New excavations along the Panama Canal have yielded a growing Early Miocene assemblage of mammals referred to as the Centenario Fauna. Despite the area’s proximity to South America, the mammals of the Centenario Fauna have entirely North American affinities. The Centenario Fauna is distributed throughout a ∼115-m stratigraphic interval encompassing the uppermost Culebra and Cucaracha Formations within the Panama Canal basin. Previously published ages constrain the age of the lower limit of the Centenario Fauna to no younger than ∼19 Ma, but the upper limit has remained problematical. A fresh exposure of the Cucaracha tuff, a prominent marker horizon within our measured sections, has yielded two new radioisotopic determinations: (1) an 40Ar/39Ar age of 18.96 ± 0.90 Ma and (2) a U-Pb zircon age of 18.81 ± 0.30 Ma. In addition, magnetostratigraphic data indicate that the Centenario Fauna occurs within chron C5Er, from 18.78 to 19.05 Ma on the geomagnetic polarity timescale of Gee and Kent. These correlations further confirm the calibration of the latest Arikareean (Ar4) to early Hemingfordian (He1) transition in Nebraska, at the base of chron C5Er, at about 19.05 Ma. The Centenario Fauna occurs at the beginning of the Hemingfordian North American Land Mammal Age, i.e., He1. A broad faunal province existed during the early Hemingfordian that can be recognized across a north-south range of 5000 km throughout North America, with the southernmost limits defined by the Centenario Fauna of Panama.

The Miocene was an important time in mammalian evolution and biogeography in the Americas. It was the time of the Middle Miocene Climatic Optimum, the spread of grasslands, and the beginnings of the Great American Biotic Interchange, to name a few of the important events (Woodburne 2010). Little is known about the fossil record at low latitudes in what are now the New World tropics. If we can use modern ecosystems as a model, this region potentially played a large role in the production and maintenance of biodiversity. With a few exceptions, such as the Middle Miocene La Venta of Colombia (e.g., Kay et al. 1997) and numerous Cenozoic localities in Mexico and Central America (e.g., Ferrusquía-Villafranca 1978; Webb and Perrigo 1984; Tedford et al. 2004; Laurito and Valerio Z. 2010), the dearth of knowledge of the ancient New World tropics mostly results from the lack of suitable exposures and paucity of paleontologists working in this region. Nevertheless, since 2009, the Republic of Panama has undertaken a massive expansion of the Panama Canal, and in so doing it has uncovered fresh exposures of Miocene fossil-bearing deposits, having thus provided a “once-in-a-century” opportunity to advance knowledge of ancient biodiversity in the New World tropics. During this same interval of time, crews led jointly by the Florida Museum of Natural History and the Smithsonian Tropical Research Institute have worked along the Canal, logging more than 2500 person-days in the field, to recover fossil vertebrates from the Miocene exposures.

One of the most productive zones for the discovery of terrestrial vertebrate fossils comes from a 115-m stratigraphic interval that includes the uppermost part of the Culebra Formation and the overlying Cucaracha Formation. The fossils from this interval have been referred to as the Centenario Fauna (MacFadden et al. 2010), which takes its name from the Centenario Bridge that spans the Gaillard (Culebra) Cut along the southern reaches of the Panama Canal north of the Pedro Miguel locks (fig. 1). As a result of the field activities, including surface prospecting and screenwashing for microfauna, we are continuing to add to the diversity of extinct mammals from the Centenario Fauna. For example, in the original description of the fossil mammals from the Cucaracha Formation, Whitmore and Stewart (1965) described five species from four families; we now have at least two dozen species from 15 families from the Centenario Fauna. The paleobiogeography of the mammals from the Centenario indicates entirely North American affinities and thus a significant expansion of the North American Land Mammal Age (NALMA) biochronology into the low latitudes of the ancient North American tropics. Previous attempts to resolve the actual age of the Centenario Fauna were hampered by relatively poorly preserved material, differing taxonomic and biochronological interpretations of the more common taxa, and a lack of suitable, unweathered samples of volcanic units within the stratigraphic sequence that produces the Centenario Fauna. New fossil discoveries over the past several years resulting from additional fieldwork and screenwashing activities, coupled with focused studies of relevant museum collections, have refined and revised our taxonomic interpretations and increased the alpha-level mammalian biodiversity.

In addition, here we report independent calibration of the Centenario Fauna, using new magnetostratigraphic data and radioisotopic age determinations that use both 40Ar/39Ar and U/Pb methods from a fresh exposure of a welded tuff interbedded within the Cucaracha Formation. Our results have ramifications not just for a relatively precise age constraint for the Centenario Fauna but also for an understanding of the calibration and transition between the Arikareean (Ar4) and the Hemingfordian (He1) NALMAs, as these are known from temperate North America. This study also demonstrates the power of using multiple sources of geochronological data, i.e., both 40Ar/39Ar and U/Pb methods combined with magnetostratigraphy, in developing a chronology independent of biochronological estimates using stage of faunal evolution. The purpose, therefore, of this article is to present the independent geochronology of the Centenario Fauna and to discuss biochronological ramifications for our understanding of NALMAs both in Panama and from higher-latitude localities and faunas in North America.

Whitmore and Stewart (1965) were the first to publish a faunal list and describe the significance of Miocene mammals collected from along the Panama Canal. They described five taxa, i.e., a protoceratid artiodactyl, the oreodont Merycochoerus or Brachycrus, the rhinoceros Diceratherium, and the equids Anchitherium clarencei and Archaeohippus blackbergi, belonging to four families. (Stirton [in Woodring 1957, p. 37] reported the presence of a metapodial collected from the Canal, but it was of unknown provenience and uncertain taxonomic affinity, although it likely represents a rhinoceros [B. J. MacFadden, pers. obs., 2013].) Patton and Taylor (1973) subsequently indicated that the protoceratid from Panama was probably a new species. On the basis of screenwashing of about a metric ton of sediment from the Cucaracha Formation from the vicinity of the Pedro Miguel locks, which is near our Centenario Bridge and Cartagena Hill localities, Slaughter (1981) described the geomyoid rodent Texomys stewarti. MacFadden (2006) described the fauna from the Cucaracha Formation on the basis of collections made by Whitmore and Stewart. Two additions to the mammalian fauna were recognized, i.e., the canid Tomarctus brevirostris and an amphicyonid or hemicyonid, indeterminate. MacFadden (2006) referred the oreodont to Merycochoerus matthewi, known from the Arikareean and early Hemingfordian of the midcontinent North America, the protoceratid to Paratoceras wardi, known from the early Barstovian of the Texas Gulf Coastal Plain, and the rhinoceroses to two taxa, the larger Floridaceras whitei and the smaller Menoceras barbouri, known from the early Hemingfordian Thomas Farm locality in Florida.

