A comparison of plant-insect associations in the middle Eocene Green River Formation and the Upper Eocene Florissant Formation and their climatic implications
Published:January 01, 2008
- PDF LinkChapter PDF
Dena M. Smith, 2008. "A comparison of plant-insect associations in the middle Eocene Green River Formation and the Upper Eocene Florissant Formation and their climatic implications", Paleontology of the Upper Eocene Florissant Formation, Colorado, Herbert W. Meyer, Dena M. Smith
Download citation file:
The fossil plants found in the Eocene Florissant Formation and Green River Formation are preserved with a level of detail that allows one to closely examine traces of insect feeding damage. Levels (amounts) and patterns (abundance of various types) of fossilized insect feeding damage from Florissant and the middle Eocene Green River Formation were compared. This allowed for a detailed examination of feeding damage and provided an opportunity to examine long-term patterns of change in insect herbivory during a period of climate fluctuation. Samples including 624 fossil leaves from Florissant and 584 fossil leaves from the Green River Formation were examined to document overall damage levels, the presence/absence of specific feeding guilds (i.e., hole-feeding, skeletonization, leaf-mining), and host-specific damage types.
Florissant insects show host specificity in their feeding preferences as evidenced through the distribution of feeding damage on plants and through the presence of identifiable host-specific interactions. Some of these interactions appear to be long lasting as they are also apparent on the same, or closely related, leaf species found in the Green River Formation. Insect damage levels declined from the middle to late Eocene. This decline is correlated with a cooling event during this time interval and is in concurrence with the findings of other authors who have examined fossilized herbivory and climate change patterns. There is also an increase in the abundance of galls during this same interval, which also may be related to climate change.
The early Cenozoic fossil record provides an ideal setting for examining the evolution of plant-insect interactions and for evaluating how climate change in particular affects ecosystems at basic trophic levels. The early Cenozoic is an important period because it provides a window into the early evolution of many insect and plant groups that make up our modern temperate ecosystems. As such, this period has begun to receive much attention from researchers interested in examining long-term patterns of change in plant-insect interactions.
Many researchers have been able to establish the presence of long-lasting associations between a variety of herbivorous insects and their host plants through an examination of the fossil record (Opler 1973; Hickey and Hodges 1975; Hickey and Doyle 1977; Larew 1986; Upchurch and Dilcher 1990; Labandeira et al. 1994; Wilf et al. 2000). Some studies have documented assemblage-level patterns of herbivory (Ash 1997; Beck and Labandeira 1998, Scott and Taylor 1983; Smith 1998), whereas others have also attempted to examine changes in herbivore damage over time (Wilf and Labandeira 1999; Smith 2000; Wilf et al. 2001). The latter studies examined changes in insect herbivory during periods of climate change in the Paleocene and Eocene and found evidence for an increase in herbivory as temperatures increased and a decrease in herbivory as temperatures decreased, suggesting that climate may have played an important role in the evolution of insect-plant associations.
Today, ∼75% of all phytophagous (plant-feeding) insects are either monophagous (specialize on one species of plant) or oligophagous (specialize on a few species of plants in a single family) (Bernays and Chapman, 1994). It is likely that many of these associations originated during the Cenozoic, when the first appearances of many modern insect and plant genera occurred (Wing, 1987; Carpenter, 1992). However, insects possessed the mouthpart morphology to create the types of damage that we see on modern leaves prior to the Cenozoic (Labandeira and Sepkoski, 1993; Labandeira, 1997), and a few examples of modern associations have been documented from the Cretaceous (Labandeira et al., 1994).
