Paleobiology of a large mammal community from the late Pleistocene of Sonora, Mexico

Abstract A paleontological deposit near San Clemente de Térapa represents one of the very few Rancholabrean North American Land Mammal Age sites within Sonora, Mexico. During that time, grasslands were common, and the climate included cooler and drier summers and wetter winters than currently experienced in northern Mexico. Here, we demonstrate restructuring in the mammalian community associated with environmental change over the past 40,000 years at Térapa. The fossil community has a similar number of carnivores and herbivores whereas the modern community consists mostly of carnivores. There was also a 97% decrease in mean body size (from 289 kg to 9 kg) because of the loss of megafauna. We further provide an updated review of ungulates and carnivores, recognizing two distinct morphotypes of Equus, including E. scotti and a slighter species; as well as Platygonus compressus; Camelops hesternus; Canis dirus; and Lynx rufus; and the first regional records of Palaeolama mirifica, Procyon lotor, and Smilodon cf. S. fatalis. The Térapa mammals presented here provide a more comprehensive understanding of the faunal community restructuring that occurred in northern Mexico from the late Pleistocene to present day, indicating further potential biodiversity loss with continued warming and drying of the region.


INTRODUCTION
Conservation paleobiology aims to use knowledge of the past to make informed predictions about the future of Earth's threatened biodiversity (Dietl and Flessa, 2011, Dietl et al., 2015, Barnosky et al., 2017. These studies provide a baseline of what community structures looked like in the past, how extant communities might respond to impending changes, and how we should approach conservation. For example, in the Greater Everglades Ecosystem, paleoecological data from pollen and invertebrates have been used to establish pre-anthropogenic trends and cycles, predict future changes, and identify restoration targets for anticipatory resource management efforts (Wingard et al., 2017). Other similar work has documented the presence of Bison in the Grand Canyon region (Martin et al., 2017), inspired interagency conservation and management plans for Bison today (Plumb and McMullen, 2018), and encouraged restoration of threatened plant species in Hawaii (Burney and Burney, 2007).
Increasingly xeric landscapes, such as those in the southwest US and northwest Mexico, are challenging for natural resource management (Sayre et al., 2013). Integration of fossil sites with climate patterns may reveal potential future areas for faunal refugia and migratory corridors as fauna must shift their geographic ranges or risk extinction (Blois and Hadly, 2009). Additionally, studying past faunal change may reveal trends in restructuring of faunal communities, which may enhance anticipatory natural resource management decisions and strategies. In particular, large animals and plants are disproportionately important for ecosystem function (Enquist et al., 2020), and changes in these communities can indicate environmental change (Blois and Hadly, 2009).
Paleobiology of mammal communities has commonly focused on body size and diet because they can be inferred from fossil remains. Smith et al. (2018) demonstrated a widespread 95% decline in mean body mass of North American mammal communities (e.g., from 100 kg to 5 kg) associated with the loss of the Pleistocene megafauna. Stegner and Holmes (2013) investigated mammalian community structure over 16 million years and demonstrated static diversity of dietary functional groups except in times of major environmental pressures. In the Pleistocene of Mexico, mammal communities were more taxonomically diverse, with more large-bodied taxa and a more similar ratio between herbivores and carnivores as compared to modern communities .
Overall, our knowledge of the fossil mammals of Mexico is biased toward taxa of larger sizes and younger geological ages, possibly because of collection methods (Montellano-Ballesteros and Jiménez-Hidalgo, 2006;. Of 64 Neogene sites in Sonora, Mexico, only three sites have detailed faunal records: El Golfo, San Clemente de Térapa (hereon, Térapa), and Rancho la Brisca . El Golfo is the only known Irvingtonian LMA site in northern Mexico (Lindsay, 1984;Croxen et al., 2007;Sussman et al., 2016). Rancho La Brisca is from the Rancholabrean LMA but has a slightly different fauna than is preserved at Térapa and other sites in the region because its record is biased toward small and medium fish, amphibians, and reptiles (Van Devender et al., 1985). Additional small and isolated records in the region, including La Botana, Llano Prieto, and Chinobampo, consist of mostly large mammals, such as Bison, Equus, Mammuthus, Glyptotherium, and Camelops Cruz-y-Cruz et al., 2018). West of Térapa, La Playa has fossils of Bison, Platygonus, Mammuthus, Equus, camelids, and other large mammals, but also has the first record of Cynomys in Sonora, which suggests more small mammals may be found at other sites .
