Mammut pacificus sp. nov., a newly recognized species of mastodon from the Pleistocene of western North America

A new species of mastodon from the Pleistocene of western North America, Mammut pacificus sp. nov. is herein recognized, with specimens identified throughout California and from two localities in southern Idaho. This new taxon differs from the contemporaneous M. americanum in having narrower teeth, most prominently in M3/m3, as well as six sacral vertebrae, femur with a proportionally greater mid-shaft diameter, and no mandibular tusks at any growth stage. All known Pleistocene Mammut remains from California are consistent with our diagnosis of M. pacificus, which indicates that M. americanum was not present in California.


INTRODUCTION
The American mastodon (Mammut americanum) is one of the iconic megafaunal mammals of the North American Pleistocene, with a widespread distribution across nearly every US state, Canada, and Mexico. This ubiquitous distribution played a central role in the formation of an American identity and the founding of North American vertebrate paleontology (Semonin, 2000).
There have been surprisingly few detailed studies of North American mammutids since Osborn's (1936) original and seminal work, likely owing to a mistaken perception that a common taxon that has been recognized for over 250 years must also be well understood. Most studies of American mastodons, post Osborn (1936), have either been occurrence reports or focused on environmental context. In recent decades, our knowledge of mastodon anatomy has made strides (Green, 2006;Fisher, 2008Fisher, , 2009Hodgson et al., 2008a) as has our understanding of their disappearance within the context of the late Pleistocene extinction of the North American megafauna (Dreimanis, 1968;Alford, 1974;Barnosky et al., 2004;Saunders et al., 2010;Zazula et al., 2014;Widga et al., 2017). While the genus Mammut was divided into several different species during the first half of the 20th century (summarized in Osborn (1936)), these taxa have not withstood detailed separate species designations for the two populations (see De Queiroz, 2007 for a discussion of morphospecies concepts and their relationship to lineage splits).
For inclusion in this study, all teeth needed to include at least state-or province-level locality information, and reasonable confidence as to geological age. Where radiometric dates were not available, specimens were grouped by North American Land Mammal Age (NALMA) following Bell et al. (2004). Only specimens that could be reliably assigned to either the Irvingtonian or Rancholabrean NALMA were included; some premolars from Florida were assigned as either Rancholabrean or Irvingtonian by the repository. When reliable measurements for the left and right tooth, in the same position from the same individual (in the case of in situ specimens), were obtained, either tooth was considered to be satisfactory for the statistical analyses. In such cases, one of the two teeth were chosen at random for measurement.
The electronic version of this article in portable document format will represent a published work according to the International Commission on Zoological Nomenclature (ICZN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank Life Science Identifiers (LSIDs) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/. The LSID for this publication is: urn:lsid: Location: Diamond Valley Lake West Dam (Locality number 95Q10-16.1), near Hemet, Riverside County, California. Found in a fluvial deposit, approximately 4.05-4.31 m below the original ground surface. Associated vertebrate remains from the same locality and stratum include a juvenile partial humerus from Mammut and a partial tooth from Mammuthus columbi, as well as Urocyon, Sylvilagus, Scapanus, Dipodomys, Thomomys, Neotoma, Microtus, Callipepla, colubrids, and anurans. This specimen was collected from October 17 to November 7, 1995 during the relocation of the San Diego canal in advance of construction of the West Dam (field notes are on file at WSC).
Age: Pleistocene, Rancholabrean. Radiocarbon dating of charred organic material associated with holotype specimen WSC 18743 yielded six calibrated 14 C ages that range from 15.87 ± 0.19 to 16.79 ± 0.28 ka (Springer et al., 2009(Springer et al., , 2010. Referred specimen WSC 9622 from Diamond Valley Lake, East Dam (locality number 97SK6-27.1), yielded three calibrated 14 C ages from charred organic material, plant fibers, and eggshell that are all greater than 46.5 ka. Referred specimen WSC 8932, also from Diamond Valley Lake, East Dam (locality number 98NB1-27.7), yielded one calibrated age from a Succineidae shell of 37.8 ± 1.8 ka. Additional information on 14 C ages is provided in Table 3. Etymology: The specific name, pacificus, refers to the fact that all currently recognized specimens of this taxon were collected less than 1,000 km from the coast of the Pacific Ocean.

Diagnosis
A species of Mammut differing from Mammut americanum in the following characteristics: M3/m3 significantly narrower relative to length; six fused sacral vertebrae in later ontogenetic stages (usually five in M. americanum, with a range of four to six); femur with a greater midshaft diameter relative to length; absence of mandibular tusks and associated alveoli (variably present in M. americanum); smaller basal diameter of tusks in males for a given age.

Description
This description relies primarily on the holotype specimen, WSC 18743, but includes additional referred specimens to provide information on elements that are absent in the holotype, as well as to provide insight into intraspecific and interspecific variability in characters.

