Rosaceous Chamaebatiaria-Like Foliage from the Paleogene of Western North America

Chamaebatiaria and Chamaebatia, two characteristic genera of the Californian floristic province, are traditionally placed in different subfamilies of Rosaceae , Spiraeoideae and Rosoideae, respect ively. Analysis of the foliar and reproductive characters of the extant species of these genera indicates that the two genera could be closely related and the assignment of Chamaebatia to Rosoideae invalid. Fossil leaves of lineages of both genera occur in the Paleogene montane floras of the Rocky Mountain region and provide evidence that the two lineages diverged from a common ancestor in the Eocene . The common ancestor probably was adapted to sunny habitats in mesic coniferous forest , and , during the post-Eocene, the two lineages were able to adapt to progressively drier climates. A third extant genus, the east Asian Sorbaria, also appears to be closely related to the California genera and to have been derived from the same common ancestor. New taxa and combinations proposed are: St onebergia columbiana. n. gen. and n. sp.; Salmonensea prefoliolosa (R. W. Br.), n. gen. and n. comb.; Stockeya creedensis (R. W. Br.), n. gen. and n. comb.; Stockeya montana, n. sp.; and Sorbaria wahrhaftigii,


INTRODUCTION
Similarities in foliage to Chamaebatia foliolosa Benth.led Maximowicz to rename Spiraea millefolium Torr.as Chamaebatiaria millefolium (Torr.)Maxim.Despite the similar aspect of foliage of the two genera (e.g., pinnatisect, markedly viscid and with stipitate glands, and sweetly aromatic), however, the two genera are placed in different subfamilies.The monotypic Chamaebatiaria (Porter) Maxim., because of the presence of several ovules that ripen into follicles is placed in the putatively primitive Spiraeoideae, whereas the bitypic Chamaebatia Benth., because ofthe single ovule that ripens in an achene, is placed in the more advanced Rosoideae; within this subfamily, the single pistil places Chamaebatia in the Dryadeae.Are, in fact, the two genera subfamilially distinct and thus only distantly related?
Another genus of Spiraeoideae whose species were originally placed in Spiraea is the Asian Sorbaria (DC.) A. Br., of which approximately 10 species are recognized.Indeed, Chamaebatia millefolium was once placed in Sorbaria by Foche (1888).Like Chamaebatiaria and Chamaebatia, Sorbaria has foliage that is pubescent and fundamentally pinnatisect (although typically described as com-pound).Spiraea contrasts with these other three genera in having an undissected (although toothed) lamina and in not having stipules.
The oldest previously described fossil leaves of the Chamaebatiaria-Chamaebatia type are Chamaebatia prefoliolosa R .W. Br., based on material from the late middle Eocene (ca.37-40 million years or Ma) Salmon flora ofIdaho (Brown 1935).This taxon is also reported to occur in the late  to earliest  Ma) Ruby flora of southwestern Montana (Becker 1961) and the latest Miocene (ca.5-8 Ma) Mulholland flora of west-central California (Axelrod 1944).Brown (1937) also described a second taxon, Chamaebatiaria creedensis, from the late Oligocene (ca.25-27 Ma) Creede flora of southern Colorado, and the same species was detemined by Axelrod (1939) from the early Miocene (ca.18 Ma) Tehachapi flora of southern California.Brown (1935) considered the Salmon and Creede species to be indistinguishable from their extant relatives, and a recent treatment of the Creede taxon (Axelrod 1987) offered no morphologic criteria for separation of the Creede from the extant species.If so, other than establishing a minimal age for their respective genera, the fossil species have no bearing on the phyletic relationships of these genera.However, as discussed below, these fossils have not been critically examined previously; the Eocene and Oligocene fossils cited are unquestionably distinct from any extant species and sufficiently distinct from extant relatives to merit description as extinct, although probably ancestral, genera.
The previous fossil record of Sorbaria includes undescribed foliage from the late Eocene of Hokkaido (T.Tanai, written comm., October, 1987).In North America, S .hopkinsii (Wolfe) Wolfe & Tanai is known from the Miocene of Alaska (Wolfe and Tanai 1980), although reference has been made to an undescribed, pre-Miocene species (Wahrhaftig, Wolfe, Leopold, and Lanphere 1969).
