Peculiarities of the Messel fish fauna and their palaeoecological implications: a case study
- First Online:
- Cite this article as:
- Micklich, N. Palaeobio Palaeoenv (2012) 92: 585. doi:10.1007/s12549-012-0106-4
- 244 Views
General aspects and some peculiarities of the Lake Messel fish fauna are presented and discussed with special focus on the palaeoenvironmental framework. The overall composition of that fauna is analysed including details such as age and growth. Palaeopathological information is derived from scale regeneration, and selected aspects of mortality and taphonomy are also investigated. Special emphasis is placed on analyses of the horizontal and vertical distribution patterns of the fishes in comparison with those of plant and arthropod records on the one hand and the orientation patterns of fish carcasses on the other hand. In this context, long- and short-term, and also local, differences and modifications are discussed. All results indicate a very particular environmental scenario. Lake Messel cannot have been steadily isolated from external water bodies during the period of time that is represented by the investigated fossils. There must instead have been various opportunities for a renovation of the lake’s fish fauna. Probably, the peculiarities of that fauna were predominantly triggered by a selective influx, which also changed during more extended periods of time. The selection could have taken place during active immigration events and by modified interactions with different types of external catchment areas. There need not necessarily have been locally fixed inlets and outlets. It is probable that there were more flexible control mechanisms, like an exchange of water with other bodies of water during occasional high water periods, and in places with a partial (or complete) erosion of the tephra wall shelter. The latter may also have varied as a function of the intensity of the respective high water events.
KeywordsPalaeoecologyLake MesselMiddle EoceneFish fauna
The oil shale deposits of the World Heritage Monument Messel Pit near Darmstadt (Hesse, south Germany) are about 47 million years old. Stratigraphically, they therefore belong into the Lower Geiseltalian section of the middle Eocene, which corresponds to the Mammal Stratigraphy Level MP 11 (Franzen 1987, 2005; Franzen and Haubold 1986; Mertz and Renne 2005). They contain a well-preserved fish fauna, which was first mentioned in literature by Kinkelin (1884) and Andreae (1893). Additional information concerning certain taxa was then published by Weitzel (1933), and detailed morphological descriptions together with approaches towards their palaeoecological significance were added by Micklich (1985). Presently, the described Messel fish fauna consists of eight nominal genera and species.
During the last decade, considerable progress has been made towards a better understanding of the palaeoenvironmental situation within and around the ancient Lake Messel (e.g. Felder and Harms 2004; Mertz et al. 2004; Richter and Baszio 2001a, b, 2002, 2006; Richter and Wedmann 2005; Schulz et al. 2002). Fishes are well suited to refine and also test some of the respective hypotheses. Some of them, like the gar species Atractosteus messelensis Grande, 2010 [formerly Atractosteues strausi (Kinkelin, 1884)], and the bowfin Cyclurus kehreri (Andreae, 1893), are very abundant and reported from all oil shale sections that have so far been exposed by scientific excavations. In contrast to some other fossils, e.g. crocodiles (Rossmann and Blume 1999) and horses (Franzen 2007), which may have entered the ancient lake predominantly as carcasses, they were probably autochthonous elements of the Messel fossil assemblage, at least for certain periods of time, which means, they both lived and died there. Therefore, they provide us with a variety of comprehensive information and tools for the reconstruction of their palaeo-ecological and environmental framework. Some of them are presented and discussed in more detail here.
Materials and methods
Most investigated fossils and samples referred to here are from the Messel collection of the Natural History Department of Hessisches Landesmuseum Darmstadt (HLMD) and the respective collection of recent fish species. Comparative morphological studies were carried out using a Wild M3Z stereomicroscope with a direct connection to a Sony Color Video Monitor and colour printer (Mavigraph) and/or a photographic attachment with an EOS 400D digital camera. As far as the selection of recent comparative materials was concerned, species were chosen because they are either closely related to certain Messel forms or probably representative of similar ecotypes and life habits.
Age and growth
For these analyses, scales were selected that were (at least partly) free of overlying materials (or could be cleaned to that effect). To minimise disturbing light reflections, these were coated with ammonium chloride (NH4Cl). The course of their surface ridges (circuli) was investigated (as far as possible) from the scale centres to their margins. As in extant percoids (e.g. Sigler 1949: 223), the growth cessation marks (annuli) were identified by the intersection of the circuli in the anterior and both lateral fields. A disadvantage of the completely articulated fishes in Messel Pit is a strong degree of scale overlap. This in most cases does not allow a complete reconstruction of their growth patterns. In most cases, the annuli could be identified only in very restricted regions of the preserved scales, which may also differ from each other. For the identification of growth cessation marks on the vertebral centrum of HLMD-Me 5404, see Newbrey and Wilson (2005).
Horizontal and vertical fossil distribution patterns
Overall, the data analysis included more than 7,200 records, amongst them about 3,360 fishes (1,824 from HI 7, 661 from EF 8/9, CD 9/10 and D 9/10, 874 from H 13/14), as well as about 3,000 insects, plants and their respective remains (1,019 from HI 7, 791 from EF 8/9, CD 9/10 and D 9/10, 1,251 from H 13/14).
Only fossils with accurately calibrated stratigraphic positions were taken into consideration, as only these can provide reliable information concerning the abundance, respectively the presence and/or absence of particular taxa during clearly defined periods of time. In addition, only more or less complete fishes were incorporated into the investigations, but no isolated bones and other remains. For the latter, microstratigraphic positions could only occasionally be precisely defined. Therefore, they are not representative for the overall and microstratigraphic fossil distribution patterns. In addition, and except for the comparisons with reference column data, no attempts were made to correlate the latter with sedimentological or facial peculiarities.
The fill patterns of the bar charts in the microstratigraphic distribution analyses as well as those in the sketch profile drawings of the electronic supplementary materials are explained in Fig. ESM 1.
The documentation and evaluation of the fish and invertebrate “micro”-remains in the reference columns were done according to the general requirements for the basic data collection during Messel excavations (Habersetzer and Schaal 1994; Keller et al. 1991). Well-defined blocks (average size 50 × 50 × 50 cm) were extracted from the composite oil shale layers along the complete profile sections of the respective excavation sites and transported to the museum’s preparation workshop. Here, they were carefully split with thin blades and investigated under a Wild M3Z stereomicroscope. In the process, all inorganic and fossil remains were documented according to their identity, state of preservation (e.g. completeness) and stratigraphic position.
The windrose charts in the appendix (Figs. ESM 46–ESM 50a, b) are based on records which were exactly defined by the orientation of the main body axes of complete fish finds (for methods, see Franzen 1978: 26, 1979: 24). For the general synopses, all records were plotted according to their orientation patterns. For the details, the metrical distances of the totally excavated sections were divided into four equal quarters. Finer subdivision levels were dispensed because the number of fish records that remained in each section would then have become too small to be of statistical relevance.
More than 39,000 scales from 177 individuals of 20 genera/species were studied in total. These represented the following taxa and specimens: Amiidae: Amia calva Linnaeus, 1766: Specimens HLMD-SMFR 157, 247, 248, 249, 289, 290, 291; average SL 31.92 cm, 2,960 scales studied. †Cyclurus kehreri Agassiz, 1844: Specimens HLMD-Me 8079, 8286, 8917, 8921, 8922, 8926, 8933, 8947, 9055, 9097, 9110, 9740, 9741, 9755, 9759, 9766, 9768, 9799, 14954, 14956, 14958, 15440, 15587, 15600, 16947a, b; average SL 22.41 cm, 5,550 scales studied. Centrarchidae: Ambloplites rupestris (Rafinesque, 1817) 1820: Specimens HLMD-SMFR 98; 12.5 cm SL, 172 scales studied. Lepomis gibbosus (Linnaeus, 1758): Specimens HLMD-SMFR 02, 03, 115, 116, 221, 220, 222, 242; average SL 9.38 cm, 1,821 scales studied. Micropterus Lacépède, 1802: Specimens HLMD-SMFR 07 [M. dolomieui Lacépède, 1802], HLMD-SMFR 108 [M. punctulatus (Rafinesque, 1819)]; average SL 10.85 cm, 632 scales studied. Cichlidae: Hemichromis bimaculatus Gill, 1862: Specimens HLMD-SMFR 13, 14; average SL 6.37 cm, 333 scales studied. Hiodontidae: Hiodon Lesueur, 1818: Specimen HLMD-SMFR 219 [H. alosoides (Rafinesque, 1819)], plus five specimens (two H. alosoides and three H. tergisus Lesueur, 1818) without inventory nos.; average SL 21.6 cm, 248 scales studied. Moronidae: Dicentrarchus labrax (Linnaeus, 1758): Specimens HLMD-SMFR 37, 95, 243, 257, 258, 272, 279, 280, 281; average SL 18.99 cm, 7,397 scales studied. Morone Mitchill, 1814: Specimens HLMD-SMFR 31, 32, 34, 82 [M. chrysops (Rafinesque, 1820)], HLMD-SMFR 86, 90, plus one specimen without inventory no. [M. americana (Gmelin, 1789)]; average SL 10.93 cm, 1,085 scales studied. Percichthyidae: Macquaria Cuvier and Valenciennes, 1830: Specimens HLMD-SMFR 233, 234, 235 [M. novemaculeata (Steindachner, 1866), HLMD-SMFR 237 [M. ambigua (Richardson, 1845)]; average SL 12.5 cm, 1,236 scales studied. Percidae: Gymnocephalus cernua (Linnaeus, 1758): Specimen HLMD-SMFR 67; SL 10.6 cm, 302 scales studied), Perca Linnaeus, 1758: Specimens HLMD-SMFR 5, 42, 53, 245, 247 [P. fluviatilis Linnaeus, 1758], HLMD-SMFR 283 [P. flavescens (Mitchill, 1814)]; average SL 10.5 cm, 1,545 scales studied), Stizostedion lucioperca Linnaeus, 1758: Specimens HLMD-SMFR 259, 260; average SL 9.75 cm, 830 scales studied. Percoidei incertae sedis: †Amphiperca multiformis Weitzel, 19331: Specimens HLMD-Me 7569, 8063, 8092, 8914, 8955, 8956, 8957, 8960, 8970, 8971, 9105, 9761, 9770, 9772, 9773, 10602, 10365, 12552a, b, 12553, 14759, 14764, 14765, 14795, 14978, 15189, 15608, 15816, 16931, 16932, 16940, 16944; average SL 8.79 cm, 4,402 scales studied), †Palaeoperca proxima Micklich, 1978: Specimens HLMD-Me 7636b, 7859a, b, 8005, 9740, 10309a, b, 12555a, 13053, 13056, 13058, 13059, 13060, 13061, 13316, 13319a, b, 13322a, b, 13370a, b, 14614a, b, 14615, 14943, 14947, 14953a, b, 14955, 14979, 15031, 15134a, 15628a, b, 15821, 15841, 17004; average SL 16.47 cm, 3,580 scales studied. †Rhenanoperca minuta Gaudant and Micklich, 1990. Specimens HLMD-Me 9050, 10486, 10510, 10525, 12514, 12550b, 17006a; SMF-ME 2642; average SL 5.07, 274 scales studied. Serranidae: Serranus Cuvier, 1816: Specimens HLMD-SMFR 74, 75, 76, [S. scriba Linnaeus, 1758], HLMD-SMFR 78, 261, 284 [S. hepatus, Linnaeus, 1758)] HLMD-SMFR 287, 288 [S. cabrilla (Linnaeus, 1758)] plus one specimen without inventory no.; average SL 10.62 cm, 4,095 scales studied. Thaumaturidae: †Thaumaturus intermedius Weitzel, 1933: Specimens HLMD-Me 8911, 12511, 12513, 12516, 12519, 12525, 15581; SMF-ME 1600, 1631b, 2211a, 2211b, 2182, 2280, 2311, 2370; average SL 6.1 cm, 503 scales studied. Umbridae: Umbra Kramer, 1777: Specimens HLMD-SMFR 267, 268, 271, 272 [U. krameri Walbaum, 1742], HLMD-SMFR 269, 270 [U. pygmaea DeKay, 1842]; average SL 7.3 cm, 2,447 scales studied).
All investigated scales are from the body areas indicated in the respective figure captions. Those from recent taxa were removed from the specimens with fine tweezers and/or a scalpel, cleaned of adhering tissue remains, stained with alizarin (C14H8O4), flattened and dried, mounted on glass microslides, and protected with thin cover slides. Fossil scales could not be investigated the same way, because these specimens are transfer-prepared, completely articulated in most instances, and show a high amount of overlapping scales, which could not be cleaned and isolated in a similar manner. Therefore, the complete individuals were dusted with ammonium chloride (NH4Cl) and each examined and/or counted scale was marked with a prick of a fine preparation needle.
Abbreviations and definitions
Annulus: A distinctive growth cessation mark that indicates the end or a lowering of the growth rate on reaching the end of one year class; D1: first dorsal fin; D2: second dorsal fin; HLMD: Hessisches Landesmuseum Darmstadt (Hessian State Museum Darmstadt); ESM: electronic supplementary materials; g: Gon; HLMD-Me: Messel collection in the Natural History Department of HLMD; HLMD-SMFR: Extant fish reference collection of the author in the Natural History Department of HLMD; (microstratigraphic) level/unit: Layer/horizon in the microstratigraphic distribution patterns, which is defined/measured with an accuracy of at least 1 cm; SL: Standard length of fish (tip of snout to end of vertebral column); SMF: Senckenberg Institute and Natural History Museum, Frankfurt a. M.; SMF-ME: Messel collection of the Senckenberg Institute and Natural History Museum, Frankfurt a. M.; TL: Total length of fish (tip of snout to end of caudal fin); TUD: Technical University Darmstadt; Ø: Average values.