Cranium
The cranium of WSC 18743 ( Fig. 1; M37481) is largely complete. There is some dorsoventral crushing that slightly flattened the dorsal side of the braincase; this crushing does not appear to have affected the ventral half of the cranium. The alveoli for the tusks appear to be crushed as well, particularly on the left side. In general morphology, the cranium does not differ significantly from that of Mammut americanum, when allowing for individual, sexual, and ontogenetic variability. WSC 18743 falls into Laws group XXII (39 ± 2 AEY), which is ontogenetically somewhat older than AMNH 9951 (the "Warren mastodon") from Newburgh, New York (based on our images of this specimen and the written description in Warren (1852)). The frontals of WSC 18743 more fully cover the temporal fossae in dorsal view than does AMNH 9951, but this may be affected by the dorsoventral crushing in WSC 18743. In any case, this feature exhibits some ontogenetic variability (compare, e.g., AMNH 9951 and AMNH 14535 in Osborn (1936)). Although WSC 18743 and AMNH 9951 both represent male mastodons and are of roughly similar ontogenetic ages, the cranium of WSC 18743 is considerably shorter, with a length of one m (measured along the midline on the dorsal surface) (Table 1). According to Warren (1852), the cranial length of AMNH 9951 is 48 inches (1.23 m). Osborn (1936) noted that M. americanum specimens differed from each other in the position of the tusk alveoli relative to the maxillary tooth rows, a feature that he considered to be sexually dimorphic and related to the larger size of the tusk in the male. In AMNH 14292 (approximately LG XXI), which Osborn (1936) considered a female, and AMNH 17727 (LG V or VI), considered a juvenile male by Osborn (1936), there is a vertical step between the maxillary tooth row and the ventral margin of the premaxilla that forms the tusk alveolus when seen in lateral view (e.g., see Osborn, 1936, Figs. 131 and 132). In contrast, in AMNH 9951 (LG XX or XXI), YPM 12600 (LG XIX or XX), and AMNH 14535 (LG XVI or XVII), all regarded by Osborn (1936) as males, the ventral margin of the premaxilla is only slightly raised above the tooth row, with no distinct step. Two specimens, ISM 71.3.261 (the "Buesching mastodon," LG XIX) and a UMMP unnumbered specimen from Michigan (the "Fowler Center mastodon," LG XVI or XVII) are interesting intermediate stages, with a small but distinct step. In IPFW 1 (the "Routsong mastodon" LG XV), an apparent female, the premaxilla is elevated well above the tooth row, but instead of a sharp vertical step there is a more gradual transition from the tooth row to the premaxilla, showing that there is some variability in this feature. In WSC 18743 there is a distinct vertical step between the premaxilla and the maxillary tooth row (Fig. 1C). Even though this is a mature specimen (LG XXII), and interpreted as a male based on pelvic measurements, this condition is more similar to M. americanum specimens considered by Osborn (1936) to be females or juvenile males. As tusks continue to grow late in mastodon ontogeny, it might be expected that in males there would be some ontogenetic variability in the expression of this character. Indeed, referred specimen WSC 8817, which has an estimated age of 55 ± 4 AEY (LG XXVIII), has tusk alveoli that extend ventrally to just above the tooth row (Fig. 7). This suggests that M. pacificus may have reached its maximum tusk size later in ontogeny than M. americanum, although a full analysis of this possibility will require a detailed examination of tusk diameters, premaxilla morphology, and ontogenetic ages of both M. pacificus and M. americanum.
The left and right maxillary tooth rows in WSC 18743 and in referred specimens WSC 10829 (a possible male, based on tusk size) and WSC 8917 (a possible female, based on tusk size) are parallel or nearly so (Figs. 1B,11 and 18C). This is in contrast to  . 12), and UCMP 36684 (Fig. 14), in which the tooth rows are posteriorly convergent. Osborn (1936) noted that similar variability in M. americanum does not appear to correlate with sex or ontogeny. M. pacificus seems to exhibit similar variability, with both morphologies found across several ontogenetic ages, and at least the parallel tooth row form found in both males and females.

Mandible
The mandible of WSC 18743 (Fig. 2) is relatively elongate, with a smoothly downturned anterodorsal surface of the mandibular symphysis as is typical of Mammut. The ascending ramus is angled slightly posteriorly, with the anterior margin forming an angle of about 110 with the tooth row. The anterior margin of the ascending ramus extends vertically to from a somewhat recurved coronoid process. Posteriorly the ascending ramus expands transversely into the mandibular condyle, the dorsal margin of which is slightly higher than the apex of the coronoid process. There is a partially resorbed alveolus for the m1 on the right side only. There are no lower tusks, and no trace of alveoli for them. There are two pathologies on the right dentary, a bony growth near the anterior tip of uncertain origin and a deep groove with swollen margins on the lateral surface adjacent to the anterior end of m2. The more posterior pathology appears to have resulted from an impact injury. Fisher (2009) proposed that male specimens of M. americanum show frequent injuries that resulted from intraspecies combat, which is one possible source of the pathologies observed in WSC 18743. The entire mandible is asymmetrical, with the right tooth row shifted posteriorly relative to the left. Given that the degree of asymmetry is variable across the length of the dentary, it is interpreted as a response to the trauma suffered by the right dentary, rather than due to taphonomic processes. There is some variability in the shape of the ascending ramus among different specimens that has no direct correlation with age or body size.  anterior margins of the ascending rami are nearly vertical. There is also discernible variability in the angle between the dentaries in dorsal view, from nearly parallel to posteriorly divergent (Fig. 19). This divergence is most noticeable in the portion of the dentaries posterior to the tooth row and may become less pronounced with age as the mandible increases in length (compare,e.g.,Figs. 19A and 19M). However, this character may be subject to subtle taphonomic alteration, especially if the mandible is lying on its side when buried. Thirteen specimens of M. pacificus preserve the mandibular symphysis, and range from LG I to LG XXIX (WSC 18743, WSC 19730, WSC 8817, LACM 128927, LACM-HC 38, LACM-HC 87073, LACM-HC 475, LACM-HC 1631, LACM-P23 26389, UCMP 86140, UCMP 212944, UCMP RSF 0201, and SBMNH-VP-3342) (Fig. 19); all of these specimens lack mandibular tusks or any trace of tusk alveoli. Moreover, at the two California localities that have produced numerous mastodon remains (Rancho La Brea and Diamond Valley Lake), no isolated elements consistent with mandibular tusks have been recovered. This is a remarkably stable pattern when compared to M. americanum. While the geographic, temporal, sexual, and ontogenetic distribution of mandibular tusks in M. americanum is not fully understood (Laub, 1999(Laub, , 2009, there does appear to be a trend toward reduction of the mandibular tusks over time. Every specimen of M. americanum from the early Rancholabrean deposits at Ziegler Reservoir in Colorado has mandibular tusks, which are often quite large (Cherney, Fisher & Rountrey, 2017;Fisher et al., 2014). Mandibular tusks are also present in the late Rancholabrean AMNH 9951. Alveoli are present in PRI 49820, but are absent in YPM12600. Green (2006) noted that mandibular tusks were always present in Irvingtonian M. americanum (n = 11) from Florida, but only present 27% of the time in Rancholabrean individuals (n = 22). An exact binomial test of goodness-of-fit comparing the M. pacificus lower tusk occurrence rate (0/13) to the Florida Rancholabrean M. americanum occurrence rate (6/22) yielded a p-value of 0.026, indicating that M. pacificus does exhibit mandibular tusks with lower frequency than M. americanum based on these data. Unfortunately, no Irvingtonian mastodon specimen from California has a preserved mandibular symphysis.