Located in Sonora, Mexico, Térapa (29°41 ′ N, 109°39 ′ W; 605 m elevation; Fig. 1) provides an excellent case study of faunal response to environmental and climatic changes in an area vulnerable to continued alteration because of its geographic location in an ecotone between the northern Nearctic climate and the southern Neotropical climate. Modern (i.e., interglacial) Sonoran and Chihuahuan deserts are typified by scrubland ecosystems with warm tropical-subtropical climates, and at present, this area is situated on the margin of the Madrean Archipelago and the Sinaloa-Sonora Hills and Canyons (Morrone et al., 2017). However, when the fossil-bearing sediments were deposited in the Térapa basin, western Sonora was more temperate, with cooler, drier summers and wetter winters than are seen today (Nunez et al., 2010), and was typified by pinyon-juniper-oak woodlands (Van Devender, 2007). Stable isotope analyses of carbon and oxygen from ungulate teeth (Nunez et al., 2010) and ostrocodes (Bright et al., 2016) recovered at Térapa suggest that, at least in the valley, marsh and grasslands were likely present at the time of deposition. Previous studies of Térapa avifauna (Steadman and Mead, 2010;Oswald and Steadman, 2011), crocodilians (Mead et al., 2006), bats and shrews (Czaplewski et al., 2014), and ostracodes (Bright et al., 2016) indicate that the environment was an elevational mosaic of temperate to tropical-subtropical marsh and savanna/grassland with a slow-moving freshwater stream flowing from north of Térapa down to the Río Moctezuma and Río Yaqui to the Gulf of California.
Discovered in 2000, late Pleistocene (MIS/OIS 3) faunal deposits at Térapa include fossils of more than 60 taxa (Mead et al., 2006). A preliminary faunal list is provided in Mead et al. (2006), and an updated faunal list is provided in Supplemental Table S1. Recent publications have described Mammuthus and Cuvieronius (Mead et al., 2019), Glyptotherium cylindricum and Pampatherium cf. P. mexicanum (Mead et al., 2007), two shrews and six bats (Czaplewski et al., 2014), and 39 species of birds (Steadman and Mead, 2010;Oswald and Steadman, 2011). With Bison present throughout the stratigraphic profile, Térapa represents one of the few Rancholabrean North American Land Mammal Age (LMA) sites studied in detail in Sonora (Bell et al., 2004;Mead et al., 2006). However, the Holarctic genus Bison did not arrive in Mexico until much later than in the northern part of the continent. This biochronological delay complicates assigning LMA ages to Mexican sites, and also illustrates the need to further explore Mexico for fossil sites that can provide insight into North American fauna during the Pleistocene.
Fossils found at Térapa were deposited where the Tonibabi basalt flow dammed and diverted the Río Moctezuma and created a shallow lake at the northern extent of the Sierra Madre Occidental (Mead et al., 2006). Río Moctezuma sediments beneath the basalt dam date to 42.9 ± 3.3 ka using infrared stimulated luminescence, providing a maximum age estimate for the fossils (Bright et al., 2010). Also using infrared stimulated luminescence, sediments 1 m above the basalt date to 40.2 ± 3.2 ka (Bright et al., 2010). Using radiocarbon dating, charcoal near the top of the 11-m-thick sequence of medium-to fine-grained sediments dates to 41.7 ± 1.0 cal ka BP (Bright et al., 2010). Additional dates throughout the sequence using amino-acid racemization on ostracodes and radiocarbon analysis on a bivalve confirm deposition between 40 ka and 43 ka (Bright et al. 2010). Because of Térapa's geographic location and diverse fossil record at a time of extensive environmental change, ongoing research on this fauna will contribute to a better understanding of the late Pleistocene in northwestern Mexico.
Here, we use the fossil record at Térapa to investigate changes in large mammal community structure in response to environmental change. To do this, we explore community structure, as represented by feeding strategy and body size, by describing Térapa specimens and their traits, and discussing their geographic distributions in the North American desert region. We hypothesize that changes in this site's large mammal community structure at the end of the Pleistocene are because of a loss of herbivores and large-bodied taxa. We expect that the fauna described in this paper will help to further refine paleoecological interpretations as well as provide a long-term record of faunal change in a region that is vulnerable to ongoing climate change.