MacFadden and Higgins (2004) analyzed the carbon isotope ratios preserved in the mammalian herbivores from the Cucaracha Formation. Although all of the taxa analyzed indicated a C3 ecosystem (and no evidence for C4 grasses), differences in niche preferences were discovered that represented habitats ranging from woodlands to forests. Kirby and MacFadden (2005) analyzed the tooth size, as a proxy for body size, of the taxa from the Cucaracha and found no difference relative to these species from higher-latitude North America, thus arguing for a continuous extension of Central America southward (and the lack of an archipelago) during the Early Miocene. MacFadden (2009) described a well-preserved maxilla of A. clarencei collected from the Cucaracha Formation. MacFadden et al. (2010) described the presence of the peccary “Cynorcaoccidentale from a 115-m stratigraphic interval encompassing the upper part of the Culebra Formation and extending well into the Cucaracha Formation. As the field context of the Miocene mammals had been documented from numerous individual localities and encompassed a stratigraphic interval, the name “Centenario Fauna” was applied to this assemblage. Since the designation of the Centenario Fauna, recent additions have included descriptions of plants (Herrera et al. 2010), crocodilians (Hastings et al. 2013), turtles (Cadena et al. 2012), and boid snakes (Head et al. 2012).

With regard to the GPTS, we prefer to use that of Gee and Kent (2007), derived from that of Cande and Kent (1995), rather than the more recent compilation by Gradstein et al. (2012). These two timescales are based on different assumptions (i.e., calibrated seafloor spreading rates vs. astronomically tuned chronology), resulting in different ages for magnetochrons. During the Early Miocene, the interval of geological time of relevance to our study, the two timescales differ by about 0.25 m.yr. for the same points in geological time. We decided to follow the recommendation of Tauxe (2010) in her comparative analysis and preference for the Gee and Kent (2007) GPTS. Nevertheless, as we show below, both of these timescales are consistent with the geochronological framework in Panama during the Early Miocene that we have developed.

The stratigraphic sequence exposed along the Panama Canal encompasses one of the most complete and well-exposed Eocene to Miocene volcanic, volcaniclastic, and clastic sequences in Central America (Woodring and Thompson 1949; Woodring 1982; Kirby et al. 2008; Montes et al. 2012b). Although the Panama Canal Basin is a structurally complex area, two distinctive lithostratigraphic regions can be distinguished in the Gaillard Cut (Stewart et al. 1980; fig. 1). Along the southern part of the Panama Canal, exposures in the northwestern half of the Culebra (i.e., Gaillard) Cut (from the Lirio Norte region to Gatun Lake) mainly include the volcanic and volcaniclastic deposits of the Late Oligocene Bas Obispo Formation, the layered tuff and tuffaceous sediments of the Early Miocene Las Cascadas Formation, and a transgressive sequence of the Early Miocene Culebra Formation. Strata from the upper Las Cascadas Formation provide definitive evidence for the earliest terrestrial paleoenvironments in the Gaillard Cut, including preserved in situ tree roots, palm fronds, and a mammalian assemblage indicating the arrival of a North American paleocommunity by the late Arikareean NALMA (Rincon et al. 2012, 2013). A heterogeneous package of carbonates, conglomerates, litharenites, and carbonaceous mudstones onlap the upper Las Cascadas Formation (Montes et al. 2012b), marking the Culebra Formation marine incursion, which culminates with the Emperador Limestone and shallow shelf litharenites and lithic wackes of the lowermost portions of the upper Culebra Formation (Kirby et al. 2008; Pimiento et al. 2013).

Of relevance to this report, the southeastern half of the Culebra Cut, more specifically, the region between Lirio Norte and the Pedro Miguel Locks (fig. 1), includes exposures of the Culebra Formation, continental deposits of the lower Miocene Cucaracha Formation, and the subaerial volcanic and volcaniclastic rocks of the Middle Miocene Pedro Miguel Formation (Stewart et al. 1980; Kirby et al. 2008). The sedimentary exposures constitute the main fossil-bearing horizons yielding the Centenario vertebrate fauna (Whitmore and Stewart 1965; MacFadden 2006; MacFadden et al. 2010), with occurrences in both the upper Culebra and Cucaracha Formations (fig. 2). The upper Culebra Formation, represented by a transition from predominant lithic wackes (lowermost upper Culebra Formation) to either amalgamated lenses of bioturbated litharenites or fining-upward sequences of conglomeratic to medium-grained litharenites, carbonaceous mudstones, and lignites (uppermost upper Culebra Formation), has been interpreted to indicate a regressive transition from shelf marine environments to estuarine and deltaic-front environments (Kirby et al. 2008; Pimiento et al. 2013). Although this sequence has produced an abundant marine invertebrate fauna (Woodring 1982), ichthyofauna (Pimiento et al. 2013), and occasional marine mammals (Uhen et al. 2010), the conglomeratic litharenite strata have also yielded significant dental and postcranial elements of terrestrial mammals, including peccaries (tayassuids; MacFadden et al. 2010), rhinocerotids, and extinct protoceratids.

The overlying Cucaracha Formation consists of amalgamated conglomeratic lenses, litharenites, pedogenically altered mudstones, lignites, and occasional tuffaceous deposits, representing progradational deltaic and nearshore wetland deposits (lower Cucaracha Formation) and fluvial and associated floodplain deposits (upper Cucaracha Formation; Retallack and Kirby 2007; Kirby et al. 2008; Head et al. 2012; Montes et al. 2012b). Tilted fault blocks of the Cucaracha Formation preserve variable thicknesses, ranging from 60 m in the Hodges Hill area to 85 m in the Centenario Bridge area. A 1-m-thick tuff, here referred to as the Cucaracha tuff, is a prominent marker horizon that allows unambiguous correlation between the measured sections at Centenario Bridge and Hodges Hill. This tuff takes on further significance, as described below, because it yielded the radioisotopic age determinations used to calibrate the Centenario Fauna.

Lower Miocene faulting, structural deformation, and erosion of the Cucaracha Formation in these individual fault blocks are likely responsible for the inconsistent stratigraphic relationships with the overlying Pedro Miguel Formation, especially in the Cartagena Hill area, where the thickness of the fossiliferous sequences varies substantially over an area of less than 1 km2 (Head et al. 2012). Despite these structural and erosional complications, the lithostratigraphy is remarkably consistent across isolated exposures (fig. 2), potentially indicating similar sequences of depositional environments over a nearly 4-km unreconstructed southeast-northwest transect. As a result, many of the Centenario vertebrate localities within and between the Centenario Bridge and Hodges Hill areas can be consistently correlated at a scale of ∼5–10-m-thick packages solely on the basis of lithology.