General composition of the Messel ichthyofauna
The Messel ichthyofauna is characterised by a clear predominance of gars and bowfins, whose extant relatives are able to exist under rather unfavourable environmental conditions such as low levels of oxygen during very hot summers (e.g. Lambou 1962; Pflieger 1975). In addition, and although some of them are very rare, there are taxa that probably entered the lake from external water systems. Extant relatives of the Messel eel Anguilla ignota Micklich, 1985, which, however, is presently only known from one specimen, are catadromous fishes, which enter fresh waters as early-stage juveniles and return to the sea for breeding as fully grown, mature individuals (Micklich 1983). Some of the more common Messel fish species must also be suspected to be migratory and only occasional “guests” within the lake (Micklich 2002a, Micklich and Klappert 2004: 158). Like the eel, the short-snouted gar species Masillosteus kelleri Micklich and Klappert, 2001 is also represented by only a few specimens. It may not have been a constant inhabitant of ancient Lake Messel because of a lack of appropriate prey species (Micklich and Klappert 2001). Interestingly, these gars are more frequent in the nearby Prinz von Hessen fossil site (Klappert and Micklich 2007), which may represent a different type of Lagerstätte (Felder and Harms 2004; Felder et al. 2001) with a different stratigraphic position (Franzen 2006) and other palaeoenvironmental conditions.
Aside from a few exceptions, all Messel fish taxa are represented by both comparatively small and larger specimens. In the past, these were regarded as juveniles and adults, respectively. The comparison of length–frequency distributions, as well as of particularities concerning the development of growth cessation marks on the scales, with extant species that are either closely related to the fossils or considered to represent similar ecotypes, clearly demonstrated that the majority of the Messel specimens were juveniles of rarely more than 3 years old (Micklich 2002a). By contrast, and with the exception of Thaumaturus intermedius and Rhenanoperca minuta, “real” baby individuals, especially of Amphiperca multiformis and Palaeoperca proxima, are very rare. Respective records are mainly restricted to certain excavation areas,2 especially “turtle hill” (grid square HI 7). Some smaller individuals of Amphiperca multiformis (HLMD-Me 9060, 4 cm SL) are also known from profile sections at some distance above marker bed α on the western slope of the pit (grid square D 10), where, in turn, the largest known Rhenanoperca minuta (HLMD-Me 15348, 7.5 cm SL) was also found.
The predominance of physiological “hardlines” in the Messel fish fauna suggests rather unfavourable palaeoenvironmental conditions over extended periods of time. The occurrence of migratory species in combination with the predominance of subadult (and very probably immature) individuals in excavation areas that represent a rather wide range of different profile sections, as well as the rarity of real baby individuals in large areas and profile sections of Messel Pit, cannot be explained by internal control mechanisms in an isolated system like a maar lake. It is strongly indicative of external mechanisms, which means an influx of different kinds of fishes during different periods of time. The only option for fish to enter such a (geomorphologically) closed system may have been by means of water birds, which theoretically could have brought in fertilised eggs with their plumage. However, even for extant fish species, this is an object of controversial discussions and likely to only work under very particular and highly favourable circumstances (e.g. Riehl 1991). Another objection is the general rarity of water birds. Even “rails” (Messelornis cristata Hesse, 1988), which are rather abundant Messel bird fossils, can be discarded as supposed “suppliers”, as they were probably more land-bound and lived close to the shoreline (Mayr 2000: 369, 2005: 41). This assumption is further corroborated by their gut contents, which consisted predominantly of seeds and fruit, whereas only one known specimen had also fed on (probably dead) fish (Morlo 2004: 31–32). In addition, Messelornis cirstata is not common enough in all excavation areas and oil shale sections (Schaal and Möller 1991: 136) to be a likely important supplier of materials for new fish populations to the ancient lake. Furthermore, water birds cannot have played a major role in the dispersal of certain Messel fish species for other reasons. First of all, extant bowfins (Amia calva), which must be considered closely related to the most abundant Messel fish species Cyclurus kehreri, e.g. have a special breeding behaviour. Their adhesive eggs are attached to materials at the bottom of a nest built in a shallow water body. The larvae then remain attached to these materials or lie on the bottom until the resorption of their adhesive organ is completed (Mansueti and Hardy 1967: 23; Pflieger 1975: 73). The nests, larvae, and even the leaving school of fry are guarded by a very aggressive male that will viciously attack approaching predators (M. Newbrey, personal communication). Moreover, when looking at the different frequencies of “baby” fishes in different areas and profile sections of the pit, water birds would have needed to bring in fish fry in a highly selective manner, which means eggs of different species in different quantities during different periods of time, to create the existing patterns, and this is not a very likely hypothesis. Furthermore, such a scenario would exclusively work with the adhesive eggs of substrate breeders but the presumed relatives of at least some Messel fish species have free-floating eggs that are not adhesive at all.
A more plausible alternative explanation than the renovation of the Messel fish fauna by water birds is that of a habitat compartmentation. This possibility has already been considered by Lutz (1990: 112–113), but he assumed that juveniles were lacking because they may have lived mainly in bays that were separated from the main lake basin. Present investigations suggest that this also refers to the majority of fully-grown and sexually mature adults (see next section), which may have lived predominantly in different places.
Age and growth
The rather homogenous surface structures in combination with the somewhat irregular growth line patterns of the Palaeoperca proxima scales may be interpreted such that the respective individuals were not regularly exposed to strongly negative and/or periodically varying environmental conditions. This is in contrast to the results of the microstratigraphic fossil distribution patterns (see respective section below), which indicate repeated switches between hostile and favourable environmental conditions inside the lake. But this is only an apparent contradiction which could be excluded if Palaeoperca proxima would not have been a steady inhabitant of the lake. Referring to the body proportions and the general morphology, it has already been suggested (Micklich 1985: 77, 2007: 61) that these percoids may have cruised in a more widespread habitat and only occasionally came into Lake Messel. Apart from other explanations, which may in principle be possible, this would correspond nicely to the extreme rarity of baby individuals of that species in Messel, the absence of truly large individuals, and some particularities in the composition of the palaeopopulation (e.g. length–frequency distributions; Micklich 2002a: 114–116) of that species. Furthermore, it is another argument against Lake Messel being an isolated system.
The new (strictly speaking, rediscovered) percoid remains clearly show that some individuals must have existed in and around Lake Messel that were distinctively larger than any previously known respective records from Messel. The Grube Prinz von Hessen fossil site is not exactly contemporaneous with Messel Pit (Franzen 2006). However, it represents a rather similar fauna and flora and also an at least roughly similar habitat. Therefore, it was probably also inhabited by percoids with similar lifestyles as the Messel ones, so that the respective percoid fragment may also be used here for the interpretation of the situation in Lake Messel. As it has already been stated, it is obvious that, in the latter, really large individuals of fish are similarly as rare as the babies. This can be explained most simply by the assumption that they lived mostly in different places. The sizes of the new percoid records correspond well with those of some mature adult individuals of certain extant Moronidae (Carlander 1997: 1–62). This family is (among others) represented by certain somewhat anadromous species whose adults migrate to the upper reaches of their home waters for spawning and breeding (e.g. Pflieger 1975: 244–249).4 Later on, the adult individuals return to their original home ranges, leaving the juveniles behind. It is very easy to imagine that these would congregate in depressions and other deep structures with declining water levels. Such a deep structure would be a maar lake, like Messel, if it were not isolated by a tephra wall at the respective periods of time.
Selective size filters that may have played a role during the the transport of the primeval Messel horses into Lake Messel were suggested by Franzen (2007: 70). Fishes may also have been subjected to selection mechanisms if they immigrated into the lake via tributary systems. The largest reported sizes of Messel bowfins, for example, indeed fit into a critical size range as in the manner postulated for the horses. They, of course, have a much more streamlined body than the latter, and even larger individuals should have in principle been able to immigrate or be washed into the Lake Messel. The predominantly aquatic crocodile species Asiatosuchus depressifrons, which—alive or dead—probably also came into the lake on the waterway, is, in Messel, also represented by records and/or at least by respective remains that exceed the critical size of horse carcasses. The different SLs of fishes in different excavation areas can scarcely depend on the consistency of such inlet systems.
General fossil distribution patterns
The general distribution patterns of the fossils as they have been described above correspond to the published information such that counts of small fossil remains in test quadrants (Lutz 1991a) also indicated an increase of plant and insect remains (except Coleoptera) towards the more central areas of the pit. Franzen et al. (1982: 53–69) also reported an increase of insect remains with increasing distance from his supposed inlets (SMF 13 vs. SMF 2, 6 vs. SMF 13), but also, by contrast, a decrease of plants and plant remains (leaves). Both authors, however, registered comparatively many fishes in the older oil shale section in the NW and central parts of the pit as was also found to be the case in the present study.
With regard to the horizontal distribution patterns of the various fish taxa, the published information concerning Cyclurus kehreri corresponds well with the present results. Franzen et al. (1982: 53) registered an “increase of records” with increasing distance from their supposed inlets, which means records were more numerous in the central excavation areas (e.g. SMF 13, 15) than in the marginal ones (SMF 6, 8). As for Thaumaturus intermedius, the present data match published information well as this species has also been reported to be rather common in the NW excavation sectors and very rare in those of the NE (Franzen 1979; Franzen et al. 1982). The HLMD results, however, also contradict the literature, as many Thaumaturus have also been recorded from the central parts of the pit where they had been completely lacking in previous investigations. As for Amphiperca multiformis, a comparison with literature data is rather difficult, because representatives of Rhenanoperca minuta are often confused with juvenile Amphiperca multiformis, which has resulted in a massive overestimate of the numbers, respectively the records; in the NNE excavation areas (SMF 1, 2). Last, but not least, for the two other percoid species (Palaeoperca proxima, Rhenanoperca minuta5), the present results are once more in good agreement with the literature. Franzen (1979: 60) and Franzen et al. (1982: 53) also reported both to be most abundant in those excavation areas (SMF 1, 2) that were close to their supposed NE inlet. They also mentioned Rhenanoperca minuta from younger layers of the more central parts of the pit (e.g. SMF 13, grid square LM 13/4); however, whilst it was completely absent from a nearby excavation area of HLMD-1 (HLMD-1; grid square HI 13/14) that exposed somewhat younger oil shale sections (−660 to −700 cm below M vs. +143 to −401 cm below M).
Concerning the relative lengths of the various fish taxa in the different excavation areas, the possibilities for comparisons with published data are poor. Franzen et al. (1982: 83) referred the largest Cyclurus and Atractosteus specimens to the excavation area SMF 5 (grid square E 15, from ca. +320 to −250 below marker bed α). Comparatively large individuals of Cyclurus were also reported from SMF 9 (grid square E 10, ca. +90 to −60 cm below marker bed α). This agrees with the present investigations insofar as the largest Ø SLs of HLMD specimens were recorded from excavation areas HLMD-2, 2a and 2b (grid squares CD 9/10, EF 8/9), which are comparatively close to SMF 5 (E 15) and also exploited similar oil shale sections (from +115 to −157 cm and from +90 to −115 cm below marker bed α in the HLMD localities vs. ca. +320 to −255 cm below α in the latter). Interestingly, a relative abundance of comparatively large (more than 30 cm SL) bowfins close to marker bed α was also noticed in recent (2011) Messel excavations of the HLMD, but in a NNE position (grid squares H 8/9) rather than in the more western slope sectors. In contrast to the SMF results, the gars with the largest Ø SL (21.71 cm) were found in the younger layers of the central parts of the pit (HI 13/14) in the present investigation. For Thaumaturus intermedius, Franzen (1979: 68) noticed an increase in length towards the central areas of the pit, a result that exactly matches the present data.
Different relative abundances of fossil taxa were noticed and described for different excavation areas inside Messel Pit a long time ago. Together with other peculiarities of the general fossil assemblage, they became the object of various discussions concerning their palaeoenvironmental implications. One favourite idea was a slightly open system with two inlets that entered the lake in different places and during different periods of time (Franzen 1978, 1979; Franzen et al. 1982). Here, most deviations of the faunal assemblages at the various excavation sites were ascribed to alternations in the predominance of the one or the other of these two inlets. The other hypothesis was a maar lake, which was more or less completely isolated from major tributary systems (except for an outlet, Rietschel 1988: 180; or small creeks and/or rainwater runoffs, Lutz 1990, 1991a) for most of the time. Here, the accumulation of particular fossil groups in certain areas is not referred to the proximity of tributaries but to the stratigraphy of the oil shale sections that were exploited layers in these particular places (e.g. Goth 1990: 74). With regard to the fishes, Schaal and Möller (1991: 143) assumed that modifications of the water quality were controlling the composition of fish associations in different excavation areas. One main problem surfacing in all previous discussions is the fact that the excavation areas do not only represent different topographic positions in the ancient lake’s basin but also different time windows (“Zeitfenster”), which probably also means different developmental stages of the ancient ecosystem with different biocoenoses. This problem is rather difficult to control, as it would require comparing excavation areas that are almost identical in their stratigraphic range but differ distinctly from each other in their topographic positions. Contemporaneous oil shale sections are, however, either not accessible in such different places of Messel Pit, or, in those cases where solid oil shale crops out in distant areas, it is of unknown stratigraphic affiliation. Franzen et al. (1982: 53–59) made some attempts to analyse the fossil distribution patterns of two roughly contemporary series (“younger layers” vs. “older layers”) of excavation areas with different topographic positions. However, even within each of their two series, they clustered together excavation areas, some of which were apart in their stratigraphic positions by several meters of vertical distance, which means temporally separated by tens of thousands of years.