METHODS
Taxa described here were presented by Mead et al. (2006) as: Equus, Tapirus, cf. Platygonus, Camelops-sized camelid, Hemiauchenia-sized camelid, Canis dirus, Lynx rufus, and Procyon. Mead et al. (2006) also included cf. Odocoileus, Capromeryx, Stockoceras, and Bison on the preliminary list of fauna, but these additional artiodactyl specimens have been excluded from this analysis because they either require additional detailed revision or are part of another review.

Systematic paleontology
To determine taxonomic assignments, we primarily used comparative reference specimens from the Florida Museum of Natural History (FLMNH) and the East Tennessee Museum of Natural History (ETMNH). When possible, linear measurements and the related citations are provided within the taxonomic descriptions. Geographic occurrences were determined using The Paleobiology Database (https://paleobiodb.org) and literature. Records were downloaded from The Paleobiology Database on 27 July 2017 for: Taxonomy = "Equus, Palaeolama, Camelops, Platygonus, Canis, Lynx, Procyon, Smilodon," Time = "Rancholabrean," and Location = "North America." Species identifications for equid post-crania were determined using a quadratic discriminant analysis implemented in the R statistical computing environment using the package MASS (R Core Team, 2016). Training data used in the analyses are from Sertich et al. (2014) and McHorse et al. (2016). For analysis of the phalanx, training data included Equus occidentalis, E. conversidens, E. lambei, E. scotti, and a northwestern stilt-legged taxon. For analysis of the metacarpal, training data included Equus complicatus, E. conversidens, E. occidentalis, and E. scotti. R code for this analysis is given in Supplementary Materials.

Community structure
To investigate how community structure has changed at Térapa through time, the fossil community of Perissodactyla, select Artiodactyla, and Carnivora presented here was compared to the same three orders from the modern community. Modern mammal range maps were downloaded from the IUCN (2018). Range maps were sampled at the geographic point of Térapa to assemble a faunal list. Taxa were limited to those that are extant and either native or reintroduced. Antilocapridae, Bovidae, and Cervidae were excluded from both communities because, although they are known from Térapa (Mead et al., 2006), they are not within the scope of the discussion within this manuscript.
Feeding strategies were obtained from Paleobiology Database (https://paleobiodb.org) on 10 August 2020 by searching for taxonomic names. Strategies were grouped into herbivore, herbivorous omnivore, carnivorous omnivore, and carnivore. Body mass (BM; kg) averages were extracted from the MOM database (Smith et al., 2003) for each species. Mass estimates include males and females across distribution ranges and are not precise values for the southern geographic location of Térapa. However, they provide a measure to estimate temporal changes in the fauna.

RESULTS
From the mammalian fauna found at Térapa, we provide descriptions for eight species in seven genera, including Equus, Platygonus, Camelops, Palaeolama, Canis, Procyon, Lynx, and Smilodon. Six of these genera appeared on the preliminary faunal list of Mead et al. (2006), and two of these genera (Palaeolama and Smilodon) are new additions to the known faunal community. Specimens from Térapa are housed temporarily at The Mammoth Site, Hot Springs, South Dakota, USA, and are curated with numbers prefixed with TERA. Specimens will be permanently housed at the Instituto Nacional de Antropología e Historia in Sonora, Mexico.
Description: The left metacarpal (TERA 313) is broken transversely across the diaphysis, but the distal end is complete ( Fig. 2A). The left partial first phalanx (TERA 320) is missing the lateral, distal portion (Fig. 2B). The metacarpal and phalanx articulate. The second phalanx is complete (TERA 319; Fig. 2C) and articulates with a sesamoid. The first and second phalanges do not articulate.
Identification: In quadratic discriminant analyses, the second phalanx (TERA 319) was identified to species level with 87.5% confidence, and the distal metacarpal (TERA 313) was identified with 100% confidence.