Florissant fossil beds and the Green River Formation are deposits that are well known for their abundance, diversity, and excellent preservation of both insects and plants. Nearly 1500 species of insects and approximately120 plant species have been described from Florissant (MacGinitie, 1953; Manchester 2001; Meyer, 2003; Drummond, 2004) and nearly 300 insects and at least 157 plant morphotypes have been identified from the Green River Formation (MacGinitie, 1969; Barclay et al., 2003: Smith, 2000, pers. observation). But, what is known of the interactions between these two groups during this time interval? With the quality of preservation found in the Florissant and Green River Formations, it is possible to examine feeding guilds and learn whether insect feeding strategies and plant defensive strategies have changed over time. It is possible to determine the plants on which insects fed and whether this involved the most abundant plants, specific families, or no selectivity at all. Finally, the assemblages can be examined for evidence of stereotyped (recognizable, host-specific feeding damage) and long-lasting interactions between specific plant and insect groups.
In this study, I examined fossil leaves from the late Eocene Florissant Formation and the middle Eocene Green River Formation of Colorado, USA. I examined the insect-mediated damage that is found preserved on the leaf macrofossils and determined overall damage levels, the distribution of feeding guilds, and the presence of specialized damage types. The comparison of herbivory from the two fossil floras provides a rare opportunity to detect long-lasting associations (over a period of 9–13 m.y.) and to examine patterns of change in plant-insect interactions during a climatic cooling event.
MATERIALS AND METHODS
Florissant Fossil Beds
Geologic and Paleoclimatic Setting
Fossil plants were collected from Florissant Fossil Beds National Monument in Teller County, Colorado (Fig. 1). The Florissant Formation has been dated at 34.07 Ma (Evanoff et al., 2001) and the fossil beds were deposited in a lake that was formed by the volcaniclastic debris flow damming of the drainage basin (McLeroy and Anderson, 1966; Evanoff et al., 2001). On the basis of modern biogeographic affinities of the fossil plant assemblage MacGinitie (1953) interpreted the ancient lake setting as a warm-temperate environment. Subsequent studies, using consideration of physiognomic and floristic criteria (Meyer, 1992, 2001; Leopold and Clay-Poole, 2001), multivariate analysis of leaf characters using CLAMP (Wolfe, 1992, 1994; Gregory and McIntosh, 1996), comparison to modern forest communities (Boyle et al., this volume), and the mutual climate range of Diptera (Moe and Smith, 2005), have estimated mean annual temperatures (MATs) within a range of 10.8–17.5 C°. A mean annual precipitation (MAP) of 50 cm was estimated from the composition of the fossil flora (MacGinitie, 1953). MacGinitie also utilized floristic comparisons to estimate the paleoaltitude of the Florissant flora at 300–900 m. More recent studies have used foliar physiognomic techniques and lapse rates or enthalpy to estimate paleoelevations for Florissant ranging from 1900 to 4133 m (Gregory and Chase, 1992; Meyer, 1992, 2001; Wolfe, 1992, 1994; Gregory 1994; Forest et al., 1999).
Collection and Identification
Fossil plants were collected from two localities (FLFO-5 and FLFO-7) in the middle shale unit of the Florissant Formation (Fig. 1) within the boundaries of the national monument. These sites are located on the eastern margin of the paleolake and are interpreted here as nearshore environments on the basis of the distribution pattern of the Florissant Formation relative to the current topography, which is bounded by Precambrian Pikes Peak Granite and is thought to reflect ancient lake boundaries (Meyer, 2003). Fossils from both localities are preserved as impressions and compressions in thinly laminated shale. The shale consists of couplets of diatoms and ash/clay (McLeroy and Anderson, 1966; O'Brien et al., this volume). Leaves are the main type of fossil found from the localities, but insects, mollusks, and fish remains also have been collected. Fossil specimens are sometimes found preserved among large amounts of organic debris and multiple leaf specimens are often found on the same piece of shale.
