Olfaction at depth: Cribriform plate size declines with dive depth and duration in aquatic arctoid carnivorans

Abstract It is widely accepted that obligate aquatic mammals, specifically toothed whales, rely relatively little on olfaction. There is less agreement about the importance of smell among aquatic mammals with residual ties to land, such as pinnipeds and sea otters. Field observations of marine carnivorans stress their keen use of smell while on land or pack ice. Yet, one dimension of olfactory ecology is often overlooked: while underwater, aquatic carnivorans forage “noseblind,” diving with nares closed, removed from airborne chemical cues. For this reason, we predicted marine carnivorans would have reduced olfactory anatomy relative to closely related terrestrial carnivorans. Moreover, because species that dive deeper and longer forage farther removed from surface scent cues, we predicted further reductions in their olfactory anatomy. To test these hypotheses, we looked to the cribriform plate (CP), a perforated bone in the posterior nasal chamber of mammals that serves as the only passageway for olfactory nerves crossing from the periphery to the olfactory bulb and thus covaries in size with relative olfactory innervation. Using CT scans and digital quantification, we compared CP morphology across Arctoidea, a clade at the interface of terrestrial and aquatic ecologies. We found that aquatic carnivoran species from two lineages that independently reinvaded marine environments (Pinnipedia and Mustelidae), have significantly reduced relative CP than terrestrial species. Furthermore, within these aquatic lineages, diving depth and duration were strongly correlated with CP loss, and the most extreme divers, elephant seals, displayed the greatest reductions. These observations suggest that CP reduction in carnivorans is an adaptive response to shifting selection pressures during secondary invasion of marine environments, particularly to foraging at great depths. Because the CP is fairly well preserved in the fossil record, using methods presented here to quantify CP morphology in extinct species could further clarify evolutionary patterns of olfactory loss across aquatic mammal lineages that have independently committed to life in water.