Remarks: Equus is widespread across the US and Mexico during the Rancholabrean , including in Sonora (Cruz-y-Cruz et al., 2018). Taxonomy of Equus introduces complex questions surrounding species identifications. Equus scotti is a stout-legged horse that has been documented across the North American desert region during the Rancholabrean (Harris, 2014), including in the Tule Springs local fauna (Scott et al., 2017). Recent efforts to revise equid taxonomy have considered E. scotti to be synonymous with E. mexicanus (see Alberdi et al., 2014) and E. excelsus (see Priego-Vargas et al., 2017), but by convention we use E. scotti to be consistent with the data sources used in the analyses. Previous work identified E. excelsus at Térapa (Carranza-Castañeda and Roldán-Quintana, 2007), and it is likely the same as our E. scotti. However, equid taxonomy needs to be thoroughly and formally evaluated before these issues can be confidently resolved. Identification: We hypothesize that the teeth are from the same species as the postcranial elements previously identified as Equus scotti because of their large size. More specific identification is difficult with isolated teeth because they lack a majority of diagnostic characters (Famoso and Davis, 2014). Description: The postcranial elements are slight compared to extant Equus caballus and the Térapa specimens conferred to Equus scotti, and these teeth ( Fig. 3E-H) are noticeably smaller and have less complex enamel patterns than those assigned to E. cf. E. scotti (Fig. 3A-D).
Identification: It is unknown if the post-cranial elements are of the same species as the dentition. At this time, there are no morphological features to confirm species designations.
Remarks: Previous work on different specimens also identified E. conversidens at Térapa (Carranza-Castañeda and Roldán-Quintana, 2007), and, because of the smaller size, it is possible that our Equus sp. refers to the same taxon. Occurrence of both E. scotti and a smaller-sized horse is common at Rancholabrean sites, and the smaller horse is often identified as E. conversidens (Harris, 2014). Again, most isolated teeth can only be identified to genus because of the lack of diagnostic characters (Famoso and Davis, 2014). Mead et al. (2006) listed Tapirus sp. among the taxa found at Térapa based on a mandibular symphysis (TERA 294; Fig. 4A). However, this specimen is now identified as Equus sp. Depth of the mandible suggests hypsodont teeth as in Equus. The narrow intermandibular space extends anterior to the second premolar as in Equus. In Tapirus, this space is closed at the anterior end of the second premolar. The mandibular foramen is located well anterior to the tooth row as in Equus. In Tapirus, the mandibular foramen is inferior to the anterior second premolar. Description: The molar fragments (TERA 167) are hypsodont and zygodont (Fig. 5A). The deciduous tooth (TERA 280) is complete and more bunodont than the molar fragments (Fig. 5B). The canine is complete (TERA 281) and has an anterior occlusal surface (Fig. 5C).
Identification: This specimen was initially identified as cf. Platygonus by Mead et al. (2006). Comparisons with fossil material at FLMNH suggest that these teeth are Platygonus because the cusps are more zygolophodont, as in Platygonus, rather than bunodont as in Mylohyus. Because Platygonus is monotypic in the middle and late Rancholabrean (Kurtén and Anderson, 1980;Wright, 1998), these specimens are assigned to P. compressus.
Remarks: Platygonus compressus is known from the Rancholabrean of Arizona (Murray et al., 2005), New Mexico, and eastern Texas, as well as the Central Plateau and Trans-Mexican Volcanic Belt . In Sonora, Platygonus sp. is known from La Playa and Bajimari  and P. cf. P. vetus is known from El Golfo (Croxen et al., 2007). Térapa provides the first record of P. compressus in Sonora, but it is not unexpected. Description: The partial distal phalanx (TERA 279) has splayed ventral trochlea (Fig. 6A). This specimen is not a metapodial because of the lack of a condylar keel (Zazula et al., 2016). The mandibular fragment (TERA 278) has the roots of large selenodont molars (Fig. 6B). At its base, the m2 measures 43.28 mm.
Identification: These specimens were initially identified as "Camelops-sized" by Mead et al. (2006). The m2 measurement is within the range of Camelops provided by Honey et al. (1998) and Baskin and Thomas (2016). The mandible is also substantially larger than comparative material of Palaeolama and Hemiauchenia. The phalanx was also compared to phalanges of Palaeolama and Hemiauchenia, but it is considerably larger than specimens within both genera. Camelops is the only other Rancholabrean camelid, and it was large enough to have mandibles and phalanges of the size found at Térapa (Baskin and Thomas, 2016). The most recent review of Camelops described two species: C. hesternus in the Rancholabrean and C. minidokae in the Irvingtonian (Baskin and Thomas, 2016). Because Camelops is considered monotypic in the Rancholabrean and the specimens are consistent with the morphology, these specimens are assigned to C. hesternus.   Remarks: Camelops hesternus is widespread across the US and northern Mexico , including in Sonora (Cruz-y-Cruz et al., 2018). In the southern US, C. hesternus is known from Rancho La Brea (Jones and DeSantis, 2017), Diamond Valley Lake (Springer et al., 2009), Tule Springs (Scott et al., 2017), and throughout Arizona  and New Mexico (Harris, 2014).