The collections were made by National Park Service staff, volunteers, and the author during the summers of 1997 and 1998. Specimens from both sites were combined into one Florissant sample for the purposes of comparative analysis. Because each locality samples only a small local area of the source forest and represents a short interval of time, pooling data from both localities provides a better representation of the source forest overall. Nearly 1200 fossil leaves were collected from Florissant, but only 624 angiosperm leaves were sufficiently preserved (preservation quality allowed for recognition of herbivore damage) to be used in this study. Leaf fragments smaller than 1 cm2 were not included in the study. In neontological studies of herbivory, sample sizes range from ∼200 to 10,000 leaves/sample, depending on the questions addressed (Bray, 1961; Nielsen, 1978; Coley, 1983; de la Cruz and Dirzo, 1987; Basset, 1991; Lowman and Heatwole, 1992; Filip et al., 1995; Barone, 2000; Williams-Linera and Herrera, 2003; Smith and Nufio, 2004). All dicotyledonous angiosperm specimens were identified to species when possible and non-angiosperm plants were not included in the analysis. MacGinitie's (1953) monograph and subsequent taxonomic revisions of the flora (Manchester and Crane 1983; Axelrod, 1986; Manchester 1987; Wolfe and Tanai, 1987; Manchester, 1989a, 1989b; Wolfe and Schorn 1990; Manchester 1992; Manchester and Donoghue 1995; Wang and Manchester, 2000; Manchester 2001; McClain and Manchester 2001) were used to identify the plants. Specimens that were not identifiable to the genus level but were identifiable at the family level were given a morphotaxon assignment. For the purposes of this study, a morphotaxon is a designation that is given to a specimen or group of specimens that are morphologically distinct from all other specimens but not yet assignable to a specific taxon. “Unidentified” specimens (17% of the specimens) were usually not identifiable because of poor preservation and/or the lack of distinguishing characteristics. All specimens are reposited in the National Park Service collections at Florissant Fossil Beds National Monument, Colorado (FLFO ACC 250–252 and 272).
Green River Formation
Geologic and Paleoclimatic Setting
Fossil plant material was collected from the Parachute Creek Member of the Green River Formation in Colorado (Fig. 2), which has been dated at 43–47 Ma (Roehler 1973; Remy, 1992; Hail and Smith 1997). MacGinitie (1969) interpreted the flora of the Green River Formation as representing a tropical to subtropical environment with a distinct dry season. Paleoclimate estimates based on nearest living relatives and leaf physiognomy characters give a MAT of 16–23 °C and MAP of ∼45–86 cm for this portion of the Green River Formation (MacGinitie, 1969; Wolfe, 1994; Smith, 2000). MacGinitie (1969) estimated that the Green River flora grew at an elevation of no more than 300 m. However, foliar physiognomic techniques and paleoenthalpy yielded more recent paleoelevation estimates for the middle Eocene Green River Formation range between 1564–2900 m (Wolfe, 1994; Wolfe et al., 1998; Forest et al., 1999).
Collection and Identification
Fossil plant material was collected from four sampling areas spaced 35 m apart laterally within the same stratigraphic interval, from UCM locality 2001058, which is located near Long Point, in Garfield County, Colorado. Collections were made by the author and field assistants during the summers of 1997 and 1998. Data from the four sample areas were combined into one Green River sample of 584 leaves. Leaf fragments smaller than 1 cm2 were not collected.
Located in the western Piceance Creek Basin, this portion of the lake has been interpreted as a marginal lacustrine environment (MacGinitie, 1969; Picard and High, 1972; Lundell and Surdam, 1975). Fossils are preserved as impressions and compressions in thinly laminated oil shale. The oil shale is dolomitic and interbedded with marlstone and claystone (Roehler, 1973; Hail and Smith, 1997). Leaf macrofossils are the predominant fossil found, but insects, seeds, and feathers have also been collected from the locality. Multiple specimens are usually found on the same piece of shale.
All dicotyledonous angiosperm specimens were identified to species using MacGinitie's (1969) classification and all subsequent revisions (Manchester, 1986; Manchester, 1989a, 1989b; Call and Dilcher, 1994, 1995, 1997; Lott et al., 1998; Manchester et al., 1998; Wang and Manchester, 2000). Some legume leaflets were not identifiable to the genus level and are listed as “undetermined leaflets” from the family Leguminosae. Other “unidentified” specimens were not identifiable because they were either poorly preserved or because they were not identified or figured in any of the Green River plant publications. All specimens have been reposited in the paleontological collections of the University of Colorado Museum, Boulder, Colorado (UCM accession no. 348).
