Staminate inflorescences with in situ pollen from Eocene Baltic amber reveal high diversity in Fagaceae (oak family)

Abstract: Eocene Baltic amber forms the largest amber deposit worldwide; however, its source vegetation and climate are much debated. Representatives of the oak family (Fagaceae) were abundant in the Baltic amber source area based on numerous inclusions of staminate inflorescences or individual florets, previously assigned to Castanea and Quercus. However, the actual generic and infrageneric diversity of Fagaceae from Baltic amber remained unknown. Using flower characteristics and section-diagnostic in situ pollen of staminate inflorescences and detached floret inclusions, we describe 18 fossil-species of Fagaceae making this family by far the most diverse plant family preserved in Baltic amber. We substantiate the occurrence of the Castaneoideae, Quercoideae (Quercus sect. Cyclobalanopsis/Lobatae; Q. sect. Lobatae; Q. sect. Protobalanus), Trigonobalanoideae and the extinct genus Eotrigonobalanus. Among the 18 fossil-species, six are described as new: Q. aimeeana, Q. casparyi, Q. multipilosa, E. campanulata, E. conwentzii, E. longianthera; and one new combination is published: Q. brachyandra (≡ Castanea brachyandra). In addition, a lectotype is designated for the name Quercites meyerianus and neotypes are designated for the names Castanea inclusa and Quercus longistaminea (≡ C. longistaminea). Members of the Fagaceae probably inhabited azonal and zonal vegetation types of the amber source area, including bottomland flood-plains and stream banks (Q. sect. Lobatae), dry habitats (Q. sect. Lobatae, Q. sect. Protobalanus), peaty soils, riparian and swamp forests (Castanopsis, Eotrigonobalanus), as well as mixed mesophytic forests (castaneoids, Quercoideae, trigonobalanoids). Affinities to extant North American and E to SE Asian floras support the recent notion that late Eocene Baltic amber (38–34 Ma) was formed in a warm-temperate climate. Citation: Sadowski E.-M., Schmidt A. R. & Denk T. 2020: Staminate inflorescences with in situ pollen from Eocene Baltic amber reveal high diversity in Fagaceae (oak family). – Willdenowia 50: 405–517. doi: https://doi.org/10.3372/wi.50.50303 Version of record first published online on 1 December 2020 ahead of inclusion in December 2020 issue.


Introduction
The Fagaceae (oak family) are the most diverse extant N temperate woody plant family, comprising ten genera (Castanea Mill., Castanopsis (D. Don) Spach, Chryso lepis Hjelmq., Colombobalanus Nixon & Crepet, Fa gus L., Formanodendron Nixon & Crepet, Lithocarpus Blume, Notholithocarpus Manos & al., Quercus L., Trigonobalanus Forman) and more than 900 species. They are chiefly distributed in the N hemisphere, occurring from Canada to Colombia, across temperate Eurasia and extending into tropical forests of SE Asia (Soepadmo 1972;Kremer & al. 2012). Quercus and Colombobalanus occur as far south as equatorial Colombia in the Americas and Castanopsis and Lithocarpus cross the equator in Malesia (New Guinea, New Britain; Govaerts & Frodin in nearly every fossiliferous specimen of Baltic amber. The Baltic amber deposit is also famous for the exquisite and three-dimensional preservation of plant organs, which are otherwise rare in the fossil record, including Fagaceae (Weitschat & Wichard 2010). For instance,  recently reported a partial pistillate inflorescence of Castanopsis, which showed a life-like preservation of all internal structures. Staminate catkins of Fagaceae are also preserved in high fidelity, showing delicate structures such as petals, anthers and trichomes, which are significant for identification. The taxonomic affinities of these catkins in the Baltic amber flora were differently interpreted in the past. Caspary (1880Caspary ( , 1881Caspary ( , 1886aCaspary ( , 1886b described 13 species of Fagaceae (12 species of Quercus and one species of Castanea), whereas Conwentz (1886) reported four castaneoid and 10 Quercus species. Kirchheimer (1937) reviewed the Fagaceae catkins from Baltic amber and placed them into three species of Quercus and one species of Castanea. The latest review by Iljinskaja (1982) accepted two species of Castanea and four species of Quercus, which significantly reduced the number of Fagaceae species from Baltic amber. All previous studies referred amber inclusions of staminate catkins to modern European genera.
Our recent survey of historic collections of Baltic amber and newly found material indicated that Fagaceae were more diverse in the source area of Baltic amber than previously suggested and that previous records need revision. Therefore, the aim of this study was to clarify the generic and infrageneric diversity of Fagaceae preserved in Baltic amber and to use the three-dimensionally preserved fossils to investigate their evolutionary and palaeogeographic implications.
This study focuses on the staminate inflorescences of Baltic amber Fagaceae. Evaluation of the amber fossils required a thorough assessment of the morphology of extant fagaceous staminate catkins, as comprehensive comparative studies on staminate inflorescences of the Fagaceae are still not available. Up to now, there are only few studies that mention or briefly describe the morphology of fagaceous catkins (e.g. Trelease 1924;Hjelmqvist 1948). Therefore, we studied herbarium specimens of the Fagaceae to investigate the morphology of extant staminate inflorescences and to assess whether they exhibit genus-diagnostic features. We then provide a comprehensive revision of fagaceous inclusions from Baltic amber, including type material of museum collections as well as newly discovered specimens. For assignment of these inclusions we also used in situ pollen extracted from the anthers of the staminate catkin inclusions. Numerous previous studies of Fagaceae pollen (fossil and extant; Solomon 1983aSolomon , 1983bDenk & Grimm 2009;Makino & al. 2009;Denk & al. , 2012Grímsson & al. 2015Grímsson & al. , 2016aGrímsson & al. , 2016b have shown that pollen provides sufficient information for distinguishing genera and subgroups of the Fagaceae. Our investigation of extant Willdenowia 50 -2020 herbarium material, amber specimens and their in situ pollen revealed a highly diverse fagaceous flora of the Baltic amber source area, encompassing Quercus sect. Lobatae, Q. sect. Protobalanus (Trel.) O. Schwarz, trigonobalanoid and castaneoid taxa, as well as the extinct genus Eotrigonobalanus. For the latter, we report the first known fossil record of its staminate inflorescences.

Origin and age of the fossils
The majority of Baltic amber derives from opencast mines of the Sambian Peninsula (Samland, Kalinigrad Oblast, Russia) where up to several hundred tons of amber are currently mined every year. Baltic amber mainly originates from the so-called Blue Earth layer, which was dated as late Eocene (Priabonian; 37.8 -33.9 Ma) on the basis of lithological and biostratigraphic studies, including pollen, spores and phytoplankton (Kosmowska-Ceranowicz 1997;Kasiński & Kramarska 2008;Standke 2017). Based on K-Ar age estimates of glauconites from the Blue Earth, a Lutetian (47.8 -41.2 Ma) age of Baltic amber was suggested (Ritzkowski 1997). However, a study by Clauer & al. (2005) questioned the reliability of glauconite-based chronometers, because contamination of the glauconite splits or reworking processes of the glauconites can lead to older age estimations (Clauer & al. 2005;Grimaldi & Ross 2017). The Lower Blue Earth (Lutetian) and the Lower Gestreifte Sande (upper Oligocene) also yield Baltic amber, but in smaller quantities. Hence, all amber-bearing strata possibly cover an age range of 48 to 23 million years (Kosmowska-Ceranowicz & al. 1997;Kasiński & Kramarska 2008;Standke 1998Standke , 2008; see Sadowski & al. 2017a for further age discussion). However, the amber concentrations of the Lower Blue Earth and the Lower Gestreifte Sande are low. Therefore, the Lower Gestreifte Sande are treated as overburden in the opencast mine and is not further processed (pers. comm. with Dr. Gerda Standke, Freiberg). In the lithological section, the Lower Lutetian Blue Earth layer is located about 20 m beneath the Blue Earth and has been reported only from a single drilling core (Standke 2008: fig. 5;Störr & al. 1998, cited from Standke 2008. Due to the great depth of this layer and its low amber concentration, the Lower Blue Earth layer is not mined either (Standke 2017;pers. comm. with Dr. Gerda Standke, Freiberg). Baltic amber that eroded from the amber-bearing sediments can be found along the Baltic and North Sea coasts, where it is occasionally washed ashore (Weitschat & Wichard 2010). It is not documented whether historic Baltic amber collections contain these "sea ambers", but because the layers of the Blue Earth are exposed on the coast of the Samland Peninsula (Ūsaitytė 2000), they are likely the source layers of these "sea ambers". All in all, it is most likely that the vast majority of Baltic amber collections derive from the Priabonian Blue Earth and are therefore 33.9 to 37.8 million years old.

Examination, preparation and imaging of specimens
For reviewing historic publications of fagaceous inclusions from Baltic amber, the type specimens described by Göppert & Berendt (1845), Caspary (1881Caspary ( , 1886 and Conwentz (1886) were required. However, due to losses in World War II, the location of many specimens is unknown (Kosmowska-Ceranowicz 1990). Therefore, we screened historic amber collections of the Museum für Naturkunde Berlin and of the Geoscientific Collection of the University of Göttingen (Germany) for type material. Furthermore, we studied recently collected amber specimens from the Hoffeins Amber Collection housed at the University of Göttingen. All amber specimens used in this study, including repository information, are listed in Table S1 (see Supplemental content online). If type material could not be found, we reviewed specimens by using original descriptions and illustrations. Most specimens did not require preparation because they were already ground into shape and polished by previous preparators, curators and/or collectors. Therefore, in most cases, inclusions were located close to the amber surface with their anthers exposed (e.g. Fig. 16I, J; 23F, G; 49I -K). If available, pollen was carefully removed from the anthers using a scalpel and placed on carboncovered SEM mounts, sputtered with gold/palladium (11 -14 nm coat thickness), using an Automatic Sputter Coater (Canemco Inc.). The samples were then examined with a field emission scanning electron microscope (Carl Zeiss LEO 1530).
Two specimens from the Hoffeins Amber Collection (GZG.BST.21992 and GZG.BST.21991) needed further preparation. In both cases, wet silicon carbide paper was used (manufacturer Struers) to grind the specimens in stages (grit sizes FEPA 800, 1200, 2400, 4000) for creating even and parallel facets closely located to the inclusions. In a final step, the specimens were polished using a toothpaste suspension, which was applied on a leather cloth (for more details on amber embedding and preparation see Nascimbene &Silverstein 2000 and. All amber specimens and the extant herbarium material were examined with a Carl Zeiss AxioScope A1 compound microscope, a Carl Zeiss Discovery V8 dissecting microscope and a Carl Zeiss Stemi 508 Microscope. We used incident and transmitted light simultaneously and photographed each specimen with digital Cameras (Canon EOS 5D), which were installed on both microscopes. We took up to 170 images from numerous focal planes, which were then merged to photomicrographic composites using the software package Helicon Focus 6.3.3 Pro for enhanced illustration of the three-dimensional inclusions. A micrometre eyepiece was used for measuring morphological features.  (Lozano & al.) Nixon & Crepet was obtained from previous studies (Hjelmqvist 1948;Forman 1964;Soepadmo 1972;Abbe 1974;Lozano-C. & al. 1979;Nixon & Crepet 1989;Kubitzki 1993). Within Quercus, we examined staminate inflorescences of most Quercus sections (Q. sect. Cerris Loudon, Q. sect. Cyclobalanopsis, Q. sect. Ilex Loudon, Q. sect. Lobatae, Q. sect. Protobalanus and Q. sect. Quercus; see Table 2, Fig. 7 -13) with special focus on Q. sect. Lobatae, because their pollen was most abundant in the studied amber samples. We also sampled staminate inflorescences of all genera of the Castaneoideae.

Terminology
In the descriptions of fossil and extant fagaceous staminate inflorescences, we followed Simpson (2010).

Perianth morphology
The perianth of staminate florets differs among species of the Fagaceae. In the present study, we distinguished between different shapes of the perianth and the degree of its incision. We further described the shape of the perianth lobes, and whether their margins were entire, serrate or irregularly incised.
Topshaped -The perianth is shaped like a cone, but with the apex pointing downward. The base of the perianth is narrow and then expands upward (Fig. 1A).
Campanulate -Bell-shaped; the perianth is more or less tube-shaped, basally rounded and flaring toward the apex. The upper margin is incised (mostly to the upper third) and forms triangular lobes (Fig. 1B, C). Campanulate perianths commonly occur in extant red oaks, such as Quercus agrifolia Née (Fig. 9F).
Cyathiform -The perianth is shaped like a cup and shorter and wider than a campanulate perianth. The perianth is incised to the upper third or the middle, forming lanceolate or triangular lobes (Fig. 1D, E).
Deeply lobed -In perianths where the incision extends to the base, the shape is difficult to assess, as the perianth lobes are diverting from each other (Fig. 1F, G; e.g. Quercus robur L., Fig. 12A, F). In this case, we did not mention a specific shape of perianth. There is a large variation of these shapes among extant and fossil Fagaceae, and sometimes transitions between these shapes were observed, such as in Quercus imbricaria Michx. (Fig. 10E, J).