Identification: This specimen was initially identified as "Hemiauchenia-sized" by Mead et al. (2006). The groove on the p4 is indicative of the 'complex infolding' seen in Palaeolama (Fig. 6C; Kurtén and Anderson, 1980;Honey et al., 1998, p. 454). Both m2 measurements are within the range of Palaeolama provided by Honey et al. (1998). Because Palaeolama is monotypic in the Rancholabrean of North America (Honey et al., 1998) and the specimens fit the morphology, these specimens are assigned to P. mirifica.
Remarks: Palaeolama mirifica is found at Rancholabrean sites in Florida, California, and Texas (Kurtén and Anderson, 1980), South Carolina (Sanders, 2002), Costa Rica (Pérez, 2011), and in the Mexican state of Puebla (Bravo-Cuevas and Jiménez-Hidalgo, 2015). However, it has not been documented in Arizona . Palaeolama sp. is documented at Irvingtonian sites at El Golfo in northwestern Sonora (Croxen et al., 2007) and Rio Tomayate in El Salvador (Cisneros, 2005), and from the Rancholabrean of Guatemala (Dávila et al., 2019). Térapa is the first Rancholabrean occurrence of P. mirifica in northwestern Mexico.
Description: The left mandible with c1-m2 (TERA 154) and the distal humerus (TERA 155) were previously described in detail (Hodnett et al., 2009); the remaining specimens (TERA 450) are described here. All teeth in the maxilla and both mandibles are in an advanced stage of wear. The left maxilla is articulated with the left jugal, and the P3 is broken between the anterior and posterior roots (Fig. 7A, B). The left mandible is missing the coronoid process and angular process, but the condyle is complete (Fig. 7C). There is no evidence of an alveolus for a p1 on the mandible. The anterior mental foramen is inferior to the anterior p2, and the posterior mental foramen is inferior to the posterior p3. The right mandible is broken between the p4 and m1 and along the inferior masseter fossa (Fig. 7D). Only the anterior root of the m2 is present and the alveolus for a singlerooted m3 is broken. Cranial and dental measurements suggest a larger than average C. dirus (Fig. 8, Supplemental Table S2; Tedford et al., 2009).
Identification: Mead et al. (2006) initially identified Canis dirus at Térapa, and Hodnett et al. (2009) agreed for TERA 154 and TERA 155. Measurements on the left mandible indicate that this specimen is too large to be a different Rancholabrean-age canid, such as C. latrans or C. lupus (Fig. 8). Canis armbrusteri was also a large canid in the late Irvingtonian (Kurtén and Anderson, 1980;Harris, 2014) and is thought to have given rise to C. dirus (Tedford et al., 2009). The Térapa teeth are too worn to examine the cusp patterns, but the upper molars have reduced labial cingula as in C. dirus rather than C. armbrusteri (Tedford et al., 2009).
Remarks: Canis dirus is considered "one of the most common mammalian species in the Rancholabrean" and is found across the US and Mexico (Kurtén and Anderson, 1980, p. 171;Mead et al., 2005;Harris,

Family Procyonidae Procyon Storr, 1780
Procyon lotor (Linnaeus, 1758) Material: Left calcaneum (TERA 453). Description: The calcaneum (TERA 453) is complete (Fig. 9A). There is no tubercle present on the trochlear process, which has a minimal groove, and the calcaneum does not have an accessory surface between the anterior articular surface and the cuboid facet. The latter feature has a point on its dorsal edge. The greatest length of the calcaneum is 28.03 mm and the transverse breadth of the sustentaculum is 14.49 mm.
Identification: The calcaneum was previously identified as Procyon sp. (Mead et al., 2006). Following descriptions provided by Stains (1973; lack of a knob on the trochlear process), the calcaneum is now referred to P. lotor. Within Procyon, the calcaneum is assigned to P. lotor instead of P. cancrivorous because of the minimal trochlear groove, the cuboid facet, and the lack of an accessory surface between the anterior articular surface and the cuboid facet (Stains, 1973). In addition, the length and breadth measurements are within the range of P. lotor provided by Stains (1973). Additional Pleistocene species of Procyon have been synonymized with P. lotor because the morphology was within the range of intraspecific variation (Kurtén and Anderson, 1980). Kurtén and Anderson (1980) recognized only one other fossil species, P. rexroadensis, which was limited to the Blancan LMA. Emmert and Short (2018) recommended synonymizing P. rexroadensis with P. lotor because of a lack of distinct morphological characters.
