Extinction at the end-Cretaceous and the origin of modern Neotropical rainforests

Abstract

The birth of modern rainforests
The origin of modern rainforests can be traced to the aftermath of the bolide impact at the end of the Cretaceous. Carvalho et al. used fossilized pollen and leaves to characterize the changes that took place in northern South American forests at this time (see the Perspective by Jacobs and Currano). They not only found changes in species composition but were also able to infer changes in forest structure. Extinctions were widespread, especially among gymnosperms. Angiosperm taxa came to dominate the forests over the 6 million years of recovery, when the flora began to resemble that of modern lowland neotropical forest. The leaf data also imply that the forest canopy evolved from relatively open to closed and layered, leading to increased vertical stratification and a greater diversity of plant growth forms.
Science , this issue p. 63 ; see also p. 28

Full text

PAL EO EC OLO GY
Extinction at the end-Cretaceous and the origin of
modern Neotropical rainforests
Mónica R. Carvalho
1,2
*, Carlos Jaramillo
1,3,4
*†, Felipe de la Parra
5
, Dayenari Caballero-Rodríguez
1
,
Fabiany Herrera
1,6
, Scott Wing
7
, Benjamin L. Turner
1,8
, Carlos D’Apolito
1,9
, Millerlandy Romero-Báez
1,10
,
Paula Narváez
1,11
, Camila Martínez
1
, Mauricio Gutierrez
1,12
, Conrad Labandeira
7,13,14
, German Bayona
15
,
Milton Rueda
16
, Manuel Paez-Reyes
1,17
, Dairon Cárdenas
18
, Álvaro Duque
19
, James L. Crowley
20
,
Carlos Santos
21
, Daniele Silvestro
22,23
The end-Cretaceous event was catastrophic for terrestrial communities worldwide, yet its long-lasting
effect on tropical forests remains largely unknown. We quantified plant extinction and ecological change
in tropical forests resulting from the end-Cretaceous event using fossil pollen (>50,000 occurrences)
and leaves (>6000 specimens) from localities in Colombia. Late Cretaceous (Maastrichtian) rainforests
were characterized by an open canopy and diverse plant–insect interactions. Plant diversity declined
by 45% at the Cretaceous–Paleogene boundary and did not recover for ~6 million years. Paleocene forests
resembled modern Neotropical rainforests, with a closed canopy and multistratal structure dominated
by angiosperms. The end-Cretaceous event triggered a long interval of low plant diversity in the Neotropics
and the evolutionary assembly of today’s most diverse terrestrial ecosystem.
Paleontological evidence indicates that the
bolide impact at Chicxulub, 66.02 million
years ago (Ma) (1), had immediate cata-
strophic effects on plant communities and
reshaped terrestrial ecosystems world-
wide (2–4). Despite the extent of this ecolog-
ical disruption, the long-term extinction and
recovery patterns were geographically hetero-
geneous (5). As much as 90% of pre-extinction
palynomorphs reappeared during the Danian
(66 to 61.6 Ma) in Patagonia and New Zealand
(6,7), and species-rich Danian megafloral as-
semblages with diverse types of insect damage
indicate rapid recovery of diversity in Patagonia
(8,9). By contrast, palynofloral extinction was
up to 30% in the Northern Great Plains of
North America (10), and floral and insect-
damage diversity may not have reached pre-
extinction levels until the latest Paleocene or
early Eocene [(11,12); but see (3)].
Phylogenies of several plant lineages suggest
that the Cretaceous–Paleogene (K/Pg) event
marking the end of the Cretaceous played a
role in shaping modern tropical lowland rain-
forests (13–15), but the fate of tropical forests
following the K/Pg boundary is not well under-
stood. Assessing plant extinction and recovery
requires a thoroughly sampled fossil record, yet
aside from an impact-related fern-spore spike
in deep-water strata from Gorgonilla, Colombia
(16), the plant fossil record across the K/Pg
boundary in the lowland Neotropics is sparse
(17).Here,wequantifychangesinthediversity,
structure and composition of forests across
the K/Pg boundary in tropical South America
using a palynological dataset spanning the
Maastrichtian–Paleocene interval, including
39 stratigraphic sections from outcrops and
wells, 637 samples, 1048 taxa, and 53,029 occur-
rences (Fig. 1 and table S1) (18). As fossil pollen
assemblages typically integrate information
at large spatial scale (i.e., tens of square kilo-
meters), we also examined the composition
and diversity of autochthonous assemblages of
leaf fossils, which instead reflect local forest
communities. These included 2053 fossils from
the Maastrichtian Guaduas Formation and
4898 fossils from the middle-late Paleocene
Bogotá and Cerrejón formations (19). Situated
near the paleo-equator, this then-coastal re-
gion of northern South America was wet and
megathermal throughout the globally warm
Maastrichtian and Paleocene. As a result, the
effect of the end-Cretaceous event on the fossil
record is not confounded by major changes in
climate.
Extinction and turnover of tropical vegetation
We estimated diversity using the corrected
sampled-in-bin diversity (20), the shareholder
quorum subsampling (SQS) (21), origination and
RESEARCH
Carvalho et al., Science 372,63–68 (2021) 2 April 2021 1of6
1
Smithsonian Tropical Research Institute, Panama.
2
Grupo de
Investigación Paleontología Neotropical Tradicional y Molecular
(PaleoNeo), Facultad de Ciencias Naturales y Matem áticas,
Universidad del Rosario, Bogotá, Colombia.
3
ISEM, U.
Montpellier, CNRS, EPHE, IRD, Montpellier, France.
4
Department
of Geology, Faculty of Sciences, University of Salamanca,
Salamanca, Spain.
5
Instituto Colombiano del Petróleo,
Bucaramanga, Colombia.
6
Negaunee Institute for Plant
Conservation, Chicago Botanic Garden, Chicago, IL, USA.