Tusks
The complete right tusk and distal third of the left tusk of WSC 18743 are preserved (Figs. 1F and 1G). Where the tusks emerge from the premaxillae they are angled ventrally about 5 relative to the maxillary tooth row, and laterally about 20 from the midline. The right tusk spirals gently both medially and dorsally, so that the tips would have  influenced by environmental variables, evaluation of sexual dimorphism, and gender identification based on tusks must be approached with caution. That said, at least for M. americanum, at any given ontogenetic age females tend to have smaller tusks than males all (Fisher, 2008(Fisher, , 2009Smith, 2010), so it is possible to make a tentative gender identification provided tusk (or alveolus) size and Laws Group can be determined. Available data suggest that in M. americanum adult females with tusk circumferences >36 cm and males with circumferences <39 cm occurred rarely if at all (Fisher, 2008(Fisher, , 2009Smith, 2010). Remarkably, all of the tusks associated with maxillary teeth from Diamond Valley Lake appear to represent adult males, based on their sizes and ontogenetic ages. There are several smaller tusks, not associated with maxillary teeth that may represent either females or juvenile males, although we cannot rule out juvenile mammoths as a possibility. Referred specimen SDSNH 86541 ( Fig. 21A) appears to represent an adult female (LG XX), with very small (78 mm maximum diameter), straight tusks. While SDSNH 86541 is a particularly small individual, the tusk morphology is consistent with the larger WSC 3795 and WSC 3817. Another referred specimen, WSC 8917 ( Fig. 18), is also an apparent adult female (LG XVIII or XIX). While lacking tusks, the maximum diameter of the alveolus for the left tusk is approximately 71 mm.