Fossil vertebrates are usually concentrated within the coarser intervals (conglomeratic sandstones and conglomerates) of the upper part of the Cucaracha Formation, above the prominent Cucaracha tuff, and are dominated by freshwater turtles (Cadena et al. 2012), crocodilians (Hastings et al. 2013), and freshwater ichthyofauna. Mammals, although rare, are represented by dental and postcranial elements that exhibit evidence of moderate to intense hydrodynamic transport. Among these fossils, diagnostic dental elements include small artiodactyls (protoceratids, moschids, and floridatraguline camels), followed in abundance by rodents, rhinoceroses, anchitherine horses, tayassuids, carnivores, oreodonts, and bats, as described in more detail below.

Within the Centenario Bridge region, the most productive vertebrate localities occur within the lower half of a 30-m-thick sequence representing channel migration and paleosol development within an anastomosing river and floodplain system (19–49-m levels in Centenario Bridge North section; fig. 2). Water-rounded fragments of medium- to large-bodied crocodilians, turtles, and mammals (e.g., protoceratids, anchithere equids, and rhinocerotids) are often found in amalgamated conglomeratic lenses with variably directed cross stratification at the base of this sequence. Decimeter-thick, coarse-grained litharenite lenses that rapidly fine upward into organic-rich mudstones represent ephemeral stream erosion, deposition, and abandonment in a proximal floodplain, as indicated by pedogenically immature mudstones with limonite mottling. These ephemeral stream deposits, ∼5 m above the amalgamated conglomerates (25-m level in Centenario Bridge North section; fig. 2), preserve a wider size and taxonomic range of vertebrate and plant remains, including carbonized leaves, fruits, and structural elements. A distinctive paleosol with a >1-m red B horizon marks the top of the sequence (80-, 45-, and 94-m levels in the Centenario Bridge South and North and Hodges Hill sections, respectively) and represents a well-drained soil profile on the distal floodplain, potentially in a lowland dry tropical forest (Retallack and Kirby 2007). Fossils are rarely found in this distinctive paleosol.

Similarly, the Hodges Hill microsite, the most productive fossil locality in the Hodges Hill region, occurs at the base of a 10-m fining-upward sequence also associated with fluvial deposition. A 2–3-m-thick package of weakly cross-bedded sandstone lenses marks the base of this sequence and has an irregular erosional contact with an underlying immature paleosol horizon. Lenticular bodies (∼1 m in diameter) of coarse sand and lithic fragments are abundant at this lower contact, contain a wide variety of vertebrate remains in terms of both size and taxonomy, and are interpreted as channel lag deposits within a meandering fluvial system. The lithology rapidly decreases to a fine sand and then grades into a 3-m-thick package of silty mudstone that forms the C horizon of an overlying, well-developed paleosol with a distinctive >1-m-thick B horizon. Small vertebrate remains, including rodents, reptiles, amphibians, and freshwater fish, remain highly abundant in the fine sand and silty mudstone horizons.

Vertebrate fossil collecting. Larger vertebrate fossils were collected by systematic surface prospecting of the exposures in the five sections. Prospecting was periodically repeated, particularly after rains. Fossils have been found as isolated occurrences or in limited concentrations; in the latter case, excavations were undertaken, e.g., from 25 m above the base of the Centenario Bridge North section and from the Hodges Hill microsite at 88 m above the base. Small vertebrates were recovered from the Centenario Fauna primarily through standard screenwashing techniques, although several of the smaller specimens were found on the surface or during excavation. Sediment samples were washed through a series of three nested screens, including quarter-inch hardware cloth (4 mesh, 6-mm opening), standard window screen (16 mesh, 1.5-mm opening), and fine steel screen (24 mesh, 1.0-mm opening), to remove the clay, silt, and fine sand. The resulting sediment concentrate was sorted through a binocular microscope at ×10–20 magnification, and all identifiable microvertebrate fossils were removed. Our screenwashing operation has produced identifiable teeth of small mammals, primarily rodents, from localities within the Centenario Fauna, all from the Cucaracha Formation, including three sites within the Centenario Bridge South section, between 72 and 75 m above the base, and the one microsite at Hodges Hill described above. Additional microvertebrates recovered from these samples, but not discussed in detail here, include fish, frogs, lizards, and snakes. Screenwashing of more than a metric ton (1000 kg) of sediment from the Hodges Hill microsite, the richest of the microvertebrate localities yielding the Centenario Fauna, has produced more than 50 rodent teeth, whereas fewer than 10 rodent teeth and two small procyonid teeth were recovered from sediment samples totaling about 500 kg from the Centenario Bridge sites. We do not know the exact locality for the rodent Texomys stewarti named by Slaughter (1981; however, he described the site as in the “vicinity of Pedro Miguel Lock, Panama Canal” (p. 113). His description would place this site near the current Centenario Bridge, which did not exist in 1981. We have identified teeth of T. stewarti from the Centenario Bridge sites and Hodges Hill. Of the two dozen mammalian taxa recovered so far, about one-third were primarily collected by screenwashing. In addition, many of the lower vertebrates, including bony fish, juvenile crocodilians, frogs, and lizards, currently under study were recovered by screenwashing.

As part of the overall development of a framework for understanding the geochronology of the Neogene deposits of the Panama Canal basin, over the past several years recent attempts have been made to determine the age of the underlying Culebra Formation and the overlying Cucaracha Formation as they crop out along the Gaillard Cut, which in turn would constrain the age of the Centenario Fauna. With regard to a maximum (older) age range for the Centenario Fauna, Kirby et al. (2008) reported six Sr-ratio ages for marine bivalves from four superposed outcrops of the Culebra Formation within the Culebra Cut in measured sections below those presented in our study. The six age determinations are internally consistent with regard to their superpositional relationships (i.e., progressively younger upward; see fig. 6 of Kirby et al. 2008). Of relevance here, the uppermost samples within the upper Culebra Formation have Sr-ratio ages of 19.12 ± 0.42 and 19.83 ± 0.39 Ma. In addition, Montes et al. (2012b) report a U/Pb age of 19.3 ± 0.4 Ma from zircons recovered from a felsic lapilli crystalline tuff located near the base of the Culebra Formation. The Sr-ratio ages from the upper Culebra Formation seem to most accurately constrain the maximum age limit of the Centenario Fauna; the U/Pb age and older Sr-ratio ages add additional information to make a robust argument confirming this age.

The upper (younger) age limit of the Centenario Fauna would theoretically be constrained by an age determination from the Pedro Miguel Formation, a volcaniclastic series of basalt-composition flows that demonstrably overlie the Cucaracha Formation (fig. 2), in particular as exposed in the principal reference section at Centenario Bridge. Our several attempts at dating the overlying Pedro Miguel Formation at the Centenario Bridge section have been unsuccessful because the basaltic rocks have very low K2O concentrations and were highly altered as a result of contact with the water-rich Cucaracha sediments. The radiogenic argon components of both the whole-rock groundmass and the plagioclase phenocrysts are small compared to the excess argon associated with the alteration.