7
Department of Paleobiology, National Museum of Natural
History, Washington, DC, USA.
8
Soil and Water Science
Departm ent, Univers ity of Florida, G ainesville, FL, USA.
9
Faculdade de Geociências, Universidade Federal de Mato
Grosso, Cuiabá, Brazil.
10
ExxonMobil Corporation, Spring, TX,
USA.
11
Instituto Argentino de Nivología, Glaciología y Ciencias
Ambientales, CCT-CONICET, Mendoza, Argentina.
12
Departamento de Geología, Universidad de Chile, Santiago,
Chile.
13
Department of Entomology, University of Maryland,
College Park, MD, USA.
14
College of Life Sciences, Capital
Normal University, Beijing, China.
15
Corporación Geológica Ares,
Bogotá, Colombia.
16
Paleoflora Ltda, Zapatoca, Colombia.
17
Department of Earth and Atmospheric Sciences, University of
Houston, Houston, TX, USA.
18
Instituto Amazónico de
Investigaciones Científicas SINCHI, Leticia, Colombia.
19
Departamento de Ciencias Forestales, Universidad Nacional de
Colombia, Medellín, Colombia.
20
Department of Geosciences,
Boise State University, Boise, ID, USA.
21
BP Exploration
Operating Company Limited, Chertsey Road, Sunbury-on-
Thames, Middlesex, UK.
22
Department of Biology, University of
Fribourg, Fribourg, Switzerland.
23
Department of Biological and
Environmental Sciences, University of Gothenburg and
Gothenburg Global Biodiversity Centre, Gothenburg, Sweden.
*These authors contributed equally to this work.
†Corresponding author. Email: jaramilloc@si.edu
N
1000 Km
10˚N
0˚
20˚N
90˚N 80˚N 70˚N
Chicxulub
crater
Caribbean Sea
Pacific Ocean
Accreted oceanic terrane
Continental marginal uplifts
500 Km
G1
P1F
DK
C1
G2
M
G1
CS
TS
R
B15
RC1
B3 LM1
F3
T2
A2
A3
A1
A4
GI
ULE 1
R1
C1
RL
G3
Z1
M3
G1
T182
R14
L3
V2
D1E
N
Shoreline
Palynological site
Macrofossil site
Continental, marginal plains, swamps
Shallow marine
/A1
Legend
Equator
AB
Fig. 1. Location of stratigraphic sections and macrofossil localities in northern South America.
(A) Map showing modern-day distance to Chicxulub crater. (B) Paleogeographic reconstruction of northern
South America [area delimited by dotted rectangle in (A)] during the late Maastrichtian, based on (64).
on April 1, 2021 http://science.sciencemag.org/Downloaded from

extinction rates using the second-for-third
method (22), and PyRate (23)[seematerials
and methods (18)]. Palynofloral diversity was
higher in the Maastrichtian (72 to 66 Ma) in
tropical South America than in the early and
middle Paleocene (66 to 60 Ma) (Fig. 2B; mean
of Maastrichtian bins 172.3 versus Paleocene
bins 84.1, ttest, df = 10.9, P<0.001;tableS3),
regardless of differences in sampling size
(Fig. 2B, SQS estimates 27 versus 12.7, ttest,
df = 3872.7, P< 0.001) or depositional envi-
ronments (table S4). This marked decrease in
diversity coincides with a peak in extinction
rates at 66 Ma (66 to 66.5 age bin; log Bayes
factors >6 with a 95% credible interval be-
tween 66.4 and 65.7) that diminishes palyno-
morph diversity by 45% and significantly exceeds
Maastrichtian or Paleocene background extinc-
tion (Fig. 2C, extinction rate of 0.44 versus a
mean of 0.04 for all other bins, SQS esti-
mates 0.53 versus 0.03, ttest, df = 309.08, P<
0.001; PyRate extinction rate 0.75; credible
interval (CI): 0.45 to 1 versus a median rate of
0.07; CI: 0.04 to 0.09 in the Maastrichtian and
0.05; CI 0.03 to 0.07 in the early Paleocene). As
a result, most Maastrichtian cohorts (groups of
palynomorphs that coexist at a given time)
decline in the first bin of the Paleocene (65 Ma
bin), well above the mean cohort reduction
obser ved throughout the Paleocene (Fig.
2E, mean slope of all cohorts at 65 bin 0.24
versus mean slope of all other cohorts, 0.05,
ttest, df = 13, P< 0.001).
Following the K/Pg boundary, palynomorph
diversity did not recover to pre-extinction
levels until after 60 Ma (Fig. 2B) and further
increased beyond pre-extinction levels through-
out the Paleocene–Eocene Thermal Maximum
and early Eocene (24,25). The second-for-third
estimates identify a peak in origination during
the 59- to 59.5-Ma interval (Fig. 2D; mean
origination rate 0.38, mean at all other inter-
vals 0.09; SQS estimates 0.95, 0.28 versus 0.08,
ttest, df = 332.54, P< 0.001), whereas PyRate
found support (log Bayes factors >6) for a drop
in origination in the earliest Paleocene (from
0.23; CI: 0.2 to 0.27 to 0.04; CI: 0.01 to 0.08)
and a strong increase between 60.7 and 60.2 Ma
(rate 0.37; CI: 0.28 to 0.47). A reanalysis of the
data allowing the PyRate algorithm to search
for rate shifts at a higher temporal resolution
resulted in similar patterns of origination and
extinction overall (fig. S3). However, the analy-
sis detected an additional brief but strong peak
in origination rates between 59.6 and 59.2 Ma,
when the origination rates increased from 0.26
(CI: 0.18 to 0.38) to 1.30 (CI: 0.90 to 1.72).