The results of the present study can only partially show a way out of this dilemma. On the one hand, the fact that Palaeoperca proxima is obviously absent in the excavation areas around reference layer α along the NW slopes, but has recently been recorded from stratigraphically almost identical oil shale sections in the north, might lead to the conclusion that local factors played a role in the composition of the fossil assemblages of certain excavation areas. The same might be suggested by the presence of Rhenanoperca minuta in oil shale sections about 15 m above α in certain north-western excavation sectors (HLMD-1985/15; grid square D 10) and its absence in nearly contemporaneous layers that are exposed in more south-eastern areas (SMF 15; grid square F 13; ca. +1,900 to 1,770 cm above α). The local co-occurrence of species that are otherwise separated from each other in the pit, which was explained as habitat-dependent peculiarites by Franzen et al. (1982: 32), could be interpreted in a similar manner. On the other hand, the co-occurrence of comparatively large-sized bowfins (Cyclurus kehreri) in some of the first-mentioned western excavation and contemporaneous layers in the north could also indicate that these associations are a result of time. Some other peculiarities in, and modifications to, the fish fauna of certain excavation areas are suggestive of the same, as are the microstratigraphic analyses discussed below. Probably, the local fish and fossil associations of the various excavation areas were determined by both stratigraphy (time) and topography (location). The leading factor could have been water quality,6 but from the described facts, it cannot be convincingly concluded what were the triggering mechanisms. The general differences in the composition of the fish assemblages of certain excavation areas might result from long-term effects and in fact represent general traits of the palaeoenvironmental conditions within Lake Messel. They could just as well be related to other peculiarities, such as geomorphological ones (e.g. steep shore lines vs. more shallow areas; presence/influence of occasional inlets or outlets), which could also have been prevailing in an ancient maar lake for longer periods of time. The comparison of fossil distributions from excavation areas that are far apart from each other but represent the same stratigraphic sections, as well of those that are located in similar places (identical positions, ideally) and expose oil shale sections that represent extended periods of time, could be key to resolving these problems and should therefore be given more consideration in the planning of future excavations.
Microstratigraphic fossil distribution patterns
Marker bed M, grid squares HI 13/1 (excavation area HLMD-1)
In these comparatively young oil shale layers in the more central parts of the pit, records from the uppermost profile sections (+143 to +99 cm above marker bed M; Figs. ESM 3a–4a) are rather sparse. 20 microstratigraphic levels are occupied by fish. There are large gaps in the vertical distribution pattern, which become smaller towards the lower parts of this profile section. The fauna is rather monotypic, and consists mainly of the primitive teleostean fish Thaumaturus intermedius (14 records), in combination with Cyclurus kehreri (14 records), one percoid, and a few gars (3 records). The records are also comparatively sparse as to their relative frequencies. There is a maximum of 5 fishes at level +102 cm.
From +99 to +25 cm above M (Figs. ESM 4a–7a), 101 levels contain fish records. Once again, the fauna is dominated by bowfins (Cyclurus kehreri, 135 records) and Thaumaturus intermedius (82 records), which clearly surpass the gars (Atractosteus messelensis, 29 records). Nevertheless, there are also some sparse percoid records (Amphiperca multiformis at +95 cm, an unidentified record at +92 cm, and a Palaeoperca proxima at +85 cm). There are only minor gaps in the vertical fish distribution pattern (e.g. from +92 to +90 cm and from +89 to +87 cm). The maximum frequency (11 records) was documented from level +65 cm; however, there are also comparatively high values (8 records) at levels +81 and +80 cm.
Fishes, as well as plants and arthropod remains, become very scarce in the profile section between +25 and 0 cm above M (Fig. ESM 7a). They show large gaps in their vertical distribution patterns, which are comparatively rarely covered by plant and arthropod remains. In addition, there are low maximum frequencies that do not exceed three records per level in the case of fish. The fauna is still dominated by Cyclurus kehreri (9 records) and Thaumaturus intermedius (10 records), with only one gar inbetween.
The fish records become distinctly more frequent in the subsequent profile section between −1 and −51 cm below marker bed M (Fig. ESM 8a), with 24 levels being occupied. Once again, Cyclurus kehreri (22 records) and Thaumaturus intermedius (4 records) are the most common species. Nevertheless, there also are four percoids (at −24, −26, −34 and −48 cm) and three gars (at −16, −45 and −50 cm). There are still some larger gaps in the respective vertical distribution patterns (e.g. from −1 to −5 cm, from −26 to −32 cm, and from −40 to −45 cm), which are fairly well covered by plants and arthropod remains.
The profile section from −52 to −101 cm below marker bed M (Fig. ESM 9a) is characterised by numerous fish records, which occur at 46 levels. The vertical distribution pattern has only small gaps (e.g. from −53 to −56 cm). Furthermore, it is almost monospecific, in that it is dominated by Cyclurus kehreri (76 records vs. 7 of Atractosteus messelensis, 8 records of percoids, and 1 of Thaumaturus intermedius). The maximum frequency (5 fishes) is reached at level −60 cm.
From −101 to −253 cm below marker bed M (Figs. ESM 10a–12a), the fish fauna becomes continuously poorer, with correspondingly larger gaps (maximum of 3 empty levels in Fig. ESM 9a, 4 empty levels in Figs. ESM 11a and 21 empty levels in Fig. ESM 12a). A total of 67 levels are occupied, 18 of which are shared with plants and arthropods. Cyclurus kehreri still dominates the fish fauna (76 records), especially in the uppermost part of this profile section (Fig. ESM 10a). This time, it is followed by percoids (12 records of Amphiperca multiformis, 5 of Palaeoperca proxima, 13 of of Percoidei indet.), which are almost as common as the bowfins in the lowermost parts of this particular profile section. Atractosteus messelensis is represented by 10 records, and Thaumaturus intermedius by 2. Fish are most common at levels −110 and −120 cm (7 records each).
The profile sections from −253 to −401 cm below marker bed M (Figs. ESM 13a–15a) are characterised by another increase in fish records, which in total occur at 82 microstratigraphic levels. Generally speaking, they reach their maximum abundance between −301 and −351 cm below M (four microstratigraphic levels, with four records each; see Fig. ESM 14a) even though their absolute maximum (six records) is at −380 cm. Interestingly, the fish fauna is dominated by Cyclurus kehreri and percoids from – 253 to −300 cm below M, where no gars at all occur (Fig. ESM 13a). The latter almost completely replace the percoids from −301 to −401 cm, where only one unidentified percoid record was made (at −384 cm). Therefore, the fish fauna in these lowermost profile sections almost exclusively consists of bowfins and gars, amongst which—aside from the one exception mentioned before—Cyclurus kehreri is the only fish species that is represented below −351 cm.
Marker bed α, grid squares CD 9/10 and EF 8/9 (excavation areas HLMD 2, 2a, b)
From +116 to +50 cm above marker bed α, the fish fauna is comparatively sparse. Respective records are restricted to only 29 microstratigraphic levels (Fig. ESM 16a). In addition, there is one large gap (from +112 to +98 cm) and some smaller ones (e.g. from +97 to +93 cm and from +68 to +63 cm). The maximum diversity is three fishes, which is reached at four different microstratigraphic levels (i.e., at +82, +73, +63 and +53 cm). Once again, Cyclurus kehreri is the most common species, followed by Atractosteus messelensis, which is slightly more common than Thaumaturus intermedius. There are no records of percoids. From +49 to 0 cm above marker bed α, fishes become more abundant and occur at 46 microstratigraphic levels (Fig. ESM 16b). The gaps also become distinctly smaller, with a maximum width of three microstratigraphic levels. The maximum is four fishes per level (at +15 cm). Cyclurus kehreri is still the predominant species, but Thaumaturus intermedius here becomes a little more frequent than the gars. In addition, the first occurrences of Amphiperca multiformis are to be found here. Maximum density (and also diversity) of the fish records is encountered in the profile sections from −1 to −50 cm below marker bed α (Fig. ESM 17a), where respective records were made from 79 microstratigraphic levels only separated by very small gaps. The maximum frequency is seven records per level, and it is reached eight times in this profile section, especially in the deeper parts between −38 and −48 cm. Bowfins (Cyclurus kehreri) are the dominant species by far, but the relative numbers of Thaumaturus intermedius exceed those of Atractosteus messelensis almost three times now. Percoids (Amphiperca multiformis) are rather scarce with only five records. The fish fauna still is comparatively dense and diverse from −51 to −100 cm below marker bed α (Fig. ESM 17b), but not to the same extent as in the preceding profile section. Fishes occur at 49 microstratigraphic levels, and the gaps become a little larger once more. The maximum is five records per level and is reached three times. Cyclurus kehreri still dominates, but Thaumaturus intermedius and Atractosteus messelensis have switched their positions again. The latter is now about two and a half times more common than the first mentioned species. Fish records are rather sparse from −100 to −158 below marker bed α (Fig. ESM 18), and only occur at 23 microstratigraphic levels. There are large gaps (e.g. 13 “empty” levels between −127.5 and −140.5 cm, 7 “empty” levels between −142.5 and −150 cm), and the maximum frequency is reduced to two fishes per level. In addition, the fish fauna is greatly impoverished and only consist of bowfins (Cyclurus kehreri) and gars (Atractosteus messelensis).
Marker bed β, grid squares HI 7 (excavation areas HLMD-4a, b)
From +211 to +150 cm above marker bed β, the fish fauna is rather poor (Fig. ESM 19a). Records exist only from 32 microstratigraphic levels, and there are numerous gaps with five to three “empty” levels. The maximum number of records (six) is reached at 170 cm above β. Aside from three records of Rhenanoperca minuta between +173 and +170 cm, the fish fauna is greatly impoverished and only consists of bowfins (Cyclurus kehreri) and gars (Atractosteus messelensis), with the first being five times more abundant than the latter. Fishes are extremely rare between +150 and +100 cm above marker bed β (Fig. ESM 19b) and only occur at five microstratigraphic levels (a maximum of two per level). There are no other species than bowfins (four records) and gars (two records). Fish records become only a little more frequent (10 occupied levels) from +50 to 0 cm above marker bed β (Fig. ESM 20a), but still show some larger gaps (e.g. from +33 to +10 cm). The maximum is still two per level; however, the composition of the fauna becomes distinctly more diverse. Rhenanoperca minuta (four records) is now the most common species, closely followed by Cyclurus kehreri and Atractosteus messelensis (three records each). In addition, there is one record of Palaeoperca proxima. This trend increases from 0 to −51 cm below marker bed β (Fig. ESM 20b), where the fishes become more abundant (22 occupied microstratigraphic levels, smaller gaps) and diverse, at least in the uppermost part (from 0 to −21 cm), which is characterised by the presence of Cyclurus kehreri (nine records), Atractosteus messelensis (one record), Rhenanoperca minuta (seven records), and Palaeoperca proxima (three records). By contrast, the deeper sections are almost monospecific. Aside from one gar, fish are here exclusively represented by Rhenanoperca minuta. The fauna becomes continuously poorer and less diverse in the lower profile sections (Fig. ESM 21a, b). There are only 17 fish-bearing microstratigraphic levels between −51 and −100 cm, and this number decreases to 6 from −100 to −133 cm. Cyclurus kehreri again becomes the dominant species, and the percoids disappear completely from −111 cm below β.
Marker bed γ, grid squares HI 7 (excavation area HLMD-3)
In the profile section from +118 cm above to −25 cm below marker bed γ (Figs. ESM 22a–26a), 129 microstratigraphic levels were found to be occupied by fish. Fish records are comparatively sparse and uniform in that they consist mainly of Cyclurus kehreri and Rhenanoperca minuta. In addition, there are a few Atractosteus messelensis (except in the section from +83 to +50 cm, where they are comparatively abundant), and one Masillosteus kelleri. There are also some comparatively large sections that are entirely free of fishes (e.g. +117 to +103, −20 to −24 cm above/below γ).
The fish fauna then becomes more diversified and also more abundant in the middle profile sections from −26 to −100 cm below marker bed γ (Figs. ESM 27a–28a), with 89 fish-bearing levels recorded. In addition, profile sections without any fish become smaller. The maximum density of fish records per level is seven (at −42 cm below marker bed γ). Rhenanoperca minuta (105 records) is the most abundant species here and clearly dominates Cyclurus kehreri (59 records) and the gar Atractosteus messelensis. With 38 records, the latter is distinctly more frequent here than in the preceding section. In addition, there are the first occurrences of Palaeoperca proxima.
From −101 to −150 cm below marker bed γ (Fig. ESM 29a), 30 levels are occupied by fish. The maximum frequency (four) is reached at −131 cm below marker bed γ. Gars (Atractoseus messelensis, 3 records) and also Rhenanoperca minuta (14 records) become much rarer than before, and there are also somewhat larger gaps in both the general and particular fish distribution patters (e.g. from −113 to −144 cm for the gars and from −108 to −129 cm for Rhenanoperca minuta).
In the profile section from −151 to −200 cm below marker bed γ (Fig. ESM 30a), 37 fish-bearing levels are found. The fauna becomes distinctly more diversified (15 Atractosteus messelensis, 8 Palaeoperca proxima, first occurrences of Amphiperca multiformis), but with fewer records of Rhenanoperca minuta (9). Once again, there are somewhat larger gaps in the occurrence of individual taxa, but these are smaller than in the preceding section. The maximum density (seven documented records) is reached at −194 cm below marker bed γ.
From −201 to −250 below marker bed bed γ (Fig. ESM 31a), 37 levels are occupied by fish. Although their vertical density distribution pattern is almost the same as in the preceding profile section, the fish fauna again becomes much more one-sided (59 bowfins, 10 gars, 1 unidentified percoid), and the relative abundance of fish records is lower than before. A singular maximum of six records was found at −225 cm.
From −250 to −301 cm below marker bed γ (Fig. ESM 32a), the vertical density distribution pattern of fish records as well as their relative frequency decrease. 34 levels are occupied, with a maximum of four records at −276 cm. The fauna becomes less uniform than before. Aside from Cyclurus kehreri (25 records) and Atractosteus messelensis (3 records), there are some occurrences of Amphiperca multiformis, one Thaumaturus intermedius, and also a very few Rhenanoperca minuta. The fishes show larger gaps in their vertical distribution patterns (e.g. −254 to −260 cm and −269 to −275 cm).
In the lowermost profile section studied (−302 to −338 cm below marker bed γ, Fig. ESM 33a), fish become rather sparse in their vertical distribution patterns. Only 27 levels are occupied, which are somewhat clustered together in the middle part of the profile section (−312 to −320 cm). There is a maximum of four records (at −320 cm below γ). The fish fauna is clearly now dominated by Amphiperca multiformis (30 records). In addition, there are a few bowfins (Cyclurus kehreri), two gars (Atractosteus messelensis), and one Palaeoperca proxima. Rhenanoperca minuta is completely absent.