On the other hand, Wegner et al. (2011) report an 40Ar/39Ar whole-rock isochron age of 18.4 ± 1.07 Ma (sample PAN-05–056) for an outcrop of the Pedro Miguel basalt somewhere within the Panama Canal basin, although the exact geographic position and stratigraphic context relative to our localities are uncertain (Montes et al. 2012b). Calculating the 40Ar/39Ar data of Wegner et al. (2011) by using only the concordant steps of the age spectrum (the last four of seven steps, with >60% of the 39Ar released) gives an age of 18.90 ± 0.59 Ma that is more appropriate for this sample. Considering the upper bounds based on the younger 1σ error (i.e., 18.31–18.90 Ma), this age is consistent with the correlations that we make in this article and further confirms that the Cucaracha Formation was deposited in a relatively short interval of time.

Several attempts to date the Cucaracha tuff, a prominent marker horizon that crops out in our measured sections (fig. 2), via the 40Ar/39Ar and U-Pb methods by our group and others have been unsuccessful. This is largely because the plagioclase phenocrysts derived from this unit have low K2O concentrations, the welded groundmass is largely altered devitrified glass, and mineral separations of four large samples did not yield zircons. An 40Ar/39Ar analysis of plagioclase phenocrysts separated only from flattened pumice fragments gave better results ( fig. 4; table 1). The age spectrum from the plagioclase is discordant with the four lowest temperature-release steps, giving apparent ages of less than 15 Ma or with very large errors. The final seven steps give an error plateau age of 18.96 ± 0.90 Ma (fig. 4), with a concordant, but less precise, inverse-isochron age of 18.87 ± 1.94 Ma and an 40Ar/36Ar intercept within error of atmosphere. The atmospheric 40Ar/36Ar ratio suggests that excess argon is not an issue with these seven release steps and that the error plateau age records the time of ash eruption and deposition. After several failed attempts to separate zircons from the tuff, we were able to extract two clear euhedral zircons. The U-Pb analyses of these two zircons gives a mean 206Pb/238U age of 18.81 ± 0.30 Ma (fig. 5; table 2). While the 40Ar/39Ar plagioclase age spectrum is not ideal and the number of zircons analyzed by U-Pb is small, these ages are identical and stratigraphically consistent with ages of the volcanic horizons in the Culebra and Pedro Miguel Formations.

In order to further constrain the age of the Miocene mammals from Panama, 38 paleomagnetic sites were collected from two measured sections that span the interval producing the Centenario Fauna (fig. 2). Three to seven separately oriented hand samples were collected from each paleomagnetic site. Individual sites were positioned into our measured stratigraphic sections and also located by GPS.

Centenario Bridge South. This section crops out along an east-west transect that started at lat 9.030484°N, long 79.635032°W (Google elevation, 104 ft. [32 m]) at the water level of the Panama Canal, passing under the Centenario Bridge to lat 9.030255°N, long 79.636574°W (Google elevation, 212 ft. [65 m]). Stratigraphically, this measured section extends from above the base of the Cucaracha (i.e., the contact with the underlying Culebra is not exposed here) to the upper part of the Cucaracha (fig. 2), where it disconformably underlies the Pedro Miguel agglomerates. Within this section, 32 paleomagnetic sites were taken from the Cucaracha Formation and one was taken from the overlying Pedro Miguel agglomerate.

Lower part of Hodges Hill. The section from which we collected five paleomagnetic sites, which is part of a larger measured section (fig. 2), crops out in a short transect at lat 9.047°N, long 79.653°W. It is important because it correlates laterally with most of the lower fossil localities from the interval that comprises the Centenario Fauna (fig. 2). This 16-m-thick section includes the conglomerates that define the contact between the Culebra and Cucaracha Formations.

In order to determine the magnetic mineralogy and hence the appropriate demagnetization regime, four pilot samples from the Cucaracha Formation representing different lithologies were subjected to 12–15 stepwise isothermal remanent magnetization acquisition increments, as shown in figure 6. Some of the samples, e.g., 10a, were saturated in relatively low applied fields (∼500 millitesla [mT]), indicating that magnetite is the dominant carrier of the remanence. In contrast, sample 30a continued to increase magnetization in incrementally higher applied fields, indicating a component of the characteristic remanent magnetization (ChRM) likely resulting from hematite. Given this variation in interpreted magnetic mineralogies, and also the fact that the surficial colorations of outcrops varied from greenish to reddish, we subjected all sedimentary paleomagnetic samples to incremental thermal demagnetization, typically in 12–15 steps ranging from natural remanent magnetization (NRM) to 590°C. The two volcanic units in our Centenario Bridge South section, i.e., the Cucaracha tuff and the Pedro Miguel agglomerate, were subjected to alternating-field demagnetization, typically in at least 10 steps up to 1000 mT.

The vector demagnetization diagrams presented in figure 6 demonstrate representative behavior of individual samples upon stepwise demagnetization. Once the magnetization component of the NRM represented in low fields was removed (typically <200°C), then most of the stable samples demonstrated decreasing decay to the origin up to the highest demagnetization step (i.e., typically between 575° and 590°C). These linear demagnetization segments were analyzed in 58 of the 95 samples (see the supplementary table) that ultimately produced interpretable results by the least squares principal-component method (Kirschvink 1980), to determine the mean value of the declination and inclination for each paleomagnetic sample. In the other 37 samples, stepwise demagnetization and the least squares method produced divergent directions; in these cases, individual declinations and inclinations were selected at one step (between 250° and 560°C) to represent the ChRM.

Samples from each site were then analyzed with Fisher statistics (Fisher 1953) and were accepted if the R (“resultant” statistic; Irving 1964; Tauxe 2010) satisfied the acceptability criteria established by Irving (1964), where R ≥ 2.62 for N = 3, R ≥ 3.10 for N = 4, and R ≥ 3.50 for N =5. Sites that satisfy these criteria are termed Class 1 (sensu Opdyke et al. 1977; Opdyke and Channell 1996). Two other kinds of site data were used in the magnetostratigraphy, i.e., Class 2, in which only two samples per site were available (either from breakage or because of poor demagnetization characteristics), and Class 3, in which three samples indicated a polarity but did not pass the Fisher and Irving criteria. Of the 38 sites that were originally sampled, 32 yielded interpretable polarities and six (7–9, 12, 19, and 102) did not. Of the 32 sites with interpretable results, and with the eight superposed sites whose results were combined (2 and 3, 13 and 14, 15 and 16, and 28 and 29), the resulting sites included 11 Class 1, 5 Class 2, and 12 Class 3 sites ( supplementary table). The plots in figure 7 indicate that after demagnetization and apparent removal of a normal overprint that affected the declination directions, the site mean values cluster in the Southern Hemisphere—i.e., they have declinations between 90° and 270° and negative inclinations—with formational mean values (two normal sites inverted to Southern Hemisphere) of declination 178.9°, inclination −52.8°, k = 4.8°, and α95 = 13.9°, where k is the precision parameter and α95 is the 95% cone of confidence (at P = 0.05), respectively, of paleomagnetic data (Tauxe 2010). The mean inclination of −52.8° (table 3) is steeper than would be predicted from the dipole field of 37.2° calculated for the site latitude (the overlap of the demagnetized data and the dipole in fig. 7 is an artifact of the large α95, resulting from dispersed individual data points). Although the dispersion of these data is more than would be expected from high-quality data from a single formation (and would not be good for calculating a paleomagnetic pole, because of the large α95), the polarities for these sites are, with one exception, unambiguous, with 26 reversed sites, one normal Class 3 site (32), and another possible normal Class 3 site (33; supplementary table).

