Linked canopy, climate, and faunal change in the Cenozoic of Patagonia

See allHide authors and affiliations

Science  16 Jan 2015:
Vol. 347, Issue 6219, pp. 258-261
DOI: 10.1126/science.1260947

Fluctuations revealed in fossil forests

The reconstruction of past vegetation unlocks the door to understanding ecological changes associated with climatic change. But it is also difficult. Dunn et al. developed a method for assessing changes in vegetation openness based on epidermal cell morphology from conserved plant tissue. Applying this method to fossil assemblages from Patagonia, they show how vegetation structure changed over the Cenozoic era (49 to 11 million years ago). These changes map onto the known climate changes over this period and can also be used to track how the evolution of herbivorous mammals responded to vegetation structure.

Science, this issue p. 258


Vegetation structure is a key determinant of ecosystems and ecosystem function, but paleoecological techniques to quantify it are lacking. We present a method for reconstructing leaf area index (LAI) based on light-dependent morphology of leaf epidermal cells and phytoliths derived from them. Using this proxy, we reconstruct LAI for the Cenozoic (49 million to 11 million years ago) of middle-latitude Patagonia. Our record shows that dense forests opened up by the late Eocene; open forests and shrubland habitats then fluctuated, with a brief middle-Miocene regreening period. Furthermore, endemic herbivorous mammals show accelerated tooth crown height evolution during open, yet relatively grass-free, shrubland habitat intervals. Our Patagonian LAI record provides a high-resolution, sensitive tool with which to dissect terrestrial ecosystem response to changing Southern Ocean conditions during the Cenozoic.

Vegetation structure—the degree of canopy openness—is a fundamental aspect of ecosystems, influencing productivity, hydrological and carbon cycling, erosion, and composition of faunal communities (1, 2). However, methods to quantify ancient vegetation structure have eluded paleoecologists. Here, we present a method with which to reconstruct vegetation openness, specifically leaf area index [LAI = foliage area (m2)/ground area (m2)], using the morphology of leaf epidermal cells preserved as phytoliths (plant biosilica) (Fig. 1). LAI quantifies vegetation structure in ecological and climate modeling studies (1, 3). In modern ecosystems, LAI relates primarily to soil moisture (4), by which vegetation becomes more closed with increasing soil water availability; ultimately, soil moisture is determined by temperature, precipitation, and atmospheric partial pressure of CO2 (Pco2) (4, 5). Disturbance in the form of fire and herbivory can offset this relationship, resulting in open habitats in areas with relatively high rainfall (6).

Fig. 1 Leaf epidermis and examples of epidermal phytoliths.

(A) Nothofagus leaf and epidermis. (B to E) Fossil phytoliths from Patagonia. (F to I) Modern soil phytoliths from Costa Rica.

Using this paleobotanical archive, we reconstructed a LAI record for the middle Cenozoic [49 million to 11 million years ago (Ma)] of Patagonia to test predictions about vegetation response to Cenozoic climate fluctuations and how changes in vegetation structure relate to the evolution of high-crowned (hypsodont) and ever-growing (hypselodont) tooth morphologies in South American herbivores (7).

Climatic cooling beginning in the middle Eocene (49 Ma) and major warming events in the late Oligocene (~26 Ma) and middle Miocene (17 to 14 Ma) (8) have been linked to tectonics, ocean circulation (9), atmospheric Pco2 (10), and ice volume after the onset of extensive Antarctic glaciation at the Eocene–Oligocene Transition (EOT; 33.9 Ma) (8). A narrow landmass, Patagonia is sensitive to Southern Ocean climate and provides an ideal test case for the influence of global climate on vegetation structure.

It has long been assumed that hypsodonty in endemic South American herbivores beginning in the middle Eocene (~40 Ma) evolved in response to the spread of Earth’s first grasslands (11), but recent work found that grasses were rare (12). When grasses are rare, “traditional” phytolith analysis cannot resolve habitat openness (13, 14), so it has remained unclear whether hypsodonty evolved in forests or in open but grass-free habitats, possibly in conjunction with tooth abrasion during ingestion of exogenous grit (12).

Our proxy for reconstructing ancient LAI [reconstructed LAI (rLAI)] is based on the well-known influence of sunlight on the size and shape of pavement cells in leaf epidermis (Fig. 1A). In shade, these cells are larger and have more undulated outlines than those of cells exposed to full sun (15, 16). Silica mineralization produces a precision cast of pavement cells in living plants that can be preserved as fossils (Fig. 1B). Because sunlight filtering through a canopy is a function of LAI, we hypothesized that leaf epidermal cells and their phytoliths are on average larger and more undulated in closed forests than in open habitats and that the relationship is linear across a canopy density gradient. Because these phytolith types are taxonomically nondiagnostic we cannot control for phylogenetic variation in cell morphology. Instead, we tested our hypotheses using modern assemblages of phytoliths extracted from soils collected across an LAI gradient from 0 (completely open) to 5 (dense forest) in Costa Rica (Fig. 1C).