To start with, it must be mentioned that these analyses were subject to particular basic conditions, which may partially subtract from their significance in certain aspects. In general, it must be considered that they are only small pinholes in the total oil shale deposit and that they, therefore, can only elucidate very brief moments in the lake’s history and might not properly tell the complete story. Respective investigations furthermore depend on a wealth of variables that are difficult to assess, but have the potential of impacting significantly on the number of detected records. Restricted fissibility of the oil shale in certain profile sections may influence this number, as well as weathering, especially in the uppermost sections of the investigated profile columns. Both effects must be carefully considered in the evaluation of data, as the resulting low numbers of records in such cases, for example, cannot necessarily be referred to unfavourable environmental conditions. The situation is almost the same when it comes to the general frequency distribution of fossils along a given vertical profile section. Results depend heavily on the intensity of the respective excavation works (e.g. the number of people who participated in a particular excavation as well as the timespan of the respective mining activities). Gaps and sparse general frequencies of records in the microstratigraphic distribution patters, even when they are not statistical artefacts and/or do not result from one of the above mentioned effects, must also not mean unfavourable living conditions by default. First of all, they can also be due to poor preservation conditions that may have substantially diminished the number of fossils during certain periods of time. Alternatively, living conditions might in fact have been excellent so that fewer fishes died in the first place. In addition, it always must be considered that—except for the reference column—disarticulate fossil remains are generally less frequently measured and documented than complete ones. By contrast, however, it should also be considered that preservation conditions in Lake Messel were in general probably good to excellent, so that gaps in otherwise well-documented profile sections could indeed be representative of suboptimal environmental conditions rather than of artefacts in the fossilisation process. Last, but not least, the size of an excavation area in combination with the maximum depth of local mining cannot be used for estimates concerning the relative numbers of fossils per volume unit; in most cases, it is not clear whether the respective area was everywhere excavated with the same care.
As far as microstratigraphic fossil distribution patterns are concerned, this means that the numbers and relative densities of records at different levels and sections of the profile columns are less representative for faunal and palaeoenvironmental changes than the general presence and/or absence of particular types of fossils, with respect to the poverty and/or richness of the taxonomic spectrum in particular profile sections. Referring to the data presented here, it can be concluded that the most reliable information stems from the excavation area HLMD-1, especially from those oil shale sections below marker bed M, where a well-defined volume body and profile section was carefully and consistently investigated and documented during a special excavation project of the HLMD (Micklich 2002b). As for their reliability, second place should be awarded to the data from the middle profile sections of HLMD-1 (above M), HLMD-2, 2a and 2b and HLMD-3 as microstratigraphic excavations were consistently conducted in these areas for expanded periods of time.
Gars (Atractosteus messelensis) and bowfins (Cyclurus kehreri) are physiologically more robust than the other Messel fish taxa.
The latter probably also differed from each other in their physiological capabilities and specific lifestyles. With these being unknown, they are here supposed equivalent.
Profile sections with sparse fish records and numerous gaps are indicative of poor environmental conditions.
Profile sections with a dense fish population are indicative of better environmental conditions.
Profile sections with a diverse fish fauna are more indicative of favourable environmental conditions than those with a monotypic one.
Looking at the microstratigraphic fish distribution pattern around marker bed α, it firstly needs to be stated that the previous assumption that a diverse fish fauna may be correlated with profile sections in which the fish fauna is generally sparse cannot be confirmed. This is not all that surprising and agrees well with hypotheses 4 and 5 of the general part of this section. Another general trend that was indicated by the distribution patterns around M, however, is confirmed. There appears to be some kind of negative correlation between the occurrence of Thaumaturus intermedius und Atractosteus messelensis insofar as their respective predominance switches round and then back in the profile section from +116 to −100 cm below α. Nevertheless, both generally co-occur in the same profile sections.
With regard to the fish distribution patterns around marker bed β, it must be emphasised that the respective excavation areas were not as intensively excavated as the other ones, which diminishes the relative significance of the respective data. As far as the presumable relationship between the poverty/richness of fish records and their taxonomical diversity is concerned, it can be stated that some sections that are generally poor in fishes are also rather biased in their species diversity, whilst others are comparatively diverse. Those sections that are generally comparatively rich in fish records are here also taxonomically rather diverse. The presumed negative correlation between Thaumaturus intermedius and Atractosteus messelensis is confirmed insofar as the former is completely lacking whilst the latter is present throughout the investigated profile.
As far as the literature is concerned, more or less detailed analyses of microstratigraphic fossil distribution patterns are rare. Franzen (1979) did so for a profile section of 280 cm total depth (grid square E 15, several metres above marker bed α), which he subdivided into 13 smaller units of well-defined horizontal and vertical extensions. He noticed distinct changes in the relative abundance of the various fish species along these profile sections, e.g. mutually exclusive frequency trends of Amphiperca multiformis and Thaumaturus intermedius as well as between the latter and the “small percids” (Rhenanoperca minuta) as well as Cyclurus kehreri. Moreover, he could not find a correlation between the fossil content and any particular type of facies. Schaal and Möller (1991) also divided various profile sections (mainly above and below marker bed α) into smaller subunits (thickness between 25 and 130 cm),8 and (among others) noticed a negative correlation between the relative abundances of Cyclurus kehreri and Thaumaturus intermedius. A negative correlation between the relative abundances of Thaumaturus and percoids (Pararhenanoperca eckfeldensis Micklich and Wuttke, 1988) and also of Cyclurus (by means of scales), and both the other taxa are also mentioned by Lutz and Kaulfuss (2006: 437), who furthermore suggested a predator–prey relationship in the latter case.
Trying to summarise the results of the preceding discussions, it can firstly be concluded that the M “event” had the most remarkable impact on composition and microstratigraphic fish distribution patterns. There are also some other striking modifications (e.g. the Cyclurus/Atractosteus “event” below −201 cm below γ), which, however, cannot be correlated with other peculiarities at the moment (e.g. changes in the type of facies, which are not investigated here). The microstratigraphic fish distribution patterns confirm the general traits that have already been mentioned and discussed. In addition, they strongly suggest there were periods of time during which the ancient lake appears to have been inhabited by fish rather continuously, or offered particularly favourable conditions for their fossilisation, and that these were repeatedly interrupted by some longer and shorter phases, during which it may have been “dead” or less favourable, at least insofar as no articulated fish remains have been found.9 There are also alternations between periods with highly uniform fish associations and those with more diversified ones, and there are abrupt as well as gradual changes in the relative frequencies of certain species, which once again clearly discards the possibility of colonisation of the lake by water birds. Many of the minor modifications of vertical fish distribution patterns that are evident from these documentations probably result from short-term effects. Gradual modifications, such as oscillations, may have occurred repeatedly in different areas of Lake Messel during different periods of time and might thus have affected the fossil record in a similar manner.
In principle, these “long-term”, as well as the “short-term”, modifications may be referred to a variety of causes. They, could, for example, be due to external effects, like a different “input”. It appears to be reasonable to presume that the different species were a priori represented in and/or immigrated into the lake in varying numbers of individuals at different periods of time. Internal effects, like selective living (or, less probably, fossilisation) conditions and/or mortality of certain individuals or species could have secondarily affected and regulated stock populations or the original input in different ways. Species that were originally represented by high numbers may have been more affected by unfavourable environmental conditions than others, and higher mortality rates could have revealed their under-representation within the fossil record. In contrast, certain environmental conditions could also have caused a prolif-eration of other species, which could then have become over-represented within the fossil record. Short-term oscillations in the faunal composition may also result from predator–prey relationships. The negative frequency correlation between Atractosteus messelensis and Thaumaturus intermedius matches this scenario nicely, and to some extent also the one between the latter and the percoids in certain profile sections. However, this has not been proven by any direct evidence until now (see “Discussion”). In addition, Cyclurus kehreri, as another presumed (and directly proven) predator of Thaumaturus and the percoids (of which smaller individuals can also be considered well-suited prey) is present almost everywhere, but does not show such clear oscillations. Last, but not least, frequency oscillations also occur between species (Thaumaturus intermedius and percoids) that can scarcely be assumed to have been in any predator–prey relationship. Predator–prey relationships cannot be an explanation for the mutually suppressive distribution patterns of Cyclurus and Thaumaturus either, because other parts of the referred profile sections just show opposing frequency characteristics (Schaal and Möller 1991: 139). In addition, mutually suppressive frequency trends have also been reported for Atractosteus and Cyclurus (Schaal and Möller 1991: 136), which can scarcely be regarded as representing a predator–prey relationship.
Generally, it is almost impossible to paint a conclusive picture from all these confusing details. Selective living or fossilisation conditions and secondary overlays of original frequency distributions may be ruled out to some extent, however. Most species were represented by juveniles (not babies!) and different proliferation rates cannot essentially have affected their fossil record within the lake’s sediments. They can also scarcely explain the numerical predominance of individuals of species that could be assumed to have been more physiologically “robust” than others. Bowfins and gars, for example, are in many layers associated with representatives of smaller-sized and presumably physiologically “inferior” species (Thaumaturus, percoids; even if some extant species, especially of the latter group, are known to live and survive in extreme environmental conditions). This is almost the same concerning the different frequencies of bowfins and gars in the same profile sections. Both must be assumed to have had similar survival abilities, and both have been recorded from the same layers/horizons in similar size ranges and qualities of preservation; some of them are even known to lie close together on the same bedding plane (e.g. HLMD-Me 15835). Last, but not least, the general objections that were just mentioned above should not be forgotten, and once again there is considerable doubt that profile sections representing considerable periods of time can be indicative of the existence of such highly particular interspecific relationships.
Most of the peculiarities of the vertical and horizontal distribution patterns described and discussed here appear to be indicative of control mechanisms that are not exclusively based on physiological ecological preferences and capabilities, respectively, of the involved fish taxa. They rather result from a combination of various effects: local particularities that could be due to the relative positions of the excavated areas, the regions they may have represented within the ancient lake basin, and general developmental traits of the lake’s history. As has already been indicated, general shifts in the “input” appear to be well suited as a basic motor for most of these modifications. If this input took place from external water bodies, it probably also improved the general water quality in Lake Messel so that immigrating fish species should have been able to live and survive in it for certain periods of time.
Comparison with reference column data
Only more or less complete fish records were considered in the preceding analyses. Therefore, it is important to know whether smaller remains, like isolated bones and shed scales, follow the same microstratigraphic distribution patterns or show major deviations from the complete finds. Furthermore, it is important to know whether there is any relationship between the consistency of the oil shale (e.g. lamination, content of siderite) and fish distribution patterns. Therefore, the latter (Figs. ESM 34c–41c) need to be compared with a detailed documentation of the fish remains distribution pattern (Figs. ESM 31b–41b) and profiling data (Figs. ESM 34a–41a), which was done by means of a reference column study. The latter was conducted during a pilot excavation project by the HLMD in 2000 and 2001 (Micklich and Drobek 2007: 17) focusing on oil shale sections below the reference layer M.10
First of all, it can be stated that, in this investigated profile section (from 0 to −401 cm below marker bed M), the microstratigraphic distribution pattern of more or less complete fishes agrees very well with the one of fish remains in most profile sections: 376 more or less complete fishes were recorded, as opposed to 321 fish remains in the reference column. The latter consisted mainly of Cyclurus kehreri scales (86.6 %). Sections with many complete fishes are also comparatively rich in fish remains (e.g. −288 to −300 cm and −302 to −350 cm below M). In contrast, the more or less complete fishes quite often cover the gaps in the microstratigraphic fish remain distribution pattern in the reference column (e.g. −256 to −288 cm below M, −368 to −440 cm below M), a peculiarity that almost never occurs in a reversed situation.11
Compared with the consistency of the oil shale, it can be stated that fishes occur in all of its types. They are present in the “regular” oil shale as well as in the one rich in siderite layers (e.g. from −73 to −90 cm below M). The situation is the same in oil shale bearing siderite shred (e.g. form −91 to −93 cm below M and from −132 to −135 cm below M) and graded layers of siderite (e.g. from −65 to −70 cm below M and from −194 to −200 cm below M). Furthermore, no peculiarities are evident concerning the microstratigraphic distribution patterns of the individual taxa. Every species appears to be represented in every type of oil shale. The percoids (Amphiperca multiformis, Palaeoperca proxima), for example even occur in the definite layers of siderite (e.g. at −23 and −24 cm below reference layer M).
As shown by the comparisons with the microfish remain distribution in the reference column, the “normal” microstratigraphic fish distribution patterns provide sufficiently reliable data for an evaluation of the development of the fish fauna inside Lake Messel for given periods of time. Of particular interest is also the fact that, in Messel, the relative abundance of fishes is not negatively correlated to the occurrence of siderite. Such a relationship was described for the Lake Eckfeld maar (“siderite event”) by Lutz (1994) and Lutz and Kaulfuss (2006), who furthermore correlated the amount of siderite precipitation in the monimolimnion as an indicator for the water level and quality (low when there is much siderite precipitation, high otherwise) in the lake and/or the presence or absence (low precipitation during connected phases, high precipitation during isolation periods) of a connection to a local fluvial system. Similarly, Franzen (2007: 70–72) also noted a correlation between accumulations of siderite and the lack of fossils of the primeval horses Propalaeotherium and Eurohippus for Lake Messel and referred it to modifications of the water level, the presence/absence of heavy rainfalls, and the inflow of water via an inlet. Such mechanisms might of course in principle have affected Lake Messel; however, it was obviously not a main driving element for the composition of its fish fauna. Interestingly, the mass mortality layer of Rhenanoperca minuta is situated on a siderite-covered oil shale surface, which may indicate hostile living conditions during the time of precipitation.