Comparison Between Fossil Sites
Certain aspects of the taphonomic character of leaves found in the Green River and Florissant floras appear to be very similar in terms of quality and detail of preservation, and in terms of size sorting of the leaves in both samples. Fossilized leaves were collected from relatively nearshore environments from both sites, and both assemblages appear to have an overrepresentation of small leaves derived primarily from the canopy of the source forest. This is based on the size of leaves from these assemblages and expectations from the plant taphonomy literature (Ferguson, 1985; Greenwood, 1992; Spicer, 1981).
According to floral lists, 68% of the families and 39% of the genera in the Florissant flora are also found in the Green River flora. MacGinitie (1969, p. 72–73) noted the similarities in taxonomy and preservation between the two floras and suggested that Florissant might contain some of the same species and/or descendants of species found in the Green River assemblage, including “29 identical or closely similar species.”
Despite these similarities, there are definite differences between the two localities. For example, Florissant was a small lake relative to the lake environment of the Green River Formation. In addition, diatoms were present in Florissant and may have helped to facilitate preservation of the macrofossils (see O'Brien et al., this volume), whereas Green River has no record of diatoms and was a much more carbonate-rich environment.
Assessment of Herbivory
Each leaf was examined for the amount and type of insect damage using the criteria outlined by Labandeira (1998) and Beck and Labandeira (1998). Damage was categorized as belonging to one of five functional feeding groups (Fig. 3): (1) hole feeding—external foliage feeding in which an insect feeds through the leaf leaving behind a hole; (2) margin feeding—external foliage feeding in which an insect feeds on the margins of the leaf; (3) skeletonization—external foliage feeding in which an insect feeds on the soft tissues of the leaf, but does not feed on the veins (includes scraping leaf tissue layers); (4) leaf mining—internal feeding in which an immature insect lives and feeds within the leaf layers, leaving behind a blotch or serpentine shaped mine; and (5) galling—internal feeding in which an insect feeds and lives between the leaf tissue layers and the plant responds by developing histologically anomalous leaf tissue around the site. Plant galls can be induced by non-insect arthropods as well and sometimes it is difficult to distinguish between them. Specific gall types and potential inducers will be discussed in greater detail in the results section. Any distinguishable, stereotyped feeding damage found on leaves also was recorded. Insects that feed exclusively on one plant group often leave a pattern of damage that is unique to that insect and is consistent and clearly identifiable. This type of identifiable damage pattern is referred to as stereotyped feeding damage. All of the fossil leaves were photographed using a manual camera and stereo-microscope with a camera attachment. The photos were then scanned with a flatbed scanner and saved as digital images. All leaves and insect damage were measured using NIH Image software for the Macintosh (http://rsb.info.nih.gov/nih-image/).
All statistical analyses were performed with the use of JMP IN 3.2.1 (SAS Institute, 1996). Chisquare analysis of contingency tables was used to compare the relative abundance of plant taxa with the distribution of insect feeding damage in the Florissant assemblage. A t-test was used to compare the leaf area measurements between the Florissant and Green River floras. The amount of damage area was also compared between the two sites using a Wilcoxon signed-rank test. For families and genera that had at least ten leaves preserved in both localities, a Wilcoxon signed-rank test was used to compare the amount of damage made by insects between the two floras by family. Wilcoxon signed-rank tests were used because of the large number of leaves that had no feeding damage, necessitating the need for using a non-parametric test. Families that were compared are Anacardiaceae, Leguminosae, Salicaceae, Sapindaceae, and Ulmaceae. Genera that were compared are Cedrelospermum, Populus, and Rhus. Comparisons between Florissant and Green River with respect to overall incidence of herbivory, herbivore intensity, and feeding-guild structure were performed using chisquare analyses of contingency tables.