| INTRODUC TI ON
Mammals rely on their sense of smell to varying degrees and their olfactory systems have evolved to operate in distinct ecological contexts. As lineages, foraging landscapes, and chemical stimuli change over evolutionary time, species acquire and lose olfactory capacities (Gittleman, 2013;Hayden et al., 2010;Van Valkenburgh et al., 2011). For example, it is widely accepted that obligate aquatic mammals such as odontocete cetaceans, and to a lesser extent mysticetes, have lost some degree of olfactory anatomy, genes and behaviors relative to their living terrestrial relatives and ancestors (Kishida, Thewissen, Hayakawa, Imai, & Agata, 2015;Liu et al., 2019;Oelschläger, 1992;Oelschläger & Buhl, 1985). There is less agreement on the relative role smell plays in the life of aquatic mammals with residual ties to the land, such as marine arctoid carnivorans, the pinnipeds (seals, sea lions, and walrus) and sea otter (Enhydra lutris). Some studies (Harrison & Kooyman, 1968;Van Valkenburgh et al., 2011) have suggested that the olfactory apparatus of pinnipeds is generally reduced relative to their terrestrial carnivoran relatives, while another study found no significant difference (Pihlström, 2008). Support for a keen sense of smell in pinnipeds and sea otters comes from field observations of scent-driven behaviors, such as nose-to-nose nuzzling, genital sniffing, alarm responses to upwind biologists, and aversive reactions to con-specific carcass odors (Lowell & Flanigan, 1980;Peterson & Bartholomew, 1967;Riedman & Estes, 1990;Ross, 1970), all of which are also observed in terrestrial carnivorans. However, there is one olfactory dimension missing from this discussion. Unlike terrestrial species, aquatic carnivorans capture prey exclusively underwater and do so "noseblind." With nostrils closed, diving mammals are shut off from all chemical cues except those they detect at the surface (Reidenberg, 2007;Riedman & Estes, 1990). It is thought that foraging pinnipeds use surface odors, such as dimethyl sulfide (DMS), to locate areas of high marine productivity in the same way mysticete whales and sea birds do (Bouchard et al., 2019;Kowalewsky, Dambach, Mauck, & Dehnhardt, 2006;Nevitt, 1999); however, once underwater, these diving carnivorans can no longer use the landscape of chemical cues relied on by terrestrial species to locate and capture prey (Smith, 1980;Ylönen, Sundell, Tiilikainen, Eccard, & Horne, 2003). For this reason, we pose a first, general hypothesis that aquatic carnivorans rely less on olfaction than closely related terrestrial species and predict that this will be manifested in reduced olfactory anatomy.
Secondly, dietary regimes vary widely across aquatic carnivorans and include pelagic and mesopelagic cephalopods and fish, benthic invertebrates, coastal zooplankton, penguins, and pinniped pups, among others (Bowen & Siniff, 1999;Pauly, Trites, Capuli, & Christensen, 1998). Coupled with this ecological diversity, pinnipeds and sea otters have evolved a wide range of diving behaviors, both the depth at which they pursue prey and the length of time spent diving (Ponganis, 2011;Schreer & Kovacs, 1997). For example, sea otters' dives average ca. 12 meters and last a little over a minute (Bodkin, Esslinger, & Monson, 2004;Tinker, Costa, Estes, & Wieringa, 2007), while northern elephant seals' dives average over 500 meters and can last up to two hours Robinson et al., 2012). We hypothesize that this diversity in diving behavior influences olfactory capacity for two reasons. First, because deeper and more extended dives remove underwater foragers from informative surface odorant cues that might be present at a dive's initiation site (Davis, Fuiman, Williams, Horning, & Hagey, 2003;Davis, Fuiman, Williams, & Le Boeuf, 2001;Harcourt, Hindell, & Bell, 2000), we hypothesize that among aquatic carnivorans, selection for keen olfactory performance is further reduced in more extreme divers. Additionally, cranial adaptations to the challenges of diving in low light (Welsch et al., 2001) under fluctuating pressure (Kooyman, 1973) include enlarged orbits and the reduction of air-filled skull cavities (Curtis, Lai, Wei, & Van Valkenburgh, 2015;King, 1983). These aquatic specializations likely constrain the space available for olfactory structures and tend to be more extreme in species that dive deeper (Debey & Pyenson, 2013). Consequently, we predict that within the aquatic carnivorans, reductions in olfactory anatomy will be inversely related to diving depth and duration.
Previous work on nasal turbinals in carnivorans suggested that aquatic species had reduced olfactory turbinal surface areas relative to their terrestrial relatives (Van Valkenburgh et al., 2011). However, only five aquatic species were sampled, making this conclusion tentative, and the authors did not examine any correlations with diving behavior. To further test the impact of aquatic foraging on olfactory anatomy, we expanded the number of aquatic species sampled to 19 and examined a different metric of olfactory anatomy, the area of the cribriform plate (CP). The CP is a bone in the posterior nasal cavity of mammals that is perforated with passageways for olfactory nerve bundles crossing from the periphery to the olfactory bulb of the brain (Negus, 1958) (Figure 1).
We chose to study the CP for several reasons. First, because its size varies with the amount of peripheral olfactory innervation found in a mammal's snout (Pihlström, Fortelius, Hemilä, Forsman, & Reuter, 2005), quantifying the CP provides an opportunity to gauge and compare relative olfactory investment across aquatic and terrestrial species (Bird, Amirkhanian, Pang, & Van Valkenburgh, 2014).
Second, earlier work found that, across all superorders of mammals, relative CP size is closely correlated with the number of olfactory F I G U R E 1 Nasal anatomy of grizzly bear (Ursus arctos). Left half of a sagittally sectioned skull. Pink, perforated cribriform plate bone separating nasal cavity from the brain case. Green, olfactory (ethmo-, fronto-and naso-) turbinals. Blue, respiratory (maxillo-) turbinals receptor genes in an animal's genome, thereby establishing CP morphology as an informative metric of relative reliance on the sense of smell (Bird et al., 2018). Third, CP area is tightly correlated with the surface area of the ethmoturbinals, the bony plates that bear olfactory epithelium (Bird et al., 2014). Finally, CP area can be quantified in some fossil skulls (Bird et al., 2018), and so will enable future studies into the evolution of olfactory anatomy in extinct mammal lineages that have transitioned from land to water. Here, we perform the first extensive comparative and quantitative study of the CP morphology of arctoid carnivorans, a clade that has seen multiple independent invasions into the marine habitat and includes species at the intersection of terrestrial and aquatic life.
Our sample group, the arctoid carnivorans, is an ecologically rich clade that includes ursids (bears), mustelids (e.g., weasels, otters, and badgers), procyonids (e.g., raccoons and kinkajous), mephitids (skunks), and pinnipeds, among others ( Figure 2) (Upham, Esselstyn, & Jetz, 2019). Within the arctoids, there were multiple independent, secondary entries into aquatic habitats (Berta, Sumich, & Kovacs, 2015), resulting in a diversity of closely related species from disparate ecologies (aquatic, semi-aquatic, and terrestrial) along a spectrum of olfactory demands. According to the recent comprehensive mammalian phylogenetic analysis (Upham et al., 2019), pinnipeds diverged from the lineage leading to Musteloidea ca. 24-33 ma., and otters diverged from terrestrial mustelids more recently, ca. 8.5-12 million years ago. Studying Carnivora is advantageous, as the group has a fairly well-resolved phylogeny, allowing the application of comparative methods that account for phylogenetic relatedness in our study of ecological influences on olfaction.