Remarks: Pleistocene-age Procyon has been found across North America and into northern South America (Kurtén and Anderson, 1980), but the fossil record is sparse (Harris, 2014). In Mexico, Procyon sp. is found at the Irvingtonian-age El Golfo (Croxen et al., 2007), and Procyon lotor is known from the Rancholabrean in California and New Mexico (Harris,   2014) as well as the Chihuahua-Coahuila Plateaus and Ranges, the Sierra Madre Oriental, the Trans-Mexican Volcanic Belt and the Yucatan Platform , so it is expected in the northwest of Mexico, although it is not reported from the Rancholabrean of Arizona .
Identification: This radius was initially reported as Lynx rufus by Mead et al. (2006), and the identification is confirmed here. The distinct features are as in Felidae rather than Canidae, and the anteroposterior compression excludes Felis (Kelson, 1946). As in L. rufus, there is a distinct horizontal ridge superior to the distal articulation on the posterior surface, and a lack of mediolateral constriction between the diaphysis and epiphysis. Lynx rufus is known from the Rancholabrean of Mexico  and Arizona , whereas Lynx canadensis has not been found south of Utah (Lavoie et al., 2019).
Remarks: Lynx rufus is frequently found at North American Pleistocene sites (Kurtén and Anderson, 1980), including at Rancho La Brea (McHorse et al., 2012), Diamond Valley Lake (Springer et al., 2009), and Tule Springs (Scott et al., 2017). L. rufus is known across northern and central Mexico during the Rancholabrean . In Mexico, L. rufus is also known from the Irvingtonian of Cedazo in central Mexico (Mooser and Dalquest, 1975), and the latest Pleistocene or early Holocene of Jimenez Cave in Chihuahua (Messing, 1986).
Description: The tooth fragment (TERA 452) is mediolaterally compressed and has a distinct parastyle and ectoparastyle (Fig. 10B). The tooth fragment is lacking a distinct protocone and preserves no evidence of any lingual flaring that would indicate a protocone had been present. There is a minimal anterior cingulum on the tooth. Diagnostic measurements are not possible because of the fragmented nature of the specimen.
Identification: This specimen was initially identified as Canis latrans by Mead et al. (2006). However, the mediolateral compression and parastyle suggest this tooth is from Felidae and not Canidae. The size suggests a large cat, possibly Panthera, Puma, or Smilodon. The lack of a protocone excludes Panthera and Puma (Cherin et al., 2013;Babiarz et al., 2018), and the ectoparastyle and cingulum are as in Smilodon (Christiansen, 2013). Therefore, we confer the tooth fragment to Smilodon. Because Smilodon fatalis is common during the Rancholabrean, S. populator has not been found in North America, and S. gracilis is from the early Irvingtonian, we confer the specimen to S. fatalis.
Remarks: Smilodon is known throughout much of North America from the Irvingtonian and Rancholabrean LMAs (Kurtén and Anderson, 1980;Babiarz et al., 2018), but this is the first record from the northwest of Mexico. In Mexico, S. fatalis has been found at Pleistocene sites across the Central Plateau, the Trans-Mexico Volcanic Belt, and the Sierra Madre Oriental . In the United States, S. fatalis is known from the Rancholabrean of eastern New Mexico and southern California (Kurtén and Anderson, 1980;Morgan and Lucas, 2001;Harris, 2014), including at Rancho La Brea (McHorse et al., 2012), Diamond Valley Lake (Springer et al., 2009), and Tule Springs (Scott et al., 2017). The presence in northwestern Mexico is novel but not wholly unexpected, although it has not been reported from the Rancholabrean of Arizona .

Community structure
The late Pleistocene and modern mammalian communities vary in their composition. The fossil community has two equids, two camelids, one tayassuid, one canid, two felids, and one procyonid, whereas in the present day, the Térapa region has one tayassuid, two canids, three felids, four mephitids, two mustelids, and three procyonids. The fossil community presented here consists of four herbivores from two orders of ungulates (Fig. 11). The four carnivorans are equally divided into carnivores and carnivorous omnivores. Yet, the modern community is largely dominated by 11 carnivorous omnivores, with three carnivores, one herbivorous omnivore, and no herbivores (Fig. 11). Lynx rufus (carnivore) and Procyon lotor (carnivorous omnivore) are the only taxa to occur in both fossil and modern communities.