Of the 624 leaves from Florissant, 514 were identified to 48 plant species and morphotaxa within 17 families (Table 1). There were 110 unidentifiable specimens. The four most abundant plant species (Cedrelospermum lineatum, Fagopsis longifolia, Staphylea acuminata, and Rhus sp.) accounted for 50% of the leaves in the assemblage, and 44% of the insect damage in Florissant is found on these plants. The rank order of plant abundance and the rank order of plants with feeding damage are significantly different (χ2 = 91.22, p < .0001), indicating that insects were not feeding on plants in proportion to the plant's relative abundance. Twenty-two percent of the 48 plant species at Florissant have no insect-mediated damage. Four of five feeding guilds were present in the assemblage. No leaf-mining damage was found on the leaves observed in the study sample from Florissant.
Green River Formation
Of the 584 fossil leaves that were examined from the Green River Formation, 522 of the specimens were assigned to one of 22 plant species in 16 families and the remaining 62 specimens remain unidentified (Table 2). Four species (Cardiospermum coloradensis, Cedrelospermum nervosum, Parvileguminophyllum coloradensis, and Syzigiodes americana) accounted for 57% of the leaves in the assemblage, and 59% of the insect-damaged leaves were from these species. Nine species did not have any insect damage on them. These species occurred at low frequencies—only one to four specimens per taxon (Table 2).
Comparison of Florissant and Green River Formation
Overall leaf sizes in the Florissant and the Green River floras were not significantly different (F = 1.49, p = .221), but the amount of area damaged by insects was significantly greater (Wilcoxon signed-rank test, Z = 4.61739, p < .0001) in the Green River flora (2.5% of original leaf area) than in the Florissant flora (1.4% of original leaf area) (Table 3).
Comparisons of damage levels by plant family indicated significantly higher damage levels for Green River leaves from the Salicaceae (Wilcoxon signed-rank test, Z = −2.68160, p = .0073) and Sapindaceae (Wilcoxon signed-rank test, Z = −2.79887, p = .0051). All other plant families did not have significantly different damage levels (Wilcoxon signed-rank test, p > .05). When damage levels were compared by genus, only Populus had significantly different damage levels (Wilcoxon signed-rank test, Z = −2.17680, p = .0295), with herbivory being greater in the Green River Formation.
Only 23% of the leaves were damaged in the Florissant flora, compared to 34% in the Green River flora (Fig. 4A). This difference is statistically significant (χ2 = 16.39, p < .0001). There is no significant difference (χ2 = 3.4, p = .183) in the number of feeding types per leaf between the Florissant and Green River floras (Fig. 4B), and damaged leaves usually have only one type of feeding damage. No leaf from Florissant or the Green River has more than three types of damage.
Feeding-guild structure (Fig. 4C) is slightly different (χ2 = 9.97, p = .041) between the two floras. The main differences between the two are the dominance of leaves with hole-feeding damage and the greater proportion of leaves with galls in the Florissant assemblage.
There are several examples of feeding damage found on particular Florissant plants that appear to be created by an insect with a host-specific relationship with that plant. These damage patterns tend to be found on the rarer taxa in the assemblage. By contrast, Fagopsis longifolia is the most abundant plant at Florissant, but it has no specialized damage and very little generalized damage of any sort.
Cedrelospermum lineatum and Staphylea acuminata are the only abundant species with stereotyped, specialized feeding damage. Staphylea acuminata has two different examples of feeding damage that are likely produced by a specialist insect: distinctive skeletonization damage and gall damage (Fig. 5A). Distinctive galls and rounded hole-feeding patterns are also found on Cedrelospermum lineatum leaves (Fig. 5B). Cedrelospermum is an extinct genus; however, one of its closest living relatives, Zelkova, has gall damage on its leaves that is similar in appearance to the fossil galls. Galls on modern Zelkova are produced by aphids. Galls are also found on the leaves of the confamilial Ulmus tenuinervis at Florissant, including an example of two galls on a single leaf (Fig. 5C and 5D), and galls have been found on a specimen of Rhus and a specimen of Robinia. Aphids are known to produce host-specific galls on the leaves of Ulmus (Tribe Eriosomatini) and Rhus (Tribe Melaphini) in modern forests (Wool, 2005) and are good candidates for the producers of the galls found on the fossil leaves.