We used detrended correspondence analysis
(DCA) and cluster analysis to evaluate changes
in palynofloral composition across the K/Pg
boundary. Rapid change through time in the
first axis scores of samples (Fig. 2F, first axis
explains 57% of variation) and a distinct clus-
tering of Maastrichtian and Paleocene plant
communities (Fig. 2G and fig. S1) reflect a
major and permanent change in floristic com-
position. Although the Maastrichtian contained
roughly equal proportions of angiosperm
(47.9%) and spore grains (49.5%), angiosperm
grains dominated in the Paleocene (mean abun-
dance 84% versus 16% of non-angiosperms,
Wilcoxon test, W-statistic = 14,552, P< 0.001; fig.
S2). Gymnosperms (mostly Araucariaceae) are
2.5% of Maastrichtian grains but only 0.4%
of Paleocene grains (Mann-Whitney test, U-
statistic = 17,509, P< 0.01). Gymnosperms also
occur in 75% of Maastrichtian samples but
only in 24% of Paleocene samples having >100
grains. Sediments from the Maastrichtian
Umir Formation (central Colombia) are rich
in gymnosperm lipid biomarkers (26), sup-
porting the abundance of gymnosperms prior
to the K/Pg extinction. Living species of
Araucariaceae occur as large trees and are
often underrepresented in the palynologi-
cal soil record and do not disperse long dis-
tances (27), such that their low abundance in
Maastrichtian deposits is likely to be an under-
estimation of their true abundance.
Leaf physiognomy and forest types
We recognize 41 angiosperm and 4 fern
morphotypes in the Maastrichtian Guaduas
macroflora. In the Paleocene, we found 46
angiosperms and 2 ferns in the Bogotá flora
and 58 angiosperms, 5 ferns, and 1 conifer
leaf morphotype in the Cerrejón flora. The
foliar physiognomy of nonmonocot angiosperm
leaves (ANA-grade angiosperms: Amborellales,
Nymphales, and Austrobaileyales; magnoliids;
and eudicots) in both the Maastrichtian and
Paleocene assemblages resembles that of mod-
ern tropical rainforests, characterized by leaves
of large size, untoothed margins, and elongated
drip tips (Fig. 3). Of the 36 species of non-
Carvalho et al., Science 372,63–68 (2021) 2 April 2021 2of6
Age (Ma)
72
71
70
69
68
67
66
65
64
63
62
61
60
59
58
A
Pollen/Spores Taxa 010 30 50
0 100 200 300
B
Morphospecies
0 0.5 1.0 1.5
C
Alroy's
Total
Subsampled
00.51
PyRate
Mean
CI 95%
Extinction rates
012
D
01
Origination rates
E
0 −0.1 −0.2
Slopes
−1.0 0.0 1.0
F
DCA Axis 1
71
70
69
68
67.5
67
66.5
66
65
63.5
62
61
60
59.5
59
58.5
58
G
Cluster
H
0 0.5
71
70
69
68
67.5
67
66.5
66
65
63.5
62
61
60
59.5
59
58.5
58
Angiosperm
proportion
PyRate
Mean
CI 95%
1
LADs
FADs
b
b
Alroy's
Total
Subsampled
0.5 1.5
Time
Bins
Total
Subsampled
Fig. 2. Changes in diversity and composition of Maastrichtian-Paleocene
palynofloras in northern South America. (A) Stratigraphic ranges of taxa across
the Maastrichtian-Paleocene interval. Shown in green are the taxa that became
extinct and in orange, the taxa that originated during thistime period. (B)Corrected
sampled-in-bin diversity. (C)PyRate(23) extinction rate mean and 95% credible
interval (orange shadow) and Alroy’s second-for-third (22)extinctionrate.
(D) PyRate origination rate mean and 95% credible interval (blue shadow), and
Alroy’s second-for-third origination rates with 0.95 confidence interval (SQS = 0.95;
gray shadow). (E) Boxplot of slopes from the survivorship analysis performed
on 1-million-year bin cohorts. (F) Change in floral composition shown by scores
of samples on DCA axis 1 plotted against time. (G) Sørensen Cluster showing
two distinct clusters, Maastrichtian (green) and Paleocene (orange); see fig. S1 for
individual samples cluster. (H) Boxplot of the proportion per bin of angiosperm
grains versus total flora; see fig. S2 for proportion of individual samples.
RESEARCH |RESEARCH ARTICLE
on April 1, 2021 http://science.sciencemag.org/Downloaded from

monocots in the Guaduas flora, 89% have
leaves larger than 45 cm
2
(mesophylls), 81%
have untoothed margins, and 11 of the 25 species
with preserved apices have drip tips (44%). In
the Paleocene assemblages, 63 and 76% of
nonmonocot species haveuntoothedmargins
(Bogotá and Cerrejón, respectively), and 30
to 35% have elongated drip tips. Estimates of
mean annual rainfall based on Leaf Area Analy-
sis (18,28,29)indicateannualprecipitation
of 234 to 293 cm year
−1
for the Guaduas flora,
182 to 184 cm year
−1
for the Bogotá flora, and
240 to 308 cm year
−1
for the Cerrejón flora
(Table 1 and table S8).
Leaf mass per area (LMA) values, estimated
on the basis of the scaling relationship between
leaf mass and petiole diameter observed in
living plants (30), were consistent with modern
Carvalho et al., Science 372,63–68 (2021) 2 April 2021 3of6
Fig. 3. Representative leaf taxa. (A to K) Taxa from Paleocene Bogotá and
(L to W) Maastrichtian Guaduas floras. (A) Menispermaceae (BF6). (B) Salicaceae
(BF5) with midrib gall. (C) Fabaceae leaflet (BF38) with surface feeding
damage. (D) Euphorbiaceae (BF37) with hole and margin feeding. (E) Fabaceae,
Caesalpinioideae (BF21). (F) Water fern, Salvinia bogotensis, Salviniaceae
(BF22). (G)Malvaciphyllum sp. Malvaceae (BF4). (H) Example of drip tip in
Salicaceae (BF23). (I) aff. Eleaocarpaceae (BF13). (J) Fabaceae leaflet
(BF21, 5 mm) with hole feeding damage. (K) Arecaceae (BF27). (L) Arecaceae
(GD47, 10 cm). (M) aff. Lauraceae (GD54). (N) aff. Hamamelidaceae (GD56).