Dentition
Like other crown proboscideans, mammutids exhibit horizontal tooth replacement, progressing through a total of six teeth in each quadrant. Although there is a considerable amount of variation in dental morphology, in both M. pacificus and M. americanum dP2/dp2 and dP3/dp3 are bilophodont (although in both taxa dP3/dp3 often develop a posterior shelf/partial third loph, as described in Green & Hulbert (2005) Tables 4-15. WSC 18743 includes both left and right M2/m2 and M3/m3. Both the upper and lower second molars are heavily worn with only the margins of the lophs remaining. The M3/m3s show lighter but still heavy wear, especially on the anterior lophs and lophids The cingulum is generally poorly developed on the M3/m3 of M. americanum as well, except the anterior margin of the tooth, but there are some individuals with a strongly developed cingulum (e.g., Fig. 24F). In M. pacificus teeth anterior to M3/m3 frequently have some development of a cingulum and more rugose enamel, although the expression of these features varies across individuals. All known M. pacificus teeth are more similar in texture to the "smooth variety" teeth of M. americanum associated with spruce/deciduous forests, rather than the "rugged variety" associated with pine-parkland habitats (King & Saunders, 1984). However, "smooth variety" M. americanum teeth tend to be larger than both "rugged variety" M. americanum teeth and M. pacificus teeth (King & Saunders, 1984). Only limited data (e.g., femoral measurements) provide body size estimates for M. pacificus, but available specimens suggest that tooth size may not correlate closely, if at all, with body size. For example, the m2 of WSC 18743 (an apparent male) has a length of 103.6 mm, while the m2 of SDSNH 86541 (Fig. 21C) (a presumed female) is only slightly longer at 104.6 mm. Yet the distal femur width of WSC 18743 is 288 mm, compared to only 203 mm in SDSNH 86541. Given the inability to determine gender from isolated teeth, comparative body size estimates between Mammut specimens should be approached with caution if based on isolated teeth.
The most remarkable feature of the dentition of M. pacificus is the narrowness of the crowns, particularly in the third molars (Figs. 24I-24P and 26; Tables 4 and 5; Table S1). The length:width (L:W) ratio of the m3 in WSC 18743 is 2.44 (left) and 2.46 (right). This pattern of narrow m3s is seen in multiple specimens of M. pacificus, with an average L:W ratio of 2.24 (N = 24; SD = 0.13; max = 2.44, min = 1.95). In contrast, the L:W ratio in M. americanum is only 1.91 (N = 121; SD = 0.11; max = 2.23, min = 1.63) (Figs. 26C and 26D; Table S2). While there is overlap in the L/W ranges between the taxa, the differences are significant (T-test value p < 0.001).
The same pattern is apparent in the M3s (Figs. 24A-24H), in which the M. pacificus average L:W ratio is 1.98 (N = 39; SD = 0.14; max = 2.33, min = 1.69) while the M. americanum average L:W ratio is 1.77 (N = 79; SD = 0.10; max = 1.95, min = 1.59). As with the m3s, these differences are significant (T-test value p < 0.001). The variation in loph count in M3 does affect L:W ratio in M. pacificus. In the 24 pentalophodont M3s L/W = 2.02, while in the 15 tetralophodont M3s L/W = 1.91 (T-test value p = 0.007). However, the frequent presence of a fifth loph alone is insufficient to explain the difference in L:W ratios between M. pacificus and M. americanum. Even when considering only   Table 5). The average length of M3 is significantly greater in M. americanum (174.68-166.94 mm, Mann-Whitney test p = 0.038; Table 4), apparently due to several very large M. americanum M3s (the longest M. americanum M3 is 17.5 mm longer than the longest M. pacificus specimen, while the shortest M. americanum specimen is only 0.3 mm longer than the shortest M. pacificus specimen) ( Table 4). The pattern of narrow crowns in M. pacificus is also apparent in Irvingtonian specimens. Four Irvingtonian m3s referred to M. pacificus, including two from Idaho and two from California, have L:W ratios ranging from 2.23 to 2.38 (average = 2.32), even more narrow than the Rancholabrean average (2.24) (Table S1). Only a single Irvingtonian M. pacificus M3 has been identified, for which L/W = 2.00. Green (2006) provided measurements of a number of Irvingtonian M. americanum specimens from Florida. Four m3s had an average L/W of 1.93, and four m3s had an average L/W of 1.77, very similar to the Rancholabrean M. americanum averages of 1.91 and 1.77 (Table S2). While the data are limited, it appears that the disparity in M3/m3 proportions between M. pacificus and M. americanum was present through the entire Pleistocene.
The narrow M3/m3 crown of M. pacificus relative to M. americanum is not as obvious in other tooth positions (Figs. 24Q-24LL and 27-31; Tables 6-15), although in some cases small sample size is an issue. Even with large sample sizes, M2 and M1 of M. pacificus do not significantly differ from those of M. americanum in L:W ratio (1.32-1.29 for M2, and 1.27 to 1.23 for M1) (Figs. 27B and 28B). Nevertheless, within the data set used in this study, for 11 out of 12 tooth positions, the highest L:W ratio measured was a California tooth attributed to M. pacificus. In the lone exception, dP3, there are only two known teeth from California, and one of these had the second-highest L:W ratio.

Vertebrae
WSC 18743 includes portions of seven vertebrae (Fig. 5), including the fifth cervical, two posterior thoracic, two lumbar, and two poorly preserved specimens that are probably additional posterior thoracics. Morphologically, these vertebrae do not differ in any substantial way from corresponding elements from M. americanum, and the key features indicated by Olsen (1972) and Hodgson et al. (2008a) as distinguishing Mammut from Mammuthus are all present. One of the lumbar vertebrae (Figs. 3G and 3H) is a close morphological match for the first lumbar of PRI 8829, a somewhat larger specimen of comparable ontogenetic age (LG XXI) (Hodgson et al., 2008a). Based on comparison to PRI 8829, both thoracic vertebrae (Figs. 3C-3F) appear to be posterior to the 13th thoracic (Hodgson et al., 2008a). Several other specimens include larger numbers of preserved vertebrae, particularly among the Diamond Valley Lake sample. WSC 7561, tentatively identified as an adult male (based on an associated tusk) includes the first 11 vertebrae in articulation (Fig. 20). The first four thoracic vertebrae have long, slender, and posteriorly directed spinous processes, increasing in height from the first to the fourth vertebra, that are expanded dorsally, a feature also seen in Mammut americanum. SDSNH 86541 (Figs. 21D-21F) and  LG XXIV, 45 ± 2 AEY Armitage Heights Orange California Figure 19C LACM-HC 475 Partial mandible with right dp2-dp4, and unerupted m1 LG II or III, 2-3 ± 0.5 AEY

Rancho La Brea
Los Angeles California Figure 19L LACM 128927 Partial mandible with left and right m1 and m2 LG XII; 18 ± 1 AEY  Uncertainties are calculated as the difference between the midpoint and either the upper or lower limit of the calibrated age range, whichever is greater (reported at the 95% confidence level; 2s). Multiple ages are reported when the probability of a calibrated age range exceeds 0.05. 3 Sample 14 C age or sample 14 C age plus uncertainty is beyond the limit of the IntCal13 dataset and therefore, cannot be calibrated (Reimer et al., 2013). 4 p, probability of the calibrated age falling within the reported range as calculated by CALIB.