Recent collecting from the early middle Eocene (ca.48-50 Ma) One Mile Creek locality of the Princeton flora of southern British Columbia has produced leaves ofa markedly pinnatifid to pinnatisect rosaceous taxon, described below as Stonebergia columbiana Wolfe & Wehr.Analysis of these leaves indicates that they have strong phenetic similarities to Chamaebatiaria, Chamaebatia, and Sorbaria.Although leaves of the extant species of these two genera have only a few synapomorphic characters, Eocene representatives of the lineages strongly indicate that they diverged from "a common ancestor in the Eocene after other putatively related genera in Spiraeoideae and Rosoideae were recognizable.

EXTANT TAXA
The leaves ofChamaebatia and Chamaebatiaria have been considered by some botanists as pinnately compound, and hence terms such as "bipinnate" and "tripinnate" are applied.We consider that true compound leaves are formed by discrete laminar units that dehisce; thus, true pinnately compound leaves occur in, for example, Leguminosae, Juglandales, and Rutanae.The leaves of both Chamaebatia and Chamaebatiaria, however, dehisce only at the stem and are thus basically simple.The foliage of Sorbaria appears to be truly compound, but again the unit that dehisces is typically the entire "compound" leaf rather than the "leaflets."Pinnate leaves that are compound in having discrete laminar units that dehisce also typically have lateral leaflets that are arranged opposite to one another, whereas in Chamaebatiaria, Chamaebatia, and Sorbaria, the "leaflets" are typically subopposite.Terms such as "pinnatifid" (i.e., deeply lobed) and "pinnatisect" (i.e., with lamina or laminar segments completely dissected to primary, secondary, or tertiary veins) are hence appropriate (Fig. I).We suggest that leaves that have the appearance of compound leaves but that dehisce as entire units be termed paracompound.Whereas in pinnatisect leaves (or laminar segments) the dissected segments have a broad base of laminar attachment, in paracompound leaves attachment is only by vasculature.
Whether true pinnately compound leaves occur in Rosaceae is questionable, although some species of Sorbus L. may have achieved this state.Because pinnately compound leaves are probably basic for subclass Rosidae (Hickey and Wolfe 1975), the same might be expected to be true for Rosaceae or any rosid family that appears to have compound foliage.However, saxifragalean families most closely related to Rosaceae (Cronquist 1981;Thome 1983) and presumably having a sister relationship to the stock from which Rosaceae arose are simpleleaved.Thus, within Rosaceae a reasonable assumption is that the trend is from pinnatifid to pinnatisect to paracompound (and probably to compound) leaves.In this context, the pinnatisect leaves of Sorbaria and the bipinnatisect leaves of Chamaebatiaria are obviously specialized relative to most other members of Spiraeoideae, as are the paracompound, thrice-divided leaves of Chamaebatia relative to other members of Dryadeae.
In the systematics section below, diagnoses for foliage of Chamaebatiaria, Chamaebatia, and Sorbaria are included.Some characters in common between the two genera are probably plesiomorphies, e.g., the basically pinnate venation and the glandular vestiture, because these characters are found in many genera of Rosaceae and allied families.On the other hand, the totally paracompound primary lamina and pinnatisect secondary laminar segments as in Chamaebatiaria and Chamaebatia are not found in other genera of Rosaceae; in some instances (e.g., Poteridiumi, secondary laminar segments are deeply pinnatifid but they are not pinnatisect. In inflorescence, floral structure, and fruit, Chamaebatiaria and Sorbaria have the characters common in, or defining of, Spiraeoideae and, in particular, the tribe of Sorbarieae.The flowers are arranged in a panicle, which has numerous flowers.The hypanthium is campanulate (or turbinate), the floral formula is K5/ C5/Aoo, and five pistils are present.Several ovules are present and the fruits are follicles.
The flower and fruit characters just listed are generally considered to be primitive for Rosaceae.From a cladistic standpoint, presence of only plesiomorphic characters should not be taken as evidence of close affinity.Indeed, the foliage of Chamaebatiaria and Sorbaria is highly specialized and indicates a long divergence from other members of Spiraeoideae.
For a member of Dryadeae, Chamaebatia is anomalous in having a manyflowered panicle, which is putatively plesiomorphic for Rosaceae.Cowania, Pur- shia, and Dryas, for example, typically have solitary flowers.The stamens of Chamaebatia are also in several series (a presumably primitive feature), whereas the stamens in other members of Dryadeae are in one to three series.These primitive characters contrast with the advanced foliar characters of Chamaebatia.
A single ovule is present in Chamaebatia, and primarily this character, plus the achene and single pistil, have been used to place the genus in Rosoideae.Although achenes are, in Rosaceae, generally considered to be restricted to Rosoideae, Holodiscus (Spiraeoideae, Holodisceae) also has achenes.An achene has clearly evolved more than once in Rosaceae.Only the single ovule and the single pistil are left as synapomorphic characters shared between Chamaebatia and other members of Dryadeae.