Two galls with a characteristic arrangement were found on a leaf of Cercis parvifolia (Fig. 5E and 5F). Specialist hole-feeding damage is found on leaves of Trichilia florissantii and highly specialized hole-feeding patterns are found on Cardiospermum terminalis (Fig. 6C). One Populus crassa specimen has distinctive skeletonization damage and one unidentifiable leaf has distinctive hole-feeding damage.
Three examples of specialized insect damage are found in both the Green River and Florissant floras. Populus crassa from Florissant has skeletonization damage that is also found on Populus cinnamoides (Fig. 6B) and Salix sp. leaves from the Green River Formation. The rounded hole-feeding patterns that are found on Cedrelospermum nervosum leaves in the Green River flora (Fig. 6A) are also found on the same species in Florissant. Finally, there are the distinctive, hole-feeding patterns with parallel-sided margins that are found on the leaves of Cardiospermum terminalis at Florissant and Cardiospermum coloradensis in the Green River Formation (Fig. 6C and 6D) This damage pattern is characterized by elliptical to slightly rounded holes. The holes typically occur between the second-order veins and have very well developed reaction tissue around the rim of damage. The holes range from 2 to 8 mm in length and are usually found in clusters of several elliptical holes.
Levels and Patterns of Herbivory
The majority of insect damage at Florissant was not found on the most abundant leaves in the assemblage. Some plant species seemed to have been avoided, whereas others seemed to have been preferred and included several examples of specialized feeding damage. This suggests that insects were exhibiting feeding preferences for specific host plants. This type of selectivity is common in modern insects (Bernays and Chapman, 1994) and could be due to a variety of factors, including differing phenology, varying physical and chemical plant defense, varying nutritive value of plants, and many other possibilities (reviewed in Coley and Aide, 1991). In addition, some of these interactions are likely to be examples of long-term associations between plants and insects and could be indicative of coevolved relationships (Ehrlich and Raven, 1964).
It is intriguing that no leaf miners were found on over 600 leaf specimens in the study sample. All leaf-mining families of insects had already made first appearances in the fossil record prior to the late Eocene (Labandeira and Sepkoski, 1993) and leaf mines have been documented from as early as the Late Triassic of South Africa (Scott et al., 2004; Labandeira and Anderson, 2005). In addition, eleven insect genera that are known to leaf mine on modern plants have been described from the Florissant deposits. Although not found in the study sample, there are two known examples of leaf mining damage at Florissant, one on a leaf of Trichilia (Fig. 4 in Meyer, 2003; specimen FLFO-3514) and another on a leaf of Koelreuteria alleni (pers. observation, one leaf mine found among ∼1950 leaves examined; specimen UCM 34992). So, although leaf mining is rare, it did exist at Florissant. This also suggests that one must collect or examine very large sample sizes from assemblages in order to capture any evidence of leaf-mining. There are a few possible explanations for the rarity of leaf mining damage on Florissant leaves. The first possibility is that leaf miners were less common in the Eocene than in modern communities, where leaf-mining levels have been documented to be as high as 13.4–27% of leaves in a sample having leaf mines (Faeth et al., 1981; Basset, 1991). Second, it is possible that leaf mining insects had a preference for herbaceous plants. Lower levels of herbivory have been documented on canopy leaves than on understory leaves in some modern communities (Lowman and Heatwole, 1992; Barone, 2000). If this was the case, we might expect leaf mines to be rare to absent in fossil assemblages, because herbaceous, understory plants are known to be underrepresented in fossil assemblages (Spicer, 1981; Ferguson, 1985; Greenwood, 1992). Finally, leaf mines have been shown to be more apparent on the lower surface of modern leaves, (Smith, 2004) and even if fossil leaves are preserved in a 1:1 ratio of adaxial to abaxial surfaces (Gastaldo et al., 1996), this could decrease the sample of leaf mines in fossil assemblages by as much as 50% if both halves of specimens are not collected. However, at Florissant, where specimens typically have both parts and counterparts and often preserve both upper and lower surfaces of leaves, miners should be more fairly represented.