The magnetostratigraphic results described above argue that on the basis of 22 sites, the ∼85-m-long portion of the Centenario Bridge South section is almost entirely of reversed polarity. Two Class 3 sites, one of which (32) has a ChRM direction that would be expected for a normal polarity and the other of which (33) is questionable (with a declination in the Southern Hemisphere but positive inclination; supplementary table), are discussed below in terms of whether they represent a normal-polarity zone or an excursion of the ancient magnetic field. Likewise, the four sites from Hodges Hill indicate a reversed polarity.

The observation that the Cucaracha Formation is of dominantly reversed polarity is consistent with the results recently reported by Montes et al. (2012a). They found that from two sampling sites within the Canal at Hodges Hill (see our fig. 1), one yielded a mean declination of 186.3° and inclination of −34.5° (N/n = 7/7, α95 = 10.3°, k = 35.02) and the other a mean declination of 149.6° and inclination of −40.3°, (N/n = 7/5, α95= 4.4°, k = 298.56), thus indicating a reversed polarity. Their results also indicate a positive tilt test, indicating that the timing of magnetization is prefolding, and they also confirm a higher inclination than would be predicted from this paleolatitude.

The correlation of the Centenario Fauna is based on two lines of evidence, i.e., the dominantly reversed magnetostratigraphy and the calibration of the two interbedded radioisotopic determinations, i.e., an 40Ar/39Ar age of 18.96 ± 0.90 Ma and a U-Pb age of 18.81 ± 0.30 Ma (fig. 8). Given these constraints, there is a unique and unambiguous correlation of the sedimentary interval measured at the Centenario Bridge South and lower Hodges Hill sections that brackets the Centenario Fauna collected from this locality to C5Er. According to the Gee and Kent (2007) GPTS, this interval occurs between 19.05 and 18.78 Ma. The questionable normal-polarity zone represented by paleomagnetic site 32 and possibly site 33, both of which are Class 3, is therefore considered to be either a spurious ChRM or perhaps a previously unrecognized excursion or cryptochron (Tauxe 2010). With regard to the age of the overlying Pedro Miguel Agglomerate, given its reversed polarity, as recorded from the Centenario Bridge South section, and the age reported by Wegner et al. (2011), recalibrated here to 18.90 ± 0.59 Ma, this volcanic flow also likely occurred within C5Er. These data indicate a relative short hiatus represented by the disconformity above the Cucaracha Formation.

Although the upper limit of the Cucaracha Formation previously has been ambiguous, the lower limit of this formation has been well constrained by (1) Sr-ratio ages of marine mollusks from the lower portion of the Culebra Formation ranging from 20.99 to 19.12 Ma (Kirby et al. 2008) and (2) a U-Pb age (zircon) of 19.3 ± 0.04 Ma (2σ) from a felsic tuff near the base of the Culebra Formation. These ages together constrain the lower age of the Cucaracha, which is also consistent with being younger than the Culebra, as is also demonstrated from the superpositional field relations.

Miocene mammals have been reported from the Panama Canal since the original paper published by Whitmore and Stewart (1965), which identified five taxa (Merycochoerus, a protoceratid, Anchitherium, Archaeohippus, and Diceratherium) within four families (Oreodontidae, Protoceratidae, Equidae, and Rhinocerotidae). From this same collection, housed in the US National Museum (USNM), MacFadden (2006) added two species, one within the Amphicyonidae (or Hemicyonidae) incertae sedis and Tomarctus brevirostris (Canidae). Our systematic efforts to collect fossils from the Panama Canal started in 2002, during the excavations for the Centenario Bridge, but then these were intensified by the new excavations and expansion of the Canal starting in 2009. Originally focusing on surface prospecting at new exposures in the Culebra Cut, we have since diversified our fieldwork by identifying potentially rich horizons for microfauna and then screenwashing at these localities. Since the description of the original USNM collections (MacFadden 2006), the Gaillard Cut local fauna (LF) has been expanded and renamed the Centenario Fauna (MacFadden et al. 2010), and the number of mammalian taxa recovered has doubled. Since MacFadden et al. (2010), we have recognized a dozen new species from six new families, i.e., Camelidae, Phyllostomidae, Sciuridae, Heteromyidae, Moschidae, and Procyonidae (also see table 4). The obvious additions to the Centenario Fauna are mostly within the smaller taxa and have resulted from the screenwashing and subsequent microfaunal picking efforts. Given the number of taxa at a comparable well-sampled site in North America, e.g., Thomas Farm, Florida, with 25 families and ∼60 species of Miocene mammals, the most likely explanation for the relatively lower diversity in the Centenario Fauna is that we have not sufficiently sampled the actual fauna, particularly those taxa that are small and/or rare. Consequently, we realize that the faunal list presented in table 4 is a snapshot of our knowledge at a point in time and is likely to increase via continued fieldwork. Nevertheless, even though the Centenario Fauna is not nearly as rich as the Thomas Farm Fauna, it is still considerably more diverse than any other Miocene fauna in Central America and, along with the underlying Lirio Norte LF from the Las Cascadas Formation, is the only Miocene fauna in Central America that contains small mammals.

Despite its evolving nature, the Centenario Fauna contains meaningful biochronological information to calibrate the stage of evolution of the mammalian fauna, and as developed below, it has significance with regard to comparisons with similar-aged faunas in higher-latitude North America. In assessing the individual biochrons of each taxon, with one exception (the two rhinocerotid species; see below) we used its temporal range from North America to establish its biochronological signal. If a species occurs within a subdivision of a NALMA, e.g., He1, it is graphically depicted as extending within the entire interval, although we realize that in some cases it may have a duration for only part of this interval. Nevertheless, this is the level of resolution that our biochronological analysis will allow at the present time. If a species or genus is qualified with a “cf.” (confer, i.e., “compares with”; Lincoln et al. 1982), then we used the range of the taxon, realizing that additional phylogenetic analysis might revise the biochron. If a species is new, then we used the biochronological range of the genus. Given this method, the two taxa listed in table 4 as “new genus and sp. of large insectivorous phyllostomine” and “Larger tayassuid, incertae sedis” have no biochronological utility and are therefore not depicted in figure 9.