Anther morphology
Anthers of Fagaceae differ in the shape of their apices and in the way they are attached to the filament. To distinguish between the different morphologies, we applied the following terms: Mucronate -The anther is apically pointed ( Fig. 2A Fig. 8G, H). This "peak" can be very acute or obtuse. However, we do not differentiate between different types of apical "peaks".
The attachment of the anthers to the apex of the filament differs among Faga ceae. Quercoideae possess anthers that are basifixed (Hjelmqvist 1948), meaning that the distal end of the filament connects to the base of the anther connective (Fig. 2E). In contrast, all Cas taneoideae possess stamens where the anthers are dorsi fixed (Hjelmqvist 1948). In this case, the filament apex is attached dorsally or medially to the anther (Fig. 2F). In Trigonobalanoideae, anthers are basifixed with a pronounced heart-shaped (cordate) base ( Fig. 2G; e.g. Tri gonobalanus verticillata, Fig. 6C).
Bifurcate -The trichome consists of two rays, which are fused at their base. Both rays possess an acute apex (Fig. 3C). Deng & al. (2014) suggested that bifurcate trichomes were reduced stellate trichomes and should therefore be incorporated into the category "simplified stellate". However, "simplified stellate" also comprises trichomes with up to eight arms and is therefore a category too unspecific for our study. For this reason, we kept the category "bifurcate".
Uniseriate -Singular trichomes, consisting of numerous cells, which are arranged in a simple row. They vary in length and terminate in an acute apex or a rounded, enlarged cell (Fig. 3E). The basal cells are similar to the terminal ones and are often bulbous (Type 15, "simple uniseriate" sensu Jones 1986).
Branched uniseriate -A basal cell branches into numerous multicellular arms. In each arm, the cells are arranged in a simple row (Fig. 3F). In our study, the basal cell was not as large, as observed by Deng & al. (2014). However, we consider this difference to be a variation of this trichome type, which was previously observed by Jones (1986;Type 18, "branched uniseriate").
Peltate -For this trichome type we follow Jones (1986). This trichome type consists of a short stalk, which apically terminates into a round, flat cap. This cap is composed of numerous cells, which radiate from the centre (Fig. 3G, H). Jones (1986) distinguished between "thick-walled" and "thin-walled" peltate trichomes. This is, however, not discernible in our fossils.

Phylogenetic framework
We used the molecular phylogeny by

Staminate inflorescences of extant Fagaceae
In previous studies on staminate inflorescences of the Fagaceae from Baltic amber, Caspary (1880Caspary ( , 1881Caspary ( , 1886aCaspary ( , 1886b, Conwentz (1886), Kirchheimer (1937) and Iljinskaja (1982) used different morphological features of the florets to distinguish between species and genera. The taxonomic value of these characteristics in extant Fagaceae has not been comprehensively studied.
Although several studies provide information on staminate inflorescences of modern taxa (e.g. Camus 1929Camus , 1936Camus -1938Hjelmqvist 1948;Soepadmo 1972), they did not picture the florets and entire catkins of all genera or sections of Castaneoideae, Trigonobalanoideae and Quercus. In the following, we describe and compare the different morphological characters of staminate inflorescences of the Fagaceae, mainly based on our own observations (see Table 1, 2). In Castaneoideae, the staminate inflorescences are spikes, which are ascending, rigid and sometimes branched ( Fig. 4A, E, I, K; 5A -F, I, L). They can be exclusively staminate or androgynous with pistillate flowers at their bases, such as in Castanea (Camus 1929) or Notholithocarpus densiflorus (Hook. & Arn.) Manos & al. (Camus 1952-1954b. In contrast, staminate inflorescences of Quercus are always a pendulous catkin with a lax rachis, which may be branched (e.g. Fig. 12A -D), but which is never androgynous. The staminate florets of the Castaneoideae are mostly clustered together in dichasia of three to 11 florets (Fig. 4I, K, M; 5B, E), whereas in Quercus, the florets are mainly arranged singularly along the rachis (e.g. Fig. 9A -D; except for Q. sect. Cyclobala nopsis, Fig. 7C -J). Another major difference between both groups is the morphology of stamens. In the Castaneoi deae, the filaments project far beyond the perianth and are about twice as long (or more) as the perianth (Fig. 4A, B, F, H, M -Q; 5D, K). In contrast, most staminate florets of Quercus exhibit stamens with short filaments, which are mostly covered by the perianth (Fig. 8F, G, H, J). Anthers of Castaneoideae are round and dorsifixed (Fig. 4B, H, N,    Table S2 [Supplemental content online] for affiliations). -A: overview of staminate catkin of Trigonobalanus; rachis is rigid; B, D: dichasia of florets; note central trichome tuft (B, arrowhead); C: anthers with mucronate apex and cordate base; E: overview of staminate catkin of Formanodendron; rachis is lax; F: dichasium of numerous florets; G: rachis with different trichome types; H: detached floret; arrowhead indicates dense, central trichome tuft; I: anthers with mucronate apex (arrowhead) and cordate base; J, K: "glandular" peltate trichomes of rachis. -Scale bars: A, E = 1 mm; B, D, F, H = 500 µm; C, G, I = 300 µm; J, K = 50 µm. Willdenowia 50 -2020 Fig. 7. Staminate inflorescences and florets of extant Quercus sect. Protobalanus (A, B) and Q. sect. Cyclobalanopsis (C -J), sampled from herbarium specimens (see Table S2 [Supplemental content online] for affiliations). -A, B: Quercus chrysolepis; note acute anthers (B); C -E: Quercus glauca; note elongated linear bract subtending dichasium (C) and large ovate subtending bracts (D, E); F, G: pubescent, staminate catkin (F) and dichasium (G) of Quercus salicina Blume; H -J: Quercus kerrii Craib; note short filaments and dense arrangement of florets in dichasium. -Scale bars: A, C, D, F, H = 1 mm; B, E, G, I, J = 500 µm.        O), whereas in Quercus, anthers are often quite large and elliptic with a basifixed attachment to the filament apex (e.g. Fig. 9F, G, I). As far as we observed in the herbarium specimens, none of the Castaneoideae possessed staminate florets with mucronate anther apices; however, this often occurs in Quercus (Q. sect. Ilex and Q. sect. Loba tae, Fig. 8H; 9F). Most staminate florets of Quercus are subtended by narrow and acute bracts (Fig. 8E, L; 12D), which differ from the large and ovate bracts subtending staminate florets of the Castaneoideae (Fig. 5E). Due to these differences between staminate inflorescences of Cas taneoideae and Quercus, it is possible to distinguish staminate inflorescences of both groups from another. Among the Trigonobalanoideae, only Trigonobala nus verticillata possesses staminate inflorescences that are spike-like and ascending (Fig. 6A). However, T. ver ticillata is distinct from the Castaneoideae in the morphology of the stamens: anthers are basifixed, the anther base is distinctly cordate and the anther apex is mucronate ( Fig. 6C), anthers are larger, and filaments are not as elongated as in the Castaneoideae.
The bigger challenge is to differentiate between staminate inflorescences of genera and species within the Cas taneoideae, because the staminate inflorescences of castaneoids share many features, such as the morphology of stamens (Table 1). However, among the studied species, we found a few distinct characteristics of each genus. Staminate florets of Castanea are more densely clustered and helically arranged along the rachis (  Table 1 for references). Staminate inflorescences of Chrysolepis and Notholithocarpus are the most pubescent spikes of all Castaneoideae observed in this study . In contrast, staminate inflorescences of the remaining genera possessed trichomes (Fig. 4C, D, I, J; 5G, H), but never in such a dense cover as in these two genera. Despite the named features, it remains difficult to distinguish between castaneoid staminate inflorescences. Pollen of Castaneoideae is uniform and therefore cannot be used for identification to genus level (Praglowski 1984;Grímsson & al. 2015).
Although sections of Quercus can be differentiated by their distinct pollen morphology (Denk & Grimm 2009), the distinction between sections of Quercus based on the morphology of their staminate catkins is difficult (Table  2). Staminate catkins of Q. sect. Cyclobalanopsis are most distinct from all other Quercus sections. The staminate florets are arranged in dichasia of three  florets are singular in all other sections) and possess up to 17 stamens per floret (which is the highest stamen number within quercoids). Due to the very dense arrangement of the florets, the perianth is almost invisible (Fig. 7C, G), but according to Hjelmqvist (1948) and Kaul (1985) it is deeply incised, forming ovate to lanceolate lobes. The anthers are apically notched and round in shape (Fig. 7G, J), whereas in all other quercoids they are mostly elliptic. Another striking feature of Q. sect. Cyclobalanopsis is the large bract subtending the dichasia (Fig. 7C, D, I, J;Hjelmqvist 1948;Kaul 1985). In addition, we observed a very slender and elongated bract located beneath the flower dichasium of Q. glauca Thunb. (Fig. 7C).
Quercus sect. Protobalanus exhibited widely open cyathiform perianths that are incised to the upper third or to the base (Fig. 7B). Despite the widely open perianth, filaments are not discernible, because they are particularly short. The anther apex is acute, which was not observed in the other sections (Fig. 7A, B). Subtending bracts were still attached to the florets and lanceolate in shape (Fig. 7B).   Table S2 [Supplemental content online] for affiliations). Note short filaments, large anthers and cyathiform perianth, which are typical features of Q. sect.  Table S2 [Supplemental content online] for affiliations). Note short filaments, large anthers and cyathiform perianth, which are typical features of Q. sect. Quercus sect. Quercus differs from the other sections by the deeply incised perianths, forming linear slender lobes ( Fig. 12F -H). However, some species of Q. sect. Quercus also possess campanulate perianths with triangular lobes (Fig. 12D, I). Some species possess catkins with few trichomes (Fig. 12A -C), whereas other species are more pubescent (Fig. 12D, E). The anthers of all species of Q. sect. Quercus examined in this study are notched ( Fig. 12F -J). Fig. 11. Staminate inflorescences of extant Quercus sect. Lobatae, sampled from herbarium specimens (see Table S2 [Supplemental content online] for affiliations). Note short filaments, large anthers and cyathiform perianth, which are typical features of Q. sect.         In summary, some sections of Quercus exhibit distinct features in their staminate inflorescences, such as the high number of stamens per floret in Q. sect. Cyclo balanopsis. However, many features, for instance the campanulate perianth shape, are not exclusive for one section and are shared between sections (Table 2). Therefore, to distinguish staminate inflorescences of Quercus sections, additional information from pollen ornamentation is needed.
Conwentz (1886) chose the perianth shape and trichomes to distinguish between species of Quercus. When Kirchheimer (1937) reinvestigated Baltic amber inclusions of staminate inflorescences of the Fagaceae, he stated that these features were not reliable because they depended on the maturity of the inflorescence and, therefore, may differ within one individual inflorescence.
When examining extant herbarium specimens of the Fagaceae, we noticed that staminate florets of one individual inflorescence mostly looked similar. However, in a specimen of Quercus grisea Liebm. (Q. sect. Quercus; Fig. 13C, F, G, H), some florets had elongated filaments (Fig. 13G, H), whereas others of the same specimen had short filaments (Fig. 13C, F). Because the anthers of the short filaments were mostly closed (Fig. 13C, F), we interpreted this particular catkin as "immature" (before anthesis), whereas the other catkin with elongated sta- mens had widely open anthers (Fig. 13G, H) and therefore was likely more "mature" (at or after anthesis).
Despite the different maturities of florets, their morphology was alike and we did not observe differences in the perianth shape or trichome distribution. We further compared two herbarium sheets of Q. candicans Née (Q. sect. Lobatae) from Mexico to test whether different morphologies of staminate inflorescences occurred within one species of Quercus ( Fig. 13A, B, D, E). We again observed that the staminate inflorescences of both specimens possessed filaments of different length ( Fig.  13A vs. 13B), which is likely related to the immaturity of the catkin, because the shorter filaments possessed closed anthers (Fig. 13D). Besides the differences in filament length, all other features, such as the perianth shape, were alike. Based on this observation, we concluded that different stages of maturity of staminate catkins not necessarily lead to a marked morphological change. However, the maturity of anthers needs to be considered when discussing the length of filaments as a distinguishing feature.
Our comparative study of extant Fagaceae showed that it is possible to distinguish between the staminate inflorescences of Quercus, Castaneoideae and Trigono balanoideae. To differentiate between sections of Quer cus, additional information from pollen morphology is needed, because morphological features of the staminate catkins are not always distinct enough within each section (Table 2). Based on staminate inflorescences only, it is difficult to distinguish between genera of the Cas taneoideae; however, the density of pubescence and the arrangement of florets along the rachis are indicative for some genera (Table 1). Within Trigonobalanoideae, the morphology of the staminate inflorescence in combination with the pollen sculpturing are very helpful to differentiate between species/genera (Table 1).
Based on this comparison, we considered the following features to be reliable for distinguishing between staminate inflorescences of Quercus, Castaneoideae and Trigonobalanoideae: rachis (pendulous or rigid), shape and degree of incision of the perianth, trichome morphology and their presence or absence, length of filaments  when anthers were open, meaning they were "mature"), anther shape and apex (notched, mucronate, obtuse or acute), fixation of anthers (dorsifixed or basifixed), and shape of the subtending bract. If available, we further included pollen morphology in the study of the Baltic amber fossils.
The locations of the holotypes of these species names (except for C. brachyandra) are unknown. Therefore, we compared staminate inflorescence and floret inclusions to illustrations and descriptions of the lost holotypes, published by Conwentz (1886), Kirchheimer (1937) and Iljinskaja (1982). Based on this comparison, we accommodated several specimens into the fossil-species Castanea inclusa and C. longistaminea (although their placement in the genus Castanea could not be determined; see under these two species below). Because the type specimens of these names of species from Baltic amber are lost, newly discovered inclusions were used to designate neotypes and to emend the diagnoses. Castanea brachyandra was accommodated into Quercus (see Q. sect. Protobalanus), while the true generic identity of C. subvillosa remains unresolved, because neither detailed descriptions nor illustrations of this species exist. Besides these neotypes, we found further amber inclusions of castaneoid affinities that have not been reported from Baltic amber before. Emended diagnosis -Staminate florets with campanulate to top-shaped perianth, incised to lower third forming 6 lanceolate to linear lobes. Apex of lobes obtuse, margins of lobes slightly curved inward, densely ciliate with 2 trichome types (uniseriate; solitary). Outer surface of perianth with few trichomes of both types. Stamens up to 10 per floret, projecting far beyond perianth, filiform. Anthers almost round, dorsifixed, apically notched. Centre of floret with dense trichome tuft, composed of solitary trichomes.
Remarks -From all fagaceous inflorescences from Baltic amber described and pictured by Conwentz (1886), specimen GZG.BST.21989 most closely resembles Cas tanea inclusa in the following aspects: campanulate glabrous perianth, ciliate perianth margin, deeply incised with elongated lanceolate lobes, filaments longer than perianth (2 -2.5 × as long as the perianth), small and round anthers (Table 4). However, Conwentz (1886) did not mention the central trichome tuft and the different trichome types. Castanea inclusa was first described by Conwentz (1886), but later Kirchheimer (1937) included C. inclu sa within C. longistaminea, arguing that the absence or presence of trichomes was not a valuable morphological feature to distinguish species of Castanea (Table 3). However, in extant castaneoid genera, species clearly differ by their indumentum (compare Fig. 4A -F). Furthermore, Jones (1986) showed that trichome types are diverse and distinct among certain genera of Fagaceae. Hence, we followed Iljinskaja (1982) and distinguished two species of Castanea from Baltic amber. The floret inclusion shares some features with a castaneoid staminate inflorescence (see below; specimen GZG.BST.21991, subfamily Castaneoideae, genus indet.), such as the elongated stamens, the small anthers and the dense trichome tuft; however, GZG.BST.21991 does not show uniseriate trichomes along the perianth margin, but has solitary and bifurcate trichomes (Table 4). Therefore, it remains unclear if C. inclusa and GZG.BST.21991 belong to the same genus. Comparing the floret inclusion with extant genera of Castaneoideae, the following similarities occur: elongated filaments, dorsifixed and roundish anthers, presence of trichomes, and perianth shape are found in all modern genera (Table 1). Uniseriate trichomes mainly occur in Castanea and Lithocarpus (Jones 1986); however, Lithocarpus may exhibit a pronounced pistillode in the floret centre (Soepadmo 1970 ; Fig. 5J), which is missing in C. inclusa. At this point, it is impossible to definitely assign C. inclusa to a particular genus of the    Emended diagnosis -Staminate florets with campanulate to top-shaped perianth, incised to lower third forming 6 lanceolate to linear lobes. Apex of lobes acute-obtuse, margins of lobe and outer surface of perianth glabrous. Stamens 9 -12 per floret, longer than perianth, filiform. Anthers round to slightly elliptic, dorsifixed, apically notched. Centre of floret with dense trichome tuft, composed of solitary trichomes.
Description -Four detached florets in one specimen (neotype is Fig. 15B, E, G). Perianth: campanulate, 1280 -1680 µm long × 1200 -2200 µm wide, glabrous, incised to lower third, connate at base, lobes linear to lanceolate, each with acute-obtuse apex, margins entire and glabrous ( Fig Remarks -From all fagaceous inflorescences from Baltic amber described and pictured by Conwentz (1886), specimen GZG.BST.24371 most closely resembles Castanea longistaminea in the following aspects: campanulate glabrous perianth, perianth deeply incised with elongated lanceolate lobes, up to 9 stamens per floret, filaments much longer than perianth. In the original description, Conwentz (1886) did not illustrate or mention a central trichome tuft, but he observed few trichomes on the perianth surface, which we did not find. Castanea longistaminea differs from C. inclusa in the glabrous perianth margin, the longer anthers and the larger perianth size (Table 4). In Caspary's (1881) original description, Castanea longistaminea was described as a species of Quercus. Later, Conwentz (1886) transferred the species to Cas tanea due to the elongated filaments and elliptic-round anthers; this was accepted by Kirchheimer (1937) and Iljinskaja (1982 ; Table 3). Comparing the inclusion with extant Castaneoideae, Chrysolepis and Notholithocarpus differ by a densely pubescent perianth. The perianth of Castanea is pubescent as well, especially along the margin and the upper third of the lobes, which is dissimilar to − basifixed the inclusion. Only Castanopsis and Lithocarpus exhibit perianths that can be glabrous on the outside, but possess a dense trichome tuft in the floret centre. In our opinion this is not sufficient morphological evidence for assigning the amber specimen to one of the mentioned castaneoid genera. Therefore, we retain the name C. longi staminea with the question mark highlighting its general castaneoid affinities.          Close to endoaperture, 2 or 3 pairs of parallel rugulae framing apertural area (Fig. 17F, G). Footlayer, tectum, and columellae of same thickness in mesocolpium (Fig.  17D, I).