Cenozoic-aged floras from Patagonia contain a nonanalog mix of mesic and xeric taxa of tropical and sub-Antarctic lineages (such as Arecaceae, Anacardiaceae, Fabaceae, Zingiberaceae, Proteaceae, Nothofagus, Podocarpaceae, and Araucariaceae). We chose to sample phytoliths from modern tropical soils in Costa Rica because it has wet and dry forests, savanna, and shrubby areas containing many of the reported fossil genera (41%) and families (85%) (table S1). We assume that relative change in epidermal cell morphology is based on canopy density and is independent of taxonomy and latitudinal variation in light regime. Using light microscopy, nongrass epidermal phytoliths in extracted samples were photographed and measured for the calibration data set. Phytolith undulation was standardized by using an undulation index (UI = circumference of cell/circumference of a circle with cell area) (16), and mean site values for phytolith UI (PUI) and phytolith area (PA) were compared with field measurements of LAI from hemispherical photographs (Fig. 2A). Measurements of fossil phytoliths followed the same protocol.

Fig. 2 Modern soil phytolith morphology and LAI.

(A) Hemispherical photographs from Costa Rica illustrating LAI values. (B) Linear regressions for mean PUI and LAI (rLAI = 13.92 × PUI – 17.31); and (C) mean Phytolith Area (PA) and LAI (rLAI = 0.0028 × PA + 0.531). (D) Plot of simulated versus measured values of LAI. Simulated values are derived from Eq. 1.

In the modern training data set of 45 sites (table S2), LAI correlates with PUI [coefficient of determination (R2) = 0.59, P < 0.0001] (Fig. 2B) and PA (R2 = 0.44, P < 0.0001) (Fig. 2C). A linear multiple regression analysis including both variables improves the correlation (Fig. 2D and table S3):rLAI = 0.0012 × PA(μm2) + 10.4118 × PUI – 13.1621 (1)(R2 = 0.63, P < 0.0001, SE = 0.695, F2,42 = 39). Using Eq. 1, we reconstructed rLAI for 46 fossil phytolith assemblages from Patagonian paleosols spanning 49 to 11 Ma (Fig. 3A, fig. S3, and table S4). Data from different times should be comparable because all samples are from the same region with the same moisture resources (for example, similar vegetation occupy all sites today). From the middle Eocene to early Oligocene (49.0 to 32.3 Ma), rLAI values decline from ~4 to 0.6, indicating an opening of the landscape, from dense vegetation (such as broad-leaf forest; LAI = 4 to 3) to progressively more open vegetation (such as dry forest, woodland, and scrub; LAI = 2 to 1); and last, to very open habitats (such as shrubland or desert; LAI < 1) (Table 1 and Fig. 3). High rLAI values during the middle Eocene correspond in age with the highly diverse megathermal floras 380 km farther north (17), and middle Eocene–early Oligocene decreases in rLAI correspond with increased abundances of sub-Antarctic taxa such as Nothofagus (18, 19).

Fig. 3 Middle Cenozoic rLAI and comparisons to climate and biotic records.

(A) rLAI reconstruction with 95% confidence intervals. Shown are raw values (open circles) and binned values by age or geologic unit (red diamonds) (table S4). (B) Foraminiferal δ18O records for sea surface (red) (21) and deep sea temperature (blue) (8). (C) MAT (blue), MAP (green), and dry-season precipitation (orange) estimates from megafloras (27, 33), with corrected ages (table S5). Length of the solid bars indicate age uncertainty. Yellow boxes are MAP estimates from paleosols (34). (D) Proportion hypsodont+hypselodont taxa of notoungulates, with 95% confidence intervals (table S6).

Table 1 Modern habitat LAI ranges from literature.

Dashes indicate no reported data.

View this table:

Between 38.5 and 38.0 Ma and 35.0 and 32.2 Ma, habitats were maximally open (rLAI <1). Diagnostic phytoliths from these assemblages indicate abundant palms, woody eudicotyledons, and sparse grasses (<1 to 15%) (12). We interpret these habitats as a nonanalog palm shrubland with a discontinuous canopy. During the late Oligocene, rLAI increases, indicating dry forest, woodland, or scrub until the earliest Miocene (~21.1 and 18.8 Ma), when rLAI drops again. Early Miocene rLAI fluctuations (<1 to 2.4) suggest shifts between palm shrubland and open forest without a continuous grassy understory (12). The middle Miocene (~15.7 to 14.6 Ma) saw a brief spike in rLAI values (>3.5) at 14.6 Ma, after which (~14.2 Ma) rLAI values trend downward again. Pollen records from late Oligocene–middle Miocene marine strata of Patagonia indicate humid forest conditions dominated by Nothofagus and podocarps, with low abundances of arid-adapted (inferred as open habitat) taxa before 10 Ma (20). These pollen data do not contradict our findings because they reflect regional vegetation and are biased toward prolific pollen producers, whereas our phytolith samples represent in situ vegetation of the central Patagonian lowlands.