Sacrum and pelvis
WSC 18743 includes a nearly complete pelvic girdle (Fig. 4). The pelvic elements are fully fused, consistent with the maturity of the specimen based on dental eruption and wear (Haynes, 1991). The sacrum consists of six fused vertebrae. A second nearly complete pelvic girdle from Diamond Valley Lake, WSC 9642 (Fig. 22), represents a younger individual than WSC 18743, with only partial fusion of the pelvic elements.   and USNM 8204. PRI 8829 includes five vertebrae in the sacrum, which is fused anteriorly to the third lumbar vertebra (Hodgson et al., 2008a). There is, however, some variation in the M. americanum sacral count. According to Woodman & Branstrator  (Lister & Stuart, 2010). Green (2006), Hodgson et al. (2008a), andFisher (2008) found that the ratio of the pelvic aperture width to the minimum width of the ilial shaft is a reliable indicator of sex in M. americanum, a relationship that was first recognized in mammoths by Lister (1996). Females have a ratio of greater than 2.6, while in males it is less than 2.6. In WSC 18743, this ratio is 2.29, indicating it is a male specimen, consistent with its large overall size and robust tusks. Likewise, WSC 9642 is also apparently a male, with a ratio of 2.45.

Femur
The only preserved limb element from WSC 18743 is the distal end of the left femur (Fig. 5). The distal femoral epiphysis is fully fused to the shaft of the femur, which is consistent with the tooth wear state and vertebral epiphyseal fusion in indicating a fully mature animal. The femur is noteworthy for its size. The maximum width across the distal end is 288 mm, far larger than any other California specimen. There are, however, several specimens of M. americanum that are larger in this dimension. For example, DMNH 1496 from Indiana has a distal width of 299 mm with a length of 1,133 mm, while DMNH 61338 from Colorado has a distal width of 323 mm, with a length of 1,180 mm.  from an adult, which is consistent with the wear on m2 and indicates LG XXII. Based on the presence of a small diameter tusk combined with the ontogenetic age, this specimen is interpreted as representing a female animal. Yet this femur has a distal width of only 203 mm, with a length of just 817 mm, remarkably small for an adult mastodon. Nor is this unique among specimens of M. pacificus. LACM-HC 1266 is not associated with any dental material, but it does exhibit complete epiphyseal fusion, with a distal width of 222 mm and a length of 820 mm. The nearest comparable M. americanum femur we have examined is G25656, a specimen from Carters Bog, Ohio, of an apparent young adult female (LG XII; 18 ± 1 AEY), with a distal femoral width of 220 mm and a length of 899 mm. WSC 9622, an apparent adult male M. pacificus (LG XX) with complete femoral epiphyseal fusion has a distal femoral width of 261 mm and length of 956 mm. This is only slightly larger than PRI 49618, an apparent adult female (LG XX-XXI; Hodgson et al., 2008b;Fisher, Beld & Rountrey, 2008) with a distal width of 240 mm and a length of 950 mm (Hodgson et al., 2008a). In fact, seven California femora (Fig. 23) from five different Rancholabrean localities spread across at least 35,000 years have an average maximum length of 918.8 mm, compared to the M. americanum average of 1,041.2 mm, based on nine Rancholabrean specimens, suggesting that femora of M. pacificus may have been shorter on average than those of M. americanum. A Mann-Whitney test on the length data showed a significant difference in the medians of the samples (U-value = 10.5, p = 0.050). In a slightly smaller subsample for which data are available (six M. pacificus and eight M. americanum), femora of M. pacificus femora tend to be more robust, with a length:midshaft diameter ratio of 6.16 compared to 6.75 in M. americanum (Mann-Whitney test U-value = 1, p < 0.001) (Fig. 25). published records from Boney Spring and Trolinger Spring (Saunders, 1977). The much larger sample examined in this study confirms this trend and demonstrates that the California/Idaho population is unique in this regard. The average L:W ratio of m3s of M. pacificus (2.24, SD = 0.13) is nearly three standard deviations above the average of M. americanum (1.91, SD = 0.13), and for M3s the M. pacificus average (1.98, SD = 0.14) is just over two standard deviations higher (1.77, SD = 0.10) (Tables 4 and 5).