Reduction in number of ovules has also clearly occurred more than once in Rosaceae.Holodiscus, for example, has two ovules and Lyonothamnus four; ovular number is variable in Physocarpus (2-4) and Spiraea (2-several).The Prunoideae also have a single ovule.Similarly, number ofpistils has also been reduced more than once in Rosaceae.Indeed, Prunoideae characteristically have a single pistil, and in Spiraeoideae Physocarpus has one to five.We cannot, therefore, find any substantive characters that unquestionably relate Chamaebatia to Rosoideae and to Dryadeae in particular.On the other hand, the foliar similarities between Sorbaria, Chamaebatiaria, and Chamaebatia, (and particularly the last two) although few, are fundamental.The advanced floral and fruit characters in Chamaebatia are all reductions that could be derived from the plesiomorphic conditions in Chamaebatiaria or Sorbaria.Stonebergia, which is the oldest known fossil of the Chamaebatiaria-Chamaebatia type, has conspicuous, toothed, bilobed stipules; along the margin of the stipule (as well as the petiole) are numerous, simple hairs, a few of which are capped by glands (some other hairs are flat-topped and thus probably were also glandular).On the lamina proper, hairs were simple or stellate.The primary lamina is pinnatifid to pinnatisect, and parts ofthe primary lamina can be paracompound; the secondary laminae (or primary lobes) are pinnatifid, with large teeth on the lobes (Fig. 2-5).The teeth are obtusely rounded with both apical and basal sides convex (AI of Hickey 1973).In the pinnatisect parts of the lamina, the intersecondary laminar segments can be entire-margined but are typically toothed.The primary lamina has a maximum of nine divisions per side.Venation tends to be obscure (probably related to immersion of the venation in a thick and/or hairy lamina), but the lobes of the secondary laminar segments are entered medially by a strong tertiary vein.The tertiary veins branch close to the point of departure from the secondary vein, with the apical branch extending toward the sinus between lobes.In specimens in which the third-order laminar segments are deeply lobed, the quartary veins have a pattern similar to the tertiary veins (Fig. 2).The tertiary veins are craspedodromous and give off thin quartary veins that form brochidodromous loops.
The stipules of Stonebergia (Fig. 48) are similar to, but larger than, the bilobed stipules of Chamaebatia, in which the lower part is also fused to the petiole; the stipular lobes in Stonebergia are also much more prominent.In Chamaebatiaria and Sorbaria the stipules are entire-margined.In all three extant genera, as in Stonebergia, the foliage is densely stipitate glandular.The hairs of Stonebergia are either simple or stellate as in Sorbaria and Chamaebatiaria; the hairs in Chamaebatia are exclusively simple.The primary laminae of Chamaebatiaria and Chamaebatia are invariably paracompound; the ultimate lam inar divisions are always entire-margined in Chamaebatiaria (Fig. 17) and Chamaebatia (Fig. 9) rather than toothed as in Stonebergia (Fig. 2, IS) and Sorbaria (Fig. II).The teeth of Stonebergia and the ultimate laminar segments of Chamaebatiaria and Chamaebatia are rounded, whereas the teeth of Sorbaria are sharp.Further, the tertiary and quartary veins are unbranched in Chamaebatiaria and Chamaebatia, and the fifth-order venation of Chamaebatia is open (Fig. 9), rather than forming a closed reticulum.Stonebergia is rare but two specimens were deposited together (Fig. 19, 20) suggesting that, like the extant Sorbaria, Chamaebatiaria, and Chamaebatia, the leaves adhered to one another because they were resinous and/or had glandular pubescence.Thus, while sharing certain similarities with leaves of Chamaebatiaria and Chamaebatia, the leaves of Stonebergia are distinguished by what are probably primitive characters for Rosaceae: a less dissected lamina, numerous teeth, and a more elaborate venation pattern.The degree of dissection of the primary lamina is less and the venation pattern is less rigid in Stonebergia than in Sorbaria, but these two genera have about the same number of teeth.In the late Eocene Salmonensea prefoliolosa (R. W. Br.) Wolfe & Wehr, which was previously assigned to Chamaebatia, the primary lamina is consistently paracompound, and the secondary laminar segments are deeply pinnatifid (Fig. 6).Unlike Chamaebatia, in which the tertiary laminar segments can also be pinnatisect (Fig. 8), the tertiary segments in Salmonensea are only pinnatifid.Salmonensea also has a closed system of fifth-order veins (Fig. 7).The branching of the quartary veins that enter the lobes in Salmonensea is similar to the branching of the quartary veins in Stonebergia.The stipules in one specimen of Salmonensea are represented by a vague outline that has small lobes; apparently the stipule was almost entirely fused to the petiole.Salmonensea is specialized relative to Stonebergia in having a more highly dissected lamina and reduced stipule but is clearly less specialized than Chamaebatia.
