guild of fugitive
prairie plants. Amer. Nat. III: 479-513.
RABINOWITZ, D., AND J. K. RAPP. 1981. Dispersal abil-
ities of seven sparse and common grasses from a Mis-
souri prairie. Amer. J. Bot. 68: 616-624.
RIDLEY, H. N. 1930. The dispersal of plants throughout
the world. L. Reeve and Co., Ashford, England.
SHARPE, D. M., AND D. E. FIELDS. 1982. Integrating the
effects of climate and seed fall velocities on seed dis-
persal by wind: a model and application. Ecol. Model.
17: 297-310.
SHELDON, J. c., AND F. M. BURROWS. 1973. The dispersal
effectiveness of the achene-pappus units of selected
Compositae in steady winds with convection. New
Phytol. 72: 665-675.
SIGGINS, H. W. 1933. Distribution and rate of fall of
conifer seeds. J. Agr. Res. 47: 119-128.
VOGEL, S. 1981. Life in moving fluids. Willard Grant
Press, Boston.
WERNER, P. A., AND W. J. PLATT. 1976. Ecological re-
lationships of co-occurring goldenrods (Solidago:
Compositae). Amer. Nat. 110: 959-971.
WHITMORE, T. C. 1984. Tropical rain forests of the Far
East. Oxford Univ. Press, Oxford.
WOODALL, S. L. 1982. Seed dispersal in Melaleuca quin-
quenervia. Florida Sci. 45: 81-93.
ZAR, J. H. 1974. Biostatistical analysis. Prentice-Hall,
Englewood Cliffs, N.J.