Florissant Compared with the Green River Formation
In the interior Rocky Mountain region of North America, levels of herbivory declined from the middle to the late Eocene. Not only did the amount of leaf area removed by insects decrease, but the number of leaves that were fed upon also declined. Further evidence of the decline in levels of herbivory can be found in the family and genus comparisons, where damage levels decreased over time in the families Salicaceae and Sapindaceae. However, changes in the Salicaceae are predominantly due to significant decreases in damage levels in the genus Populus over time. All other groups analyzed showed no significant difference in damage levels.
If many of the modern insect-plant associations were established, or were becoming established, in the middle Eocene, why were herbivory levels lower in the late Eocene Florissant flora? One possible explanation may involve differences in paleoelevation (which likely reflects differences in climate) between the two sites, although interpretation of this is complicated by differences in global climate between these times, as discussed below. As cited earlier, elevation estimates for Green River range from 1564 to 2900 m and for Florissant from 1900 to 4133 m, generally suggesting that Florissant was several hundred meters higher than Green River. Although not tested directly, modern herbivory levels may decline with higher elevation, as is seen in comparisons of high-elevation wet tropical sites, which have lower levels of herbivory (4.3% of leaf area removed), with low-elevation wet tropical sites, which have higher levels of herbivory (5%–10% leaf area removed) (Smith, 2000). An examination of the relationship between elevation and herbivory levels in modern forests would help to resolve this issue—although plant and insect diversity and temperature are usually correlated with elevation change as well. This may confound an understanding of the factors that most strongly influence herbivory.
Another related explanation for the differences in levels of herbivory between the Green River and Florissant floras may be cooling temperatures in the middle to late Eocene (Wolfe, 1994; Zachos et al., 2001). A correlation between climate change and herbivory levels has been documented for the Paleocene through the middle Eocene (Wilf and Labandeira, 1999; Smith, 2000; Wilf et al., 2001). These studies have documented an increase in levels and types of herbivory as temperatures increased and a decrease in levels and types of herbivory as temperatures decreased. A similar relationship has been identified in modern ecosystems, where decreases in latitude appear to be correlated with increases in temperature and increases in insect feeding damage levels (Coley and Barone, 1996).
Although climate is likely to play a role in plant-insect interactions, the simple pattern of increasing herbivory levels with increasing temperature is probably not the whole explanation. For example, modern increases in insect herbivory with decreasing latitude are correlated not only with changes in climate, but also with a general increase in insect (Erwin, 1982) and plant diversity (Chazdon and Denslow, 2002) and with changes in levels of plant defense (Coley and Aide, 1991). In the fossil record, we see a general increase in insect diversity from the early Cenozoic onward at both the familial (Labandeira and Sepkoski, 1993) and generic (Farrell, 1998) levels, with a corresponding general trend of increasing angiosperm diversity (Niklas et al., 1985). The interplay between insect herbivory and other biotic as well as abiotic factors (such as temperature, precipitation, seasonality, levels) is likely to become clearer with soil chemistry, and CO2 the addition of other studies that focus on insect feeding damage in Cenozoic assemblages, and through a greater understanding of how these variables interact in modern ecosystems.