Within the rodent families Sciuridae and Heteromyidae, the three taxa represent unidentified or new species, and thus the generic ranges used in figure 9 indicate long biochrons that are not useful in assessing a precise age for the Centenario Fauna. The rodent family Jimomyidae (sensu Flynn et al. 2008) is represented by the genus Texomys, which Slaughter (1981) allocated to the late Hemingfordian (He2) through early Barstovian (Ba1) of the Texas Gulf Coastal Plain. (Slaughter judged the stage of evolution of Texomys stewarti from Panama to be more primitive and inferred it to be of early Hemingfordian [He1] age. But given our biochronological method, this inference about relative stage of evolution is not used here.) Albright (1999) identified Texomys sp. from the early late Arikareean (Ar3) Toledo Bend LF of the Gulf Coastal Plain of Texas and the correlative Ar3 Buda LF of Florida. Both of these occurrences therefore extend the range of the genus back in time considerably, relative to Slaughter’s (1981) original description.

Three taxa of Carnivora are listed in table 4 and figure 9. The canid T. brevirostris is otherwise known from the late Hemingfordian (He2) through late Barstovian (Ba2) of higher-latitude North America (Wang et al. 1999). This occurrence is out of synch, i.e., it does not overlap, with most of the other biochronological signals from the Centenario Fauna, and we consider it to be problematical at present. This identification was based on poorly preserved material (MacFadden 2006). It is conceivable, therefore, that the original identification may be incorrect and that this taxon may represent another carnivoran. This hypothesis will have to be tested by further study in the future. Additional carnivoran fossils have been collected, including a mandibular ramus with dentition that is possibly referable to the amphicyonid Daphoenodon. This is likely the same taxon that was originally reported from the Cucaracha Formation as either Amphicyonidae or Hemicyonidae, incertae sedis (MacFadden 2006). After a questionable range in Ar2, Daphoenodon ranges from Ar3 through He1. The later part of this range is consistent with the general biochronological signal of the Centenario Fauna.

The Centenario Fauna includes one species of procyonid referable to the genus Bassaricyonoides. On the basis of the fragmentary material recovered so far by screenwashing, it is unclear whether this represents a new or a previously described species. The genus Bassaricyonoides is known from two localities elsewhere in North America, i.e., from the He1 Suwannee River Miller site in central Florida and the He2 Massacre Lake site in Nevada (Baskin 2003). Thus, an He1–He2 biochron is indicated for the presence of Bassaricyonoides from the Centenario Fauna.

With regard to the ungulates, the Artiodactyla is the most diverse order collected so far and is represented by eight taxa within five families (table 4). The family Tayassuidae is represented by two taxa, including a fragmentary, relatively large, and heavily worn tooth representing an indeterminate taxon (which is not considered to be biochronologically informative and therefore is not depicted in fig. 9) and “Cynorcaoccidentale (MacFadden et al. 2010). “Cynorcaoccidentale is represented by a suite of relatively well-preserved specimens (which has increased from additional fieldwork since the publication of MacFadden et al. 2010) collected from most of the stratigraphic interval from which the Centenario Fauna has been recovered, i.e., ranging from the Upper Culebra to near the top of the fossiliferous interval within the Cucaracha Formation. Taking into account the reported occurrences in higher-latitude North America, MacFadden et al. (2010) list the biochron of “C.” occidentale as ranging from He1 to Ba1. The oreodont Merycochoerus matthewi is known from localities in South Dakota, Nebraska, and Wyoming of Ar4 to He1 NALMA age, and this biochron is used for the Centenario Fauna, as it pertains to this occurrence.

The Moschidae, a family of relatively small artiodactyls, is represented by two genera that are listed in table 4 as cf. Parablastomeryx sp. and cf. Machaeromeryx sp. (At the present time, it has not been determined whether these species are new or referable to previously named species.) On the basis of the ranges reported in Webb (1998), the genus Parablastomeryx first appears in He1, or possibly Ar4, and extends into the late Clarendonian, and Machaeromeryx is reported from the Hemingfordian (He1–He2) in higher-latitude North America.

With three dozen catalogued specimens in the University of Florida Vertebrate Paleontology collections, plus those in the USNM and Southern Methodist University collections, the Protoceratidae currently is the most abundant family of medium- to large-sized mammals represented in the Centenario Fauna. MacFadden (2006) recognized one taxon of protoceratid and referred it to Paratoceras wardi, which was otherwise known from Ba1 of Texas Gulf Coastal Plain (Patton and Taylor 1973). From the collection of more specimens and a rigorous phylogenetic analysis, A. Rincon et al. (unpublished manuscript) recognized two new species of Paratoceras from the Centenario Fauna. Paratoceras has a long biochron, ranging from He1 to Cl1 (sensu Prothero 1998a, but with the addition of Webb et al. 2003; table 4; fig. 9). The family Camelidae is represented by a single taxon referred to cf. Floridatragulus nanus. According to Honey et al. (1998), this genus is known from He1 through Ba2 elsewhere in North America, and this is the biochron that we use here, realizing that it likely will be refined with future studies of this potentially important taxon.

The order Perissodactyla is represented in the Centenario Fauna by two families, i.e., Equidae and Rhinocerotidae, each with two taxa (table 4). The equid Anchitherium clarencei is known from several localities in higher-latitude North America, and the biochron of this species as it is currently recognized is from He1 through Ba1 (MacFadden 2001, 2006, 2009). The other equid, Archaeohippus sp., is based on limited material and thus difficult to diagnose at the species level. Thus, it is a long-ranging genus in higher-latitude North America, and this is represented by a biochron that ranges from He1 through Ba1 (table 4; fig. 9). The two rhinocerotid species, i.e., the larger Floridaceras whitei and the smaller Menoceras barbouri, are represented by well-preserved and diagnostic material in the Centenario Fauna (MacFadden 2006). Floridaceras is known from only the species F. whitei, which was originally named from Thomas Farm, Florida (Wood 1964). Tedford et al. (2004) indicate that Floridaceras has a first appearance datum (FAD; Woodburne 1987) that is part of the definition of the beginning of He1, and it extends throughout this interval but probably not into He2, depending on the age of its occurrence at Thomas Farm, which is depicted as straddling He1 and He2 in Tedford et al. (2004). With regard to M. barbouri, in addition to the type locality at Thomas Farm, Florida, it has been otherwise reported from numerous localities in North America, including New Mexico, Texas, South Dakota, and New Jersey (Prothero 1998b). If, however, these reported occurrences from Texas (Toledo Bend LF and Castollon LF) are discounted (respectively, Albright 1999; Tedford et al. 2004), then the FAD and subsequent biochronological range of M. barbouri are likewise restricted to the He1, depending on the age of the occurrence at Thomas Farm, Florida. (The genus has a much earlier range, i.e., with a FAD in the late Arikareean; Prothero 1998b.)