(Oand P) Fertile and sterile fragments of Polypodiaceae (GD22). (Q) aff.
Salicaceae (GD6). (R) Lauraceae (GD7) with drip tip. (S) aff. Urticaceae
(GD52). (T) Zingiberales (GD46, 5 cm). (U) aff. Cucurbitaceae (GD8).
(V)Bernhamniphyllum sp. Rhamnaceae (GD1). (W) aff. Dilleniaceae (GD3).
Scale bars: 1 cm except where noted in parentheses after taxon.
RESEARCH |RESEARCH ARTICLE
on April 1, 2021 http://science.sciencemag.org/Downloaded from

evergreen rainforest environments across all
three floras (Guaduas: 36 to 206 g m
−2
; Bogotá:
52 to 206 g m
−2
; Cerrejón: 44 to 126 g m
−2
), yet
LMAs of the Guaduas and Bogotá floras are
lower than those of Cerrejón (ttest, P< 0.001,
tables S9 and S10) (18). Evergreen trees tend
to have higher LMAs when living under drier
climates (31), which is consistent with the
relatively lower precipitation of the Bogotá
flora compared with Cerrejón. The Guaduas
and Cerrejón floras had similar precipitation
(>200 cm year
−1
), so it is possible that the
higher Guaduas LMA may reflect a higher
irradiance related to canopy structure (see
below) or poorer soils (31).
A notable feature of the Paleocene Cerrejón
flora is its resemblance to modern Neotropical
rainforests in terms of family-level composi-
tion of angiosperms (19). To examine this, we
compared the natural affinities of leaf taxa in
the Guaduas with those at Bogotá, Cerrejón,
and modern Neotropical rainforests. Some
Maastrichtian angiosperms have confirmed or
tentative affinities to families that are widely
distributed in (but not necessarily restricted to)
the lowland tropics, including Lauraceae (t wo
or three morphotypes), Araceae (two morpho-
types), Theaceae (one or two morphotypes),
Arecaceae, Rhamnaceae (32), Piperaceae
(33), Salicaceae, Canellaceae, Dilleniaceae,
Urticaceae, and Monimiaceae (one morpho-
type each), among others (Fig. 4A, table S6,
and data S4). The flowering plants of the
Paleocene Bogotá flora closely resemble those
of the Cerrejón flora (19) and include the dom-
inant tree families in modern Neotropical rain-
forests. The Bogotá flora has two leaflet types of
Fabaceae, one of these representing the earliest
record of Caesalpinioideae (with abundant
legume pods) (34), Euphorbiaceae, Lauraceae,
Salicaceae, Violaceae (two morphotypes each),
Malvaceae, Melastomataceae (35), Rhamnaceae,
Arecaceae, Eleaocarpaceae, and Araceae (one
morphotype each; table S7 and data S5).
Fossil seeds of Annonaceae, Icacinaceae,
Menispermaceae, and Passifloraceae are also
present in the Bogotá flora. Because nearly
autochthonous leaf assemblages reflect tree
biomass as a combination of stem abundance
and stem diameter (36), we compared the
family-level composition of five unbiased cen-
sus sites (two from Guaduas, one from Bogotá,
and two from Cerrejón) with permanent plots
in two living Neotropical rainforests: Barro
Colorado Island (BCI), Panama (37), and
Amacayacu, Colombia (38). In living tropical
rainforests, samples of leaf litter that are an al-
ogous to single fossil quarry sites can represent
most of the standing vegetation (90% biomass)
in a 12.5-m radius (36). The Paleocene census
sites are more similar in family composition
to the living forest at BCI (Fig. 4B; Wilcoxon
test, W = 46882, P< 0.001) and Amacayacu
Carvalho et al., Science 372,63–68 (2021) 2 April 2021 4of6
0
5
10
15
20
Fabaceae
Arecaceae
Rubiaceae
Moraceae
Bignoniaceae
Lauraceae
Myristicaceae
Annonaceae
Euphorbiaceae
Meliaceae
Melastomataceae
Sapindaceae
Malvaceae
Sapotaceae
Burseraceae
Lecythidaceae
Salicaceae
Clusiaceae
Myrtaceae
Chrysobalanaceae
Violaceae
Monimiaceae
Nyctaginaceae
Celastraceae
Apocynaceae
Mean percent of plot
species in family
A
0.00
0.25
0.50
0.75
1.00
Amacayacu BCI
CDI
Within living plots
Maastrichtian vs. Living
Paleocene vs. Living
B
Guaduas
Bogotá
Cerrejón
Guaduas
Bogotá
Cerrejón
0
20
40
60
80
Total
Damage
Specialized
Damage
DT Frequency (%)
C
Guaduas
Bogotá
Cerrejón
Guaduas
Bogotá
Cerrejón
0
20
40
60
DT Richness
D
Bogotá Cerrejón
0510 0510
0510 0510
Guaduas
0510
0510
0.00
0.25
0.50
0.75
1.00
Probability Density
Frequency
CDI
E
Pg
K
Total
Damage
Specialized
Damage
Families present at Paleocene sites
Families present at Maastrichtian sites
Fig. 4. Forest composition and insect-feeding damage of fossil floras. (A) Percentage of tree species in
72 extant Neotropical forest plots that belong to the 25 plant families that together account for 75% of
diversity. Half of the stems belong to the 10 families shown as dark bars. Orange circles indicate families
present in the Bogotá or Cerrejón floras (Paleocene), and green diamonds indicate families present in the
Maastrichtian Guaduas flora. (B) Density plot of dissimilarity in family composition between fossil
assemblages and samples of living Neotropical forests (see materials and methods for details). Chao-
Sørensen dissimilarity (CDI) was calculated between randomly selected subregions of the 50-ha plot at Barro
Colorado Island (Panama) and the 25-ha plot at Amacayacu (Colombia). Gray areas depict the distribution of
dissimilarities of the randomly selected subregions within each site. (C) Average frequency of damaged
leaves in 400 randomly selected leaves from each fossil flora. (D) Richness of total and specialized
insect-mediated damage types, rarefied to 95 and 90% sample coverage, respectively. Gray lines indicate
95% confidence intervals. (E) Histogram of leaf damage beta-diversity across host plant species with
more than 20 leaves at the Guaduas, Bogotá, and Cerrejón floras. Pairwise beta-diversity was quantified using
CDI and is depicted in solid bars. The blue curve indicates the probability density for the null expectation that
the observed DTs are randomly distributed across host plant species.