Additional specimens investigated
Remarks -The pollen sculpturing of GZG.BST.21991 suggests affinities to Castaneoideae, because the smooth striation is typical for all genera of the Castaneoideae.
However, based on pollen morphology only, genera of Castaneoideae are nearly indistinguishable, because pollen size and shape of Castanopsis, Chrysolepis, Lithocar pus and Notholithocarpus are very similar (Grímsson & al. 2015;Praglowski 1984). As possible exceptions, pollen of Castanea can be distinguished from other castaneoids by its smaller size (Grímsson & al. 2015) and by the smooth tectum in the polar region (Bouchal & al. 2014); however, the preservation of pollen of GZG.BST.21991 does not allow evaluating the sculpturing of the polar region. Besides pollen morphology, features of staminate inflorescences of the Castaneoideae are similar to specimen GZG.BST.21991 in that they are all rigid, spike-like and erect; thus they can be clearly distinguished from Colombobalanus, Formanodendron and Quercus, which all possess pendulous and lax catkins (see Table 1 for references). Only Trigonobalanus exhibits spike-like and erect staminate inflorescences as well (Forman 1964),  but shows a fine granular and rugulate pollen sculpturing (Denk & Grimm 2009;Nixon & Crepet 1989), which is different from the rugulate-fossulate pollen of specimen GZG.BST.21991.
Specimen GZG.BST.21991 is not as densely pubescent as staminate inflorescences of Notholithocarpus and the tomentose perianth of Chrysolepis. Species of Casta nea, Castanopsis and Lithocarpus, however, vary in the morphology of their staminate inflorescences, all combining features that occur in specimen GZG.BST.21991, such as the number of florets per dichasium, the shape and number of lobes, the glabrous perianth with ciliate margin, the central, dense trichome tuft and the protruding stamens (Table 1). Within castaneoid Fagaceae, peltate trichomes occur only in Chrysolepis, Castanop sis and Lithocarpus (Jones 1986). Among these taxa, peltate trichomes of GZG.BST.21991 are most similar to thin-walled peltate trichomes sensu Jones (1986) of Castanopsis and Lithocarpus, because these exhibit multicellular centres with stellate to discoidal caps (unlike Chrysolepis, which possesses radially fused peltate trichomes; Jones 1986).
These shared synapomorphies support the assignment of specimen GZG.BST.21991 to the Castaneoideae. Based on the presence and morphology of peltate trichomes, and the absence of a dense indumentum covering the entire inflorescence (as it is present in Chrysolepis and Notholithocarpus), we suggest closest affinities of GZG. In the 19 th century, Caspary (1881), and Conwentz (1886) described species of the Castaneoideae from Baltic amber: Castanea brachyandra, C. inclusa, C. longi staminea and C. subvillosa. Except for C. brachy andra, these fossil-species represent floret inclusions, detached from the inflorescence axis. Because all specimens (except for C. brachyandra) are currently missing, we use descriptions and figures by Conwentz (1886), as well as the neotypes, for comparison (Table 4). Casta nea inclusa has an almost glabrous perianth with uniseriate trichomes along the margin, different from the solitary trichomes along the margin of specimen GZG. BST.21991.
Castanea longistaminea has an almost glabrous perianth with sparse trichomes on its outer surface (Conwentz 1886), but does not possess a ciliate perianth margin. Castanea brachyandra and C. subvillosa were not illustrated by Caspary (1881) and Conwentz (1886). Caspary (1881) and Conwentz (1886) distinguished C. subvillosa from other fagaceous species by the strongly tomentose perianth surface and glabrous apices of the lobes. Further morphological details were not given; also, Conwentz (1886) mentioned that the specimen of C. subvillosa was rather poorly preserved and therefore C. subvillosa still needs verification. Anyhow, the indumentum described for C. subvillosa does not match specimen GZG. BST.21991, which has a glabrous perianth and hairy margins. Following Conwentz's (1886) description, C. brachyandra is a dichasium of seven florets with elliptic, short to elongated, mucronate anthers dissimilar to the round, emarginate anthers of specimen GZG.BST.21991 (see text below for more discussion on C. brachyandra). Based on these differences, we do not assign specimen GZG.BST.21991 to one of the previously described Cas tanea species from Baltic amber. Crepet & Daghlian (1980) described fossil staminate inflorescences of Castaneoidea puryearensis Crepet & Daghlian (Castaneoideae, Fagaceae) from the middle Eocene Claiborne Formation (Tennessee, U.S.A.). This species differs from the amber fossil in the presence of dichasia of three florets and in the absence of peltate trichomes and ciliate perianth margins.