Except for one fragment, the Ruby leaves previously referred to Chamaebatia prefoliolosa represent a genus and species distinct from the Salmon Salmonensea prefoliolosa.In the Ruby taxon, which is described below as Stockeya montana Wolfe & Wehr, the tertiary laminar segments are totally untoothed, as are the intersecondary laminar segments.The branching tertiary veins in Stockeya (Fig. 13) resemble the branching tertiary veins in Stonebergia.Stockeya shares with Salmonensea a consistently paracompound primary lamina.Of the four known specimens of Stockeya montana, three were deposited adhering to one another (Fig. 27, 28); as with Stonebergia, this suggests foliage that was resinous and had glandular pubescence.
The Creede species formerly assigned to Chamaebatiaria is below reassigned to Stockeya as S. creedensis (R. W. Br.) Wolfe & Wehr; the Ruby and the Creede species have the same degree and order of dissection, both typically lack teeth on the lobes , and both have branching tertiary veins.In S. montana, however, the apical branch of a tertiary vein is as thick as the more basal, craspedodromous branch (Fig. 13), and the intersecondary laminar segments are contiguous and numerous (Fig. 51); in S. creedensis, the apical tertiary branch is weak (Fig. 15), and the intersecondary laminar segments are widely spaced and few.One specimen of S. creedensis (Fig. 50) has some lobes (tertiary laminar segments) that have a few teeth.Although teeth are not known in S. montana, obviously this condition was rare in S. creedensis and could also have been rare in S. montana, which is known from only four specimens.
The two species of Stockeya can be viewed as forming a transition from Stonebergia to Chamaebatiaria.The apical branch of the tertiary vein, strong in Stonebergia and Stockeya montana, is weak in S. creedensis and absent in Chamaebatiaria.The tertiary laminar segments are toothed in Stonebergia but are typically entire-margined in Stockeya and are invariably entire-margined in Chamaebatiaria.The intersecondary laminar segments are large and toothed in Stonebergia , untoothed but contiguous and numerous in Stockeya montana, and untoothed but separated and few in S. creedensis and Chamaebatiaria.
Foliage of the latest Eocene to earliest Oligocene Sorbaria wahrhaftigii Wolfe & Wehr has no characters that separate the species at the generic level from extant species of Sorbaria.Although areoles in the latter are typically large, the areoles are small in S. wahrhaftigii, as in Stonebergia.The tertiary vein that braces the major sinuses between groups of teeth in Sorbaria wahrhaftigii (and most extant species of the genus) originates from the apical side of a secondary vein, as in Stonebergia (Fig. 10, II, IS).The extant species of Sorbaria examined typically have one subsidiary tooth (a second tooth can be present) on the abmedial side of a primary tooth, but in Sorbaria wahrhaftigii, as in Stonebergia, up to three subsidiary teeth are present.Of the 23 specimens of Sorbaria wahrhaftigii collected, only 5 represent isolated secondary segments, and the other 18 have a total of 53 laminar segments attached.Sorbaria wahrhaftigii was clearly paracompound.
However, in some characters, Sorbaria wahrhaftigii is specialized relative to some extant species of the genus.In sharpness of the apices of the secondary laminar segments and overall outline of groups of teeth, S. wahrhaftigii (Fig. 42) is unlike Stonebergia and thus probably specialized.Less specialized, for example, are the secondary laminar segments of the extant Sorbaria pallasii (G .Don.) A. Poj ., which tend to have apices and groups of teeth with a rounded outline (Fig. 40); this morphology is particularly similar to one specimen of Stonebergia (Fig. 21).Sorbaria wahrhaftigii, like most extant species of the genus, has no intersecondary laminar segments, but these are present in the extant S. sorbifolia.Except for these last two advanced characters, S. wahrhaftigii could be considered as a direct ancestor of all other species of the genus.The presence of these advanced characters indicates that Sorbaria had already undergone some diversification by about 35Ma.