The faunal turnover of the taxa in this study corresponds with a 97% decrease in mean body mass (BM; kg) from 289 kg to 9 kg. Odocoileus virginianus is the largest ungulate at Térapa today (BM = 55 kg), but, in the past, the larger Equus scotti (BM = 555 kg), Camelops hesternus (BM = 1100 kg), Palaeolama mirifica (BM = 80 kg), and Platygonus compressus (BM = 110 kg) also occurred at the site. Similarly, with the loss of Smilodon cf. S. fatalis (BM = 400 kg), the largest carnivoran today is Puma concolor (BM = 52 kg). Canis dirus (BM = 50 kg) was nearly the same mass as P. concolor. Procyon lotor (BM = 5.5 kg) is the smallest taxon in the fossil community presented here, but the modern community has eight species of Carnivora that are smaller than P. lotor.

DISCUSSION
Térapa provides the first Rancholabrean occurrences of Palaeolama, Procyon, and Smilodon in northwest Mexico and the first records of Platygonus compressus and Canis dirus in Sonora. Equus, Camelops, and Canis are well-represented at sites across the North American desert region and, although Platygonus and Lynx are sparsely represented in northern Mexico, they have extensive records in the southwestern US, making them expected at Térapa during the Rancholabrean.
Térapa lies along the Pleistocene Sonora-Central America Pacific lowlands corridor and the Rocky Mountains-Sierra Madre Occidental corridor (Ceballos et al., 2010). The former allowed dispersal of tropical taxa north and the latter permitted the movement of temperate taxa south (Ceballos et al., 2010). Additional 'holding pen' areas would have been inhabited by taxa until more preferred environments allowed further migration (Ceballos et al., 2010;Woodburne, 2010). With the presence of water and diverse vegetation, Térapa likely acted as an area of faunal exchange. For instance, Palaeolama originated in South America (Webb, 1974), and researchers postulate that P. mirifica used tropical corridors along the Sonora-Central America Pacific lowlands and the Tamaulipas-Central America Gulf lowlands to move north across Mexico (Bravo-Cuevas and Jiménez-Hidalgo, 2015).
The fauna at Térapa documents a shift in feeding strategy from a community of primarily herbivores during the late Pleistocene to one of primarily carnivorous omnivores at present. Ranges of carnivoran species and the richness of carnivoran communities have been shown to be affected by climatically driven habitat changes during the transitions of the glacial and interglacial stages of the Pleistocene (Arias-Alzate et al., 2017. In this region of mosaic complexity and shifting ecosystems, these dynamic environmental changes may have driven faunal community change in this area, including the transition to a carnivorandominated community. Previous studies have documented large declines in community average body mass at the transition between the late Pleistocene and Holocene (Stegner and Holmes, 2013;Smith et al., 2018). The mammalian fauna presented here indicates a similar change, with a 97% decline of community average body mass from 289 kg to 9 kg. The considerable change is due to the loss of megaherbivores and the reduction in body mass of the felids present. Average body mass decline has been linked to rises in global temperature (Gardner et al., 2011;Martin et al., 2018;Martin and Barboza, 2020) as well as shortages in food availability (Huston and Wolverton, 2011;McNamara et al., 2016;Westover and Smith, 2020). Although Térapa provides a broad scope of temporal change at a single locality (with several sub-localities therein), it nevertheless captures broad environmental change as recorded by the reorganization of the terrestrial fauna.
If antilocaprids, bovids, and cervids had been included in the study of community structure, the fossil community would add four artiodactyl species (Capromeryx sp., Stockoceros sp., cf. Odocoileus, and Bison sp.; Mead et al., 2006) while the modern community would only add one species (Odocoileus virginianus). The inclusion of these taxa would only exacerbate the differences between the faunal communities during the two time periods. While it is likely additional species lived at Térapa and have not been recovered, it would require finding 59 additional species of Carnivora at Térapa to match the proportional community structure of carnivores and herbivores that is present currently. The dramatic shift from herbivores to carnivores at Térapa follows a similar pattern documented across Mexico .