In addition to helping clarify overall patterns of herbivory, long-term studies of herbivory in the fossil record allow us to track host-specific interactions in space and time. For example, the host-specific interactions seen at Florissant and Green River on Cardiospermum, Cedrelospermum, and Populus provide excellent examples of specialized interactions that existed for at least a 9 m.y. period. Ehrlich and Raven (1964) predicted that specialized plant-insect associations would be maintained for long periods of time. One must wonder, how long do these types of interactions last? Do any or all of the host-specific associations between Florissant and Green River plants and their insect herbivores still exist today? If not, how much longer did these association last? If plant and insect interactions promote the diversification of these groups as Ehrlich and Raven suggested, then should we expect to find an increase in specialized feeding damage that coincides with increases in plant and insect diversity? These questions can be answered through further investigation of the Cenozoic fossil record.
There were four examples of specialized plant-insect associations that were maintained from the middle to the late Eocene and there are several examples of host-specific associations in the Florissant assemblage. However, these values based on the fossil record should be viewed as underestimates, because more specialist associations were likely to have been maintained over these periods. Host-specific associations are probably underestimated in the fossil record because it is nearly impossible to identify specialist feeders that make non-stereotyped damage. Only the damage made by insects that feed in a consistent and distinctive manner, on the same or similar plant groups, can be recognized as specialist damage. Even though many externally feeding insects, especially margin feeders, specialize on only one plant species, they do not make a distinctive and recognizable damage pattern. For this reason, the number of specialists found in the fossil record is likely to be undercounted. Only highly stereotyped damage patterns, such as those made by galling and leaf-mining insects, should be compared through time.
The diversity of galling damage in the Florissant flora is higher than in the Green River flora. Today, there is a pattern of higher species diversity of galling insects in drier habitats that have nutrient-poor soils, and it has been hypothesized that galling insects have lower survivorship in wetter environments due to increased fungal and pathogenic attack (Fernandes and Price, 1991). The increase in the diversity of insect galls in the Florissant flora may be attributable to a relatively drier environment at Florissant than at Green River. Even very slight differences in precipitation levels or seasonality can result in dramatic differences in patterns and levels of insect herbivory (Smith and Nufio, 2004). One might therefore predict that galling insects would be even less diverse in the early Eocene, when conditions were comparatively warmer and wetter, than in the middle or late Eocene. Ongoing studies of middle Eocene insect assemblages (e.g., Green River Formation, Colorado, and Republic, Washington, USA; Messel, Germany) will make clear whether this prediction holds true.
The quality of preservation in the Florissant and Green River Formations provides an excellent opportunity for studying plant-insect interactions. Cenozoic insects do appear to express some level of host-specificity in their feeding preferences. In addition, there is evidence for long-lasting associations between insect herbivores and their host plants, which may indicate coevolution between these groups. Comparisons of insect damage from the Florissant and Green River Formations show a decline in insect damage through time that may be attributable, in part, to global cooling. Evidence of an increase in galling diversity during the same time interval may also be related to climate changes that occurred during this time interval.
I thank the National Park Service for permission to collect and for the use of their collections, facilities, and the paleontologic database. Many thanks to H.W. Meyer for his support, advice, and help with fossil plant identifications. A. Kinchloe helped with many phases of the Florissant project. I thank C.R. Nufio, F.A. Smith, M. Lyon, and R. Crain for assistance with field collecting and the curation of the Green River fossils. K.W. Flessa, H.W. Meyer, J.T. Parrish, C.C. Labanderia, C.R. Nufio, and two anonymous reviewers provided valuable suggestions on the early drafts of this manuscript.
Financial support for this project was provided by the American Museum of Natural History, the Colorado Natural Areas Program, the Department of Geosciences and the Graduate College of the University of Arizona, the Geological Society of America, a National Science Foundation Predoctoral Fellowship, the Research Training Grant for the Analysis of Biological Diversification–University of Arizona, and Sigma Xi.
Figures & Tables
Paleontology of the Upper Eocene Florissant Formation, Colorado
- Florissant Fossil Beds National Monument
- Florissant Lake Beds
- Garfield County Colorado
- Green River Formation
- herbivorous taxa
- statistical analysis
- Teller County Colorado
- United States