EVOLUTIONARY AND SYSTEMATIC CONSIDERATIONS
Both the Salmonensea-Chamaebatia and Stockeya-Chamaebatiaria lineages are morphologically transitional back into Stonebergia.We do not suggest that Stonebergia columbiana is unquestionably the direct and distant ancestor of Chamaebatia and Chamaebatiaria (as well as of Sorbaria), although this could be valid.What we are stating (Fig. 18) is that on morphological grounds, the leaves of the common ancestor of these three extant genera would be assignable to Stonebergia and would represent S. columbiana or a related species.Further, as conceptualized here, Stockeya and Salmonensea would contain the ancestors of Chamaebatiaria and Chamaebatia, respectively, although the described species of these extinct genera may not be in the direct lines ofdescent.Insofar as known, the described species of Stockeya and Salmonensea have no specialized characters that would remove these taxa from the direct lines of descent.
Based on known diagnostic features, a strict cladistic viewpoint would place Stockeya in Chamaebatiaria and Salmonensea in Chamaebatia, because these fossil genera are segregated from their closest modern relative only on inferred primitive features.We are , however, operating in a classificatory framework based on typology and morphologic similarities when terms such as "Rosaceae" are used.We suggest that were the species of Stonebergia, Stockeya, and Salmonensea still extant, systematists dealing with Rosaceae would, on morphologic distinctiveness, make the same generic separations that we have made.
On the other hand, we emphasize the desirability of making typologie-morphologic categories correspond to monophyletic groups.As currently defined in rather simplistic characters such as pistil/ovule number and general type of fruit, Rosoideae contain more than one lineage that arose from a common ancestor that would not also be a member of Rosoideae; Rosoideae are, therefore, an unnatural or polyphyletic group.The present subfamilial classification ofRosaceae needs to be reexamined, especially using cladistic methodology and the fossil record.
Whether any other extant genera of Rosoideae are closely related to the Chamaebatiaria-Chamaebatia group is uncertain.Some species of Potentilla also can have a paracompound primary lamina and a pinnatifid secondary lamina; in the species examined, however, intersecondary laminar segments are absent.Further, bracing of the sinus between lobes is typically accomplished by a tertiary vein separate from the tertiary vein that enters the lobes.Such differences, added to the fact that a transition from a pinnatisect to paracompund lamina occurs within Potentilla, suggest that the paracompound lamina of Potentilla arose separately Stonebergia Fig. 18.Inferred phyletic relationships of Stonebergiaand its descendant genera .I Primary lam ina only paracompound, secondary laminar segments only pinnatifid. 2 Secondary laminar segments deeply pinnatifid, tertiary laminar segments pinnatifid, stipule size reduced .3 Secondary laminar segments pinnatisect, tertiary laminar segments pinnatisect, quartary veins freely ending.4 Intersecondary laminar segments discrete and untoothed, stipule lanceolate.5 Secondary lam inar segments not pinn atifid and strongly serrate.6 Tertiary laminar segments mostly untoothes.7 Secondary laminar segments pinnatisect, quartary veins irregularly brochidodromous.
from that of the Chamaebatia, i.e., the relationship between Potentilia and Chamaebatia predates the origin of Stonebergia.Capitate glandular hairs also occur in Dryadeae such as Cowania, but the presence of such hairs is probably plesiomorphic (e.g., such hairs occur in some species of Physocarpus of Spiraeoideae and in Grossulariaceae).Irrespective of the possible relationships of other Rosoideae to Chamaebatia, this genus clearly belongs with Chamaebatiaria and Sorbaria in any classificatory scheme.
Note, however, that one ofthe suggested synapomorphic foliar characters shared by Chamaebatiaria and Chamaebatia, namely the pinnatisect secondary laminar segments, was independently achieved in each lineage.Stonebergia is uncanalized (cf.Upchurch 1985) in regard to laminar dissection.Some specimens (e.g., Fig. 19b,20a,21,45) have a deeply pinnatifid primary lamina, whereas others (e.g., Fig. 19a,20b,21,23,49) have a totally pinnatisect primary lamina.The first group also tends to have only moderately pinnatifid secondary laminar segments, whereas the second group tends to have deeply pinnatifid (almost pinnatisect) secondary laminar segments.However, some specimens (e.g., Fig. 19a,20b,22) have a pinnatisect primary lamina but only moderately dissected secondary laminar segments.The presumed immediate descendant of Stonebergia would have had a totally pinnatisect to paracompound primary lamina and probably uncanalized moderately to deeply pinnatifid secondary laminar segments.In the divergence leading to Salmonensea, the secondary laminar segments became only pinnatisect; this occurred within the Eocene.In the divergence leading to Sorbaria and Stockeya, on the other hand, the secondary laminar segments were first only moderately pinnatifid (Stockeya montana), became deeply pinnatifid by the late Oligocene (Stockeya creedensis), and subsequently became pinnatisect during the Neogene, when Chamaebatiaria is presumed to have evolved.In this regard , what both lineages share is a strong tendency toward laminar dissection, a tendency developed in Stonebergia and one perhaps based in genetic structure.