According to the Wood Committee report (Wood et al. 1941), and as subsequently elaborated upon and refined (e.g., Tedford 1970; Woodburne 1987), the NALMAs are complex assemblage zones consisting of four kinds of biochronological data: (1) index fossils, (2) FADs, (3) last-appearance data (LADs), and (4) characteristic fossils. Within the Centenario Fauna, all four of these kinds of data are found, and taken together, they can be used to unambiguously constrain the NALMA to which it belongs. It also should be noted that these four kinds of data are not mutually exclusive; e.g., at our current level of biochronological resolution, the amphicyonid Daphoenodon sp. has its LAD and is a characteristic fossil of He1. On the basis of the occurrence data presented in table 4 and figure 9, we argue here that the age of the Centenario Fauna is early Hemingfordian, i.e., He1.

1. Index fossils. These are taxa that are unique to the biochron. Depending on the age of the Thomas Farm LF, i.e., on whether it is correlated to the latest part of He1 or straddles the He1-He2 boundary, then the rhinocerotids Menoceras barbouri and Floridaceras whitei may be index fossils within the Centenario Fauna. Depending on whether its range does not occur in the Ar4, which is questionable, the moschid cf. Parablastomeryx may also be an index fossil for He1.

2. First appearances. In addition to the index fossils listed above, the procyonid Bassaricyonoides sp., the tayassuid “Cynorcaoccidentale, the camelid cf. Floridatragulus nanus, the equid Anchitherium clarencei, and the rhinocerotids M. barbouri and F. whitei all have their first appearance in North America at the beginning of or during He1. In addition, depending on whether Parablastomeryx actually occurs in Ar4 in North America, this genus of moschid may also first occur in He1.

3. Last appearances. In addition to the index fossils listed above, the amphicyonid Daphoenodon sp. and the oreodont Merycochoerus matthewi have their LADs, i.e., they last appear, in He1.

4. Characteristic taxa. These taxa occur during He1 but also occur earlier and/or later. From the Centenario Fauna, these include the rodents Nototamias sp., Petauristodon sp., Proheteromys sp., and Texomys spp., the amphicyonid cf. Daphoenodon sp., the oreodont M. matthewi, the moschid cf. Machaeromeryx sp., the protoceratid cf. Paratoceras n. sp., the camelid cf. F. nanus, the equids A. clarencei and Archaeohippus sp., and possibly the moschid cf. Parablastomeryx. Many of these ranges are artificially long because of our current taxonomic uncertainty, e.g., in the rodents, or because of the use of generic ranges for taxa not identified to species.

In our current state of knowledge, one taxon does not occur within the He1, i.e., the canid Tomarctus brevirostris. On the basis of its occurrence in North America, T. brevirostris has a range restricted to Ba1, which does not overlap in time with the rest of the biochronologically diagonistic taxa within the Centenario Fauna. There are at least two plausible explanations for this asynchronous (nonhomotaxial) occurrence within the Centenario Fauna: (1) it represents a range extension, so that T. brevirostris actually occurred first in Panama and then in higher-latitude North America, or (2) the taxonomic identification of T. brevirostris, which is based on fragmentary material (MacFadden 2006), is incorrect. Our ongoing research suggests the latter, and we may revise our taxonomic identifications in light of additional comparisons and phylogenetic analyses.

Many authors (e.g., Woodburne 1987; Tedford et al. 2004) have argued that immigration events (FADs) of nonnative (allochthonous) taxa should be used to define NALMA boundaries. Furthermore, Woodburne (1987) posits that using a single taxon, e.g., of a commonly occurring genus, is in many cases the best method, unless it is known that a group of taxa immigrated at the same time, i.e., within the available level of geochronological resolution. Tedford et al. (2004, fig. 6.3) list two genera of allochthonous genera that occur in the Centenario Fauna, i.e., the rodent Petauristodon to indicate the beginning of He2 and rhinocerotid Floridaceras to indicate the beginning of He1. With regard to the first FAD, Petauristodon sp. is now known from the underlying Ar4 Las Cascadas Formation in Panama (G. S. Morgan et al., unpublished manuscript; also see discussion about this genus in Pratt and Morgan 1989), so its utility for indicating the He1-He2 boundary can no longer be considered correct. Nevertheless, the presence of the rhinocerotid Floridaceras sp. in the Centenario Fauna indicates He1. Regardless of whether one uses the overall assemblage-zone biochronology or the FAD of the allochthonous immigrant Floridaceras, the biochronological age of the Centenario Fauna is unambiguously in the early Hemingfordian NALMA, i.e., He1.

The stratigraphic sequence that contains the Centenario Fauna is now well constrained by multiple lines of geochronological and biochronological evidence. Our research has yielded two new radioisotopic determinations: (1) an 40Ar/39Ar age of 18.96 ± 0.90 Ma and (2) a U-Pb zircon age of 18.81 ± 0.30 Ma. In addition, magnetostratigraphic data indicate essentially a reversed polarity at the Centenario Bridge section that, under the constraints of the radioisotopic ages, correlates to chron C5Er (18.78–19.05 Ma in the Gee and Kent [2007] GPTS). Given these geochronological and paleomagnetic constraints, the age of the Centenario Fauna from the Bridge section occurs within the 0.27-m.yr. interval of chron C5Er and correlates to the early Hemingfordian (He1) in figure 10. These results have significance when this correlation is compared to other established ages for similar-aged sediments and faunas from higher-latitude North America.

Nebraska. The type sections and faunas for the late Arikareean and Hemingfordian NALMAs from the equivalent interval during the Early Miocene come from northwestern Nebraska in the adjacent Sioux and Dawes Counties (Tedford et al. 2004). Critical sections from Sioux County include a superpositional sequence in the upper part of the Harrison Formation, the Anderson Ranch Formation (previously Upper Harrison, renamed by Hunt 2002), and the Runningwater Formation. The faunal transition between the late Arikareean (Ar4) and the early Hemingfordian (He1) occurs within this transition between the Anderson Ranch and Runningwater Formations. The Anderson Ranch Formation, which constrains the upper temporal limit of Ar4, is calibrated by a fission-track (zircon) age determination from the Eagle Crag ash of 19.2 ± 0.5 Ma (Tedford et al. 2004). The magnetostratigraphy of this transition is characterized by a zone of reversed polarity that was correlated to chron C5Er on the GPTS (MacFadden and Hunt 1998), the same zone identified from the Culebra to Cucaracha Formations in Panama that contains the Centenario Fauna.

Florida. The Thomas Farm LF, from northern peninsular Florida, is a well-sampled and highly diverse fauna, many elements of which are found at either the species or the generic level in the Centenario Fauna. Unfortunately, the Thomas Farm fossil locality represents an isolated sinkhole fill that lacks a meaningful stratigraphic section, and there are no means of independent dating other than mammalian biochronology. Thomas Farm is nevertheless important because, according to Tedford et al. (2004), the Thomas Farm LF straddles the He1-He2 boundary between about 18 and 17 Ma. If the Centenario represents the early part of He1, then Thomas Farm is at least 1 m.yr. younger and shows the extent of faunal similarity through time.