| MATERIAL S AND ME THODS
Using high-resolution CT scans and 3D imaging software and methods developed in previous studies (Bird et al., 2014(Bird et al., , 2018, we measured the surface area of the perforated region of the CP as well as the cumulative cross-sectional area of the CP foramina as proxies for relative olfactory innervation found in individual arctoid species. F I G U R E 2 Time-calibrated phylogeny for arctoid carnivorans. All taxa except the gray wolf were sampled for this study. Topology and divergence estimates are taken from Upham et al. (2019)

| Specimen collection
We sampled 65 skulls from 31 species representing eight families of arctoid carnivorans (Figure 2) (Upham et al., 2019). Specimens and their source museums are listed in Table A1 in Appendix. All species are extant with the exception of the tropical monk seal (Neomonachus tropicalis). Body sizes span several orders of magnitude from <1 kg (long-tailed weasel, Mustela frenata) to at least 1,275 kg (male southern elephant seal, Mirounga leonina) (Irvine, Hindell, Van Hoff, & Den, 2000). Where possible, we sampled two wild-caught adult specimens, one male and one female, for each species.  foramina were calculated and tallied. While total foramina area may be the most direct estimate of the cross-sectional area of an animal's olfactory innervation, it cannot be resolved from low resolution scans, damaged skulls, or fossils. Therefore, because foramina area is closely correlated with CP surface area (r 2 = .92; pgls-r 2 = .9, p < .001, Figure A3), we used the latter to maximize sample size.

| Morphological data
As a body size proxy, we used the skull metric, occiput-to-orbit length (OOL), defined as the distance between the posterior extent of the occipital condyles and the anterior most extent of the orbit ( Figure 3). The correlation between OOL and body mass is similar to that between full skull length and body mass (Van Valkenburgh, 1990), and OOL offers advantages over skull length. First, OOL excludes the confounding influence of snout length, a feature that varies widely among arctoids independent of body size. Indeed, in our sample, relative snout lengths are shorter in aquatic species than terrestrials (p = .036, inclusion of skulls with broken premaxillae and is better suited to analyze incomplete fossil skulls in the future. Using the skull metric OOL instead of body mass reduced the excessive influence of large fat stores in pinnipeds on body size estimates. For all specimens, in our sample, the skull metric OOL was measured from 3D skull reconstructions using Mimics or from skulls directly using digital calipers.

| Habitat groupings
We grouped the arctoids into three ecological categories, terrestrial, aquatic, and semi-aquatic. We defined terrestrial species as those that live and forage exclusively on land. These include ten species of ursids, procyonids, mephitids, and mustelids ( Figure 2, Table A1).
Species in the aquatic group (n = 19) forage exclusively underwater but also spend some time hauled out on land or pack ice and include eighteen pinniped species and the mustelid sea otter (Enhydra lutris). Semi-aquatic species forage both underwater and on land and include two mustelid species. Although the polar bear (Ursus maritimus) is often referred to as a semi-aquatic marine mammal and sometimes swims to stalk its prey (Berta et al., 2015), we chose to classify it as terrestrial, as it does not seek and capture its prey underwater (Stirling, 1974).