DISCUSSION
We considered other explanations than species-level separation for the disparity in tooth proportions between California/Idaho and non-California populations of Mammut but ultimately rejected each one. For example, one possibility is that the narrow tooth morphology was an outlier in the normal range of Mammut americanum morphology that became dominant when certain environmental conditions were met for which this morphology was advantageous. Indeed, even with the most extreme tooth positions (M3 and m3), there are a small number of specimens of M. americanum in which the L:W ratio approaches the mean of M. pacificus (none actually reach or exceed the mean of M. pacificus). However, these occurrences are infrequent and do not exhibit any clustering; each occurrence is found in an area in which the narrow tooth is an isolated outlier in a population more typical of M. americanum. Moreover, the narrow tooth morphology observed in M. pacificus is found over a large geographic area encompassing an area ranging from the Pacific coast to at least 1,000 km inland, and at least 1,200 km north-south. These localities include low-lying coastal areas, inland valleys, and mountainous areas in northern California and Idaho. Indeed, allowing for the spatially discontinuous nature of known deposits, the only parts of California that appear to have been completely devoid of M. pacificus (or, indeed, mammutids of any species) in the Pleistocene are the Sonoran and Mojave Deserts. It is unlikely that a critical environmental condition would occur in all of these areas yet not be found anywhere else in North America. Moreover, the narrow tooth morphology was present in this region over a lengthy temporal span, a period of at least 190-135 ka years and, based on the limited number of Irvingtonian specimens, likely considerably longer. Given this, we consider it unlikely that Mammut pacificus is a locally adapted morph of M. americanum and instead conclude that they in fact represent different taxa. Another possibility that we considered and ultimately rejected is that the California/Idaho population arose via the isolation of an M. americanum subpopulation at a relatively late date. Such an isolated population of a widespread, morphologically diverse species can diverge through a Founder's Effect (Mayr, 1942), leading to an anomalously high occurrence of characters that only appear as outliers in the parent population, even with a In fact, we observe the opposite; the SD for length:width in M. pacificus is slightly greater than that in M. americanum (0.14-0.10 in M3, 0.13-0.11 in m3). Moreover, most California and Idaho teeth are not simply outliers within the range of M. americanum proportions, but in fact are completely outside the range of M. americanum specimens examined in this study.
Finally, the small number of Irvingtonian specimens available suggest that this morphological pattern has been present for at least 190-135 ka, and perhaps much longer. Irvingtonian M3s/m3s from California and Idaho show proportions consistent with Rancholabrean specimens M. pacificus, while Irvingtonian specimens from Florida are similar to Rancholabrean specimens of M. americanum.
A third potential explanation for the disparity in tooth proportions continentwide is that the narrow-toothed, western populations of Mammut lie within the range of morphological variability expected for a genus as cosmopolitan and morphologically variable as Mammut. However, statistical comparisons between disparate Mammut populations suggest that M. pacificus is a consistent and significant statistical outlier. No other geographic subpopulation examined in this study differs so dramatically from M. americanum as do the California and Idaho populations (Table 16). T-test values for California M3 and m3 L:W are p < 0.001 when compared to non-California specimens. These p-values hold when comparing California to other individual states with large samples (N 9) and normal sample distributions, including Florida and Missouri (M3) and Colorado,Indiana,and Louisiana/Mississippi (m3). No other pairing of these states has such a low p-value, ranging from 0.031 (Indiana-Missouri m3) to 0.909 (Colorado-Louisiana/Mississippi m3). Mammut americanum maintains remarkably consistent L:W ratios over an enormous area, further undermining the hypothesis that the M. pacificus morphology is a regional variation of a single highly variable species, while bolstering the idea that midwestern and eastern specimens really do belong to a single taxon, M. americanum.