Subfamily: Quercoideae
Inclusions of staminate inflorescences and singular florets of Quercus are easily distinguishable from other genera of the Fagaceae due to the combination of the following characters: rachis pendulous, florets mostly singular, short filaments (covered by the perianth) and basifixed anthers (see Table 1 for references). If available, pollen sculpturing supports the assignment to the Quercoideae (Denk & Grimm 2009).
Remarks -Quercus casparyi is different from the other amber specimens of Quercus in the elongated filaments, which project beyond the perianth, the mucronate anthers and the arrangement of 2 or 3 florets in dichasia. The fossil catkin therefore combines characteristics found in Q. sect. Cyclobalanopsis (high number of stamens per florets, 2 or 3 florets per dichasium, elongated subtending bract exceeding the perianth, mucronate anthers) and Q. sect. Lobatae (campanulate perianth; Table 2).
Likewise, the pollen sculpturing of Quercus casparyi shares features with pollen of Q. sect. Cyclobalanopsis, e.g. Q. acuta Thunb. and Q. glauca (Denk & Grimm 2009;Makino & al. 2009), with virtually identical pollen sculpturing. At the same time, the relatively thin footlayer as seen in the fossil specimen (Fig. 19C) appears to be more typical of Q. sect. Lobatae (Denk & Tekleva 2014). Also, the relatively large size of the pollen fits better with Q. sect. Lobatae than with Q. sect. Cyclobalanopsis.
Conwentz (1886) (Fig. 20B). Furthermore, the rachis of MB.Pb.1979/0813 is rather glabrous, except for the trichome tufts at the base and pedicel of each floret. The perianth margin of MB.Pb.1979/0813 is entire with few solitary trichomes, whereas the other specimens possess irregularly serrate perianth margins. Anther apices in MB.Pb.1979/0813 are obtuse, whereas they are clearly notched in MB.Pb.1979/0817 (Fig. 20E). Based on these differences, we treat both specimens as distinct species and follow Caspary's (1881) (1886) and Caspary (1880Caspary ( , 1881Caspary ( , 1886aCaspary ( , 1886b, Q. limbata, Q. nuda, Q. piligera Casp. and Q. taeniatopilosa Conw. exhibit a perianth that is not incised to the base. However, they differ from specimens of Q. subglabra in the following aspects: Q. limbata has a ciliate margin, elongated filaments overarching the perianth and densely pubescent subtending bracts; Q. nuda has a widely open perianth and a glabrous rachis; Q. piligera has broadly ovate subtending bracts and a pubescent perianth surface; and Q. taeniatopilosa has six stamens and trichomes on the outer surface of the perianth. Based on the pollen morphology of GZG.BST.24535 and GZG.BST.24418 (weakly microverrucate perforate tectum, thin footlayer; Fig. 22F; 24D, E) both specimens are assigned to Quercus sect. Lobatae. Among modern species, pollen very similar to Q. subglabra occurs, for instance, in Q. imbricaria (see Solomon 1983b: fig. 10a).
Etymology -The specific epithet refers to the high number and morphological diversity of trichomes located on the perianth margin.
Comparing the type specimen MB.Pb.1979/0660 ( Fig. 33) with GZG.BST.24385 (Table 8, Fig. 34), simi-larities occur, including: numerous florets clustered in one dichasium, which is subtended by large bracts (Fig.  34B); perianth divided into distinct lobes, which have ciliate margins (Fig. 34C -E); stamens overarching the florets (Fig. 34B); anthers mucronate and elongated (Fig. 34F, G). However, florets in GZG.BST.24385 are more open and wider (Fig. 34A), which is probably due to maturity. Specimen MB.Pb.1979/0660 has only up to seven stamens per floret, but we observed several malformations of stamens, because filaments of neighbouring stamens were fused (Fig. 33F). Therefore, it is difficult to determine the exact number of stamens per floret. Also, anthers of MB.Pb.1979/0660 are shorter than in GZG. BST.24385, which might be related to the deformed filaments as well. The holotype did not contain any pollen; however, based on the very distinct morphology shared between MB.Pb.1979/0660 and GZG.BST.24385, we think that they represent the same taxon.
Conwentz (1886) stated that Castanea brachyandra is likely not affiliated with Castanea, because the anthers were too elongated and the filaments too short. In his revision of plant inclusions from Baltic amber, Kirchheimer (1937) also had difficulty assigning this specimen of C. brachyandra. The pollen morphology of GZG. BST.24385 supports affinities with Quercus. The rod-like masked, microechinate sculpturing of pollen grains was reported by Denk & Grimm (2009) and Bouchal & al. (2014) and assigned to Q. sect. Protobalanus. In addition to the pollen morphology, several other features of C. brachyandra justify assignment to Quercus: basifixed, elliptic, large (1.2 -2 mm long) and mucronate anthers; perianth deeply incised and connate at the base. Because of these morphological similarities, we propose the new combination Quercus brachyandra. Although the pollen morphology of C. brachyandra unambiguously proves its affinities to Q. sect. Protobalanus, extant Q. sect. Pro tobalanus examined in this study (Table 2, Fig. 7A, B) is different from the amber fossils in the singular florets, the indumentum, short filaments and acute anther apex. However, we studied only Q. chrysolepis of Q. sect. Pro tobalanus, which may not reflect the overall morphology of staminate inflorescences within this section.
Up to now, there was no staminate catkin of Quer cus from Baltic amber that exhibited a dichasium with at least four florets, as in Q. brachyandra. However, a singular floret of Quercus with a deeply incised perianth and lanceolate-linear lobes was described as Q. mucro nata (Conwentz 1886). The holotype of Q. mucronata (MB.Pb.1979/0868; Fig. 37) resembles Q. brachyan dra in the following features: the shape and size of the perianth; the presence and distribution of long, simple, acute trichomes along the perianth margin; the central trichome tuft; the mucronate anthers. However, Q. mucronata can be clearly distinguished from Q. brachyandra in the shorter length of filaments and in the significantly smaller anther size (600 -920 µm long × 380 -560 µm wide).

474
Sadowski & al.: Staminate inflorescences of Fagaceae from Eocene Baltic amber Quercus meyeriana is another species with singular florets that resemble Q. brachyandra, with specimens of both species sharing the following features: perianth deeply incised with oblanceolate, elongated or lanceolate lobes; perianth glabrous with ciliate margins; stamen number 6 -10; filaments elongated; size of entire floret 2 -3.5 mm long (Conwentz 1886). The only distinguishing feature is the apically emarginate anthers in Q. meyeriana, which are mucronate in Q. brachyandra. Hence, in view of the different anther morphology, we refrain from assigning Q. brachyandra to Q. meyeriana.

Remarks -Quercus meyeriana was first published as
Quercites meyerianus (Göppert & Berendt 1845 (Turland & al. 2018), a later homonym is published if the type of an existing name is definitely excluded (Art. 48.1); however, in 1886 Quercus meyeri ana did not have a type, so that all three syntypes would have had to be excluded (Art. 48.2), which Conwentz (1886) did not do; he retained one syntype in Q. Unfortunately, the specimen that Conwentz intended to serve as the type of his Quercus meyeriana is lost.
The illustration of this specimen shows features that are rather typical for Eotrigonobalanus (such as the projecting filaments; for further discussion see Eotrigonobala nus). Moreover, Conwentz's (1886) description of his Q. meyeriana differs greatly from the diagnosis of Quercites meyerianus given by Göppert & Berendt (1845; see Table  6), for instance in the campanulate perianth (cyathiform in Göppert & Berendt 1845), the elongated filaments more than twice as long as the perianth (filaments as long as the perianth in Göppert & Berendt 1845) and in the presence of simple trichomes (stellate in Göppert & Berendt 1845).
In summary, all original specimens of Quercites meyerianus (MB.Pb.1979/0813, MB.Pb.1979/0817, MB.Pb.1979/0815) are morphologically distinct and are recognized here as separate taxa. Only MB.Pb.1979/0813 corresponds to the original diagnosis of Q. meyerianus and is therefore designated as lectotype of that name. As explained above, MB.Pb.1979/0817 was previously designated as the holotype of Quercus subglabra, and MB.Pb.1979/0815, the holotype of Q. meyeriana var. denticulata, is also excluded from Q. meyeriana because its affinities to Quercus are unresolved.