Table 1. Leaf physiognomy and precipitation of the Maastrichtian-Paleocene floras. Numbers in parentheses indicate numbers of quarries (Total
specimens), number of census localities (Census), and number of morphotypes with preserved apices (drip tips). MAP, mean annual precipitation.
Formation Age Total
specimens
Census
numbers Leaf taxa Nonmonocot
taxa
Non-monocots
with entire margins
Nonmonocots
with drip tips
Leaves mesophylls
or larger
MAP
(cm year
−1
)
Guaduas Maastrichtian 2053 (12) 1650 (2) 45 36 29 (81%) 11 (25) 32 (89%) 234–293
.................................... ....................................................... ..................................................... ....................................................... ....................................................... ........................................................ ......................
Bogotá Paleocene 2416 (19) 1370 (1) 48 40 25 (63%) 6 (20) 25 (63%) 182–184
.................................... ....................................................... ..................................................... ....................................................... ....................................................... ........................................................ ......................
Cerrejón* Paleocene 2482 (18) 1190 (2) 65 46 35 (76%) 12 (34) 44 (68%) 240–304
.................................... ....................................................... ..................................................... ....................................................... ....................................................... ........................................................ ......................
*Data reported by (19).
RESEARCH |RESEARCH ARTICLE
on April 1, 2021 http://science.sciencemag.org/Downloaded from

(Wilcoxon test, W = 7806, P< 0.001) than they
are to the Maastrichtian census sites (fig. S5).
Canopy structure is reflected in the distri-
bution of leaf vein length per area (VLA) and
stable carbon isotope ratios (d
13
C) within in-
dividual taxa (39,40). Most nonmonocots
from Guaduas have relatively high VLA values
(39,41), yet the unimodal distribution of VLA
within single taxa in the Guaduas flora (39)
and the low range of d
13
C measured in leaf
cuticles (40) suggest that these forests did not
have the range of light environments seen in
modern multistratal rainforests. By contrast,
leaves of the Paleocene Cerrejón flora show
the same bimodal distribution of single-taxon
VLA and the wide range of cuticle d
13
Cob-
served in modern closed canopy, multistratal
forests (39,40). Maastrichtian wet tropical for-
ests, therefore, likely had an open canopy that
promoted mixing of respired and atmospheric
CO
2
and a small light gradient between the
understory and the canopy compared to mod-
ern Neotropical forests. These open canopy
forests may have recycled less rainfall through
transpiration than their multistratal Paleocene
equivalents, potentially influencing regional
and global climate (42).
Diversity of plant-insect interactions
The diversity of insect-feeding damage on
leaves reflects the richness of insect herbi-
vores (43). We quantified insect damage in the
Guaduas and Bogotá floras following a stan-
dard damage type (DT) system and compared
it with damage from the Paleocene Cerrejón
flora (18). Over 50% of leaves in all three floras
show insect herbivory (Fig. 4C), indicating in-
tense biotic interactions in both Maastrichtian
and Paleocene forests. The richness of insect
DTs in the Guaduas flora is comparable to that
at Bogotá and greater than at Cerrejón, both
for total DT richness resampled at 95% cov-
erage (Guaduas versus Bogotá: 46.5 versus
47.8 DTs, ttest one-tailed, t= 0.325, df = 579,
P= 0.745; Guaduas versus Cerrejón: 46.5 versus
16.0 DTs, ttest one-tailed, t=6.23,df=483,P<
0.001) and for specialized damage only (at 90%
coverage: Guaduas versus Bogotá: 30.53 versus
29.48 DTs, ttest, t= 0.20, df = 802, P= 0.841;
Guaduas versus Cerrejón: 30.53 versus 11.43 DTs,
ttest, t=5.09,df=1121,P<0.001;Fig.4D).
Because insect-feeding damage reflects in-
flicting herbivores, DT beta-diversity across
host species provides evidence of host speci-
ficity among insect herbivore communities.
Leaf damage beta-diversity across host taxa
in the Maastrichtian Guaduas flora is higher
than expected by chance (Wilcoxon test, W =
41615, P< 0.001) and higher than that observed
at either Bogotá or Cerrejón (Guaduas versus
Bogotá: Wilcoxon test, W = 1322, P<0.001;
GuaduasversusCerrejón:Wilcoxontest,W=
1410, P< 0.001) (Fig. 4E). This distribution of
DTsintheGuaduasflorasuggestsgreaterher-
bivore community specificity than at either
Paleocene site.
The end-Cretaceous shaped modern
Neotropical rainforests
Prior to the end-Cretaceous, Neotropical rain-
forests had relatively open canopies; contained
a mixture of angiosperms, ferns, and conifers
(mostly Araucariaceae); and suffered intense
and host-specific insect herbivory. Paleocene
forests, by contrast, were more similar to mod-
ern Neotropical rainforests in having closed,
multistratal canopies, biomass dominated by
angiosperms, and a similar plant family com-
position. Yet, Paleocene rainforests were less
diverse than Maastrichtian, Eocene, or mod-
ern rainforests (19), and the low plant diversity
seen throughout the Paleocene shows a long
lag in the recovery of diversity following the
P/Kg event.