The large mammalian fauna at Térapa aligns with previous faunal environmental analyses that described an elevational mosaic of a temperate to tropical/subtropical marsh, an adjacent semiarid savanna-grassland, a slow-moving freshwater stream, and a riparian forest area along the water (Mead et al., 2006;Nunez et al., 2010;Bright et al. 2016). The presence of both browsing and grazing herbivores implies a diverse assemblage of trees, forbs, and grasses at Térapa in the past. Currently, there are no grazing ungulates present at Térapa, and the region lacks expansive grasslands.
Additionally, the presence of Lynx rufus suggests a more restrictive paleoenvironmental interpretation and may be indicative of a mosaic-edge habitat associated with the nearby mountain foothills (∼16 km). Extant L. rufus occupy strictly temperate habitats of savanna/grassland/chaparral but also occupy relatively large home ranges (3-96 km 2 for males and 1-38 km 2 for females; Lariviere and Walton, 1997). The diet of L. rufus includes rodents, small ungulates, large ground birds, and reptiles (Lariviere and Walton, 1997), which are all present in the fossil assemblage at Térapa (Mead et al., 2006).
During the glacial period when the fossils at Térapa were deposited, cooler global temperatures produced a more constricted Intertropical Convergence Zone (ITCZ) and increased precipitation in northwestern Mexico (Metcalfe, 2006). As temperatures warmed into the modern interglacial period, the ITCZ expanded poleward, and shifted the mid-latitude jet stream and associated precipitation poleward (Metcalfe, 2006). The poleward shift of the ITCZ and mid-latitude jet stream also shifted the biodiversity-rich savanna habitats northwards into the southwestern US and as a consequence, left behind the present-day scrubland habitats of northern Mexico (Metcalfe, 2006). As temperatures increased by ∼6°C and precipitation decreased, Sonora became more arid, and species were forced to shift their ranges, often northward (Metcalfe, 2006).
With impending climate change and anticipated warming of an additional 4°C globally by the end of the 21 st century, paleoecological studies of past faunas can have great implications for wildlife conservation because of the associated biodiversity shifts (Walther et al., 2002;Foley et al., 2005;Lipton et al., 2018). The Sonoran Desert is expected to continue expanding further northwards in Arizona and New Mexico as climate projections suggest an increasingly arid climate (Magaña et al., 2012). With shifting environments, species may have to alter their geographic ranges, but only 41% of natural areas (i.e., areas with low effects of human modification) in the US are suitably connected to allow for species movement (McGuire et al., 2016). Conservation efforts on the margins of species ranges, especially those in or adjacent to the desert southwest of North America, may also help facilitate the preservation of unique genetic adaptations to harsh climatic pressures (Plumb and McMullen, 2018). Sustaining habitats and their connectivity for conservation may be difficult to facilitate, but we aim to provide a broad scope of anticipated biodiversity change in the presence of continued warming.

CONCLUSIONS
Relatively few fossil sites are well known from the Rancholabrean LMA of northern Mexico (Ceballos et al., 2010;White et al., 2010), and Térapa provides an extensive faunal record in an area that was once a marsh savanna but is now xeric desert chaparral. At this site, we describe mammalian community restructuring due to the loss of the large-bodied, herbivorous ungulates in response to environmental change. Whereas the fossil community was nearly evenly split between carnivorans and herbivorous ungulates, the present community is dominated by carnivorans (Fig. 11). There is also an associated 97% decrease in mean body mass through time because of the loss of the largest taxa. We postulate that this change is largely due to rising temperature and shifting precipitation regimes and the resulting climatically driven changes in vegetation, such as the loss of grasslands. As the climate continues to warm and the deserts shift north, today's fauna in the southwestern US and northwestern Mexico will be similarly affected.
Long term records of faunal change, such as that at Térapa, provide valuable information for guiding modern conservation practices. The presence of eight taxa of Perissodactyla, Artiodactyla, and Carnivora in a community that existed approximately 40-43 ka (MIS/OIS 3) provides a critical spatial and temporal record, including the first Rancholabrean occurrences of Palaeolama, Procyon, and Smilodon in northwest Mexico, and the first records of Platygonus compressus and Canis dirus in the state of Sonora. It is probable that more sites exist in this region and can contribute to a more complete understanding of the area but have yet to be found, fully analyzed, and reported upon. Future fossil recovery will provide much needed details about the fauna in and around Térapa and in Sonora, and will also enable further study of regional habitat and biodiversity shifts in the southwestern US.
Supplementary material. The supplementary material for this article can be found at https://doi.org/10.1017/qua.2020.125