Quercus mucronata
Remarks -Quercus mucronata was first described by Caspary (1881) and later illustrated by Conwentz (1886, t. II, fig. 15, 16;herein Fig. 37C, D). Quercus mucronata differs from other Quercoideae from Baltic amber in the deeply incised perianth with linear lobes and long, solitary trichomes along the lobe margins (Table 9). Only Q. brachyandra possesses a deeply incised perianth, but differs from Q. mucronata by its larger anthers (up to 1920 µm long in Q. brachy andra, up to 920 µm in Q. mucronata), the dense, central trichome tuft, the higher number of stamens per floret, and the longer filaments (2000 µm in Q. brachyandra, 1600 µm in Q. mucronata).
The remaining species of Quercus from Baltic amber do not share the combination of features of Q. mucronata (perianth incised to the base, linear lobes, long trichomes along lobe margins, stamens not projecting beyond perianth, anthers mucronate and round-elliptic). Therefore, we treat Q. mucronata as a separate fossil-species.
Remarks -Quercus trichota is distinct from all other quercoids from Baltic amber in the pubescent upper third of the perianth and the mucronate anthers. However, Q. emanuelii from Bitterfeld amber is similar to Q. trichota in the mucronate anthers (Fig. 38D), the presence of solitary trichomes on the perianth (Fig. 38B Table 9). Unlike in Q. emanuelii, trichomes of Q. trichota are mostly concentrated on the upper third of the perianth. The perianth of Q. trichota is not as deeply incised as in Q. emanuelii (Fig. 38A, C). Based on these differences (Table 9), we decided to keep Q. trichota as a separate fossil-species. Quercus trichota was first described by Caspary (1881), but never illustrated. Conwentz (1886) was the first to publish illustrations of the particular floret inclusion, i.e. MB.Pb.1979/0814, mentioning that this specimen was previously treated as Quercites meyerianus by Göppert, but that it was not a type specimen of Q. meyerianus; Göppert & Berendt (1845) did not mention the specimen clearly. Kirchheimer (1937) and Iljinskaja (1982) included Q. trichota in Q. mucronata (Table 3). Comparing both holotypes, they clearly differ, because Q. mucronata possesses a deeply incised perianth with linear lobes, whereas in Q. trichota the perianth is campanulate and incised only to the middle. The pubescent perianth in Q. trichota further differs from the glabrous perianth in Q. mucronata (Table 9). Therefore, we see enough evidence to maintain two distinct taxa. Conwentz (1886) depicted two different specimens, both of which he assigned to Quercus trichota. He mentioned that only one of them possessed mucronate an-thers (herein Fig. 39H, I), whereas the other specimen (Fig. 39F, G) had rounded anthers. Because this particular specimen is lost, we cannot confirm this observation. However, we found further inclusions of detached florets (not figured) that exhibited a dense trichome cover (restricted to the upper third of the perianth) and mucronate anthers. Therefore, the mentioned characteristics are considered key features of Q. trichota. Conwentz (1886) also described a variety of Quercus trichota, Q. trichota var. macranthera. However, illustrations of this specimen clearly differ from the type of Q. trichota (MB.Pb.1979/0814, i.e. in the larger, apically rounded anthers), which is why we excluded Q. trichota var. macranthera from Q. trichota (for a detailed discussion see Unresolved affinities).  Description -Detached floret with glabrous rachis remains at base (Fig. 40E). Floret: singular, 3.9 mm long × 2 mm wide. Perianth: campanulate, 2.2 mm long × 2 mm wide; 5 perianth lobes, triangular, margin densely pubescent ( Fig. 40A -C, H), perianth surface sparsely covered with trichomes (Fig. 40E). Stamens: 5, filaments mainly covered by perianth (length not measurable), only slightly overtopping perianth (Fig. 40B); anthers basifixed, elongated, 1100 -1240 µm long × 560 -600 µm wide, apex obtuse (Fig. 40F). Subtending bracts: 2 bracts present, both subtending floret (Fig. 40E); 1 bract as long as perianth, 2.2 mm long × 0.6 mm wide, ovate, apex obtuse, margins entire, upper half of abaxial surface and margin tomentose, adaxial side glabrous (Fig. 40E); other bract 920 µm long × 120 µm wide, linear, acute, slightly enrolled and twisted, with few trichomes (Fig.  40G). Trichomes: 2 types: solitary and bifurcate. Solitary trichomes often helically curled, located at base of floret, along perianth margin (Fig. 40H), rarely on perianth surface, on bracts (Fig. 40I); bifurcate trichomes located along margin of perianth and bract, as well as on abaxial surface (mainly upper half) of ovate bract (Fig. 40I).
Comparing Quercus limbata to other fagaceous inflorescences from Baltic amber, similarities to Q. piligera occur (Table 10). The latter species was also first introduced by Caspary (1881) and further described and illustrated by Conwentz (1886). However, the holotype of Q. piligera is not available. Comparing Conwentz's (1886) (Fig. 40A); elongated anthers with round apex (Fig. 40F); stamens about twice as long as perianth (Fig. 40A, B; 41A, C); trichomes on perianth margin (Fig. 40A, B, H; 41C). As opposed to Q. piligera, Q. limbata possesses an additional small, linear subtending bract ( Fig. 40A -C, G), which Conwentz (1886) interpreted as remains of the rachis. These linear bracts are often caducous, so it is possible that they were already shed in Q. piligera. The perianth of Q. limbata is sparsely covered with trichomes (Fig. 40C) and seems not as densely hairy as the perianth of Q. piligera. Furthermore, Q. limbata possesses only five stamens (seven to nine in Q. piligera; Fig. 41A, C).
An additional species similar to Quercus limbata is Q. taeniatopilosa (Fig. 41E -G), which also possesses a similar perianth shape (Fig. 41F, G), trichomes on the perianth surface and along the margin, as well as six stamens with similar length and morphology (Fig. 41F, G; Table 10). However, in Q. taeniatopilosa the trichomes are arranged in rows that proceed only along the margin and between the perianth lobes (Conwentz 1886;Fig. Willdenowia 50 -2020 41F, G; Table 10). This specific arrangement was considered diagnostic of Q. taeniatopilosa, distinguishing it from other Quercus species from Baltic amber (Conwentz 1886). According to Conwentz (1886), Caspary suggested that this particular specimen might belong to Q. piligera. Because the type specimens of Q. piligera and Q. taeniatopilosa are lost, it is impossible to clarify their definite affinities with Q. limbata. However, certain similarities are present (Table 10), which is why we tentatively accommodate them into one species.
Comparing Quercus limbata with other Quercus species from Baltic amber, Q. aimeeana, Q. casparyi, Q. mul tipilosa and Q. subglabra resemble Q. limbata in the shape of the perianth. However, Q. limbata differs from the aforementioned species by the tomentose, ovate bracts (which are linear, elongated, very narrow and glabrous in those species), the tomentose perianth (which is glabrous in all those species), the rounded elongated anthers (mucro nate in Q. casparyi), and the number of stamens (more than nine stamens in Q. aimeeana and Q. multipilosa).   Description -According to Conwentz (1886) and Caspary (1880), Quercus nuda is described as follows: Staminate inflorescence: pendulous, 1.5 cm long; rachis filiform, glabrous (Fig. 42A, B). Florets: 1.5 -2 mm in size, shortly petiolate, in dichasia of 2 or 3 florets (Fig.  42A). Perianth: widely open, campanulate to cyathiform (Fig. 42C, D, E); margin dentate; perianth lobes widely ovate to triangular (Fig. 42E, H), longitudinal median with "keel" (Fig. 42C). Stamens: 7 -10 stamens per floret, entire stamens twice as long as perianth, anthers basifixed, elongated with slightly acute apex (Fig. 42C (1880), including a detached floret with ten stamens, a campanulate perianth that appears "keeled", and acute anthers overarching the perianth. Therefore, it is likely that MB.Pb.1979/0656 is the type of Q. nuda. In contrast to Caspary (1880Caspary ( , 1881, we observed several trichomes on the surface of the perianth and at the base (Fig. 42H, I); because they are very inconspicuous, they might have been overlooked by Caspary (1880Caspary ( , 1881, who mentioned that the perianth surface was entirely devoid of hairs. Conwentz (1886) assigned an additional specimen to Quercus nuda, a catkin with numerous florets, arranged in dichasia of two to three (Fig. 42A -C). Conwentz (1886) also described the anthers as acute, although this is not clearly visible in his illustration. Because the specimen illustrated in Conwentz (1886) is lost, it is impossible to compare it in detail to MB.Pb.1979/0656 (Table  10). However, based on the mentioned similarities, it is likely that they both represent Q. nuda.
Remarks -This specimen is unique among all fagaceous inflorescences from Baltic amber described and depicted by Conwentz (1886). It differs from Quercus by its stiff rachis with florets arranged in dense dichasia. This morphology occurs only in the Castaneoideae and Trigonobalanoideae. However, in contrast to the castaneoids, the specimen possesses basifixed, mucronate anthers, which are also larger than in castaneoid florets (anthers are 0.2 -0.5 mm long in all extant genera of Castaneoideae; see Table 1 for references). This anther type is known from extant Trigonobalanoideae, of which only Trigonobalanus verticillata has a spike-like, rigid staminate inflorescence (Forman 1964; Table 1). Additional features shared between the amber specimen and T. verticillata (Fig. 6A -D) are: florets arranged in dense dichasia; perianth deeply incised with oblanceolate, ciliate lobes; stamens projecting beyond perianth; anthers broadly elliptic to round, basifixed; anthers with mucronate apex and cordate base.
Remarks -Among all staminate inflorescences investigated here, specimen GZG.BST.21996 resembles only the trigonobalanoid sp. 1 in the densely pubescent and rigid rachis, the mucronate, basifixed anthers with a cordate base and the deeply lobed perianth (Table 11). Both specimens differ from each other in the number of florets per dichasium, the arrangement of florets on the rachis, the filament length, the shape and number of subtending bracts and the size of the anthers (Table 11).
Several features of specimen GZG.BST.21996 resemble extant Trigonobalanus verticillata: the rigid rachis, the arrangement of florets in dichasia, the deeply lobed perianth with ciliate margins, the central trichome tuft and the presence of several subtending bracts ( Fig.  6A -D, see Table 1 for references). However, extant T. verticillata possesses longer stamens and its rachis is not as densely pubescent and not as broad (Forman 1964) as in the amber specimen. Formanodendron doichangensis exhibits mucronate anthers (Fig. 6I), a deeply incised perianth (Fig. 6H) and a pubescent rachis (Fig. 6E, G); but the rachis of F. doichangensis is not as broad as in the amber specimen, is rather lax and exhibits peltate trichomes (Fig. 6J, K), which are absent in GZG.BST.21996. In the absence of in situ pollen information, we therefore assign the specimen to the Trigonobalanoideae, highlighting its similarities with T. verticillata.  Diagnosis -Pendulous staminate inflorescence, rachis glabrous. Florets singular or in clusters of 2 or 3, sessile. Perianth cyathiform, incised to lower third, forming triangular, acute lobes with irregular serrate margin. Outer perianth surface and margin only rarely with solitary trichomes. Each floret with central, loose trichome tuft composed of solitary trichomes. Stamens numerous, 6 -11 stamens per floret, filaments elongated and projecting beyond perianth. Anthers narrowly elliptic, apically notched, basifixed. Florets subtended by single bract, lanceolate to linear, terminating in acute apex; bracts mostly glabrous, scarcely with few solitary trichomes on margin and abaxial lamina. Pollen tricolporate, elliptic, microrugulate, perforate, microrugulae twisted and interwoven, forming aggregates; footlayer moderately thick in mesocolpium.
A further inclusion of a staminate inflorescence that possesses similar pollen to that of E. conwentzii is specimen GZG.BST.24577 (Fig. 47). However, GZG.BST.24577 differs from the other Eotrigonobalanus species from the Baltic amber (Table 12) by the short filaments (Fig. 47A, C, D), fewer anthers per floret (Fig. 47D) and the presence of three different trichome types (Fig. 47E -H). It is difficult to evaluate if these differences are caused by specific degradation processes (maybe due to fungi; Fig. 47D), by the immaturity of the inflorescence (anthers are closed; Fig. 47B -D) or if they are of taxonomic value. Comparing the pollen of the three specimens, pollen of GZG. BST.24577 is the smallest (Fig. 47I, J). The sculpturing of the pollen (Fig. 48) resembles E. conwentzii and would support inclusion of this specimen within E. conwentzii; however, based on the morphological differences of the inflorescences, we only tentatively assign GZG.BST.24577 to E. conwentzii (E. cf. conwentzii).
Specimen GZG.BST.24577 (Eotrigonobalanus cf. conwentzii) is labelled with its original collection number SB6 of the Stantien and Becker Amber Collection from Königsberg, which was later incorporated into the Königsberg Amber Collection (now housed at the University of Göttingen, Germany). In the catalogue of the Stantien and Becker Amber Collection, specimen SB6 is listed as "inflorescence of Quercus subglabra Casp. original" (Klebs 1889), meaning that this particular specimen was studied and possibly also published by Caspary. Caspary (1881) did not describe this particular specimen and only included Q. subglabra in an identification key for staminate inflorescences from Baltic amber. Conwentz (1886) described Q. subglabra in detail, but illustrated a specimen (MB.Pb.1979/0813, Fig. 36; here Q. meyeriana, see also text above) other than SB6 (= GZG.BST.24577). However, he mentioned that in total ten specimens of Q. subglabra existed, and it is not documented which of these specimens was later transferred to the Stantien and Becker Amber Collection.
Remarks -Eotrigonobalanus longianthera differs from E. campanulata and E. conwentzii in the following features (Table 12): rachis with stellate and simple acute trichomes; florets singular; filaments and anthers longer; anthers wider; apex of anthers mucronate; four trichome types present. The pollen ornamentation of E. longian thera differs from E. conwentzii in the wider rugulae, the stronger relief and the thicker pollen wall. Diagnosis -Pendulous staminate inflorescence, rachis occasionally with solitary and stellate trichomes. Florets singular or in dichasia of 3 florets, sessile. Perianth campanulate, mostly incised to upper third, sometimes also to lower third, forming triangular, acute lobes with irregular serrate margin. Outer perianth surface only rarely with solitary trichomes. Each floret with loose, central trichome tuft composed of solitary trichomes. Stamens numerous, 6 or 7 stamens per floret, filaments elongated, projecting far beyond perianth. Anthers narrowly elliptic, apically notched, basifixed. Dichasia subtended by singular, glabrous bract, sometimes lacking (caducous), lanceolate, terminating in acute apex.
Remarks -The combination of the following features supports the assignment of GZG.BST.24419 to Eotrigo nobalanus (Table 12): rachis pendulous; filaments projecting far beyond the perianth; anthers basifixed and large; florets singular or in dichasia; subtending bracts present; trichomes generally scarce; loose, central trichome tuft present. However, E. campanulata differs from E. conwentzii and E. longianthera by the smaller, campanulate perianth, the lower number of notched anthers and from E. longianthera by the florets arranged in dichasia of three (Table 12).
Etymology -The specific epithet refers to the campanulate perianth shape.
Comparison of Eotrigonobalanus species from Baltic am ber -Among staminate inflorescences and singular florets of the Fagaceae from Baltic amber, Quercus meyeri ana sensu Conwentz (here treated as ? Eotrigonobalanoid sp. 2) and Q. trichota var. macranthera (here treated as ? Eotrigonobalanoid sp. 1; see below) resemble the three Eotrigonobalanus species in their high number of stamens, the projecting filaments and the elliptic, elongated, large anthers. However, the perianth of E. conwentzii and E. longianthera is more cyathiform with triangular lobes, whereas Q. meyeriana sensu Conwentz and Q. trichota var. macranthera possess a slender, top-shaped perianth with lanceolate to oblanceolate lobes. Quercus meyeriana sensu Conwentz has a ciliate perianth margin (Conwentz 1886), which is not the case in the Eotrigonobalanus specimens. Moreover, E. longianthera has mucronate anther apices, in contrast to the rounded anthers of Q. meyeriana and Q. trichota var. macranthera. The latter taxon is also distinct by its pubescent perianth surface, which is almost glabrous in specimens of Eotrigonobalanus. Grímsson & al. (2015) compared the fossil-taxon Amentoplexipollenites catahoulaensis Crepet & Nixon from a staminate catkin of the middle to late Oligocene (Texas, U.S.A.; Crepet & Nixon 1989b) to pollen of Eot rigonobalanus based on some similarities of the sculpturing of the pollen grains. However, the staminate inflorescence of A. catahoulaensis differs from E. campanulata, E. conwentzii and E. longianthera by its robust, erect rachis and in the case of E. campanulata and E. conwentzii by its mucronate anthers. This suggests that superficial similarities of pollen ornamentation in Amentoplexipollenites and Eotrigonobalanus do not indicate a close relationship between these taxa. Amentoplexipollenites has a pollen sculpturing consisting of loosely braided (micro)rugulae lacking the regular perforations and furrows (Crepet & Nixon 1989b: fig. 16, 18). Therefore, the staminate catkins of Eotrigonobalanus reported in this study add previously unknown information to the morphology of the extinct fagaceous genus Eotrigonobalanus. Willdenowia 50 -2020  Conwentz (1886: t. II, fig. 11, 12) and currently lost. Second specimen (J, K) from Conwentz (1886: t. II, fig. 13, 14). Both specimens show affinities to Castaneoideae and Trigonobalanoideae. -A -C, I: overview of detached floret from different sides showing top-shaped perianth; outward-bent perianth lobe (A -C, arrowheads; I, a), which Conwentz (1886) interpreted as a subtending bract; D: dense, central trichome tuft, which overtops perianth lobes; E: perianth from below; white lines indicate basally merged perianth lobes; linear, slender bract subtends floret (arrowhead); F: perianth lobe with few trichomes along margin (white arrowhead) and subtending bract (black arrowhead); G: apically notched anthers (arrowhead); H: overview of amber specimen with inclusion shown in A -G and I; J, K: further amber specimen with detached floret (J) and proximal overview of respective floret (K); apex (a) of anther that is flipped over, creating acute shape; however, Conwentz (1886)  Description -Conwentz (1886) described the species as follows: detached floret (Fig. 52A, B). Perianth: campanulate to top-shaped, 2.2 mm long × 2.65 mm wide, incised to lower third, lobes 5 or 6, lanceolate-oblanceolate, each with acute apex, margins entire, outer surface of perianth pubescent (Fig. 52A). Stamens: 9 stamens per floret; filaments projecting beyond perianth, 2 -3 × as long as perianth; anthers elliptic, large, with rounded apex (Fig. 52A). Trichomes: on entire outer surface of perianth (Fig. 52A).
Remarks -Although the original specimen of Quercus trichota var. macranthera is lost, Conwentz's (1886) illustration and description clearly show that this particular specimen differs from Q. trichota in the larger size of the perianth, the greater number of stamens (nine in Q. tri chota var. macranthera, six in Q. trichota), the elongated filaments, which are two to three times as long as the perianth, and the rounded apex of anthers (mucronate in Q. trichota). According to Conwentz (1886), Q. trichota var. macranthera and Q. trichota shared the pubescent outer surface of the perianth, which, however, was not restricted to the upper third. Due to these differences, we exclude Q. trichota var. macranthera from Q. trichota. We disagree with the assignment of Quercus tricho ta var. macranthera to Q. mucronata, as suggested by Kirchheimer (1937) and Iljinskaja (1982; Table 3). The holotype of Q. mucronata (Fig. 37) clearly differs from Q. trichota var. macranthera by its mucronate anthers (Fig. 37F, G), glabrous perianth (Fig. 37H, I) and six stamens ( Fig. 37A -C).
Currently, there are no other species of Quercus from Baltic amber that resemble Q. trichota var. macranthera in all aspects (Table 13), and there is no specimen that is similar to it and could therefore serve as a neotype. The long filaments also occur in the Castaneoideae; however, the anthers in Q. trichota var. macranthera are basifixed (Fig. 52A) and elliptic, which is different to the round, dorsifixed anthers of the castaneoids. The deeply incised perianth, the large number of stamens, the elongated filaments and the basifixed, large, elliptic anthers of Q. trichota var. macranthera are similar to Eotrigonoba lanus. However, none of the species of Eotrigonobalanus exhibits a pubescent perianth surface. Because we cannot reinvestigate the holotype of Q. trichota var. macranthera, and no other specimens are known, a definite assignment to Quercus or Eotrigono balanus is impossible. Description -The specimen that Conwentz (1886) intended to serve as the type of Quercus meyeriana is lost. Therefore, we can only evaluate the descriptions and illustrations given by Conwentz (1886), who described Q. meyeriana (Fig. 52C, D) as follows: Detached floret (Fig. 52C), 2 -3.5 mm long. Perianth: campanulate to top-shaped, incised to base, lobes 6, elongated oblanceolate or lanceolate, each with rounded apex, margins entire, outer surface of perianth glabrous, margins of lobes ciliate (Fig. 52D). Stamens: 6 -10 stamens per floret; filaments projecting beyond perianth, 2 -3 × as long as anthers; anthers elliptic, large with notched apex (Fig.  52D). Trichomes: 1 type: solitary, simple, long, awlshaped, ascending from margin of perianth lobes, only very rarely on outer surface of perianth (Fig. 52D).
Remarks -As discussed above, Quercus meyeriana sensu Conwentz is inadmissible, because Conwentz (1886) chose a new specimen to serve as the type. For the distinction of Conwentz's taxon from Q. meyeriana (Göpp. & Berendt) Brongn. and Q. subglabra, see under those species above (and Table 6).
Quercus meyeriana sensu Conwentz also differs from other Quercus species in Baltic amber by having 6 -10 stamens that project beyond the perianth, notched anthers, and a deeply incised perianth with ciliate margins (Table 13). Although Q. brachyandra exhibits elongated stamens and trichomes along the margin of the perianth lobes, it clearly differs from Q. meyeriana sensu Conwentz in having mucronate anthers. The very long stamens with large, elliptic anthers, the basifixed anthers and the deeply incised perianth resemble Eotrigonobala nus. However, in other Eotrigonobalanus species recognized in the present study, the perianth margins are not as ciliate as shown in Conwentz's illustration. Therefore, we suggest affinities to Eotrigonobalanus, but we cannot unambiguously resolve the true affinities of Q. meyeriana sensu Conwentz, because the type specimen is lost. Conwentz (1886) listed Quercus ciliata as a synonym of Q. meyeriana. Caspary (1880) published a diagnosis and description of florets of Q. ciliata without an illustration. We found an amber specimen with inclusions of two detached florets (Fig. 53A, D), accompanied by a handwritten label of Q. ciliata (MB.Pb.1979/0644;Fig. 53H). One of these florets is similar to Caspary's (1880) description ( Fig. 53A -C; campanulate perianth, six-lobed, lobes lanceolate with acute apex and serrate margin, margin of lobes with elongated trichomes also occasionally located on perianth surface, six stamens), but is different to Conwentz's (1886) illustration of Q. meyeriana, for instance in the apically acute perianth lobes. The second detached floret of MB.Pb.1979/0644 (Fig. 53D -G) is different to Q. meyeriana sensu Conwentz in the mucronate anthers (Fig. 53G) and short filaments ( Fig. 53D; Table  13).
Based on these differences, it seems doubtful that the floret inclusions of Quercus ciliata (MB.Pb.1979/0644) are conspecific with Q. meyeriana sensu Conwentz. Furthermore, the morphology of Q. ciliata (MB. Pb.1979/0644) is quite similar to Q. emanuelii from Bitterfeld amber and to Q. mucronata and Q. trichota from Baltic amber. Therefore, Q. ciliata (MB.Pb.1979/0644) is too indistinct in its morphology to be kept as a separate species or to be incorporated into one of the existing Quercus species from Baltic amber.  54D); uniseriate trichomes few, rarely located along margin of perianth lobes and subtending leaf (Fig. 54F).
Remarks -The holotype of Quercus nuda var. serrulata (MB.Pb.1979/0816) was labelled as "Quercus subglabra var. denticulata" by Caspary (Fig. 55I). Conwentz (1886) was aware of this label but disagreed with Caspary, because the keeled perianth of the holotype is similar to Q. nuda. There is no diagnosis or description of Q. subglabra var. denticulata, and this varietal name is not validly published. Willdenowia 50 -2020 Conwentz (1886) (Göppert 1853), but that species name was not validly published because Göppert (1853) did not provide a description (Turland & al. 2018: Art. 38 .1(a)). Furthermore, the epithet "serrata" was already validly published for the extant species Q. serrata Murray (Murray 1784) and therefore cannot be applied to the amber inclusion. Reinvestigation of the holotype of Q. nuda var. serrulata (MB.Pb.1979/0816;Fig. 55) and of the type of Q. nuda (MB.Pb.1979/0656;Fig. 42) revealed the presence of trichomes on the perianth in both specimens (Fig. 42H, I; 55D -F). Quercus nuda has acute anthers and the same perianth shape as MB.Pb.1979/0816. However, anthers in Q. nuda are smaller ([1120 Kirchheimer (1937) and Iljinskaja (1982) did not see enough evidence to treat MB.Pb.1979/0816 as a variety of Quercus nuda and included it in Q. subglabra (Table 3). Comparing the holotype MB.Pb.1979/0816 of Q. nuda var. serrulata with Q. subglabra, they share the following features: the perianth shape and irregular perianth margin; the anther number and length of filaments; solitary trichomes scarcely distributed along perianth margin; more or less glabrous perianth surface. However, all specimens of Q. subglabra possess significantly smaller anthers with notched apices. Therefore, we disagree with Kirchheimer (1937) and Iljinskaja (1989) and would not include MB.Pb.1979/0816 in Q. subglabra.
Quercus multipilosa has anthers and stamens of similar size as in the holotype MB.Pb.1979/0816 of Q. nuda var. serrulata, a similar perianth shape and bifurcate trichomes; however, they are distinct from each other, in that Q. multipilosa possesses 9 -11 notched anthers per floret and five distinct trichome types. Because the morphology of MB.Pb.1979/0816 is intermediate between specimens of Q. multipilosa and Q. subglabra, it is impossible to assign it unambiguously to one of these taxa. No other specimens of staminate florets from Baltic amber resemble MB.Pb.1979/0816 and its morphology is not distinct enough to justify a new species name (Table 13). The overall morphology of MB.Pb.1979/0816 resembles Quercus, especially in the campanulate perianth, the large, basifixed anthers and short filaments. However, we observed in MB.Pb.1979/0816 that not all anthers are completely open, which may be a sign of immaturity of the floret. This may affect stamen length, which is a significant feature for the generic assignment. Based on the ambiguous morphology of MB.Pb.1979/0816, we cannot definitely assign this specimen to any fagaceous genus.