The differences between Maastrichtian and
Paleocene forests in floral composition and
canopy structure, but similar leaf physiognomy,
denote two fundamentally distinct ecosystems
that developed under the same wet, tropical
climate. Because of their open canopies, lower
angiosperm abundance, and a constant, albeit
minor, presence of conifers, Maastrichtian rain-
forests may have been accompanied by slower
rates of carbon fixation, transpiration, and nu-
trient cycling when compared to Paleocene
rainforests. In addition, the development of
closed canopy rainforests in the Paleocene
would have created stronger vertical gradients
in light and water use, providing opportunities
for new plant habit and growth forms and
leading to the vertical complexity seen in mod-
ern rainforests.
These notable differences raise two ques-
tions: (i) Why did Maastrichtian rainforests
lack a closed canopy? By the Late Cretaceous,
angiosperms were taxonomically and ecolog-
ically diverse (44,45) and had evolved a wide
range of growth habits, ranging from aquatic
plants to large trees (45,46), making it unlikely
that they were inherently unable to form a
closed canopy. (ii) Why did Paleocene rain-
forests establish a different plant community
composition and structure instead of return-
ing to the Maastrichtian-like rainforests? This
is particularly perplexing given the similarity
in Paleocene and Maastrichtian climates.
We offer three, non–mutually exclusive ex-
planations for the observed pattern. One is
disturbance by large herbivores. Sustained
trampling and extensive feeding by large her-
bivores, mostly dinosaurs (47), could have main-
tained an open canopy by reducing competition
for light among neighboring plants through
continuous habitat disturbance and gap gen-
eration. Such pervasive disturbance could ex-
plain the abundance of ferns in Maastrichtian
palynofloras, as they typically thrive in succes-
sional vegetation (48). The extinction of large
herbivores at the end-Cretaceous would have
reduced gap formation, triggering a “race for
light”among tropical plants, and creating
more shaded habitats in which a wider va-
riety of light and growth strategies could suc-
ceed (49). A second explanation involves soil
nutrients. Extensive and stable lowlands de-
veloped in northern South America during the
Maastrichtian (50), with a persistent humid
climate over millions of years. Maastrichtian
forests therefore must have grown on strongly
weathered soils characterized by extreme in-
fertility (51) with nutrient limitation of growth
exacerbated by the high CO
2
concentrations
and associated high water-use efficiency that
reduces nutrient uptake by mass flow (52,53).
These low-nutrient conditions would have pro-
moted an open canopy structure by favoring
the conifers, which in modern tropical for-
ests are typically associated with infertile soils
(54). Ashfall from the Chicxulub impact added
weatherable phosphorus minerals to terres-
trial ecosystems worldwide (55), instantly
resetting fertility to the high-phosphorus, low-
nitrogen period that characterizes young stages
of ecosystem development (51). This set the
stage for the diversification of nitrogen-fixing
taxa in the Fabaceae, whose rise in the Paleo-
cene (34) would have increased soil fertility,
stimulated forest productivity (56), and en-
hanced the relative advantage of high–growth-
rate angiosperms over conifers and ferns (57,58).
These proposed changes in nutrient cycling
could be tested by analyzing paleosol com-
position and isotopic signatures across the
Maastrichtian–Paleocene interval. A third ex-
planation of the observed pattern concerns se-
lective extinction. Although the Araucariaceae
were not diverse, they could have been impor-
tant in structuring the Late Cretaceous canopy
environment (59). Lineages with narrow eco-
logical ranges and tree growth forms such as
Araucariaceae are particularly susceptible to
mass extinction events (60). By contrast, high
ecological diversity within Maastrichtian angio-
sperm lineages (44,45) may have made them
more resistant to extinction (60), as might their
higher capacity for whole-genome duplication
(61–63). The near disappearance of conifer trees
from tropical rainforest canopies at the end
of the Cretaceous may have released resources
upon which the modern angiosperm canopy-
forming lineages diversified during the Paleo-
cene. This scenario could be tested by assessing
shifts in diversification rates across the K/Pg
of Neotropical canopy trees, epiphytes, and
lianas.
Although there is still much to be learned
about the Cretaceous and Paleocene tropical
forests, the changes described here show that
the end-Cretaceous event had profound conse-
quences for tropical vegetation, ultimately en-
abling the assembly of modern Neotropical
rainforests. It is notable that a single historical
Carvalho et al., Science 372,63–68 (2021) 2 April 2021 5of6
RESEARCH |RESEARCH ARTICLE
on April 1, 2021 http://science.sciencemag.org/Downloaded from

accident altered the ecological and evolution-
arytrajectoryoftropicalrainforests,inessence
triggering the formation of the most diverse
biome on Earth.
REFERENCES AND NOTES
1. T. Westerhold et al., Clim. Past 13, 1129–1152 (2017).
2. V. Vajda, A. Bercovici, Global Planet. Change 122,29–49 (2014).
3. T. R. Lyson et al., Science 366, 977–983 (2019).
4. P. M. Hull et al., Science 367, 266–272 (2020).
5. K. R. Johnson, Nature 366, 511–512 (1993).
6. V. D. Barreda et al., PLOS ONE 7, e52455 (2012).
7. V. Vajda, J. I. Raine, N. Z. J. Geol. Geophys. 46,255–273 (2003).
8. M. P. Donovan, A. Iglesias, P. Wilf, C. C. Labandeira,
N. R. Cúneo, Nat. Ecol. Evol. 1, 12 (2016).
9. A. Iglesias et al., Geology 35, 947–950 (2007).
10. D. J. Nichols, in The Hell Creek Formation and the Cretaceous–
Tertiary Boundary in the Northern Great Plains: An Integrated
Continental Record of the End of the Cretaceous. Geologic al
Society of America SpecialPaper 361,J. H. Hartman,K. R. Johnson,
D. J. Nichols, Eds. (2002), pp. 393–456.
11. D. J. Peppe, Palaeogeogr. Palaeoclimatol. Palaeoecol. 298,
224–234 (2010).