Although Sorbaria has the same tendency to dissect the primary lamina, no such tendency has occurred in regard to the secondary laminar segments.Indeed, as we have interpreted the phyletic relationships in this group, the foliage of Sorbaria reversed somewhat the trend seen in Stonebergia , i.e., the secondary laminar segments of Sorbaria are less dissected than those of Stonebergia.The less-dissected condition in the secondary laminar segments of Sorbaria could represent a more primitive state, but this would necessitate deriving the paracompound primary lamina more than once.Reversal of the trend toward dissection of the secondary laminar segments could be due to the high-latitude environments, which the Sorbaria lineage entered during the Eocene.Plants in these environments characteristically, from a variety of factors, have larger leaves than lower-latitude relatives (Wolfe 1985).Assuming that the derivation of Sorbaria from a Stonebergia descendant that had an already paracompound primary lamina is valid, then expansion of the lamina in response to high-latitude conditions would result in expansion of secondary laminar segments.

BIOGEOGRAPHIC CONSIDERATIONS
As emphasized elsewhere in relation to Acer (Wolfe 1987;Wolfe and Tanai 1987), the montane Eocene flora of western North America may have played a, if not the, major role in the origin and diversification of many "arcto-tertiary" taxa.This certainly seems to be valid for Spiraeoideae.The early middle Eocene montane assemblages such as the Republic and Princeton have many representatives of extant woody genera (or related but extinct genera) of Spiraeoidae, including those that are geographically and ecologically now distant from one another, strongly suggesting that the period of 48-50 Ma was geologically a short time after the origin and initial diversification of Spiraeoideae and that the diversification had occurred in this same geographic region.
Further diversification occurred after 48-50 Ma in this same region , as exemplified by Sorbaria, Chamaebatiaria, and Chamaebatia.Sorbaria can be considered as an east Asian vicariant of the western American Stockeya-Chamaebatiaria lineage , a vicariant relation that probably developed during the post- Eocene as Stockeya adapted to progressively drier climates in southwestern North America and became geographically isolated from the northern Sorbaria .Note, however, that vicariance in this group did not develop between Asia and North America until after the Miocene, because Sorbaria persisted in Alaska into the latest Miocene.
In Raven and Axelrod's (1978) analysis of the California flora, they recognized the basically northern relationships of Chamaebatiaria.a relationship confirmed in the present report.Raven and Axelrod, however, included Chamaebatia with a group of endemic California genera whose biogeographic relationships were ambiguous.Our analysis indicates that Chamaebatia, along with Chamaebatiaria, was originally of more northern occurrence and has subsequently entered and become restricted to California.Henrickson (1985) has also suggested that several specialized, dry-adapted genera of Spireeae (Kelseya, Luetkea, Petrophytum, Xerospiraea) similarly had an origin from a northern lineage of Spiraea.

ECOLOGICAL CONSIDERATIONS
The early middle Eocene assemblages from northeastern Washington and adjacent British Columbia represent mesic coniferous forest (Wolfe 1987;Wolfe and Wehr 1987).The One Mile Creek assemblage, which contains Stonebergia, is preserved in lacustrine rocks.Among the dicotyledons, the most abundant leaves are those of Betula leopoldae Wolfe & Wehr (a birch of the generallutea alliance; cf.Crane and Stockey 1987) and a true Cercidiphyllum; many of the birch leaves are> 10 em long .Also common are Aesculus, Prunus (3 spp.), Ulmoideae, and Tsukada davidiifolia Wolfe & Wehr (Davidiaceae).Less common are Acer (7 spp.; cf.Wolfe and Tanai 1987) , Spiraea, Pterocarya, and Fagus ; Tetracentron, Amelanchier, Rubus, and Grossularia are rare.The leaves of Aesculus are large and can occur as complete compound leaves, suggesting little transport.Although some of the leaves display insect damage and biological degradation, little mechanical breakage is evident, again suggesting little transport.The data indicate that the leaves of Stonebergia were not transported far and that the plants were in a mesic, summer-wet climate.