Diversity and biogeographic patterns of Fagaceae in the Baltic amber flora
In the present study, we reveal a remarkably high diversity of Fagaceae in Baltic amber (Table 3). In total, we distinguish 18 Fagaceae fossil-species, which derive from the Quercoideae (two sections of Quercus: Q. sect. Lobatae and Q. sect. Protobalanus and ten Quercus species), Cas taneoideae (three species) and Trigonobalanoideae (two species) and the extinct genus Eotrigonobalanus (three species). Eight species that were originally assigned to Quercus were dubious or their affinities could not be entirely resolved (Table 3). Still, the number of species greatly exceeds the numbers recognized in the most recent revisions of Baltic amber Fagaceae by Kirchheimer (1937) and Iljinskaja (1982 Valencia-A. 2004). A high number of Quercus species is also found in the U.S.A., with about 89 species (Nixon 2002). Although the extant range of Q. sect. Lobatae is confined to North and Central America, this section had a wider N hemisphere distribution during the Paleogene and Neogene. The oldest fossil record is pollen from the middle Eocene of W Greenland (Grímsson & al. 2015); fossils of leaves, cupules and acorns of Q. sect. Lobatae were described from Oligocene deposits of Texas (Catahoula formation; Daghlian & Crepet 1983) and NE Asia (see Denk & al. 2017a for references) and from the Oligocene and Neogene of Europe (Denk & al. 2017a;Barrón & al. 2017 and references therein). The broad N hemisphere distribution of Q. sect. Lobatae from the Paleogene onward fits with the presence of this section in the Eocene Baltic amber forest.
Quercus sect. Protobalanus is also endemic to North America but contains far fewer species than Q. sect. Lo batae; five species occur in the SW U.S.A. and NW Mexico (Manos 1997). It was recently suggested that Q. sect. "Protobalanus appeared in the Eocene of North America and seems to be always limited to this continent" (Barrón & al. 2017: 86). In contrast, the presence of Q. sect. Pro tobalanus or a precursor of this lineage in Eocene strata of Greenland (Grímsson & al. 2015) and in latest Eocene strata of W North America (Bouchal & al. 2014) already indicated that Q. sect. Protobalanus had a wider distribution than previously thought. Therefore, inclusions of Q. sect. Lobatae and Q. sect. Protobalanus from the Eocene Baltic amber further demonstrate a wide Paleogene distribution of these sections including Europe and the North Atlantic land bridges, which connected Greenland, the landmasses of Fennoscandia and North America (Tiffney 1985;Tiffney & Manchester 2001).
In contrast to Quercus sect. Lobatae and Q. sect. Proto balanus, extant Q. sect. Cyclobalanopsis is restricted to Himalayan to SE and E Asian broad-leaved evergreen forests, where it is a dominant and diverse section (90 -122 species; Deng 2007, cited from Deng & al. 2014). However, its oldest records are suggested to be cupule fossils from the Lutetian (48 Ma) of W North America (Manchester 1994), "but without preserved stigmas the assignment of these fruits remains ambiguous" (Denk & al. 2017a: 31). More reliable records of Q. sect. Cy clobalanopsis are known from middle Eocene and early Oligocene strata of S China (leaves and pollen; Hofmann 2010; ) and from the Eocene/Oli gocene boundary in SE Tibet (leaves; Su & al. 2018). So far, there are no unambiguous fossils of Q. sect. Cyclobala nopsis from Europe (Denk & al. 2017a). Oligocene leaf fossils from Hungary referred to Cyclobala nopsis (Andreánszky 1966) and Pliocene leaf fossils from Bulgaria, which were described as C. stojanovii Palam. & G.  Mai 1976Mai , 1989Manchester 1994;Gee & al. 2003;Wu & al. 2014; Lithocarpus saxonicus Oligocene (Lausitia, Germany) Kitan. (Palamarev & Kitanov 1988;Palamarev & Ivanov 2003; cited from Barrón & al. 2017) may belong to Q. sect. Cyclobalanopsis, but leaf morphological and anatomical characteristics found in this fossil-species also occur in Q. sect. Ilex and in other Fagaceae. Dispersed pollen from the Sarmatian of Austria (Grímsson & al. 2016b) superficially resembles Q. sect. Cyclobalanop sis, but probably belong to Q. sect. Lobatae. Likewise, Q. casparyi (GZG.BST.24402) from Baltic amber and referred in the present study to Q. sect. Cyclobalanopsis or Q. sect. Lobatae closely resembles modern species of Q. sect. Cyclobalanopsis in terms of pollen sculpturing, while the thin and discontinuous footlayer in this specimen fits better with Q. sect. Lobatae. All in all, Quercus inclusions from Baltic amber suggest wider amphiatlantic distributions of sections today confined to America and support notions of transatlantic migration patterns.