From +116 to −100 cm above and below marker bed α, corresponding profile sections were excavated in the grid squares CD 9/10 and EF 8/9, which are horizontally separated from each other only by some tens of metres (Fig. 1). The general distribution patterns are rather similar in both columns and follow the general trends of the profile description (Figs. ESM 16–18). Fish records are comparatively sparse in the upper sections, more frequent and diverse towards the middle ones, and decrease in the lowermost parts. From +116 to +50 cm above marker bed α, fish records only occur at 29, respectively four, microstratigraphic levels (Fig. ESM 42a, b). These numbers increase to 40 and 21 and 49 and 53, respectively, between +49 and −50 cm above and below α (Figs. ESM 43a, b, 44a, b) and drop to 28 and 42 between −51 and −100 cm below α (Fig. ESM 45a, b). The main frequencies as well as the gaps of fish records in both profile sections do not exactly match; however, it should be stated that layers that are comparatively rich in fishes in one of them are not deficient in the other. Also, the distribution patterns of the involved species are very close to the one described for the general profile. Bowfins (Cyclurus kehreri) are the predominant species everywhere, followed by gars (Atractosteus messelensis) in the uppermost part, which are gradually replaced by Thaumaturus intermedius towards the middle part of the profile section, in which the latter species clearly surpasses the gars in frequency. This relationship switches again in the sections below −51 cm below marker bed α.
A comparison of the microstratigraphic fish distribution patterns in corresponding profile sections of the adjacent excavation areas HLMD-2, −2a and −2b indicates that locally limited factors played a less important role than the stratigraphic affiliation of the exploited profile sections. Schaal and Möller (1991: 142) also noted that the fossil distribution patterns in the SMF excavation areas around reference layer α, corresponded rather well with each other. Nevertheless, as is shown by some deviations of the fossil content of excavation areas that are located at some greater distance from each other, a larger spatial distance sometimes also played a role.
Co-occurrence of fish species
Evidence of the contemporaneous occurrence of different species of Messel fishes (Masillosteus kelleri and Anguilla ignota not are taken into consideration as they are extremely rare). Reading of table lines starts left. Information concerning stratigraphic units (average thickness about 1 cm) is from the “turtle hill” area (grid squares HI 7, main reference layer γ; records 1996–2004) and “hydraulic shoring” excavation project (grid squares HI 13/14, reference layer M; records 2000–2001). :same bedding plane; : same microstratigraphic unit; ☺: predator of; ☹: prey of; ?? probably a predator–prey relationship, but not completely clear
Obviously, and disregarding the extremely rare Masillosteus kelleri and Anguilla ignota, all Messel fish species may in principle occur in the same oil shale layers; at least they are represented in the same microstratigraphic/stratigraphic units. Some uncertainty still remains with respect to Amphiperca multiformis and Thaumaturus intermedius, but even these species can be assumed to have partially been co-existing, as there was probably an occasional predator–prey relationship. Even when assuming that some of these co-occurrences of fish species may result from occasional colluviations, this would mean that they were principally able to co-exist during identical or nearly identical periods of time. This may indicate that they also had a certain basic overlap of their preferred environmental conditions and lifestyles. Nevertheless, and as is evident from their general as well as their microstratigraphic distribution patterns, it is obvious that some species did not occur with the same frequency in different excavation areas and during different periods of time. Modifications in the fossil contents of the oil shale, which are dependent on the excavation area, and their topographic positions, respectively, were already noted by Franzen (1979: 31) and Franzen et al. (1982: 31, 32). These authors furthermore noted differences in the relative frequencies of certain fish taxa along the profiles of one and the same excavation area. Both were mainly ascribed to different degrees of predominance of a NE and a NW tributary over the course of time (see below). For the main cause, however, they did not suppose it to be purely temporal differences, but rather assumed that the different interactions of the inflows and outflows were also correlated with biotic changes that ultimately led to the different distribution patterns within the populations (“habitat-related differences”). The disparity between an obviously not too much different general ecological/physiological tolerance spectrum of the Messel fish species and the partially substantial modifications in their horizontal and vertical distribution patterns are, once again, not easily explained. Looking at a medium time scale (e.g. comparing the fish assemblages of SMF 18 with SMF 22, which differ from each other by about 25,000 years), time-dependent modifications in the predominance of particular taxa over others are clearly recognisable over shorter time spans. Species that do not co-occur in the older sections are associated with each other in younger ones (Franzen et al. 1982: 31, 32). Looking at a large time scale, e.g. when comparing the older MP 11 Lake Messel fish assemblage with the younger MP 13 one of Lake Eckfeld maar, it shows that such modification models can also be discarded. The basic types of fishes that were present in Messel are still represented in Eckfeld, and no extinctions in the intervening period of time are apparent. Modifications of the fish fauna are instead more likely to be due to biotic rather than temporal factors. The variability of habitats could not have been very large in a rather narrow, isolated and geomorphologically stable structure such as the Messel-Maar is supposed to be. It would immediately have become much more diverse through the interaction with external water bodies, as has already been postulated in the preceding sections. Then, sensitive driving and controlling systems could have established themselves that would have been able to respond (or even initiate) temporary and spatially tightly circumscribed changes, including the water quality in the lake.
Plants and arthropods
Marker bed M, grid squares HI 13/1 (excavation area HLMD 1)
In these layers, records are rather sparse in the uppermost profile sections (+143 to +99 cm above marker bed M; Figs. ESM 2b–3b). Twelve microstratigraphic levels are occupied by plant and arthropod remains, and seven of these are shared with fish. As in the case of the latter, there are large gaps in their vertical distribution patterns that become narrower towards the lower parts of this profile section. There are a maximum of five records at +100 cm. The arthropods dominate the uppermost parts (+140 to 108 cm) of the comparative column, whereas plants become more frequent in the lower sections (+107 to 99 cm).
From +99 to +25 cm above M (Figs. ESM 4b–6b), the vertical distribution patterns of plants and arthropods show densities similar to the fish records. There are 93 versus 101 occupied levels, and 65 of these are shared by both. The gaps in the vertical distribution pattern of the plants and arthropod remains are somewhat wider than those of the fishes, especially in the lower profile sections (e.g. from +46 to +37 cm and between +36 and +33 cm). Plants and arthropods are not rare at some levels of the uppermost part of the profile (10 records at +97, +93 and +85 cm), but the maximum (up to 28 records) is reached in the middle (+70, +57, +56 cm), where respective records are clearly more frequent than those of fish. Some of the gaps in the fish distribution pattern are covered by plants and arthropod remains. In return, the large gaps in the lowermost part of the distribution pattern of the latter are covered by fish. Plants and arthropods are similarly abundant throughout the profile, with exception of the section between +46 and +36 cm that is only occupied by plants.
Plants and arthropod remains become very sparse in the profile section between +25 and 0 cm above M (Fig. ESM 7b), a trait that is similar to the one found in the fishes. Here, there are some wide gaps in the vertical distribution pattern that are covered by the fishes only in small parts. Similar to the fishes, there are low maximum frequencies that do not exceed two records per level. Plant remains generally dominate over those of arthropods. They are somewhat more concentrated in the uppermost part of this respective profile section.
As in the fishes, all records become distinctly more frequent in the subsequent profile section between −1 and −51 cm below marker bed M (Fig. ESM 8b). There are 23 levels with plant and arthropod remains (versus 24 with fishes). Nine of these are shared. There are still some larger gaps in the respective vertical distribution patterns (e.g. from −34 to −48 cm). Plant and arthropod records are more abundant in the upper part of this profile section. They generally cover the gaps in the fish distribution pattern quite well this time.
The profile section from −52 to −101 cm below marker bed M (Fig. ESM 9b) is characterised by a significant poverty in plant and arthropod remains, which is greatly contrasted (9 vs. 46 occupied layers) by the numerous fish records. Six layers are shared with the latter.
From −101 to −253 cm below marker bed M (Fig. ESM 10b–12b), the plants and arthropods remain rather poor throughout the profile section. A total of 32 levels are occupied by respective records, and, of these, 18 are shared with fish. As for their generally low frequency, plant and arthropod records do not fill the gaps of the fish distribution pattern to the same extent as before. In contrast, especially in the lowermost part of this particular profile section (−240 to −253 cm), their mean frequencies are noted at the same levels as those of the fishes, and they furthermore exactly share one wide gap of the fish distribution (−218 to −240 cm).
Plant and arthropod records remain very poor in the profile sections from −253 to −401 cm below marker bed M (Figs. ESM 13b–15b), especially from −253 to −301 cm, where they are restricted to just two microstratigraphic levels. They become somewhat more abundant in the deeper sections, but are still comparatively sparse here. In total, their occurrence is restricted to only 27 levels (of which 23 are shared with fishes). Arthropods are rarer than plants from −253 to −351 cm. They become a little more frequent from −351 to −401 cm below M, but still only reach half of the latter’s frequency here.
Marker bed γ, grid squares HI 7 (excavation area HLMD-3)12
In the profile section from +118 cm above to −25 cm below this marker bed (Figs. ESM 22b–26b), 106 microstratigraphic levels were occupied by plants and arthropod remains. Of these, 68 were shared with fish. The maximum density of records per level is seven (+35 cm above γ), equalling that of fish, and this number is reached twice (+43 and +45 cm above γ). There are distinctly more plants than arthropods. Sometimes, they are comparatively abundant in sections that also yielded more fishes (e.g. at +50, +23 and −20 cm), but sometimes they are not (e.g. at +80, +45 and −22 cm).
Plant and arthropod remains become continuously more abundant in the middle profile sections from −26 to −100 cm below marker bed γ (Figs. ESM 27b–28b). This section is furthermore characterised by a significant increase in arthropods and their remains. There are 91 occupied levels, of which 58 are shared with fish. The maximum density of plant and arthropod remains (11 records) is reached at −92 cm below marker bed γ. Once again, the maximum frequencies of fishes and plants and arthropods are not at the same levels. Sometimes, there are layers in which both are comparatively abundant (at −92 cm below marker bed γ), but sometimes they are not (at −30 and −42 cm below marker bed γ).
From −101 to −150 cm below marker bed γ (Fig. ESM 29b), the plant and arthropod remains are rarer than above. They occur at 29 levels, 15 of which are shared with fish. There are also some more and larger gaps in their distribution patterns, almost equalling those of the fish. Plants and arthropods occur together throughout the respective profile section and reach comparatively high average frequency values and an absolute maximum abundance (in total, 24 records with 12 of each group) at −150 cm below marker bed γ. This level is comparatively poor in fishes, which are here represented by only one Rhenaoperca minuta.
In the profile section from −151 to −200 cm below marker bed γ (Fig. ESM 30b), there are 23 levels with plant and arthropod remains, which means their records are a little less abundant than in the preceding profile section; 16 of the levels are shared with fish. Once again, fish and plant/arthropod maxima are at different levels and even in strong contrast to one another.
From −201 to −250 cm below marker bed γ (Fig. ESM 31b), there are 33 levels with plant and arthropod remains. Of these, 26 are shared with fish. There are distinctly larger gaps (e.g. −219 to −224 cm, below −243 cm) in the vertical distribution pattern, and the plants clearly dominate the arthropods (47 vs. 23 records). The maximum frequency of plants (6 documented records) is reached at −205 cm, a level that is also comparatively rich in fish (Cyclurus kehreri).
From −250 to −301 cm below marker bed γ (Fig. ESM 32b), the vertical density distribution pattern of the plant and arthropods becomes poorer as do their relative frequencies, and the plants continue to dominate the arthropods. Only 16 fossil-bearing levels are found, and 10 of them are shared with fish. There are some larger gaps in the distribution pattern, e.g. from −255 to −260 cm, from −262 to −270 cm, and from −277 to −288 cm. Only one gap in the fish distribution pattern is covered by plant and arthropod records and vice versa.
In the last investigated profile section from −302 to −338 cm below marker bed γ (Fig. ESM 33b), only four records were made and these exclusively involved plants. Only one level is shared with fish.
Plant and arthropod remains were primarily plotted against the fish records to derive some kind of indicator for the reliability of the microstratigraphic distribution patterns of the latter. Because gaps in the fish records were not rarely “filled” with insect and arthropod remains, these can scarcely be artefacts from an unfavourable consistency of the oil shale, which would then have been more massive and not splitting as easily as it did in the respective profile sections. In addition, the distribution patterns of fishes and arthropods do not indicate a predator–prey relationship either. The general traits of both groups are sometimes rather similar to each other rather than in a negative correlation. There are sections with comparatively abundant fish records that also yielded comparatively numerous plant and arthropod records and vice versa. On the other hand, there are also sections in which the relative frequencies of both groups diverge sharply. The situation is similar with the plant and arthropod remains. There are profile sections with a strong negative correlation between these two groups (e.g. from −151 to −200 cm below γ), but there are also others (e.g. from −51 to −100 cm below γ) in which this is not the case. Interestingly, in sections in which a dominance of one of these groups was found, it was mostly the plants and not the arthropods. Such a (at least) partly negative correlation is somewhat difficult to explain. It would be expected that both groups were rather common in the flourishing rainforest that surrounded Lake Messel during the Eocene. Differences in their relative commonness, especially in excavation areas that are at different distances from the more central parts of the pit, respectively the ancient lake’s basin, therefore probably result from freight selection (see Lutz 1990, 1991a). And with regard to other profile sections in which the plants and arthropods are represented at almost equal frequencies, the latter should then have occurred at different densities in different places and at different periods of time. This is more likely to result from short-time and/or small-scale modifications (e.g. to flow conditions) rather than to the existence of stable long-term factors.
Relative abundance of fossils
During the pilot excavation project, a well-defined volume of 82.5 m3 (5.5 × 3 × 5 m) of oil shale was carefully investigated and its fossil content was documented very thoroughly. This is an excellent base for working out estimates of the relative abundances of different types of fossils per volume unit. According to this study, the following numbers can be expected per cubic metre of excavated oil shale: 4.5 more or less complete fishes, 0.354 tetrapods (0.12 complete amphibians, 0.024 reptiles/reptile remains, 0.15 birds/bird remains, 0.06 bats), and 0.012 other mammals/mammal remains.