| Diving data
Four diving behavior variables are included in this study, maximum dive depth, mean dive depth, maximum dive duration, and mean dive duration. All dive data were compiled from published behavioral field studies (Table A2). If means were not directly reported in source literature, we derived these from supplemental raw dive data, data shared in personal communications, or in two cases by visually measuring from histogram distributions. We included as many studies as possible in calculating our means, weighting the contribution of each study by the number of animals recorded. Other potentially informative variables describing potential diving capacity or overall degree of aquatic specialization, such as magnitude and distribution of oxygen stores, at-sea durations, migration distances, haul-out durations, exist for some but not all sample species, and so could not be used for this study.

| Statistical analysis
Species means of all morphological and ecological variables were used for data analysis. To view scaling relationships between CP and body size and to derive values for size-adjusted relative CP size, we plotted log 10 absolute CP surface area against log 10 OOL using phylogenetic least squares regression (PGLS). Resulting residuals were used as relative CP size (RelCP) in all subsequent analyses. To test the influence of habitat on RelCP values, we performed pair-wise ANOVA and Tukey HSD post hoc tests. All regression plots include regression lines from PGLS as well as general least squares regression (GLS). All analyses were performed in R (Team RC, 2015). For PGLS, we used Caper Package (Orme et al., 2013) and a time-calibrated mammal tree pruned to include only the species in our study (Upham et al., 2019).

| Cribriform plate area and body size
Among all 31 species, absolute CP surface area is coupled to body size, as described here by the skull metric, occipital condyle to orbit length (OOL) (pgls-r 2 = .7, p < .001), and scales with negative allometry (y = 1.37x − 0.1097), (Figure 4, Table A3). Thus, large species have proportionally smaller CP for their body size. There is considerable scatter about the line with terrestrial species tending to fall above the line and aquatics below the line. Among the aquatic species alone the relationship between CP surface area and OOL is similar (pgls-r 2 = .69, p < .001, n = 18) and among terrestrials alone it is stronger (pgls-r 2 = .84, p < .001, n = 10).

| Cribriform plate in terrestrial, aquatic, and semi-aquatic species
To test the hypothesis that aquatic and semi-aquatic species have reduced olfactory morphology relative to terrestrial species, we performed a one-way ANOVA and Tukey HSD post hoc tests on mean relative CP size (RelCP) values from all three habitat groups. Aquatic species have significantly smaller mean RelCP than terrestrial species (p < .001). Mean RelCP of semi-aquatic species is smaller than that of terrestrials (p = .014) and does not differ significantly from that of aquatics (p = 1) (Table A3).
Similarly, when running a habitat analysis on CP surface area that is size-adjusted to full skull length (SkL) instead of OOL, comparable differences between groupings emerge (Appendix A1, Figure A2).
As per Tukey HSD post hoc tests, again aquatics have significantly larger mean RelCP than terrestrials (p < .001), and there is no significant difference between semi-aquatics and aquatics (p = .36).
The difference in mean RelCP (size-corrected to full skull length) between semi-aquatic species and terrestrials is less pronounced but significant (p = .042).
To consider whether the losses in olfactory anatomy in the aquatic mustelids occurred independently from those in the lineage leading to Pinnipedia, we analyzed RelCP in aquatics and terrestrials within the clade Musteloidea (mustelids, procyonids, and mephitid; n = 9) and within the family Mustelidae (n = 6) separate from Pinnipedia and Ursidae. A phylogenetically corrected ANOVA shows that the mean RelCP of the aquatic sea otter and semi-aquatic river otter and mink together are significantly smaller than the mean RelCP of terrestrial musteloids (p = .007) and terrestrial mustelids (p = .014). It is interesting that among the three terrestrial mustelid species, the long-tailed weasel differs from the much larger badger and wolverine by having a reduced RelCP similar to that of the three more aquatic mustelids. This suggests that a reduced RelCP might be characteristic of smaller mustelids in general. Without a larger sample size of small mustelids, the diminutive RelCP of the long-tailed weasel is difficult to interpret.