Erythrobalanus, Sequoia
Extant Castaneoideae occur in North America, S and SE Asia until New Guinea, as well as in Europe (Kubitzki 1993). Castanopis and Lithocarpus are major constituents of mid-montane SE Asian forests (Soepadmo 1972;Boer & al. 1995;Sunarno & al. 1995). In contrast, Casta nea does not extend to the tropics (Kubitzki 1993); about six species of Castanea occur in E Asia and E North America. The original distribution of the economically important C. sativa Mill. is difficult to assess, because it "has been masked by strong human impact" (Krebs & al. 2004: 145); however, it is the only native species of Cas tanea in Europe . Chrysolepis (C. chrysophylla (Douglas ex Hook.) Hjelmq. and C. sem pervirens) and Notholithocarpus densiflorus are endemic to W North America and restricted to warm-temperate forests of California, Oregon and Washington (Flora of North America 2008a, 2008b).
Corresponding to their extant distribution, fossils of the Castaneoideae are known from America, Europe and Asia (e.g. Crepet and Daghlian 1980;Kvaček & Walther 1987;Mai 1989;Kvaček & Walther 2012;Wu & al. 2014). In C Europe, several genera of the Castaneoideae (Castanopsis, Lithocarpus, extinct Castaneophyllum) occurred from the middle Eocene onward (see Table 14 for an overview). However, the definite assignments of their leaf and wood fossils to extant lineages is often challenging; therefore, so far only Castanopsis and Lithocarpus have been unambiguously proven from the fossil record of C Europe (Mai 1995). The recently discovered Baltic amber inclusion of a cupule of Castanopsis kaulii Sa dowski & al. already proved the presence of this particular genus in the Baltic amber flora ). Although we cannot unambiguously assign the castaneoid inflorescences from amber to an extant genus, it is likely that at least one of them belongs to Castanop sis. The presence of several species of castaneoids in late Eocene Baltic amber further supports the wide distribution of Castaneoideae in the Paleogene of C Europe and reflects their high diversity.
Extant Trigonobalanus verticillata occurs in montane broad-leaved forests of S China (Hainan), Malaysia and Indonesia (Sulawesi, Sumatra; Nixon & Crepet 1989;Ng & Lin 2008). In contrast, Formanodendron doichangensis is restricted to a few localities of evergreen broad-leaved forests in SW China (Yunnan) and N Thailand (Sun & al. 2007). The endemic South American Colombobala nus excelsa occurs solely in montane tropical forests of Colombia (Lozano-C. & al. 1979;Nixon & Crepet 1989). The fossil history of these three taxa is still unresolved. It was suggested that the extinct Trigonobalanopsis exa cantha (Mai) Kvaček & H. Walther of the late Eocene to Pliocene of Europe and Asia shows similarities to Colom bobalanus and may therefore be affiliated with this particular taxon (Kvaček & Walther 1988). However, pollen of T. exacantha shows closer affinities to Lithocarpus (Denk & al. 2012). Another fossil-taxon with affinities to the trigonobalanoids is Trigonobalanoidea Crepet & Nixon, which was described from the Eocene Buchanan formation of Tennessee (Crepet & Nixon 1989a). Infructescences and dispersed fruits of this taxon showed more similarities to Colombobalanus and Formanodendron than to Trigonobalanus, while fossil catkins of Trigono balanoidea were morphologically transitional between Colombobalanus, Fagus, Formanodendron and Quercus (Crepet & Nixon 1989a). A trigonobalanoid fruit inclusion from Baltic amber was initially described as Fagus succinea Goepp. & Menge (Göppert 1853;Conwentz 1886) and later Forman (1964) and Mai (1970) discussed affinities of this inclusion to Trigonobalanus. The type specimen is lost, which impedes further investigation. However, illustrations of the specimen (Conwentz 1886) show typical features of Trigonobalanus, such as the trigonous shape, triangular abscission scar and capitate stigmas (Forman 1964), but future studies are needed to confirm the affinities of this fruit. Yet, considering the trigonobalanoid staminate inflorescences of the present study, it seems likely that Trigonobalanus or a precursor of the lineage existed in the Baltic amber flora. This indicates that the trigonobalanoids were already present in the late Eocene of C Europe, supporting assumptions of an "Euro-American distribution" in the early Paleogene (Crepet & Nixon 1989a).
Eotrigonobalanus is an extinct genus, which was established by Kvaček and Walther (1989) for fossils of leaves, female infructescences, cupules and fruits deriving from middle Eocene to early Miocene strata of C Europe (Kvaček & Walther 1989; see Denk & al. 2012: table 2 for a summary and references of the fossil record of Eotrigonobalanus). Pollen clumps that were found on Eotrigonobalanus leaf fossils were described as E. eiszmannii (Walther and Zetter 1993). However, fossils of staminate catkins of Eotrigonobalanus from the European Paleogene have not been discovered so far. Therefore, the amber specimens reported here represent the first fossil records of staminate inflorescences of Eotrigonobalanus. Eotrigonobalanus was a widely Willdenowia 50 -2020 distributed, abundant component of azonal associations in Paleogene and early Neogene forests of C Europe and was even a lignite-forming taxon (Mai 1995), extending to arctic regions in Greenland (Grímsson & al. 2016a). Its occurrence in Eocene Baltic amber therefore fits with the general distribution of the genus during the Paleogene.
Although Baltic amber Fagaceae are very diverse, certain fagaceous taxa have not been found there so far, including Fagus, Quercus sect. Cerris, Q. sect. Cyclobala nopsis, Q. sect. Ilex (all Q. subg. Cerris) and Q. sect. Quercus. The oldest fossil record of Q. sect. Ilex in Europe is pollen of early Oligocene age (Saxony, Germany; Denk & al. 2012;Denk & al. 2017b) and Q. sect. Cerris had not yet evolved by the Eocene (Hipp & al. 2019). Quercus sect. Cyclobalanopsis has never been reported from W Eurasia (in particular the characteristic cycle cups) and could be an example of a taxon that migrated from North America to Asia via the Bering land bridge or indeed never occurred outside E Asia (see above). In contrast, Q. sect. Quercus had a wide N hemisphere distribution during the Neogene (Borgardt & Pigg 1999). Denk & al. (2017a) mentioned that one Baltic amber inclusion of a singular staminate flower might be affiliated with Q. sect. Quercus, based on pollen extracted from this particular fossil (Crepet 1989: 61, fig. 4.9A, B). Therefore, it is likely that future studies of male inflores-cences from Baltic amber will reveal even more taxa of Quercus and other Fagaceae.
Notably, the Baltic amber flora encompasses numerous taxa that are today most diverse and/or endemic in evergreen broad-leaved forests of mid-montane regions in SE Asia, as well as typical of warm-temperate forests of the U.S.A. and Mexico. A similar pattern was previously observed for the conifer diversity in Baltic amber, which also showed a peculiar mixture of Asian and North American taxa, including genera that are rare today, such as Cathaya Chun & Kuang, Nothotsuga Hu ex C. N. Page, Pseudolarix Gordon and Sciadopitys Siebold & Zucc. (Sadowski & al. 2016a(Sadowski & al. , 2017a. The occurrence of Asian Fagaceae in the Baltic amber flora is in congruence with the evolutionary history of numerous endemic plants from Asia. It has been shown that several extant endemic taxa of Asia, including conifers such as Cathaya and Nothotsuga, occurred in North America and Europe during their fossil history (Manchester & al. 2009). The extant distribution of these plants in Asia is mainly relictual, resulting from the cooling climate and glaciation during the Pliocene and Pleistocene (Manchester & al. 2009). Furthermore, the hyperdiverse Fagace ae from Baltic amber indicate that modern lineages of the Fagaceae "were already diversified by the Eocene" (Grímsson & al. 2015: 830), as it was shown by the middle Eocene Fagaceae flora of W Greenland (Grímsson Fig. 56. Suggested habitat types of the Baltic amber source area accommodating the diverse Fagaceae (indicated in grey boxes) from Baltic amber (modified after Sadowski & al. 2017a & al. 2015). As in a previous study reassessing the conifer diversity in Baltic amber (Sadowski & al. 2017a), the present re-evaluation of amber inclusions of Fagaceae resulted in a highly refined picture of fagaceous diversity in the Paleogene of N Europe.