However, the presence of foliar pubescence in Stonebergia may indicate that the plants occupied sunny habitats (Coley 1983), assuming that the adaptive pattern found in megathermal plants holds for microthermal plants.This inference is consistent with the dissection of the lamina, which probably decreased the effective diameter of the leaf and hence would have lowered temperature in full sunlight.Some other taxa, especially members of Rosaceae, in the Princeton-Republic type assemblages also have a tendency to laminar dissection, although not to the extent of Stonebergia.
Salmonensea and Stockeya montana from the late middle to late Eocene are also part of lacustrine assemblages that represent mesic coniferous forest (Wolfe 1987).Although Becker (1961) listed a "xeric element" (including "Chamaebatia" = Stockeya) for the Ruby flora, based largely on present ecology of supposed living equivalents, this list included some obvious presently mesic genera, such as Dipteronia and Koelreuteria.Becker's suggestion that the "xeric element" occupied a woodland climate many kilometers to the south is not consistent with either the lacustrine depositional setting or taphonomic studies (Spicer and Wolfe 1987).The known species of the Chamaebatiaria-Chamaebatia group were probably still occupying mesic, summer-wet climate as late as about 33-37 Ma .The greater dissection of the lamina in the later Eocene taxa, however, suggest further adaptation to a sunny/dry habitat.
The late Oligocene Creede assemblage largely represents a montane coniferous forest, but with a large element (including "Chamaebatiaria" = Stockeya creedensis) interpreted on the basis of supposed living analogs as woodland (Axelrod  and Raven 1985; Axelrod 1987).Strong compositional differences occur between some samples and thus may indicate the presence of both forest and woodland (Spicer and Wolfe 1987).However, leaves of Stockeya are most common in samples indicative of forest (i.e., high representation of Abies) and least common in samples indicative of woodland (i.e., high representation of Juniperus and/or Cercocarpus).Further, although Axelrod (1987) suggested annual precipitation for the Creede flora of 460-520 mm, the presence of Picea probably indicates annual precipitation> 1100 mm (Wolfe 1971).Again, we suggest that Stockeya was adapted to a relatively mesic climate, although leaf size of the Creede assemblage indicates a drier climate and possibly a more open-canopy forest than do the Ruby and Salmon assemblages.
The extant Chamaebatiaria millefolium occupies xeric habitats east of the Sierra-Cascade axis from southern California north to Oregon and occurs in vegetation ranging from sagebrush scrub to juniper woodland to open bristle-cone pine forest (Munz and Keck 1959;Thome 1982).The extant Chamaebatiafoliolosa occurs in open forest in the ponderosa pine and red fir belts along the western slope of the Sierra Nevada, and C. australis (Bdg.)Abrams occurs in chapparral in southernmost California and adjacent Baja California (Munz and Keck 1959).Members of both extant genera have vasicentric tracheids, an adaptation to drought survival in the xeric to subhumid environments occupied by these taxa (Carlquist 1985).The ancestry of these taxa, however, is in mesic Eocene vegetation to the north; the respective lineages have morphologically evolved, first adapting to sunny habitats in mesic forest and then to progressively drier climates during the post-Eocene.
Sorbaria has probably changed little ecologically from Stonebergia.In China, for example, Sorbaria can occur as an understory shrub in coniferous to broadleaved deciduous forest, although the genus is common in open areas within forested regions (Wang 1961).In the Alaskan Tertiary, the occurrences ofSorbaria probably represent similar habitats.
Nomenclature.-Type locality is in British Columbia.
Notes .-Typically the only hairs preserved are the stipular and petiolar hairs, which all appear to be simple.Fourth-and fifth-order venation is not well preserved, suggesting a thick and/or hairy lamina.One specimen (UAPC S17969A) had been infested with probable scale insects (cf.Wolfe and Tanai 1987), which reveals many laminar hair bases and some stellate hairs.
Nomenclature.-Clyde Wahrhaftig has been of invaluable assistance in the collection and stratigraphic placement of the plant assemblages from the Nenana coal field in the Alaska Range.
The Rex Creek flora was recently assigned a latest Eocene age (Wolfe and Tanai 1987).However, this flora could be of earliest Oligocene age but older than 33 Ma (we are accepting an age of 34.5 Ma for the Eocene-Oligocene boundary).