Comparison of Mammut pacificus to other named species of Mammut
Mammut oregonense Hay, 1926, was based on an isolated left M2 (USNM 4911) of uncertain age from Baker County, Oregon. Measurements provided by Osborn (1936)  include a length of 111 mm and a width of 80 mm. The length of this tooth is large compared to California specimens, and rather small compared to non-California specimens, while falling well within the range of each. The L/W ratio based on these measurements, 1.38, is quite high, and in fact higher than the average for either M. pacificus or M. americanum, but again, within the range of each (Figs. 27A and 27B). Since M. pacificus was present in both northern California and southern Idaho, its presence in Oregon would not be surprising. However, the limited numbers of specimens from Washington, Yukon, and Alaska are generally morphologically similar to M. americanum and are grouped as such in this study (Figs. 26B,26D and 27D). Given the lack of diagnostic value of M2 (Table 6), M. oregonense is here regarded as a nomen dubium. Mammut furlongi Shotwell & Russell 1963 is based upon a partial mandible with m1-m3 and a referred M3 from an assemblage dating to the Clarendonian NALMA in Oregon. While the mandibular symphysis is incomplete, there is no evidence of lower tusks. The teeth are small and quite narrow; the left m3 has an L:W ratio of 2.25, approximately the same as the average of M. pacificus. The m3 is fairly choerodont compared to both M. pacificus and M. americanum, but this feature tends to be quite variable. The referred M3 has only three lophs and a large talon at the posterior end. In this respect it is more similar to Zygolophodon cf. proavus from the Barstovian NALMA of California (Lofgren & Anand, 2011) than to Mammut; a detailed comparative study is needed to determine if M. furlongi should be referred to Zygolophodon.
Mammut nevadanus (Stock, 1936), originally placed in Pliomastodon, is based on a partial cranium with M2s, unerupted M3s, and the right tusk from the Thousand Creek beds of Humboldt County, Nevada. These beds are thought to be late Miocene to early Pliocene (Merriam, 1910) and the fauna dates to the Hemphillian NALMA (Prothero & Davis, 2008). This specimen has relatively small M3s that have four complete lophs rather than three as in M. furlongi. The M3s are slightly more narrow than the average for M. pacificus (L:W = 2.03), but well within the M. pacificus range, and more narrow than any M. americanum specimen. The single, slender tusk is straight and deflected ventrally in lateral view and curved gently medially in ventral view. This differs from both M. pacificus and M. americanum, which have large, upturned tusks in males, and upturned or straight but anteriorly directed tusks in females.
Mammut raki (Frick, 1933) is based on a partial mandible from Blancan NALMA deposits in Sierra County, New Mexico (Lucas & Morgan, 1999). This specimen has a narrow m3 in which the L:W ratio (2.30) is within the range of but greater than the average for M. pacificus (2.25). The holotype specimen of M. raki also possesses well-developed mandibular tusks, distinguishing it from M. pacificus. Rancholabrean Mammut from New Mexico have tooth proportions that generally fall within the range of M. americanum rather than M. pacificus (average m3 L:W = 1.87, n = 3).
Mammut cosoensis (Schultz, 1937) is based on a holotype partial cranium and a referred partial mandible, juvenile palate, and isolated teeth from the Coso Formation in Inyo County, California. The Coso Formation is mid to late Pliocene based on K-Ar dating of sanidine, biotite, and plagioclase as well as whole-rock samples of igneous rocks bracketing and intruding the Coso Formation sediments (Bacon et al., 1982), with a fauna that is probably Blancan. Schultz (1937) originally referred this species to Pliomastodon? as he considered it the "most advanced species of Pliomastodon yet described" (p.105). Shoshani & Tassy (1996) referred this species, and in fact the entire genus Pliomastodon, to Mammut.
The holotype cranium of M. cosoensis (LACM 1719) includes the upper second and third molars, and is approximately LG XXII, 39 ± 2 AEY. The proximal ends of the tusks are preserved, and are upturned and appear to have been rather long, with a preserved length of approximately 740 mm and a basal diameter of 78 mm. In comparison, SDSNH 86541, a presumed female M. pacificus also has a basal diameter of 78 mm, with a complete tusk length of only 780 mm, while the holotype M. pacificus (WSC 18743, a male) has a basal tusk diameter of 186 mm, with a length of 1,996 mm. Both WSC 18743 and SDSNH 86541 are also LG XXII. But while the tusk morphology of LACM 1719 is consistent with a female, the premaxilla morphology is more similar to a male specimen. The ventral edges of the tusks lie just dorsal to the tooth rows in lateral view, similar to adult male specimens of M. americanum and M. pacificus. It appears that the vertical separation between the tusks and the upper teeth is lost in adult males as they reach their maximum tusk diameter, but in adult females with their smaller tusks the spacing is maintained (e.g., in M. americanum AMNH 14292 as figured in Osborn (1936)). It is also noteworthy that in WSC 18743, a male in the same Laws Group with large tusks, there is still a substantial gap between the tusk and tooth row. Whether LACM 1719 represents a male or a female, it differed in some degree from M. americanum and M. pacificus in cranial geometry associated with tusk growth. While Schultz (1937) stated that M. cosoensis referred mandible LACM 1720 showed no indication of lower tusks, examination of both Schultz's figures and the specimen itself show that the mandible, including the symphysis, is largely reconstructed with plaster, making the status of lower tusks unclear.
Both the upper and lower third molars of M. cosoensis have an enamel structure at the anterolabial corner that appears to be a hyper-enlarged cingulum. This is most strongly developed in LACM 1719 and LACM 1720, which show wear on the occlusal surface in spite of the relatively early wear on these teeth, but is only weakly developed in referred M3 LACM 2015. No specimen of M. pacificus or M. americanum examined in this study exhibits such strong development of the anterior cingulum.
While the sample size is small, the L:W ratios of M. cosoensis are uniformly high, and close to the averages for M. pacificus. Two M3s have L:W ratios near the middle of the range of M. pacificus (2.10, 1.94), and one of these is outside the range for M. americanum. The M3s have fives lophs, which is common (but not universal) in M. pacificus and rare in M. americanum. The two known m3s have L:W ratios in the lower part of the range for M. pacificus, and the upper part of the range for M. americanum (2.21, 2.02). In almost every instance, the actual length and width measurements are smaller than the average for M. pacificus, and in most cases substantially so. Mammut raki and Mammut cosoensis represent narrow-toothed mammutid taxa from Blancan deposits in New Mexico and California, respectively. Mammut nevadanus from the Hemphillian of Nevada, and Mammut furlongi from the Clarendonian of Oregon are also narrow-toothed forms. While all four of these taxa are based on limited material that makes comparison with other taxa difficult, if the narrow-toothed taxa are indeed monophyletic they suggest an early divergence of eastern and western mammutids and a wider western distribution of narrow-toothed mammutids in the Miocene/Pliocene.
Mammut pacificus is not known from Inyo County, California (the only known locality of M. cosoensis), and Pleistocene specimens of Mammut from New Mexico have broad teeth similar to M. americanum rather than narrow teeth like M. raki. Tobien (1996) showed that over time mammutid m3 length:width ratios remain relatively constant, but mammutid m3s are distinctly wider than the m3s of gomphotheriids. The m3 L:W ratios of Mammut pacificus are intermediate between those of gomphotheriids and other mammutids, including Asiatic specimens (Fig. 32). If the apparent scarcity of narrow-toothed mammutids in Eurasia suggested by Tobien's (1996) data is an accurate reflection of the Eurasian fauna, they suggest that this morphology first evolved in North America. The presence of multiple lineages of North American mammutids should be taken into account during the much-needed reassessment of North American mammutids including Zygolophodon, Miomastodon, and "Pliomastodon."