12. S. L. Wing, J. Alroy, L. J. Hickey, Palaeogeogr. Palaeoclimatol.
Palaeoecol. 115, 117–155 (1995).
13. E. J. M. Koenen et al., Syst. Biol. 10.1093/sysbio/syaa041 (2020).
14. W. Wang et al., New Phytol. 195, 470–478 (2012).
15. S. Ramírez-Barahona, H. Sauquet, S. Magallón, Nat. Ecol. Evol.
4, 1232–1238 (2020).
16. P. R. Renne et al., Geology 46, 547–550 (2018).
17. V. Vajda-Santivanez, Palynology 23, 181–196 (1999).
18. See the supplementary materials.
19. S. L. Wing et al., Proc. Natl. Acad. Sci. U.S.A. 106, 18627–18632
(2009).
20. J. Alroy, Proc. Natl. Acad. Sci. U.S.A. 105 (suppl. 1),
11536–11542 (2008).
21. J. Alroy, The Shifting Balance of Diversity Aming Major Marine
Animal Groups, Science 329, 1191–1194 (2010).
22. J. Alroy, Paleobiology 41, 633–639 (2015).
23. D. Silvestro, N. Salamin, A. Antonelli, X. Meyer, Paleobiology 45,
546–570 (2019).
24. C. Jaramillo, M. J. Rueda, G. Mora, Science 311,1893–1896
(2006).
25. C. Jaramillo et al., Science 330, 957–961 (2010).
26. A. Rangel, J. M. Moldowan, C. Niño, P. Parra, B. N. Giraldo,
Am. Assoc. Pet. Geol. Bull. 86, 2069–2087 (2002).
27. M. A. Caccavari, Rev. Mus.Argent. Cienc. Nat. 5,135–138
(2003).
28. P. Wilf, S. L. Wing, D. R. Greenwood, C. L. Greenwood, Geology
26, 203–206 (1998).
29. D. J. Peppe et al., New Phytol. 190, 724–739 (2011).
30. D. L. Royer et al., Paleobiology 33, 574–589 (2007).
31. I. J. Wright et al., Glob. Ecol. Biogeogr. 14, 411–421 (2005).
32. E. Correa, C. Jaramillo, S. Manchester, M. Gutierrez, Am. J. Bot.
97,71–79 (2010).
33. C. Martínez, M. R. Carvalho, S. Madriñán, C. A. Jaramillo,
Am. J. Bot. 102, 273–289 (2015).
34. F. Herrera, M. R. Carvalho, S. L. Wing, C. Jaramillo,
P. S. Herendeen, Aust. Syst. Bot. 32, 385–408 (2019).
35. M. R. Carvalho, F. A. Herrera, S. Gómez, C. Martínez,
C. Jaramillo, Int. J. Plant Sci. 10.1086/714053 (2021).
36. R. J. Burnham, Rev. Palaeobot. Palynol. 81,99–113 (1994).
37. R. Condit et al.,“Complete data from the Barro Colorado 50-ha
plot: 423617 trees, 35 years, 2019 version.”(2019);
https://datadryad.org/stash/dataset/doi:10.15146/5xcp-0d46.
38. A. Duque et al., Biodivers. Conserv. 26, 669–686 (2017).
39. C. Crifó, E. D. Currano, A. Baresch, C. Jaramillo, Geology 42,
919–922 (2014).
40. H. V. Graham, F. Herrera, C. Jaramillo, S. L. Wing,
K. H. Freeman, Geology 47, 977–981 (2019).
41. T. S. Feild et al., Proc. Natl. Acad. Sci. U.S.A. 108, 8363–8366
(2011).
42. C. K. Boyce, J.-E. Lee, T. S. Feild, T. J. Brodribb, M. Zwieniecki,
Ann. Mo. Bot. Gard. 97, 527–540 (2010).
43. M. R. Carvalho et al., PLOS ONE 9, e94950 (2014).
44. K. J. Niklas, B. H. Tiffney, A. H. Knoll, in Phanerozoic Diversity
Patterns, J. W. Valentine, Ed. (Princeton University Press,
Princeton, NJ, 1985), pp. 97–128.
45. S. L. Wing, L. D. Boucher, Annu. Rev. Earth Planet. Sci. 26,
379–421 (1998).
46. N. A. Jud et al., Sci. Adv. 4, eaar8568 (2018).
47. P. M. Barrett, Annu. Rev. Earth Planet. Sci. 42,207–230 (2014).
48. S. L. Wing, in Extinctions in the History of Life, P. D. Taylor, Ed.
(Cambridge Univ. Press, 2004), pp. 61–97.
49. M. Laurans, B. Hérault, G. Vieilledent, G. Vincent, For. Ecol.
Manage. 329,79–88 (2014).
50. L. F. Sarmiento-Rojas,in Geology and Tectonics of Northwestern
South America, F. Cediel, P. Shaw, Eds. (Springer, 2019), pp. 673–747.
51. T. W. Walker, J. K. Syers, Geoderma 15,1–19 (1976).
52. L. A. Cernusak, K. Winter, B. L. Turner, Tree Physiol. 31,
878–885 (2011).
53. E. P. McDonald, J. E. Erickson, E. L. Kruger, Funct. Plant Biol.
29, 1115–1120 (2002).
54. T. Jaffré, in Conifers of the Southern Hemisphere, N. J. Enright,
R. S. Hill, Eds. (Melbourne Univ. Press, 1995), pp. 171–196.
55. D. A. Kring,D. D. Durda,J. Geophys. Res.Planets 107, 5062(2002).
56. D. Z. Epihov et al., Proc. Biol. Sci. 284, 20170370 (2017).
57. D. A. Coomes et al., J. Ecol. 93, 918–935 (2005).
58. W. J. Bond, Biol. J. Linn. Soc. Lond. 36, 227–249 (1989).
59. S. V. Wyse, N. Z. J. Bot. 50, 411–421 (2012).
60. W. A. Green, G. Hunt, S. L. Wing, W. A. DiMichele, Paleobiology
37,72–91 (2011).