Broad changes in rLAI track the Southern Ocean δ18O temperature record (8, 21); rLAI decreases during mid–late Eocene cooling and increases during late Oligocene warming. The middle Miocene (~15.7 to 14.6 Ma) regreening of Patagonia coincides with increased atmospheric Pco2 (22) and reduced Antarctic ice sheet volume (23). After ~14.2 Ma, declining rLAI values are consistent with middle Miocene cooling, ice-growth, and modeled changes in meridional heat and vapor transport (24).

Vegetation-climate decoupling occurred during pulses of maximum openness at 38.5 to 38.0 Ma and at ~35 Ma, as marine temperatures gradually declined. This second pulse predates the EOT by >1 million years; in contrast, during abrupt EOT cooling, rLAI remained relatively unchanged. Quasi-constant rLAI during the EOT is consistent with phytolith abundance data, suggesting compositionally stable vegetation (12) and isotopically inferred invariant temperatures (25). However, aeolian sedimentation rates dramatically increased at 33.6 Ma (26), which is consistent with Oi-1 glaciation.

Intervals of open vegetation likely reflect reduced precipitation, although disturbance such as fire, volcanism, or herbivory may have contributed. Megafloral records from elsewhere in Patagonia document stable mean annual temperatures (MATs) of ~18°C, but decreasing mean annual precipitation (MAP) from the middle Eocene onward (Fig. 3C) (27). By at least the late Oligocene, decreased MAP values reflect reduced dry-season rainfall (27). Locally, episodes of low rLAI correspond to the lowest MAP estimates from paleosols and shifts to aeolian sedimentation (28). Additionally, phytoliths of water-demanding gingers become very rare (0.4%) by 38.1 Ma and disappear after ~38 Ma (12). Our climate interpretation is seemingly at odds with phytolith evidence for abundant palms, which in modern South America is linked to warm, humid climates (29). However, Patagonian fossils indicate that a largely dry-adapted palm clade (Attaleinae) had diversified in South America by the Paleocene (fig. S5). We hypothesize that water-use efficiency in these palms was further enhanced under elevated Eocene atmospheric Pco2 (30).

Increasing openness (rLAI < 1) ~40 Ma coincided with initiation of tooth crown height increases in several clades of notoungulates (Fig. 3D). Proportions of hypsodont+hypselodont taxa continued to rise from 38 to 20 Ma, as rLAI remained low (between 0 and 2; average ≤ 1.5). The hypsodonty trend may have reversed during more forested middle Miocene conditions, but errors are large, and constant hypsodonty proportions cannot be ruled out. In modern South American environments, the proportion of hypsodont+hypselodont species dramatically increases under a LAI value of ~1.2 (fig. S6 and table S7). These areas experience low precipitation, frequent dust storms, and erosion of tephric materials (31).

Evidently, feeding in drier, more open Eocene–early Miocene ecosystems provided evolutionary pressure to drive hypsodonty and hypselodonty in Patagonia. The temporal coincidence of wind-blown ash, low rLAI, and increased rates of hypsodonty+hypselodonty further suggests that ash played a key role in this process (12). In low LAI habitats today (such as shrublands), sparse vegetation includes both bare ground (dust source areas) and shrubs (traps for dust) (32). Thus, ingestion of dust adhering to plants growing on highly erodible surfaces (tephra-rich soils) plausibly drove this pattern of tooth evolution in South America.

Taken together, these patterns indicate that long-term climate changes that predated the EOT drove ecosystem changes in Patagonia. Specifically, we propose that Southern Ocean instability during the protracted opening of Drake Passage beginning in the middle Eocene (9) and associated cooling sea surface temperatures resulted in reduced rainfall on land and triggered successive opening-up of landscapes during the middle-late Eocene. Our method for estimating rLAI allows for quantification of vegetation structure through time, and because it relies on microfossils, extremely high-resolution records of habitat change are possible. Additionally, because leaf epidermis is a highly conserved tissue, the method should be applicable across a broad range of temporal scales to test many outstanding hypotheses in paleoecology.

Supplementary Materials

Materials and Methods

Figs S1 to S6

Tables S1 to S6

References (4162)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: Supplementary data are included in the data file. This research was funded in part by National Science Foundation grants DEB-1110354 to R.E.D. and C.A.E.S. (Doctoral Dissertation Improvement Grant) and EAR-0819910 to C.A.E.S., EAR-0819842 to R.H.M., and EAR-0819837 and EAR-1349749 to M.J.K.; Proyecto de Investigación en Ciencias y Técnicas 1860 of the Fondo Nacional de Ciencia y Tecnología (FONCyT) To A.A.C.; the Geological Society of America; and the University of Washington Department of Biology and Burke Museum of Natural History and Culture. We thank G. Vucetich, M. Ciancio, M. Conner, A. Loeser, Pan American Energy, Organization for Tropical Studies, Área Conservación de Guanacaste, and three reviewers.
View Abstract

Stay Connected to Science

Navigate This Article