| RelCP and diving ecology of aquatic carnivorans: dive depth and duration
To investigate possible interactions between diving behavior and olfactory morphology, we tested for correlations between RelCP and each of four diving parameters, mean dive duration, maximum dive duration, mean dive depth, and maximum dive depth within the 18 aquatic species for which we had published dive data (17 pinnipeds and the sea otter). We found strong inverse relationships between RelCP and three of the variables, mean dive depth (r 2 = .75, p < .001, pgls-r 2 = .65, p < .001), mean dive duration (r 2 = .76, p < .001, pglsr 2 = .61, p < .001), and maximum duration (r 2 = .66, p < .001, pglsr 2 = .48, p < .001), respectively (Figure 5a,b,d),. This relationship is largely driven by the phocids; in all three cases, accounting for phylogeny weakens the coefficients of determination because the otariids tend not to follow the main trend and phylogeny exerts a strong influence on pinniped CP morphology independent of diving behavior. Phocids have on average smaller RelCP than either otariids alone (p = .015) and otariids and the odobenid walrus together (p = .019) ( Figure A4, Table A3). In the case of the fourth parameter, maximum dive depth, what appears to be a strong negative relationship with RelCP (r 2 = .55, p < .001), is barely significant after accounting for phylogeny (pgls-r 2 = .22, p = .051) (Figure 5c, Table A3).

| D ISCUSS I ON
Results from our study point to reduced reliance on olfaction as a secondary adaptation to marine habitats, and in particular to foraging at depth. Among aquatic arctoid carnivorans, we found a pronounced loss of olfactory anatomy, specifically a reduction in relative cribriform plate size (RelCP), that mirrors established reductions in olfactory turbinal surface area (Van Valkenburgh et al.,