Biogeography
Like most other animals, proboscidean morphotypes tend to cluster in distinct geographic areas. The most extreme examples are various insular dwarf species of elephantids and stegodontids isolated on islands (Van Der Greer et al., 2016), but even widely dispersed mainland taxa tend to have distinct ranges (Roca et al., 2001;Fisher, 2009). In North America, perhaps the best-known example among proboscideans is the northern distribution of Mammuthus primigenius, which contrasts with the more southern range of Mammuthus columbi. Recent studies have suggested that there was extensive hybridization between these species (Enk et al., 2011(Enk et al., , 2016. Indeed, it seems that among the Elephantidae, interspecific breeding is possible and perhaps relatively commonplace (Roca et al., 2001;Palkopoulou et al., 2018). Yet regardless of apparently frequent interbreeding, proboscidean populations seem to have for the most part remained Figure 32 Comparison of m3 proportions of mammutids and gomphotheriids. Graph of length vs. width of mammutids and gomphotheriids, modified from Tobien (1996) morphologically stable and distinctive. While much work remains to be done on mastodon ecology, research to date suggests that Mammut may have had fairly stringent ecological preferences in terms of water availability and vegetation when compared to elephantids (Jackson & Whitehead, 1986;McAndrews & Jackson, 1988;Whitehead et al., 1982), and so it is perhaps not surprising that Mammut could diverge morphologically in geographically isolated areas even if mammutids were capable of the relatively high mobility observed in some other proboscideans.
Mammut specimens from the Pacific Northwest present an interesting anomaly. There are very few Pleistocene mammutid remains from Oregon, and no easily identified elements such as third molars. In this study, Oregon specimens have been included as M. americanum, but it is possible that these remains represent M. pacificus, especially given the proximity of M. pacificus in northern California.
There are small numbers of Mammut known from Washington, Alaska, and the Yukon. While teeth from these specimens are quite small, the L:W ratios fall within the expected range for M. americanum and are included as such here. The specimens from Alaska and the Yukon all appear to be no younger than the early Rancholabrean, and it seems that mammutids left this region prior to the glacial maximum (Zazula et al., 2014). All three referred specimens of M. pacificus from Idaho also date from prior to the last glacial maximum. The Pacific Northwest may have represented a conspecific region for both taxa, or possibly the ranges fluctuated with each taxon occupying the region at different times. A larger sample and better chronology will be necessary to determine the exact nature of northwestern mammutid interactions.
The unexpected concentration of Mammut americanum at Ziegler Reservoir in Colorado is located more than 1,300 km east of the Pacific Coast, and over 600 km southeast of the nearest occurrence of M. pacificus in Idaho. In the vast region between Ziegler Reservoir and the easternmost occurrences of M. pacificus, there are only a handful of known specimens of Mammut. The genus is relatively rare throughout Arizona, New Mexico, Utah, Nevada, Wyoming, and Montana, with only a few if any reported specimens from each state. As in Colorado, the only records from Utah are from high elevations (Miller, 1987). This stands in stark contrast to the common prevalence of other Plio-Pleistocene proboscideans in the American Southwest and Mexico, including Mammuthus and gomphotheriids such as Cuvieronius, Rhynchotherium, and Stegomastodon. McDonald et al. (2010) suggested that limited, elevation-controlled distribution of coniferous forests in the Rocky Mountains may have limited the size of mastodon populations in this region; this may have been true of the basin-and-range topography as well.
Mammut pacificus was widespread in California west of the Sierra Nevada, and (at least in the Irvingtonian and early Rancholabrean) was present as far northeast as southern Idaho, but it was apparently absent from the Sonoran and Mojave Deserts (Fig. 33), including from heavily sampled localities such as Tule Springs in Nevada (Scott, Springer & Sagebiel, 2017). These deserts, along with the high, steep, and at times glaciated Sierra Nevada and the possible patchiness of appropriate habitats in the Basin and Range and Rocky Mountains, may have served as effective geographic barriers to mastodon dispersal. It is interesting to note that there are some indications of similar distribution patterns in Mammuthus. Widga et al. (2017) found that Mammuthus exilis from the California Channel Islands and southern California specimens of M. columbi had dental  Table S2. Note that while there are many additional M. americanum localities that were not included in this study and that are not indicated on the map, there are no known M. americanum localities in California. The M. americanum locality in Oregon is a non-diagnostic specimen that was included as M. americanum in this study, but that could represent M. pacificus.
Full-size  DOI: 10.7717/peerj.6614/ fig-33 characteristics in common with each other, but that these two populations differed from M. primigenius, M. jeffersonii, M. trogontherii, and all the other regional populations of M. columbi in their dataset. The sparse distribution of high growing, generally woody vegetation in desert environments may have limited the abundance of browsing specialists such as Mammut (but see Green, DeSantis & Smith, 2017;Smith & DeSantis, 2018), while xeric shrubs, succulents and occasional grasslands could have supported generalists such as Mammuthus and Cuvieronius. The limited abundance of high-growing woody vegetation in the high mountains and deserts of Mexico may also explain the limited reports of Mammut remains from Central America (Polaco et al., 2001) and the inability for Mammut to have traveled into South America via the Panamanian Land Bridge as did at least two species of gomphotheres (Mothé et al., 2017). Thus, while Mammut was a prominent member of the Pleistocene fauna in the eastern US, with Mammut americanum present from the Rocky Mountains east to the Atlantic Coast, it seems that there also existed a large section of North America where there were few if any mastodons, between Mammut americanum in the east and Mammut pacificus from southern Idaho and the Sierra Nevada west to the Pacific Coast.

CONCLUSIONS
This study describes and formally names a new species of mastodon from the Pleistocene of western North America, Mammut pacificus sp. nov. This new taxon is recognized by specimens throughout California and from two localities in southern Idaho. With this effort, we have documented the presence of two species of Mammut in Pleistocene North America. The newly described Mammut pacificus, characterized by narrow molars, six sacral vertebrae, a femur with a proportionally greater mid-shaft diameter, and a lack of mandibular tusks, is recognized from the Sierra Nevada west to the Pacific Coast and into southern Idaho. The long-recognized Mammut americanum, characterized by wide molars, five sacral vertebrae, a femur with a proportionally small mid-shaft diameter, and occasional mandibular tusks, is known from Alaska south through the Rocky Mountains, and east to the Atlantic Coast. There is some degree of morphological overlap between the two species of Mammut, including sparse reports of other Mammut material from the western US that have led to questionable specific assignments. We suggest the need for a continental-wide reassessment of the evolution, biogeography, and phylogenetics of the Mammutidae family to elucidate these relationships.