Oregon. One of the other important and well-calibrated sequences from midlatitude North America that correlates to the Centenario Fauna of Panama is found from the upper part of the John Day Formation in Oregon (Albright et al. 2008). Hunt and Stepleton (2006) report early Hemingfordian mammals from the Rose Creek Member. At the generic level, taxa in common between Oregon and Panama include Parablastomeryx, Merycochoerus, Anchitherium, and Archaeohippus. Unpublished paleomagnetic data from the Rose Creek Member are of reversed polarity, and in conjunction with an early Hemingfordian biochronology, Albright et al. (2008) posited a correlation to either chron C5Er or C5Dr. On the basis of the correlation with Panama, figure 10 depicts the Rose Creek Member to be C5Er and coeval with the Centenario Fauna.

In summary, so far as we know, the Centenario Fauna occurs within a narrow interval of at least 270,000 yr (between 19.05 and 18.78 Ma) during the Early Miocene at the beginning of He1. We do not know how much, i.e., what fraction, of this 270,000-yr interval is represented by this fauna, but the biostratigraphic distribution upsection supports a homogeneous assemblage with no discernible evolution during its existence. Thus, this geochronological and biochronological study has yielded relatively precise and meaningful results that allow correlations to important faunas from higher-latitude North America.

As discussed above, we use the GPTS developed by Gee and Kent (2007), which is based on progressive refinement of magnetic anomalies and seafloor spreading rates with calibrated tie points (e.g., Cande and Kent 1995). As presented in figure 11, however, during the Early Miocene, the Gee and Kent (2007) GPTS differs from that of the more recent Gradstein et al. (2012) GPTS, which is based on astrochronology, by about 0.25 m.yr. Of relevance to the Panama Canal sequence, this is exemplified by chron C5Er, which in Gee and Kent (2007) ranges from 19.048 to 18.781 Ma, whereas in Gradstein et al. (2012) it ranges from 18.748 to 18.524 Ma. Thus, the two GPTSs are calibrated differently during the Early Miocene, resulting in a nonoverlapping age for C5Er.

The radioisotopic age determinations from the Cucaracha tuff independently address this issue. The two methods presented here yielded an 40Ar/39Ar age of 18.96 ± 0.90 Ma and a U-Pb zircon age of 18.81 ± 0.30 Ma. Taking into consideration that the Cucaracha tuff occurs within a zone of reversed polarity, the mean ages for both methods are more consistent with the Gee and Kent (2007) GPTS. On the other hand, if one accepts the 2σ errors associated with the two ages of, respectively, 0.90 and 0.30 Ma, then the reversed zone found in Panama could also be correlated to the Gradstein et al. (2012) GPTS. However, as presented in the correlations with the NALMAs from Nebraska (fig. 10), the C6n-to-C5Er transition is approximated by a K-Ar age determination of 19.2 ± 0.5 Ma (MacFadden and Hunt 1998). To sum up, within the limits of the precision that we have developed for the stratigraphic sequence that includes the Centenario Fauna, either of the two GPTS calibrations are appropriate and fit our data within the 2σ uncertainty of the radioisotopic age determinations.

Our results establish a precise age of an important new terrestrial vertebrate fauna for tropical North America during the Early Miocene, in a region otherwise characterized by a dearth of fossil localities. Overall, our results so far suggest that the Centenario Fauna was fairly stable (taxonomically homogeneous) during the 0.27-m.yr. interval within which this assemblage lived. It is also interesting that our independent geochronological and paleomagnetic calibrations indicate that the Centenario Fauna is coeval with early Hemingfordian (He1) faunas, as they are known from higher-latitude North America. In fact, the calibration of the Centenario Fauna confirms that the transition from Ar4 to He1 occurs within a reversed-polarity zone representing chron C5Er, between 18.78 and 19.05 Ma. This broad paleogeographic north-south range of some 5,000 km almost doubles the latitudinal distribution of He1 faunas and confirms the efficacy of the NALMAs in the ability to recognize faunal provinciality during the Early Miocene. From a tectonic point of view, the several-hundred-meter-thick sedimentary and volcanic deposits spanning the Ar4 from the upper part of the Las Cascadas Formation (Rincon et al. 2012) through the He1 of the upper Culebra and Cucaracha Formations accumulated within a relative short interval of about 2 m.yr. This apparent rapid sedimentary accumulation is perhaps not surprising, given the complex and active tectonic history of the Panama Canal basin during the Early Miocene (Farris et al. 2011; Montes et al. 2012a, 2012b).

Understanding of the Centenario Fauna will likely evolve as our fieldwork progresses, as will knowledge of the ancient biodiversity of the North American tropics. We predict that more taxa will be added to our understanding of the ancient biodiversity in the North American tropics. Regardless, with key index fossils such as Floridaceras and other characteristic taxa, the study presented here for the Panama sequence refines our understanding of the Ar4-He1 boundary as it is known from throughout North America, ranging as far southward as the exposures along the Panama Canal.

We thank the PCP PIRE (Panama Canal Project, Partnerships in International Research and Education) participants, including C. Jaramillo, D. S. Jones, and postdocs, students, and interns for their collaborations and contributions to the paleontological field work; M. Drouillard for assistance with the geochronological analyses; and U. Denetclaw for assistance with the microfaunal screenwashing. J. Bourque and his assistants prepared the fossil specimens, and R. C. Hulbert Jr. was in charge of the curation of the Centenario collection, with assistance from graduate students. K. Huang of the University of Florida and L. Taylor of Clemson University assisted with the paleomagnetic statistics; some of the paleomagnetic analyses used the Stereonet program (ver. 8.7.0; Allmendinger et al. 2013). We thank M. X. Kirby and P. Francesca for their field contributions during the development of this project. J. Velez-Juarbe and the PCP PIRE interns measured sections, and D. S. Jones and the California teachers (PCP PIRE Teach) assisted in the collection of paleomagnetic samples. For discussion about the Centenario Fauna, we are indebted to the insights provided by R. M. Hunt Jr. and M. O. Woodburne. G. Bayona provided us important locality data for the paleomagnetic samples previously reported from the Panama Canal Group (Montes et al. 2012a). This project has been primarily supported by National Science Foundation (NSF) project 0966884. Funds for digitization and curation of fossil specimens were provided by NSF 1115210 (iDigBio project) and 120233. The ACP (Autoridad del Canal de Panamá) provided funding and access to the collection localities within the Panama Canal. Other related support of the PCP PIRE was provided by Richard Toyota, the Secretaría Nacional de Ciencia, Tecnología e Innovación (SENACYT), and the US National Science Foundation grants. This is University of Florida Contribution to Paleobiology number 659.

Supplementary data