Diagnosis.-Leaves simple; petioles markedly hairy, hairs simple, some capitate glands present; primary lamina pinnately veined, paracompound; secondary lam-inar segments deeply pinnatifid; lobes typically entire-margined but can have teeth on admedial and abmedial sides; teeth acute, nonglandular; lobes furnished with a tertiary vein that branches, the abmedial branch extending toward the lobal sinus, the admedial branch craspedodromous and placed approximately medially in the lobe; quartary veins forming a single series of brochidodromous loops on either side of admedial tertiary branch; four orders of venation; teeth narrowly rounded, nonglandular, sinuses acute; intersecondary segments typically entiremargined, contiguous or isolated.Nomenclature.-The Creede species is represented by abundant, well preserved material and is, therefore, selected as the generic type.The second species assigned to this new genus is Stockeya montana Wolfe & Wehr, which is based on material from the Ruby flora of southwestern Montana.Prof. Ruth Stockey of the University of Alberta has collected extensively from the Princeton localities and has allowed us free access to these collections.
Notes.-The only hairs preserved well are those on the petiole, which, as in Chamaebatiaria, are mostly simple and glandular.Whether hairs on the laminar , .
. , .. .segments are stellate cannot be certainly determined.We reject the assignment to this species of the Tehachapi specimen (Axelrod 1939).Although stated to be part of a "bipinnate" leaf, the specimen has no petiole preserved, and thus the rank of the laminar segments preserved can only be inferred.However, Axelrod correctly observed that the ultimate segments of the fossil are obovate, and this is particularly true of the terminal segments.In Chamaebatiaria, the ultimate segments tend to overlap and have a broad area of attachment and thus typically appear as oblong to only slightly obovate as opposed to the markedly obovate ultimate segments in Chamaebatia.Chamaebatia and Chamaebatiaria can be readily separated on venation in the ultimate segments: in Chamaebatia, the major vein is placed towards the terminus of the laminar segment that has the ultimate segments and the next order of veins end freely, whereas in Chamaebatiaria, the major vein is placed medially and the next order of veins form brochidodromous loops.However, no venation is preserved in the ultimate segments of the fossil, and we consider either generic or specific assignment as highly problematic.
Hypotypes.-USNM 396067A, B, 396068A.B (all topotypes), UMMP 36912.Notes.-Axelrod's (I 944) citation of this species based on a specimen from the Mulholland assemblage of central California must be rejected because, in part, of extremely poor preservation.The specimen has tertiary veins that have an A-A orientation (Hickey 1973).But we cannot determine whether the lamina is continuous between secondary veins or is, as claimed by Axelrod (I 944) , pinnatisect.
The specimen could, if not pinnatisect, represent a leaf similar to Alnus oregona Nutt., which has AA tertiary veins that thin markedly midway between the secondary veins.Irrespective of whether the specimen has a pinnatisect primary lamina, the lamina between adjacent tertiary veins is continuous, and therefore the specimen is not twice-divided and cannot represent Chamaebatia.
The citation for the Ruby flora is based on a single fragment that apparently represents the ultimate part of a secondary laminar segment.This specimen is in a matrix of blocky shale, which, as noted by Becker (1961), is probably stratigraphically lower than the paper-shale unit, which unit has produced specimens of Stockeya montana.
Fig.I .Terminology for foliage of Chamaebatiaria, Chamaebatia, and other pinnatifid and pinnatisect plants.
FOSSIL TAXAAt least 40 taxa of Rosaceae occur in the early middle Eocene montane assemblages of the Pacific Northwest and adjacent Canada and include members of all four subfamilies.Members ofSpiraeoideae include aff Physocarpus (i.e., an extinct genus most closely allied to this extant genus), Spiraea, Sibiraea, and Stonebergia.Members of Maloideae include Photinia, aff.Sorb us, aff.Crataegus , Crataegus, aff.Malus, and Amelanchier.Rosoideae are represented only by aff.Kerria and Rubus.while Prunoideae are represented by aff.Prunus and Prunus.By the later Eocene, other genera of Rosoideae (e.g., Rosa, Cercocarpus, aff.Duchesnea) and more advanced members of Spiraeoideae (Sorbaria, Holodiscus) and Maloideae (Heteromeles, Sorbus) appeared.Such a fossil record can be interpreted as reflecting a major, generic-level diversification of Rosaceae during the Eocene.
Type species.-Stonebergia columbiana Wolfe & Wehr.Nomenclature.-We take pleasure in naming this genus for Mrs. Margaret Stoneberg of the Princeton (B.C.) Museum and District Archives.Mrs. Stoneberg has supplied both financial support for Wehr's field work and continuing encouragement to both of us.
teeth; lobes of secondary laminar segments furnished with a tertiary vein that branches, the abmedial branch extending toward the lobal sinus, the admedial branch craspedodromous and placed approximately medially in the lobe; quartary veins also forked in a manner similar to the tertiary veins if laminar segments Kenai Group at Redoubt Point has tertiary veins intermediate in number between , p. 72, pI.22, fig.I I, 12.