61. T. E. Wood et al., Proc. Natl. Acad. Sci. U.S.A. 106,
13875–13879 (2009).
62. Z. Li et al., Sci. Adv. 1, e1501084 (2015).
63. J. D. Thompson, R. Lumaret, Trends Ecol. Evol. 7,302–307
(1992).
64. C. Montes et al., Earth Sci. Rev. 198, 102903 (2019).
65. C. Jaramillo, Dataset for “The end-Cretaceous and the origin of
modern Neotropical rainforests data,”figshare (2021); https://
doi.org/10.6084/m9.figshare.13611215.v2.
ACKNO WLEDG MENTS
We thank Carbones El Cerrejón, the Montecristo/Peñitas coal
mines, and Chechua/Cogua siltstone mines; E. Cadena, A. Rincón,
J. Moreno, S. Gómez, D. Carvalho, A. Giraldo, and A. Alfonso for
fieldwork assistance; G. Doria and C. Gómez for their work
systematizing the fossil collections; three anonymous reviewers for
their helpful comments; Universidad del Norte and Museo Mapuka;
and Parques Nacionales Naturales de Colombia and Parque
Nacional Amacayacu. Funding: This research was funded by
NSF grant EAR-1829299 (to M.R.C., F.H., C.J.), STRI-Earl S. Tupper
postdoctoral fellowship and GSA graduate student research
grant (to M.R.C.), the Oak Spring Garden Foundation (to F.H.),
Smithsonian Tropical Research Institute, the Anders Foundation,
the 1923 Fund and Gregory D. and Jennifer Walston Johnson
(to C.J.), the Swiss National Science Foundation and Swedish
Research Council (PCEFP3_187012, VR: 2019-04739 to
D. Silvestro), and CTFS-ForestGeo. Author contributions:
C.J. designed and coordinated the research program; M.R.C. and
C.J. led the writing with contributions of all coauthors; F.d.l.P.,
C.D’A., M.R.-B., P.N., M.P.-R., C.J., and C.S. performed palynological
data gathering; M.R.C., F.H., S.W., C.M., and M.G. performed
paleobotanical analysis; M.R.C. and C.L. performed herbivore
analysis; M.R.C., A.D., and D.C. performed analysis of modern
vegetation; D.C.R. and C.J. performed pollen data analysis;
B.L.T. provided soil expertise; J.L.C. performed radiometric dating;
G.B. performed palinspastic reconstruction and sedimentary
analysis; and D.S. performed PyRate analysis. Competing
interests: The authors declare no competing interests. Data and
materials availability: The data reported and code used in this
paper are deposited in figshare digital repository (65). Additional
information on samples can be accessed using STRI–identification
numbers, through https://biogeodb.stri.si.edu/jaramillosdb/web/.
The BCI forest dynamics research project was founded by
S. P. Hubbell and R. B. Foster and is now managed by R. Perez,
S. Aguilar, D. Mitre, and S. Lao under the ForestGEO program of the
Smithsonian Tropical Research Institute in Panama. Numerous
organizations have provided funding, principally the U.S. National
Science Foundation, and hundreds of field workers have contributed.
SUPPLEMENTARY MATERIALS
science.sciencemag.org/content/372/6537/63/suppl/DC1
Materials and Methods
Supplementary text
Figs. S1 to S8
Table S1 to S11
References (66–114)
Data S1 to S6
13 October 2020; accepted 3 February 2021
10.1126/science.abf1969
Carvalho et al., Science 372,63–68 (2021) 2 April 2021 6of6
RESEARCH |RESEARCH ARTICLE
on April 1, 2021 http://science.sciencemag.org/Downloaded from

Extinction at the end-Cretaceous and the origin of modern Neotropical rainforests
Santos and Daniele Silvestro
Labandeira, German Bayona, Milton Rueda, Manuel Paez-Reyes, Dairon Cárdenas, Álvaro Duque, James L. Crowley, Carlos
Benjamin L. Turner, Carlos D'Apolito, Millerlandy Romero-Báez, Paula Narváez, Camila Martínez, Mauricio Gutierrez, Conrad
Mónica R. Carvalho, Carlos Jaramillo, Felipe de la Parra, Dayenari Caballero-Rodríguez, Fabiany Herrera, Scott Wing,
DOI: 10.1126/science.abf1969
(6537), 63-68.372Science
, this issue p. 63; see also p. 28Science
closed and layered, leading to increased vertical stratification and a greater diversity of plant growth forms.
that of modern lowland neotropical forest. The leaf data also imply that the forest canopy evolved from relatively open to
Angiosperm taxa came to dominate the forests over the 6 million years of recovery, when the flora began to resemble
but were also able to infer changes in forest structure. Extinctions were widespread, especially among gymnosperms.
forests at this time (see the Perspective by Jacobs and Currano). They not only found changes in species composition
used fossilized pollen and leaves to characterize the changes that took place in northern South Americanet al.Carvalho
The origin of modern rainforests can be traced to the aftermath of the bolide impact at the end of the Cretaceous.
The birth of modern rainforests
ARTICLE TOOLS http://science.sciencemag.org/content/372/6537/63
MATERIALS
SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2021/03/31/372.6537.63.DC1
CONTENT
RELATED http://science.sciencemag.org/content/sci/372/6537/28.full
REFERENCES http://science.sciencemag.org/content/372/6537/63#BIBL
This article cites 104 articles, 30 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title
(print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience
Science. No claim to original U.S. Government Works
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
on April 1, 2021 http://science.sciencemag.org/Downloaded from