F I G U R E 4
Log-log plot of CP surface area versus Occiput-orbit length (OOL) for three ecological groupings. Green circles, terrestrial species; red triangles, semi-aquatics; dark blue circles, Phocidae; turquoise inverted triangles, Otariidae; blue diamond, Mustelidae (sea otter, Enhydra lutris); light blue square, Odobenidae (walrus, Odobenus rosmarus); Solid line, best fit from phylogenetic generalized least squares (PGLS) regression; dotted line, best fit from generalized least squares regression (GLS) 2011). It is not surprising that both CP and olfactory turbinal surface areas are reduced in aquatic species, given their common developmental origin (Rowe, Eiting, Macrini, & Ketcham, 2005), and the fact that they are strongly correlated in size across all carnivorans (Bird et al., 2014). Our results accord with initial genomic studies reporting losses in the number of functional olfactory receptor genes in five aquatic arctoid species (two otter and three pinniped species) Moreover, our findings go beyond previous work in revealing that, among aquatic carnivorans, species that dive deeper and for longer periods of time tend to have an even greater reduction in CP size.
Our finding of a smaller RelCP among aquatic arctoid carnivorans contradicts earlier work that concluded that CP size did not differ between pinnipeds and terrestrial carnivorans (Pihlström et al., 2005(Pihlström et al., , 2008. There are several likely reasons for differences in our findings. First, Pihlström et al. (2005) used linear measurements to calculate CP surface area, whereas we relied on digital quantification, a method that better captures the highly irregular shape of the CP (Bird et al., 2014). Second, their body size proxy, skull area, does not exclude snout length, which can lead to the underestimation of body size in the typically short-snouted pinnipeds and sea otter and a consequent inflation of size-adjusted CP size in aquatic species. Within aquatic carnivorans, we found marked variation in olfactory morphology that corresponds closely with diving behavior.
Although there is no clearly significant association between maximum dive depth and RelCP that persists after phylogenetic accounting, there are strong inverse correlations between RelCP and the following three dive variables: mean dive depth, mean dive duration, and maximum dive duration. These relationships are even more pronounced when considering the phocids, or true seals, alone. The absence of a significant relationship between maximum dive depth and RelCP was surprising given that maximum dive depth values exhibit F I G U R E 5 Significant correlation between relative cribriform plate size and three dive variables in the four families of aquatic carnivorans. (a) Relative CP size (RelCP) versus mean dive depth (r 2 = .75, p < .001, pgls-r 2 = .65, p < .001). (b) RelCP versus mean dive duration (r 2 = .76, p < .001, pgls-r 2 = .61, p < .001). (c) RelCP versus. maximum dive depth (r 2 = .55, p < .001, pgls-r 2 = .22, p = .51). (d) RelCP versus. maximum dive duration (r 2 = 0.66, p < .001, pgls-r 2 = .48, p < .001). Dark blue circles, Phocidae; turquoise inverted triangles, Otariidae; light blue square, Odobenidae (walrus); blue diamond, Mustelidae (sea otter). Solid line, best fit from PGLS regression; dotted line, best fit from GLS regression the widest range of the four diving parameters and because the smallest RelCP values by far belong to the most extreme divers, the northern and southern elephant seals, which have been recorded diving to 1,735 and 2,388 m, respectively Costa, Robinson, et al., 2010;Robinson et al., 2012). However, close review of the published literature reveals that recorded maximum depths are often not representative of species' overall diving patterns. For example, the California sea lion has been recorded at a depth of 575 m, and yet this otariid is generally considered a moderately shallow diver (Berta et al., 2015;Costa, Kuhn, & Weise, 2007).
Why might selection favor smaller RelCP, reduced olfactory anatomy, in aquatic carnivorans in general and species performing deeper, sustained dives in particular? We present alternative evolutionary explanations. First, although odor cues play an important role in social interactions and predator defense among aquatic carnivorans above water (Lowell & Flanigan, 1980;Peterson & Bartholomew, 1967;Riedman & Estes, 1990;Ross, 1970), below water, where pinnipeds and sea otters typically forage, odor cues are no longer detectable. Although another semi-aquatic mammal, the water shrew, is known to exhale and inhale bubbles to access scent cues from food surfaces underwater (Catania, Hare, & Campbell, 2008), this behavior has been hypothesized but not tested in the carnivoran river otter (Marriott et al., 2013). Unlike terrestrial carnivorans, which follow deposited and airborne prey scents to locate food sources (Smith, 1980;Ylönen et al., 2003), aquatic carnivorans forage with shut nostrils and locate prey without scent cues, except those detected above water upon surfacing. Utilizing scent cues at or above the water surface is a probable tool of foraging at sea, given that harbor seals (Phoca vitulina) have shown keen sensitivity to dimethyl sulfide (DMS), a volatile phytoplankton odorant and indicator of local marine productivity that is utilized by sea birds as well (Kowalewsky et al., 2006;Nevitt, Reid, & Trathan, 2004). However, deeper and more extended dives increase the diver's distance from these informative odor cues at the surface (Davis et al., 2001(Davis et al., , 2003. Consequently, over time, as selective pressure for detecting prey via odorant cues was relaxed, olfactory systems among carnivorans adapting to life in water likely decreased in size. Reduced olfactory structures were further favored because olfaction is a costly sensory system made up of millions of continually self-replacing olfactory sensory neurons (Graziadei & Graziadei, 1985).