Various values have been suggested for the relative abundances of the different types of fossils per m3 of oil shale in the literature. Goth (1990) suggested a frequency of one fish, while Franzen (1979: table 1) provided more detailed information. For the excavation area SMF 5 (grid square E 15, ca. +280 to +320 above marker bed α), his estimates per m3 are 3.23 records of bowfins (Cyclurus kehreri), 0.81 records of gars (Atractosteus messelensis), 0.11 records of the percoid Amphiperca multiformis, 0.01 records of the percoid Palaeoperca proxima, and 3.10 records of the primitive bony fish Thaumaturus intermedius. This totals 7.26 fish per m3, which is a distinctly higher estimate than the calculations from the HLMD pilot excavation project referred to before. For the excavation area SMF 8 (grid square F 9, ca. 0 to +300 cm above α), estimates per m3 are 2.6 records of bowfins (Cyclurus kehreri), 1.1 records of gars (Atractosteus messelensis), 0.06 records of the percoid Amphiperca multiformis, no records of the percoid Palaeoperca proxima, and 0.7 records of the primitive bony fish Thaumaturus intermedius. This translates into 4.46 fish per m3, which is also much more in agreement with the range of the HLMD data. For the tetrapods (complete skeletons plus fragmentary remains), Franzen (1979: table 1) calculated 0.24 records per m3 for the excavation site SMF 5 versus 0.12 for SMF 8 (only complete skeletons in the latter), which is less than in the HLMD excavation site. In the same paper, Franzen (1979: 121–124) furthermore stated that the relative frequencies of different taxa could vary distinctly in different layers and oil shale sections, respectively.13 This observation fits in well with the HLMD analyses of the microstratigraphic fossil distribution patterns as presented before. The differences in the relative abundances of fishes can clearly be referred to the fact that the HLMD and SMF values are derived from records in different excavation areas (grid squares HI 13/14 vs. E 15), which also represent different periods of time (see Discussion under “General fossil distribution patterns").
Orientation of fish carcasses
Marker bed M grid squares HI 13/14 (excavation area HLMD-1)
The general windrose diagram of fish alignments from +142 to −400 cm from marker bed M is characterised by an orientation pattern that shows a slight predominance of NE and SW directions (Fig. ESM 46a). In greater detail, this general pattern is clearly subdivided into a predominating E (NE)/W (SW) flow line between −5.5 to −199 cm below M, and a N (NE)/S (SE) component from −216 to −400 cm below M (Fig. ESM 46d, e). The fish orientation patterns are not quite so strict (transition from mainly NNE and ESE directions to NE, SW and SSO) above M (Fig. ESM 46b, c).
Marker bed α, grid squares CD 9/10, EF 8/9 (excavation areas HLMD-2a and 2b)
From +97 to −157 cm below this marker bed, the general orientation pattern is less regular. More fishes are aligned towards the NE, NW and SE than towards the SW (Fig. ESM 47a). Looking at the details, it is obvious that this trait mainly results from the fish records above marker bed α that almost exclusively adhere to this general orientation pattern (Fig. ESM 47b, c). Those below α gradually show a more irregular alignment, until almost all directions are more or less equally represented in the lowermost parts (Fig. ESM 47d, e).
Marker bed β, grid square HI 7 (excavation area HLMD-4a, b)
As to the excavated sections between +210 and −132 cm from this marker bed, the general distribution pattern is characterised by a predominance of NW and NE alignments over the southern ones (Fig. ESM 48a). Looking at the details, a slight preference of the NE and NW directions is also to be noticed in the uppermost part of the excavated profile (from +210 cm to +131 cm above β; Fig. ESM 48b). Rather clear modifications are then found in the subsequent sections (+46 to −132 cm from β). At first, there is a shift from the predominating northern elements to SW alignments (Fig. ESM 48b, c). Later on (from −5 to −65 cm below β), these are once more replaced by more fishes in a northern orientation (Fig. ESM 48d), and these are finally replaced by an increase of flow lines with SSW, S and E orientations (Fig. ESM 48e).
Marker bed γ, grid square HI 7 (“turtle hill” excavation area HLMD-3)
From +117 to −337 cm from this marker bed, the general windrose diagram is quite uniform, with a predominance of records that are orientated into NE, E, SE, S and SSW directions (50g–250g; Fig. ESM 49a). SE and SSW alignments dominate from +117 to +60 cm (Fig. ESM 49b) and are complemented by some NW and ENE elements between +59 and +1 cm above γ (Fig. ESM 49c). The pattern turns rather irregular from −0.5 to −168 cm below γ (Fig. ESM 49d), with a slight predominance of alignments in NE and SW directions. This changes substantially from +169 to −337 cm from γ. Here, the vast majority of fishes are aligned in NW, NE, and ESE directions (maximum 40g–50g) and only a few towards the SW sector (Fig. ESM 49d).
As with the microstratigraphic distributions mentioned earlier, two fish orientation patterns were compared in corresponding profile sections (from +97 to −110 cm from α) in adjacent excavation areas, which are horizontally separated from each other only by some tens of metres (HLMD-2a, grid squares CD 9/10 and HLMD-2b, grid squares and EF 8/9; see Fig. 1 for the locations of these excavation areas). Both differ markedly from each other in their overall distribution patterns (Fig. ESM 50a, b) as well as in their details (Fig. ESM 50c–j). The fish orientation patterns in different profile sections of the same excavation area were also investigated by Franzen (1978: 46–48; 1979: 90–100). In his 1978 paper, Franzen came to the conclusion that flow lines were more or less constant (mainly in NW/SE orientation) for most parts of the investigated profile (totalling 448 cm), with the exception of the uppermost 107 cm where a slight (northern) shift was to be noted. In the 1979 study, a profile column with a total depth of 280 cm was divided into 13 subunits that were compared with regard to their fish orientation patterns. Once again, it was concluded that the main flow line does not significantly change for most of the investigated profile section (flow lines mainly in NW/SE orientation) except for some 45 cm in the deepest part, in which a shift into NE/SW directions was observed.
First of all, it must be stated in general that—mainly for statistical reasons (see “Materials and methods”)—the analyses of fish alignment patterns are based on records that represent comparative long periods of time. The average time span represented by them is about 6,600 years, with a minimum of ca. 2,600 years (Fig. ESM 48c) and a maximum of ca. 13,820 years (Fig. ESM 46d). This might diminish their meaningfulness, especially when considering the fact that these alignment patterns may sometimes vary considerably over the course of time. Furthermore, it must be mentioned that the fish orientations around marker bed β are based on fewer records (111) than the other windrose diagrams (average 464 records). This might also somewhat diminish the comparative significance of these data.
As for the details, it is difficult to compare the HLMD-1 fish distribution patterns with those from SMF, as published by Franzen (1978, 1979) and Franzen et al. (1982). The latter are at some distance from the first and furthermore exploited somewhat different oil shale sections. The best approach may be SMF 13, which is in rather close proximity (grid squares LM 13/14) even though in much deeper profile sections (ca. −1,665 to −1,780 cm below M). Although details of only a few fossils with alignment information have been published until now (Franzen et al. 1982: 24), the general orientation patterns of both excavation areas agree insofar as they show a predominance of NNE/SSW flow lines, even though these are much more clearly expressed in the records of the SMF excavation site. SMF-15 is also located close to the centre of the pit, but in a more western position (grid square F 13). Although it exploited oil shale sections (ca. −570 to −690 cm from M) that are closer to the ones of HLMD-1, it shows a slightly different windrose pattern, i.e. a clear predominance of a NE/SW flow line (Franzen et al. 1982: 25). This again corresponds to the HLMD data at least insofar as distinct modifications are noted in different profile sections, some of which (especially those above M) also show a slight predominance of NE/SW directions. The modifications of the mean orientation axes of fish carcasses in the more central parts of Messel Pit were mainly referred to an overlay caused by the influence of the NW inlet/SE outlet and the Coriolis force in the quiet reaches of the lake’s basin by Franzen (1979) and Franzen et al. (1982). This is supported by the HLMD data insofar as they likewise show a mix of similar alignments, but at different periods of time. This would mean that the influences of the supposed inlet and the Coriolis force should have also varied over the course of time.
The combined general fish orientation patterns of the excavation areas HLMD-2a and 2b correspond with the more or less adjacent ones of SMF 6 (grid square D 9, ca. +330 to +75 cm above α), SMF 9 (grid square E 10, ca. +90 to −60 cm from α) and SMF 3 (grid square D 10; ca. −50 to −150 cm below α) in that they all roughly show flow line preferences in NW and SE directions (see Franzen 1978, figs. 9, 10, 1979, figs. 6, 8; Franzen et al. 1982, figs. 10, 12, 17). Nevertheless, altogether, they differ substantially from the alignments in SMF 8 (Franzen 1979: 18). This excavation area is in an almost identical topographic position (F 9) as SMF 3. It exploited oil shale sections above α, which are at least partly covered by one of the preceding excavation sites (SMF 6), but clearly show a predominantly NE/SW flow line.
Concerning the layers around marker bed β, it is once more difficult to compare these HLMD data with published information. The only option is to recalculate their stratigraphic position relative to γ. Then, the HLMD-4a, b records represent a range from +410 to −130 cm below γ, which partly matches the range of SMF-2 (ca. +300 to −100 cm from γ), which is located in almost the same grid square (Franzen 1978: 28, 29; Franzen 1979: 11; Franzen et al. 1982:18). Doing so, and even when limiting the comparison to the HLMD data above γ, different traits of the fish alignments become apparent. In the HLMD records, a slight preference of NW and NE orientations is noted as opposed to a NE/SW flow line in the SMF data.
The HLMD-3 windrose graph is also best compared with the one of the excavation area SMF 2 in the northern parts of the pit (grid square H 7, ca. +300 to −100 cm from γ, and +110 to −300 cm from β, respectively). It differs from the general HLMD one by not showing the main alignments along the NE/SW axis just mentioned, but a predominance in the ENE, SE and SSW sectors. The HLMD-3 alignments of records above γ are likewise rather different from those of SMF 2. A better agreement exists with the records below γ, as they follow the general NE and SW alignment of SMF 2 more, especially from −0.5 to −168 cm below γ. This is a little surprising because SMF 2 mainly yielded records above γ, whereas the majority of fishes recorded with their individual alignments stem from below γ in HLMD-3. Expectations are fulfilled better when HLMD-3 is compared with SMF 1. The latter is in a more northern position (grid square H 5), however, and exclusively exploited oil shale sections below γ. Here, a slight predominance of NW and SW alignments is to be found (Franzen 1979, fig. 2), which is closer to the windrose patterns of the HLMD-3 records below γ.
In Messel research, the question of the presence or absence of in- and/or outlets has for decades been leading to rather heated debates. Franzen (1978: 40–48, 1979: 19–23) as well as Franzen et al. (1982: 17–32, 39) (among others) referred to fish orientation patterns14 and micro-dislocation marks and so reconstructed two different tributaries (an older one from the NE and a younger one from the NW), which should have gradually replaced each other in their influence on the flow conditions towards a supposed southern outlet in the lake over the course of time. They substantiated their hypothesis with comparisons of the relative abundances of leaves, fruit and insects in microstratigraphically identical sections of excavation areas that are at different distances from these supposed inlets and, later on (Franzen 2007: 70–72), also with the stratigraphic and horizontal distribution patterns of the primeval Messel horses. Referring to the insect thanatocoenosis, Lutz (1990, 1991a) also accepted the existence of an NW tributary and a NW/SE-orientated flow line. He furthermore referred some Messel invertebrate remains (e.g. certain water insects, caddisfly larvae and freshwater shrimps) to creek fauna elements, which must therefore have also invaded Lake Messel from the outside. Links to external water bodies were furthermore suggested by some other fossils (see “General composition of the Messel ichthyofauna”). Lastly, certain shell fillings of some viviparous snails are also indicative of a small stream by which they may have been flushed into the lake (Neubert 1999: 169). In addition, a temporary connection to an external fluvial system was also proven for the Lake Eckfeld maar, which is roughly contemporary (MP 14 vs. MP 11) with the Messel one (Lutz and Kaulfuss 2006).
In contrast, Lake Messel was interpreted as having been a more or less isolated maar structure without significant inflows (but a steady outlet) by Rietschel (1988). The same scenario, even with more emphasis on its permanent isolation, was derived by Goth (1990: 85), who definitely excluded the possibility of any connection between the lake and external water bodies and emphasised absolutely lotic water conditions (except occasional turbidity flows) below the chemocline. He ascribed most fossil orientation patterns to parallel alignment effects, which would have taken place parallel to the shoreline of the lake basin and occurred when the carcasses sank to the bottom.
The fish orientation patterns change distinctly over certain periods of time, as they show marked differences between subsequent profile sections in one and the same excavation area.
The fish orientation patterns do not change during other periods of time, as there are also other profile sections of the same excavation area in which no significant modifications are notable.
The fish orientation patterns are also sometimes due to spatial modifications, as they can differ from each other in identical or very similar profile sections of excavation areas in adjacent grid squares.15
In contrast, they are sometimes not due to such spatial modifications, as they can be very similar in corresponding profile sections of excavation areas in adjacent grid squares.
The fish orientation patterns can distinctly differ from each other in corresponding profile sections of excavation areas that are in the same grid squares.
They can also be rather similar in corresponding profile sections of such excavation areas. The fish orientation patterns can differ distinctly between corresponding profile sections of excavation areas that are in the same grid squares.
This might altogether fit a highly sensitive and spatially and temporally very flexible driving and balancing system than it should, for example, be expected for stable long-term tributaries and a locally fixed, steady outlet in a geomorphologically constant structure like a maar lake. Occasional high water periods of variable intensity, in combination with a partial erosion of the isolating tephra wall (see above)16 that may have allowed an exchange with external water bodies during certain periods of time, appear to be a better explanation for the Messel system. Such interactions must not necessarily always have affected the “input” and resulting composition of the lake’s ichthyofauna in the same manner and to the same extent (see Micklich 2002a; Micklich and Klappert 2004).
In almost all Messel fish species, there are specimens with direct structural evidence for the recovery from a wide range of bad environmental conditions, external injuries and/or disease. Sometimes, there are healed fin ray breakages; other individuals have scales that were probably infested with ectoparasites or were altered by the marginal resorption of materials that are usually correlated with avitaminoses (Micklich 1985: 127, 128, 1992: 90).