Palaeoecology of hyperdiverse Fagaceae from Baltic amber
The Baltic amber source area was composed of a variety of habitats, which were previously reconstructed mainly using inclusions of conifers (Sadowski & al. 2016a(Sadowski & al. , 2017a, as well as graminids, dwarf mistletoes and calicioid lichens and fungi (Kaasalainen & al. 2017;Sadow ski & al. 2016aSadow ski & al. , 2016bRikkinen & Schmidt 2018). These habitat types encompass azonal vegetation, including coastal lowland swamps with areas of brackish water influence and water-saturated peat, as well as back swamps and riparian forests. Areas not affected by flooding are mixed mesophytic coniferangiosperm forests including open areas.
When Mai (1970) discussed the diversity of Fagace ae in the Baltic amber flora, he concluded that Castanea, Quercus and Trigonobalanus were the most abundant taxa of the Fagaceae. He interpreted them as constituents of Lauraceae-oak-pine-Trigonobalanus forests that grew on oligotrophic, acidic soils (Mai 1970). We partly agree with this reconstruction, but suggest that the fagaceous taxa occurred in different habitats within the Baltic amber source area and were not restricted to one singular forest type. Furthermore, we suggest that Fagaceae from Baltic amber support assumptions of a warm-temperate, humid source area.
Among castaneoid inclusions from Baltic amber, the staminate catkins showed closest affinities to Castanop sis and/or Lithocarpus, which is why we focus on these two taxa when analysing the (palaeo)ecological requirements of the Castaneoideae from Baltic amber (Table  14). Baltic amber castaneoids (excluding Castanopsis, see below) most likely grew within the mixed mesophytic angiosperm-conifer forest (Fig. 56). This is supported by the extant distribution of Castaneoideae, which mainly occur in mixed angiosperm-conifer forests with a warm-temperate, humid climate. For instance, Lithocar pus grows in mid-montane forests of Malaysia, where it is a dominant or co-dominant tree, associated with taxa such as Castanopsis, Engelhardia, Lauraceae, Quercus and conifers such as Araucariaceae and Podocarpaceae (Agathis Salisb., Phyllocladus Rich. ex Mirb., Podocar pus Labill.; Soepadmo 1972 (Kvaček & Walther 2012 ; Table 14). North American Chrysolepis and Notholithocarpus also occur in mixed forests with coniferous and broad-leaved evergreen angiosperm trees ("mixed evergreen forests", see Wainwright & Barbour 1984 for discussion) as well as in conifer forests of the W U.S.A. (California, Oregon, Washington;Flora of North America 2008a, 2008b. Chrysolepis chrysophylla also occurs in broad-leaved evergreen forests of NW North America, where it grows in dense, moist forests (pers. comm. with Paul S. Manos, Duke University).
Extant evergreen Castanopsis is a major constituent of mid-montane forests, e.g. in Thailand and W Java, and even forms pure stands (Soepadmo 1972;Sunarno & al. 1995). Its highest diversity is reached in SW China, where the tropical and temperate regions overlap (Masahiko 1993). In contrast, Paleogene Castanopsis from C Europe was typically associated with Lauraceae and various conifer taxa forming the so-called "Castanopsietum oligo-miocenicum taphocoenosis" (Mai 1970 and references therein;Mai 1989 ; Table 14) resembling more closely extant forests along the S foothills of the Himalaya (e.g. Menitsky 2005). This particular taphocoenosis of the Oligocene-Miocene occurred on acidic soils, oligotrophic dunes and on river sands. According to Mai (1989), this plant association differed from such ones containing Eotrigonobalanus, various extinct Castane oideae (such as Dryophyllum Debey) and Trigonobala nopsis and from plant associations including roburoid Quercus species. As already mentioned, a pistillate inflorescence inclusion of Castanopsis kaulii was previously reported from Baltic amber, a likely constituent of riparian forests as well as raised bogs within the Baltic amber source area . Therefore, it is likely that one of the staminate florets and/or the inflorescence from this study belonged to Castanopsis and were part of the flood-plain habitats (Fig. 56).
Habitat preferences of extant Quercus are very diverse. Quercus is most diverse in seasonally dry forests and in mild temperate seasonal forests of Mexico and Pacific Central America (Nixon 2006). In North America, Quercus can be a dominant constituent of the canopy or is shrub-like or medium-sized, especially in the chaparral vegetation (Nixon 2006). Some North American Quercus species also grow in extreme habitats, such as in swamps or on serpentine rocks (Nixon 2006). In the U.S.A., Quercus dominates open to closed Mediterranean woodlands (mainly Q. sect. Quercus and Q. sect. Protobalanus, less commonly Q. sect. sect. Lobatae), partly with a pronounced understory of annual grasses (Allen-Diaz & al. 2007), as well as humid temperate forests and flood-plains (Q. sect. Lobatae and Q. sect. Quer cus; Jensen 1997). In Californian oak woodlands, several species of Q. sect. Lobatae and Q. sect. Quercus co-occur with other angiosperm trees, e.g. Aesculus californica (Spach) Nutt., Arbutus menziesii Pursh, Notholithocar pus densiflorus and Umbellularia californica (Hook. & Arn.) Nutt., as well as conifers, e.g. Pinus L. (Allen-Diaz & al. 2007).
In Florida, oak diversity is very high (26 species in total); for instance, forest communities of N Florida comprise up to 17 species of Quercus sect. Lobatae, Q. sect. Quercus and Q. sect. Virentes Loudon, which are able to co-exist within one forest region, comprising sandhill, hammock and scrub communities (Cavender- Bares & al. 2004). The high number of locally occurring oak species is the result of phylogenetic overdispersal. This means that the three main habitats are inhabited by phylogenetically unrelated oak species, while at the same time oaks within the same clade show less niche overlap than expected (Cavender- Bares & al. 2004;Cavender-Bares 2019).
Baltic amber species of Quercus sect. Lobatae might have occurred in woodlands, as well as on flood-plains and stream banks of the Baltic amber source area, as it is the case in extant forest communities in Florida (Jensen 1997;Cavender-Bares & al. 2004). This is also indicated by the fossil record of Q. sect. Lobatae, which is known as a constituent of coniferous forests (e.g. from the late Eocene Florissant fossil beds; Bouchal & al. 2014 Table 14).
Quercus sect. Protobalanus is restricted to the seasonally dry climates of Pacific North America (Bouchal & al. 2014) and inhabits chaparral vegetation, rocky and steep slopes of montane forests, ranging from dry to moist locations (Manos 1997). It forms dense thickets (Q. pal meri Engelm., Q. vacciniifolia Kellogg), inhabits narrow canyons (Q. tomentella Engelm.) or occurs in crevices, growing on rocky or sandy substrate along slopes or in valleys (Q. chrysolepis; Camus 1952Camus -1954a. In the fossil record, Q. sect. Protobalanus was recorded from sclerophyllous forests to nemoral coniferous forests in the middle Eocene of Greenland (Grímsson & al., 2015), in the latest Eocene of Colorado (MacGinitie 1953;Bouchal & al. 2014), in the Oligocene of New Jersey (Prader & al. 2020) and probably in the Miocene Macall flora of Oregon and the Pliocene Sonoma flora of California (references in Barrón & al. 2017), co-existing with species of Q. sect. Lobatae (Bouchal & al. 2014;Grímsson & al. 2015;Prader & al. 2020;Table 14). Therefore, it would be likely that Q. sect. Lobatae and Q. sect. Protobalanus were also associated in the Baltic amber source area. A diverse fagaceous pollen flora of the late Eocene Florissant beds of Colorado showed that Q. sect. Lobatae and Q. sect. Protobalanus were already adapted to seasonally dry climate during the late Eocene (Bouchal & al. 2014; Table 14). The Baltic amber source area also encompassed drier habitats (Sadowski & al. 2016b) that could have harboured thermophilic Quercus taxa, such as Q. sect. Lobatae and Q. sect. Protobalanus.
The extant and fossil examples of habitat types and conditions show that numerous Quercus species are able to co-occur in the same habitat and that Quer cus has a large range in habitat requirements. Based on the high diversity of Quercus from Baltic amber, they likely formed mixed stands. As today, they were probably associated with other angiosperm and conifer taxa, being part of the canopy as well as the understorey.
Fagaceae from Baltic amber occurred in various habitat types within the source area: Castanopsis grew in raised bogs and riparian forests associated with Pi nus, Sciadopitys and Taxodium; Eotrigonobalanus was also part of the azonal vegetation, likely occurring in peat-forming swamps with Quasisequoia and Taxodi um. Quercus sect. Lobatae grew along riversides, but might also have been part of drier sites, associated with Q. sect. Protobalanus. Besides numerous conifer species, the mixed mesophytic conifer-angiosperm forests also included Castaneoideae and trigonobalanoids (Fig.  56).
Staminate inflorescences of the Fagaceae from Baltic amber were both insect-and wind-pollinated. The castaneoid inflorescence (subfamily Castaneoideae, genus indet., GZG.BST.21991) and inclusions of staminate inflorescences with affinities to Trigonobalanus verticil lata (GZG.BST.21995, GZG.BST.21996) exhibit typical features of entomophily, including an erect and robust rachis, long stamen filaments, as well as a thick and continuous foot layer of the pollen wall (Kaul & Abbe 1984;Culley & al. 2002;Denk & Tekleva 2014). In contrast, the diverse oaks from Baltic amber were probably anemophilous, because their axes are pendulous and foot layers of Quercus sect. Lobatae are thin, which are both features suggested to be linked with anemophily (Kaul & Abbe 1984;Denk & Tekleva 2014). Based on the pendulous rachis and the long filaments, the extinct genus Eotrigonobalanus was likely wind-pollinated as well, despite retaining the plesiomorphic thick footlayer (Denk & Teklava 2014). Basal groups of Quercus may possess thick foot layers as well, such as Q. sect. Protobalanus (Denk & Teklava 2014), which is also the case for Quer cus brachyandra (Q. sect. Protobalanus) of this study. However, this is an ancestral feature, because all species of Quercus are anemophilous (Kaul & Abbe 1984;Denk & Teklava 2014).
The climate estimates suggesting tropical or subtropical conditions are predominantly a result of the interpretation of arthropod inclusions from Baltic amber based on autecological traits of the suggested closest modern relatives (e.g. Weitschat 1997;Weitschat & Wichard 1998, Weitschat 2008. This very coarse application of the uniformitarianism principle to the ecology of now extinct groups, however, has previously been critically assessed (Baranov & al. 2019).
Several orders of insect inclusions from Baltic amber (Coleoptera, Hemiptera, Hymenoptera) were recently compared with the insect fauna from fossil assemblages of the middle Eocene Eckfeld maar (Eifel, W Germany, 44.3 ±0.4 Ma;Wappler 2003) and the early to middle Eocene maar lake of Messel (47 Ma; Mertz & Renne 2005;Lenz & al. 2015), showing that all three fossil localities shared numerous insect genera, implying that they were contemporaneous faunas (Wappler 2003;Wappler & Engel 2003). This was further supported by middle Eocene age estimates for Baltic amber, based on K-Ar radiometric analyses (Ritzkowski 1997), supporting assumptions of a subtropical to tropical Baltic amber forest (Weitschat 1997(Weitschat , 2008. However, as already discussed, K-Ar-based age estimates of glauconites were recently criticized because they can be biased toward too-old ages (Clauer & al. 2005;Grimaldi & Ross 2017). Moreover, independent data from Baltic amber inclusions of lichens (Kaasalainen & al. 2017), microfungi (Rikkinen & Schmidt 2018) and conifers (Sadowski & al. 2017a), as well as the latest Priabonian Willdenowia 50 -2020 age of the Blue Earth (Standke 2017), together with the absence of signs of reworking of amber in the marine sediment (Grimaldi & Ross 2017), strongly suggest that Baltic amber was produced in the late Eocene under a more temperate climate (cf. Zachos & al. 2001).
The highly diverse Fagaceae from Baltic amber further substantiates pronounced differences between the late Eocene Baltic amber flora and the subtropical to tropical middle Eocene fossil assemblages of Eckfeld and Messel. In the Eckfeld flora, no macrofossils of the Fagaceae have been discovered so far (Wilde & Frankenhäuser 1998), whereas Fagaceae are taxonomically diverse in the rich record of dispersed pollen (Nickel 1996), indicating that the local Eckfeld flora was probably poor in Fagaceae. Likewise, macrofossils of Fagaceae are also entirely absent in the Messel flora (Wilde 1989(Wilde , 2004Collinson & al. 2012). In contrast, the dispersed pollen record from Messel (Thiele-Pfeiffer 1988) documents abundant pollen of Tricolporopollenites cingulum (R. Potonié) Thomson & Pflug that corresponds to either Casta neoideae or Eotrigonobalanus, while in situ pollen from insect fossils from Messel and Eckfeld (Grímsson & al. 2017) comprises pollen of Castaneoideae with affinities to Castanopsis and Lithocarpus. Yet, several pollen taxa from Messel and Eckfeld (e.g. Tricolporopollenites libla rensis (Thomson & Pflug) Hochuli / T. quisqualis (Potonié) Krutzsch) are difficult to assign to modern genera of Fagaceae and would need to be studied using SEM in order to be comparable to particular pollen types found in Baltic amber. At present, it appears that the Messel and Eckfeld assemblages are dominated by extinct groups of Fagaceae, entirely lacking the high diversity of modern types of Quercus pollen. This further distinguishes the warmer middle Eocene from the more temperate late Eocene floras.
The absence of Fagaceae in the well-studied leaf and carpological record from Messel suggests that they were not common near the area of deposition or that they grew distantly from the lake as part of the hinterland vegetation (Collinson & al. 2012). Both the Messel and Eckfeld plant assemblages contain a great number of paratropical to subtropical evergreen broad-leaved elements (such as

Conclusions
Eocene Baltic amber is considered the largest amber deposit on Earth and reconstruction of the source forests and climate is crucial for evaluation of hundreds of thousands of fossil organisms preserved as inclusions. It has been widely accepted that representatives of the oak family (Fagaceae) were abundant in the source area of Baltic amber, which is due to the presence of numerous stellate trichomes in nearly every specimen of Baltic amber containing inclusions. Previous assumptions suggested that these trichomes derive from oaks and several species of Castanea and Quercus have been described based on staminate inflorescences or individual florets. However, the actual generic and infrageneric diversity of Fagaceae from Baltic amber remained unknown.
In our study, we revealed 18 fossil-taxa from Baltic amber, including the Castaneoideae (affinities with Castanopsis, Lithocarpus), Quercoideae (Quercus sect. Lobatae and Q. sect. Protobalanus), Trigonobalanoi deae (affinity with Trigonobalanus verticillata) and the extinct Eotrigonobalanus. We studied inclusions of staminate inflorescences, detached florets and in situ pollen with affinities to the Fagaceae, including historic type material and newly discovered amber inclusions. Staminate inflorescences of extant taxa were examined to identify diagnostic features distinguishing fagaceous genera and sections in order to facilitate the identification of the inclusions. Morphological features of fagaceous staminate inflorescences, such as the habit of the rachis (flexuous or stiff), stamen length and number, anther size and apex (notched, acute, mucronate, obtuse), trichomes (type, distribution), and the perianth shape constitute valuable characteristics to distinguish families of the Fagaceae. In combination with in situ pollen, it is possible to differentiate between subfamilies and genera of the Fagaceae as well as sections of Quercus. For the extinct Eotrigonobalanus, staminate inflorescences are reported for the first time. The occurrence of North American endemic Q. sect. Lobatae and Q. sect. Proto balanus show a wider N hemisphere distribution of these taxa during the Paleogene, which involved the North Atlantic land bridges to the European continent. Furthermore, the extraordinarily diverse Baltic amber Fagaceae support affinities to modern North American, E Asian, and SE Asian floras.
Comparisons of Baltic amber Fagaceae with extant and fossil analogous Fagaceae taxa further substantiates our notion of the habitat heterogeneity of the amber source area, which comprised azonal and zonal vegetation types. Castaneoids (affinity Castanopsis) and Eo trigonobalanus grew in oligotrophic flood-plain habitats (riparian forests, swamps, peat bogs). Quercus inhabited riverine habitats (Q. sect. Lobatae), as well as drier habitat patches (Q. sect. Lobatae and Q. sect. Protobala nus) within the Baltic amber source area, while trigonobalanoids and further castaneoids occurred in the mixed mesophytic angiosperm-conifer forest (Fig. 56). No additional habitat types / forest types are necessary to accommodate the various Fagaceae representatives reported in this study other than those reconstructed by Sadowski & al. (2017a) based on available conifer inclusions from Baltic amber. The hyperdiverse Fagaceae from Baltic amber further indicate a warm-temperate climate for the Baltic amber forest.