Experiments using blindfolded animals revealed that harbor seals (Phoca vitulina) use their vibrissae to track the hydrodynamic trails of swimming fish (Dehnhardt, Mauck, & Bleckmann, 1998). In addition to enhanced vibrissae, aquatic carnivorans rely on a visual system adapted for hunting in dark waters. Visual specializations may include spherical lenses (Berta et al., 2015), wide pupil size range (Levenson & Schusterman, 1999) a tapetum lucidum (Kröger & Katzir, 2008) as well as proportionally large eye orbits (Debey & Pyenson, 2013). Considering these enhanced sensory specializations, it seems likely that the reduction of olfactory anatomy in pinnipeds and the sea otter over time stems, in part, from relaxed selective pressures on olfactory acuity as aquatic species come to rely more heavily on alternative sensory systems for underwater foraging.
Why does selection favor a greater reduction in CP size in aquatic carnivorans performing longer and deeper average dives?
There are a number of alternative hypotheses, all of which require further testing. First, it is possible that shallow, short dive patterns reflect a closer tie to the land/pack ice, while deeper and longer dive patterns represent a more pronounced separation from a terrestrial ecology. Longer separation from land, that is, a more aquatic life, likely results in increased disconnection from airborne and deposited odor cues that terrestrial animals rely on for food, predator protection, social communication, and reproduction.
To further test whether deeper, more sustained diving reflects a more fully aquatic lifestyle and an increasingly remote relationship to the land, all four diving variables might be viewed in relationship to other ecological proxies for relative proximity to land/sea ice. These factors could include foraging trip duration , long-distance migration patterns Costa, Robinson, et al., 2010), pupping season duration (Stirling, 1983), haul-out patterns (Cunningham et al., 2009) and overall at-sea duration Costa, Robinson, et al., 2010), among others.
An alternative, or complementary, interpretation for the negative relationship between diving depth/duration and RelCP suggests that volatile odor cues at the water's surface emitted by underwater prey play a role in prey detection for marine carnivorans, and that reliance on these surface odorants selects for retention of a larger olfactory system in shallow as opposed to deep divers. One such volatile, mentioned above, is dimethylsulfide (DMS), an odorant emitted by phytoplankton, particularly when grazed upon by krill and other zooplankton (Dacey & Wakeham, 1986). Concentrations of DMS at the sea-air interface are variable, and "hotspots" indicate underlying primary production, including the presence of krill and krill-feeding animals, such as fish or penguins (Barnard, Andreae, Watkins, Bingemer, & Georgii, 1982). The leopard seal (Hydrurga leptonyx), a shallow and short diver that feeds primarily on krill, penguins, and crabeater seal pups (Pauly et al., 1998), likely navigates a rich landscape of scent cues at the water's surface as well as on the ice sheet.
By contrast, deep divers appear to have less access to surface cues while foraging. For example, the northern elephant seal dives in a staggered stair-step pattern, reaching its prey of pelagic squid and mesopelagic fish (Pauly et al., 1998) at depths between ~300 and 1,500 m, displaced horizontally, sometimes by hundreds of meters, from the dive initiation location (Davis et al., 2001). It is notable then, that the shallow-diving leopard seal and the deep-diving elephant seal, close relatives among the pinnipeds, have the most disparate RelCP among all phocids (Figure 6a). To strengthen the argument that surface odorant cues influence foraging behavior in some aquatic carnivorans, future behavioral experiments, such as those used to test responsiveness of seabirds and whales to variable concentrations of DMS (Bouchard et al., 2019;Nevitt, Veit, & Kareiva, 1995), may be performed on pinnipeds and sea otters. gas-filled cavities (Kooyman & Ponganis, 1998). Adaptations to these challenges include, among others, collapsible alveoli, which minimize gas exchange (Scholander, 1940), distensible venous sinuses, which are thought to reduce external and middle ear cavity volume (Odend'hal & Poulter, 1966;Stenfors, Sadé, Hellström, & Anniko, 2001), and structural reductions in skull cavities, such as the narrowing of the external auditory canal (Kastak & Schusterman, 1999) and the loss of frontal sinuses (Curtis et al., 2015). Reductions in air-filled skull cavities, while adaptive under hydrostatic pressure, may constrain the development of the olfactory recess in the aquatic mammal skull, specifically the olfactory turbinals and attendant airway fluid dynamics necessary for robust odorant deposition (Craven, Paterson, & Settles, 2009).
In terrestrial carnivorans, ethmoturbinals often extend from the nasal cavity dorsally into the frontal sinuses, increasing surface area for odorant deposition and detection (Negus, 1958) ( Figure   A5a). By contrast, without the doming of the skull afforded by large frontal sinuses, the space available for ethmoturbinals and the CP in pinnipeds and the sea otter is limited dorsally ( Figure   A5b). Moreover, a survey of snout lengths in our sample reveals that aquatic carnivorans have significantly shorter snouts than terrestrial carnivorans, further reducing the nasal air space, and thereby perhaps constraining anterior extensions of ethmoturbinals as well (Tukey, p = .036, Table A4). Two exceptions to this are the California sea lion (Zalophus californianus) and the leopard seal (Hydrurga leptonyx), both of which have ethmoturbinals that extend into relatively long anterior nasal cavities, and large RelCPs as well (Figure 3 Appendix A1, Figure A5b). Finally, because aquatic carnivorans possess visual specializations for underwater vision, including relatively large eyeballs and orbits (Debey & Pyenson, 2013), the posterior nasal cavity is relatively narrow in most pinniped species (Berta et al., 2015), and most markedly in the elephant seals, further limiting space for ethmoturbinals laterally as well as ventrally ( Figure A6a

TA B L E A 1 (Continued)
TA B L E A 2 Dive data for all aquatic species, and sources. Maximum dive depth and duration values were sourced from single studies. Mean dive depth and duration were sourced and averaged from multiple studies where available. In calculating an overall average, individual study averages were weighted by sample size.