The extent of scale regeneration can differ considerably from one to the next individual. In Cyclurus kehreri, for example, there is a rather high number of regenerated scales in specimen HLMD-Me 9097 (up to 90 %), but distinctly fewer in others (59.4 % in HLMD-Me 9741). The same applies to Messel percoids: Palaeoperca proxima HLMD-Me 10309a, b and Amphiperca multiformis HLMD-Me 16931 are examples for highly intensive scale regeneration (94–95 and >94 %, respectively), whilst HLMD-Me 15841 and HLMD-Me 8971 represent the opposite end of the range of values (38.75 and 52 % respectively). There are also individual differences in the number of cases of scale regeneration in different body regions and in the relative sizes of regenerated areas of different scales, which means the period of time during which the regeneration process took place. Sometimes, comparatively small individuals (e.g. Amphiperca multiformis HLMD-Me 15608 with 8.9 cm SL) must already have been affected by two independent regeneration events, whereas in some larger ones (HLMD-Me 8970, 10.1 cm SL), most regeneration obviously took place at a later stage in life. The extant comparative materials provides clear evidence that some individuals became victims of different events requiring subsequent regeneration in the very first year of their life. Dicentrarchus labrax HLMD-SMFR 279 obviously survived three of those events even before the formation of the first annulus, whereas in HLMD-SMFR 257, only one took place during its first year and a second followed later on.
Scale regeneration depending on the relative size of the scales in extant percoids
Number of measured scales (Nsc)
Average maximum scale size (% SL)
Total number of investigated scales (Ntotal)
Relative number of regenerated scales (Ntotal)
Looking at the extent of scale regeneration as a function of the body region (Figs. ESM 51–56a, b), it is evident that, in Cyclurus kehreri as well as in Amia calva, there are higher relative numbers of regenerated scales in the posterior regions than in the more anterior ones. Similarly, but excepting the caudal stem in Amia calva, there always appears to be a slightly greater extent of scale regeneration in the ventral regions of the body than in the dorsal ones. In contrast, scale regeneration is less intensive in the posterior sections of the body flanks in Thaumaturus intermedius, although more regeneration is noted in the ventral parts of the body flank below the vertebral column. The extant comparative species Umbra pygmaea follows the general trends observed in the bowfins, except that there is less scale regeneration in the ventral predorsal part of the body flank than in the dorsal one. In Amphiperca multiformis, the general distribution pattern of the scale regeneration also follows the trends observed in the bowfins. The extent of scale regeneration in the more posterior regions of the body flank exceeds that of the anterior ones. However, except for the caudal stem, the extent of scale regeneration of the dorsal parts of the body flank exceeds that of the ventral ones. The situation is more heterogeneous in the respective extant comparative forms. There is more regeneration in the anterior parts of the body flank than in the posterior ones in Ambloplites rupestris, Macquaria and Micropterus, with the exception that the caudal stem shows a higher degree of regeneration than the area beneath the D2 in the investigated specimens of the two latter genera. In Serranus, the extent of regeneration in the posterior portion exceeds that of the anterior part of the body flank, as was observed in Amphiperca multiformis. In all reference taxa, there is generally more regeneration in the dorsal portions of the body flank than in the ventral ones, with the exception of Macquaria in which the extent of regeneration in the dorsal part of the body flank surpasses that in the ventral part of the caudal stem and also in the area beneath the D2. In Palaeoperca proxima, the extent of scale regeneration in the cheek area is greater than that in the predorsal body flank and beneath the D1, but the latter is surpassed in extent by the area below the D2 and the caudal stem. The same applies to the extant reference specimens of Morone, while the cheek region is less extensively regenerated than the predorsal area in Dicentrarchus, in which the extent of regeneration then increases posteriorly towards the caudal region. In Rhenanoperca minuta, the extent of regeneration below the D1 is greater than below the D2, but the latter is even less affected thus than the caudal stem. In Lepomis gibbosus, there is almost the same extent of regeneration beneath the D1 and the D2, which both exceed that of the caudal stem. Both species show more regeneration in the dorsal parts of the body flank than in the ventral ones.
With the extent of scale regeneration being a function of the SL, in Cyclurus kehreri, greater extents of scale regeneration are noted in the larger size classes of above 25 cm SL, with a maximum between 25 and 29.9 cm (Fig. ESM 57). This is almost the same in Amia calva, in which the maximum value is found in the largest size class (55–59.9 cm SL). The percoids, however, obviously do not follow this tendency (Fig. ESM 58). In Amphiperca multiformis, the maximum extent is observed in comparatively small individuals (3–4 cm SL), and the lowest values are found in the largest ones (11–14 cm SL). In Palaeoperca proxima, the greatest extent of regeneration was once more found in the largest size class (23–24 cm SL), but there is also one small individual with extensive scale regeneration (HLMD-Me 13053, 9–10 cm SL), and lower values are also noted in some medium size classes. Rhenanoperca minuta also does not follow the general trend of the bowfins. Here, the maximum extent of regeneration took place in the size class between 4 and 5 cm SL, whereas it was distinctly less extensive in the largest individuals examined (7–8 cm SL, e.g. HLMD-Me 15348). Of the extant comparative taxa, the maximum extents of regeneration were noted in the largest specimens of Lepomis and Serranus. In contrast, the largest specimens of Dicentrachus had the minimum extent of regeneration.
Looking at the amount of scale regeneration in different profile sections (Fig. ESM 59), the specimens of Cyclurus kehreri and Amphiperca multiformis found around marker bed M surpass those from γ. The lowest values are noted in specimens that originate from the layers around marker bed α. Palaeoperca proxima was not reported from α in the time span considered in this paper. Here, the largest extent of regeneration is noted around marker bed γ. Rhenanoperca minuta is only known from these latter layers.
Scale regeneration is actually a widespread feature in many extant fish species. Of course, the total extent of regenerated scales always depends on the specific life history of a given fish and can therefore differ considerably from one to another individual of the same species, population, and/or size-class. In addition, there are differences in the intensity of the anchoring of the scales in the integument in different species, with some losing their scales more easily than others. Scale regeneration takes place exclusively after superficial injury, which sometimes leads to a complete loss of the squamation in the affected area (Quilhac and Sire 1999). Therefore, also in fossils, scale loss should mean that the respective body region must have been “naked” for a certain period of time. If these areas are large, this probably implies serious osmotic problems, which should have resulted in death rather than survival and regeneration.
Nevertheless, there are still mechanisms that ensure survival. The first one is that, in some extant fishes, scale regeneration can take place rather rapidly (Neave 1940; Quilhac and Sire 1999). In addition, not all scales need to have been lost at the same time. As demonstrated in the preceding section and also in Figs. 13, 15, 16 and 17, extant and Messel species clearly show scales that underwent regeneration events during which they were independently lost and replaced.
Even in view of such good chances of survival, the question remains how fish afflicted thus—especially those form Messel Pit—could have lost so many scales. Looking at extant species once more, scales are mainly lost in “rough” environments, during rival and territorial combats, as well as during predator attacks, including those by scale-eating species. Unfortunately, all of these appear to be inappropriate explanations for Messel fishes. Their scales could scarcely be lost by repeated contact with rocks and gravel grounds, which did not exist in ancient Lake Messel (see below). Rival and territorial combats occur in adult individuals, but the vast majority of Messel fishes consisted of immature juvenile individuals (Micklich 2002a). Predator attacks can sometimes result in a high degree of scale regeneration, especially in body regions that are most exposed to attack (Fig. ESM 60). However, in most cases, attacks by larger predators (e.g. crocodiles) are much more likely to cause the death of the respective victim rather than recovery and regeneration. Last, but not least, none of the Messel fish species looks like being a scale-eater. They all lack the specialisations of the jaw dentition that are typical for the latter (Fryer and Iles 1972: 473–520; Liem 1979; Liem and Stewart 1976; Witte and van Oijen 1990). In addition, Rhenanoperca minuta as well as Thaumaturus intermedius, which are comparatively small-sized and could therefore be principal “suspects” of having such a peculiar feeding habit, were proven to have preferred other types of prey (Micklich 1992: 87–89; Richter and Baszio 2001b).
Looking at the extent of scale regeneration in different body areas of the Messel species and their extant comparative taxa, the general distribution patterns agree very well in Cyclurus kehreri and Amia calva (Fig. ESM 51) as well as in Palaeoperca proxima and Morone (Fig. ESM 55a, b). There is also rather good agreement in Amphiperca multiformis and Serranus, (Fig. ESM 54a, b). These analogies may result from similar lifestyles, which would also mean that the loss of scales principally resulted from the same mechanism—with the difference being that it had a much greater impact on the Messel forms than on recent ones. The bowfins, Amphiperca multiformis and Serranus, furthermore show traits, which can best be explained as a result of mechanical stress. This can be expected to affect the more posterior and ventral parts of the body flank more intensively than the more anterior and dorsal ones. Much scale regeneration is also to be noted in the posterior parts of the body flank of Palaeoperca proxima and Morone. Both taxa furthermore show large-scale scale regeneration in the cheek and predorsal regions. This can also be very well explained as a result of mechanical stress. Extant species, for example, escape very shallow waters by lying on one side, wriggling vehemently, and violently beating their tails. Forms with comparatively wide heads will easily incur superficial injuries in the lateral skull area in the process. Unfortunately, and as was already mentioned before, this type of increased mechanical stress can hardly be expected to have existed in Lake Messel. According to most reconstruction models, it developed from an ancient maar and therefore was deep, and the ground was covered by a comparatively fine-grained bottom mud (e.g. Buness et al. 2005). This means, however, that the mechanism necessitating an increased extent of scale regeneration might not have been in the lake itself, but should be sought somewhere in the surroundings. More recent volcanological and also entomological investigations indicate, for example, that the lake might have been much larger than the present-day oil shale deposits, had coarser-grained marginal faces, and a vegetated shoreline that were not preserved because of erosion (Büchel et al. 2010: 119; Wedmann et al. 2011: 976). Alternatively, the lake might not have been an isolated system (see above). At first glance, the trigger mechanism for scale loss does not appear to have been correlated with general periodical events either, as there is too much variation in the general, regional, and temporal extents of regeneration in different individuals of all Messel species. Nevertheless, it should not be forgotten here that these individuals stem from different excavation areas, which represent different periods of time and probably also different locations within the former habitat. Both possibly have the potential of obscuring such periodical events. Migrations, depending on flood pulses, are known to play an important role in the biology of many extant fish species whose spawning coincides with them. Connections between bayous and floodplains, for example, play an important role for juvenile bowfins. They use floodplain habitats during high water periods to grow and forage, and return to the bayous when the flood recedes (Davis 2003: 12). Foraging forays and/or migration into or through shallow or “rough” areas could also have caused scale loss in Messel fish species.
As to the noted relationship between scale size and the extent of regeneration, this might result from the fact that larger scales also have a larger area that is exposed to superficial injuries and destruction. Similarly, the relationship between the extent of regeneration and the SL also suggests that larger specimens can be expected to show higher numbers of regenerated scales. They are likely to have been exposed to situations that lead to scale loss more frequently and/or for longer periods of time. Extant Amia calva nicely matches this hypothesis and so does Cyclurus kehreri to a somewhat lesser extent (Fig. ESM 57). It is difficult to explain why this is not the case in the percoids in general and the fossil taxa in particular. Once again, this may depend on differences in the individual life histories of the investigated individuals and/or also on the fact that the respective records are derived from different excavation areas representing different periods in the lake’s history and probably different habitats as well (see above).
Interestingly, in the cases of Cyclurus kehreri and Amphiperca multiformis, the lowest extents of scale regeneration are found in those profile sections and excavation areas, respectively, which also supplied comparatively large-sized (sensu average SL) individuals of both species (Fig. ESM 59, Fig. 8). In addition, the respective excavation areas also yielded large-sized gars and a comparatively high number of crocodiles (Diplocynodon darwini). According to Franzen et al. (1982: 83), the largest gars and bowfins of SMF were likewise found in excavation areas and profile sections with similar topographic positions and stratigraphic affiliations. The combination of a comparatively low extent of scale regeneration and large average sizes fundamentally negates the assumption that larger specimens show higher numbers of regenerated scales (see above). In combination with the presence of other large-sized fossils, it could instead suggest a habitat that was more spacious and less rich in obstacles in these areas or during the corresponding periods of time.
Population densities and mortality
Mortality and mortality rates provide interesting sources for obtaining ecologically relevant information on Messel fishes. Various biological key facts of populations of closely allied extant species, for example, are available from the literature and the Internet. To some extent, they can also be used for fossils, especially if the latter were collected in a thorough manner and with well-defined stratigraphic affiliation data. Within the framework of fixed boundary conditions,17 minimum estimates of population densities can be derived from the microstratigraphic distribution patterns of the Messel fish species for certain periods of time. Converted to the lateral extension of today’s oil shale deposit as a proxy for the size of the ancient lake, the population of Cyclurus kehreri could sometimes have consisted of at least 8,000 individuals (slightly corrected data from Micklich 2005).
There are also other options by which Messel fishes can (however indirectly) contribute to discussions concerning mortality. Compared to other fossil sites, the Messel fossil record is characterised by a comparatively high number of small flying animals, like bats and arboreal birds, which cannot be expected to be frequently common in the bottom deposits of an ancient freshwater lake. In addition, there are some other phenomena. Records of presumed mating couples of the soft-shelled turtle Allaeochelys crassesculpta (Harrassowitz, 1922) and those of pregnant small primeval horses [Eurohippus messelensis (Haupt, 1925)], which bear almost identical ontogenetic stages of foetuses, are also not easy to understand and explain (Rietschel 1988; von Koenigswald 1998; Maier 2000). For some period of time, the favourite explanation was death by toxic gases. Animals that approached the shoreline for drinking or flew over the water could have entered these poisonous clouds, lost consciousness, and drowned (Richter and Storch 1980; Franzen and Köster 1994). Some years ago, an alternative hypothesis was published (von Koenigswald et al. 2004, 2005). In addition to seasonal blooms of coccal green algae, (e.g. Tetraedron and, to a minor extent, the “oil algae” Bothryococcus), which provided the main mass of organic matter (kerogene) within the Messel oil shales (Irion 1977; Goth 1990), there should have been seasonal mass proliferations of blue-green algae (cyanophytes). These algae are generally known to produce considerable amounts of microcystine, which is a highly toxic substance. Dissolved in the surface water, it can cause the death of any animal drinking it (Kerr et al. 1987; Fitzgerald and Poppenga 1993; Baker and Humpage 1994). The same could also have happened in ancient Lake Messel.
General considerations concerning the ancient habitat may be conducted by comparisons of the population densities of the Messel fish species with those of closely related extant species. Such comparisons suffer greatly from the fact that most of the respective investigations in extant fish populations are based on transect electrofishing or other methods of locally restricted and momentary investigations, and the resulting data are difficult to convert for use on the Messel fossils (e.g. Killgore et al. 1998; Chick et al. 2004.). Moreover, the composition of extant fish associations is extensively affected by an armada of different environmental factors (e.g. Randall et al. 1996) so that appropriate comparative data for Messel fishes are very sparse for that reason, too. Nevertheless, the minimum population density that was calculated for Messel Cyclurus kehreri is only slightly lower than the data that can be derived from Rotenone poison-fishing in Louisiana backwater lakes and block net captures as well as electrofishing in freshwater marshlands in Florida. These indicate a population of just under 10,000 individuals (Lambou 1959) and 6,000–9,000 individuals (Chick et al. 1999), respectively. Even with regard to the presumably excellent conditions for fossilisation within ancient Lake Messel, it can certainly be excluded that all bowfins that died were also preserved. This means that the bowfin population must have been larger, may be substantially so, and denser, than extant ones. This, however, would no longer be the case if the calculated number of individuals would have represented either a local accumulation of fossils from a larger catchment area, or from a generally more hostile palaeoenvironment with higher mortality rates than in the extant reference examples. On the other hand, and most importantly, it must also be considered that the numbers of the fishes in the extant catches represent momentary snapshots, whilst the number of Messel bowfins was calculated for a period of one year. This means, if the extant numbers were also extrapolated to one year, that this would lead to much higher numbers of individuals and a much more realistic basis for comparisons with the Messel bowfin population.
Weigelt (1927) and Schäfer (1962) already clearly demonstrated that biostratonomical and actuo-palaeontological investigations can play an essential part in the interpretation of fossil assemblages. Wuttke (1983a, b, 1992a, b) and Smith and Wuttke (2012) also stressed their meaningfulness for the Messel taphocoenosis. Unfortunately, aside from the reconstruction of palaeocurrents (Franzen 1978: 30–32), fish carcasses have until now been almost completely disregarded in this respect. Only a few examples that may be useful for more intensive future studies (especially if correlations between stratigraphy and predominant types of preservation can be found) are presented here in brief.
The palaeoecological discussions and conclusions in this manuscript are mainly based on fish fossils. Other groups of fossils (and the geological and sedimentological framework) of course provide palaeoecological information that may agree with or contradict the insights provided by fish. Some information concerning the insect fauna could be indicative of rather steep shorelines in some locales (e.g. Wedmann 2005). On the other hand, a palaeobotanical point of view (Wilde 1989: 129) suggests there should also have been some areas with shallow water as well as some marshy sections in the surroundings. Other fossil groups, like the crocodiles, are much more diversified than the respective extant fauna (e.g. Rossmann and Blume 1999), which might also indicate an active immigration or passive washing-in of individuals or carcasses, respectively, which could have been favoured by the presence of some shallow margins. Analysing the enamel of Propalaeotherium teeth, Tütken (2011: 165) found out that these animals were grazing on Permian sedimentary and rocky remains, which is also indicative of an at least partial erosion of the tephra ring. Looking at the “turtle hill” fish assemblage, it is—aside from some other peculiarities—to be stated that it is associated with subadult crocodiles (Diplocynodon darwini Ludwig, 1877: HLMD-Me 10262), and with comparatively abundant soft-shell and marsh turtles of the species Allaeochelys crassesculpta (Harrassowitz, 1925) and Euroemys kehreri (Staesche, 1928), the latter of which are represented by obviously mature females ready for oviposition (Gaßner et al. 2001). Furthermore, there are baby turtles (e.g. HLMD-Me 13396, HLMD-Me 13555) and, not to forget, the latest records of freshwater shrimps (e.g. HLMD-Me 13230, 13919), which are extremely rare and otherwise only known from younger layers in the centre of the pit (Wolf 1988: 105; Lutz 1990: 116). Other characteristics (comparatively high abundance of partial skeletons of small-sized birds, and the occurrence of layers with comparatively rough arenaceous-siliceous materials) also differ from excavation areas in other parts of Messel Pit. As has already been mentioned before, the “turtle hill” area and oil shale sections were in general presumed in some former publications to have been influenced by a tributary. The details of the fish fauna as well as the associated fossils match this idea insofar as they appear to indicate the presence of a more shallow and swampy area in this part of the ancient lake.
On the other hand, the excavation areas in the NW slope area (grid squares D-F 8–10) are characterised by the occurrence of particular types of arthropods, some of which belong to groups that prefer an oxygen-rich environment and were considered as illuviation from a creek (Lutz 1991b: 119), and the complete area was thought to have been influenced by a north-western inlet (Franzen et al. 1982). The results of the present paper do not directly indicate the presence of such an inlet. Somewhere in this area, there might also have been a “weakness” in or absence of the tephra wall that faciliated an occasional influx of water.
The general characteristics of the Messel fish fauna (poor taxonomic diversification, and species represented by comparatively numerous specimens) and also the composition of the respective populations (predominance of juveniles) strongly suggest that the Messel ichthyofauna was exposed to significant selective pressures. At least occasionally, there must have been a renovation from adjacent water bodies, which probably did not take place via locally constant inlets and/or outlets but via a more flexible system. As a matter of fact, the lake cannot have been a permanently isolated system during those periods of time as is demonstrated by the known fossils. Some other information (minimum population densities, presence/absence and/or diversification of certain other fossil groups19 and great extent of scale regeneration) also appears to indicate an at least occasionally larger catchment area. Local and stratigraphic peculiarities of the vertical and horizontal distribution patterns (presence of “gaps”, and switches between monotypic and heterogeneous fish species assemblages) cannot be referred to physiological or ecological capabilities or preferences of an isolated fish assemblage or with a regeneration/renovation inside such a system. Some other particularities also corroborate temporal and spatial alternations of hostile20 and more favourable environmental conditions. Many of the long- and short-term, and possibly also local, differences in the composition of the fish fauna were probably mainly triggered by changes in the input. Eclipsing by internal effects might have played a secondary role. Nevertheless, it is difficult to find filter mechanisms that might have regulated such an input. Only a few possibilities can so far be excluded.
One scenario is an active–selective immigration by a taxonomically homogeneous fish fauna from the surroundings. Filtering was then achieved by ecological preferences and physiological capabilities. Only individuals of those species immigrated into the lake that were able (or even preferred) to live under its particular environmental conditions during the respective period of time. Such a choice would have existed when areas of shallow water in the lake were (at least temporarily) connected to surrounding water bodies. More general integrations, e.g. during extensive high waters, are not expected to result in clear polarisations between internal and external environmental conditions, and various fish species have been “washed” into the lake more or less randomly.
Another, and probably more favourable, scenario is an active or passive immigration that was selective for other reasons. This time, it was based on alternating links to different types of external water bodies.21 Their distances from Lake Messel could also have played some role. If these different sections/habitats of the surrounding (drainage?) system(s) were inhabited by different fish associations (which, however, always appear to have been dominated by bowfins), they also supplied Lake Messel with a somewhat heterogeneous fish fauna during different periods of time. As has already been stated elsewhere, these contacts need not necessarily have been fixed inlets and/or outlets that dominated the lake’s fish fauna and ecosystem during subsequent periods of time.
A third scenario is a generally passive and non-selective immigration by a taxonomically homogeneous external fish fauna. This could have taken place in “crisis” situations, for example during periods of seriously low water levels. Juveniles of various species might then have “saved” themselves by gathering in deeper structures that remained filled with water. Similarly, a multiple-species assemblage would also result if the maar lake was mainly isolated as a depression and fishes from the surroundings were occasionally washed in down steep shore lines by torrents. Such mechanisms, however, are not suited to explain the particular modifications that the Messel fish fauna obviously underwent in the course of time.
This species was assigned to the genus Properca Sauvage, 1880 by Gaudant (2000). Nevertheless, the type specimens of Properca, which were described and figured by Agassiz (1833–1844), differ from the type and most other specimens of A. multiformis in various features (e.g. general shape of the body, shape of the caudal fin, presence/absence of a clear indentation between the spiny and soft parts of the dorsal fin) and probably represent a rather different type of adaptation. Therefore, the original species name is retained here.
According to a personal communication from S. Schaal (SMF), Lutz (1990: 113) also reported comparatively frequent records of bowfins and gars of less than 10 cm in total (? he does not mention whether he refers to the total or to the standard length) length in the youngest layers around marker bed M in the centre of the pit.
This distinctively differs from extant fish populations as it consists of fishes that accumulated over extended periods of time (“time-averaged samping”).
This behaviour is not restricted to some percoids, by the way, but is also known from other species, especially the extant Amia calva (e.g. Reighard 1900, 1904), which is a very close relative of Messel Cyclurus kehreri.
This species has often been referred to as “small percid”, which is rather incorrect as it is far from being a member of the family Percidae
Theoretically, and even in an isolated system, this could also have been improved by rainfall. However, these must have been very heavy and/or long-lasting to produce this result in a maar like Lake Messel.
Franzen et al. (1982) as well as Schaal and Möller (1991) mentioned (respectively indicated) the occurrence of “percids” from other areas and oil shale sections. Of special interest is the second paper in which these are reported from excavation areas in grid squares E 9 and E 11. Both are close to HLMD 2, 2a, b (where, however, no percoids of this type have been found during the HLMD excavations for many years) and also exploited oil shale sections around marker bed α. Unfortunately, it is not clear whether they really refer to Rhenanoperca minuta. The term “Percidae” is only used in their graphs, whilst “Percoidei” (but with reference to Amphiperca multiformis and Palaeoperca proxima) appears in the text. Nevertheless, and as has already been stated elsewhere in this paper, the only reliably identified individuals of Rhenanoperca minuta (all other small percoids from this excavation areas are referable to Amphiperca multiformis) in this area are from HLMD-1985/15, which is in rather close topographic proximity and also exploited similar oil shale sections as the SMF excavation sites.
With an average sedimentation rate of 0.14 mm per year (Schulz et al. 2002) and a given profile section of 23 cm thick, the authors summarised a considerable period of time (a little less than 1,900 years) during which the fish taxa may already have experienced different density fluctuations.
For more information concerning profile details and fish remains distribution patterns in the reference column, see Belzer (2002).
Both peculiarities may be explained rather easily (see Micklich and Hildebrandt 2010). Periods during which bowfins were comparatively abundant inhabitants of the lake can also be expected to produce comparatively high volumes of materials that are available for the fossil record. Favourable (concerning preservation and/or water quality) conditions will furthermore result in a predominance of complete specimens over the disarticulate ones and isolated remains. They may also diminish the number of carcasses that are available for fossilisation.
Microstratigraphic distribution patterns of plants and arthropods above and below the marker beds α and β and their remains, respectively, were not investigated in the present study.
Schaal and Möller (1991: 133, 142) furthermore noticed that not only fish but also other vertebrates (e.g. bats, birds) could be limited to particular excavation sites and oil shale sections.
which, by the way, represented much longer periods of time than the data in this manuscript.
Obviously, sometimes, even spatial differences of a few metres were sufficient to cause such different alignment patterns. The fish carcasses probably arrived at the lake floor as some kind of skin-ligament hoses, comparable with extant fishes that are prepared with clearing and staining techniques. Such specimens are extremely delicate and will probably respond to even the smallest changes in flow.
Lenz et al. (2007) demonstrated that that the stabilisation of the tephra ring and slopes by deeply rooting trees and shrubs was a comparatively rapid process, which was already completed at the boundary between the Lower and Middle Messel formations. Nevertheless, even then, the period of time between the origin of the maar and this latter period of time was long enough for an at least partial erosion of the tephra wall.
One of them, for example, is the assumption that the number of fossils was not biased by taphonomy and that the number of carcasses of the referred fish species in the sample area is representative for the complete lake during that particular period of time. This theory cannot be completely wrong when looking at relative frequencies, e.g. of bowfins in other excavation areas, and considering that there are no mass mortality layers at all.
even if this is also somewhat questionable for other reasons (see, e.g., dicussions in Reisdorf et al. 2012).
Other studies (Richter and Baszio 2001a, b; Richter and Wedmann 2005) demonstrated, however, that an abundant and diversified fauna of water insects and a comparatively complete food chain might have existed in Lake Messel for longer periods, which to some extent contradicts the existence of longer periods of hostile conditions.
Franzen (1978: 42) has already discussed the option of a direct connection of Lake Messel with other oilshale occurances, e.g. the ones at Eppertshausen and Offenthal. Up to now, however, it is not proven (or even unlikely) that these maar lakes were exactly contemporaneous with the Messel one (Harms et al. 1999).
Sigrid Belzer (Darmstadt), Marie Bobenhausen (Darmstadt), Kerstin Hlawatsch (Halle/Saale), and Shahram Karimi (Rüsselsheim) helped and assisted in various ways und intensities over the long period of time during the preparation of this manuscript. Many thanks are also due to Mario Drobek, Sabine Gwosdek and Eric Milsom of the HLMD excavation team and the countless student interns, without whose continuous willingness, thoroughness and preparedness during fieldwork the comprehensive Messel database that has been referred to here could never have been established. Further thanks go to Michael G. Newbrey (Drumheller, Alberta, Canada) who provided essential help in the shape of comments and suggestions on the growth studies. Stephan Schaal and Sonja Wedmann (both SMF) as well as Achim G. Reisdorf (Basel, Switzerland) contributed useful information. Thomas M. Ulber (www.herprint.com) improved and corrected the English text. Michael Wuttke (Mainz), and Rodrigo Soler-Gijón (Berlin) helped with critically reviewing the text as well as with further comments and discussions. Achim G. Reisdorf (Basel, Switzerland) added useful comments to one of the reviews.