A dataset of calcium-normalized ⁸⁸Sr and ¹³⁷Ba concentrations from laser ablation transects across the cementum layer of 107 (female: n = 84, male: n = 23) Pacific walrus (Odobenus rosmarus divergens) teeth. Dataset includes spreadsheets containing elapsed time of laser ablation transect in seconds (ElapsedTime_s; laser transect speed = 5μm/s), calcium-normalized strontium concentrations (Sr_ppm_m88; values below limit of detection replaced with 1/2*limit of detection, see methods), and calcium-normalized barium concentrations (Ba_ppm_m137; values below limit of detection replaced with 0.5*limit of detection). Photos (in .tif format) of tooth cementum used to estimate the positions of cementum growth layer groups are included for each animal. For teeth where more than one laser ablation scar is visible in the photo, an arrow is included to indicate the laser ablation scar that corresponds with the included trace element data. Finally, a Word document containing metadata (Catalog Number/ID, Sex, Median Age Est., Coll. Year, Est. Birth Year), weaning age estimates produced by both a visual and mathematical method (Sr Vis. Est., Sr Math. Est., Ba Vis. Est., Ba Math. Est.), and a descriptive grouping of the patterns of accumulation of Sr and Ba (Sr Pat., Ba Pat.) are included. See methods and accompanying paper in Methods in Ecology and Evolution for more details.

Trace element analysis and data processing

Walrus teeth used for this study (female: n = 84, male: n = 23) were on loan from the University of Alaska Museum in Fairbanks, Alaska, the National Museum of Natural History, Smithsonian Institution, in Washington DC, and the Alaska Department of Fish & Game. These specimens each represented an individual animal, and were collected at various locations throughout the Pacific walrus range in the Bering and Chukchi seas. Most teeth originated from Alaska Native subsistence harvests, though a small number came from research expeditions. Dates of collection ranged from 1932 to 2016. In preparation for analysis, teeth were sectioned longitudinally using a slow-speed, water-cooled saw equipped with a diamond blade, creating a cross section of the center of the tooth with a thickness of ~1.5mm. This cross section was then polished using a rotary polishing wheel with a 3000 grit smoothing disc, rinsed with ultrapure water, and allowed to dry. Specimens were rinsed and dried again immediately prior to trace element analysis.

            Trace element analyses were conducted in the Advanced Instrumentation Lab at the University of Alaska Fairbanks (UAF), Fairbanks, Alaska. Concentrations of ⁸⁸Sr and ¹³⁷Ba were measured using an Agilent 7500ce Inductively Coupled Plasma Mass Spectrometer (ICP-MS; fitted with an Agilent 7500cs lens stack to improve sensitivity) coupled with a New Wave UP213 laser. Instrumental precision for the ICP-MS is reported at ± 5%. ⁴³Ca was used as an internal standard for these analyses, and all results are reported in parts per million (ppm). Measured element concentrations were compared to a United States Geological Survey microanalytical phosphate standard (MAPS-4), as well as a National Institute of Standards and Technology Standard Reference Material (SRM 610). Accuracy and precision were estimated by comparing element concentrations measured during ablation of the reference materials (n = 363) with reported concentrations. Sr and Ba measurements were both accurate to within 1% of reported values, with a precision (± 1 standard deviation) of ± 4% and ± 5%, respectively. Laser transects were ablated at a beam width of 25μm, at 55% power, with a pulse frequency of 10Hz, and a transect speed of 5μm/s. Dwell times were 0.02 seconds for ⁴³Ca, 0.01 seconds for ⁸⁸Sr, and 0.15 seconds for ¹³⁷Ba. Ablation was conducted at locations that maximized distance from the root of the tooth, where cementum GLGs converge and become distorted, while avoiding areas of tooth wear near the crown, where not all cementum layers are present for sampling. Each transect was ablated from the cementum-dentin interface (first year of life) to the outer edge of the tooth (final year of life), thereby measuring lifetime changes in element concentrations for each walrus (Fay, 1982).

            Trace element data were extracted and processed in Igor Pro version 6.37 using the Iolite software package version 3.0. Statistical analyses were conducted using R version 3.6.3 (R Core Team, 2020) with RStudio version 1.2.5033 (RStudio Team, 2015). Limits of detection were calculated for each analytical run using the standard method applied by Iolite (Longerich, Jackson, & Günther, 1996). Typical limits of detection were 0.18ppm for Sr and 0.34ppm for Ba. Element concentrations falling below the limit of detection were replaced with a value of one half the limit of detection (U.S. Environmental Protection Agency, 2000). Data points that were more than 4 standard deviations above or below the mean were considered outliers (Tukey, 1977). These were typically single data points believed to represent measurement errors generated during data collection and processing, rather than changes in element concentrations within the tooth, thus they were removed from subsequent analyses.

            All specimens used in this study were obtained from museum collections and/or Alaska Native subsistence harvests, thus this work is Institutional Animal Care and Use Committee (IACUC) exempt. Specimens from subsistence harvests were transferred to UAF for analysis under a Letter of Authorization from the United States Fish and Wildlife Service (USFWS) to Dr. L. Horstmann.

Weaning age assignment

After analysis on the ICP-MS, teeth were photographed under a Leica M165 C optical microscope coupled with a Leica DFC295 camera using reflected light. Level of magnification used when taking photographs varied depending on the size of the tooth, and was selected to maximize visibility of the GLGs. The first five GLGs in the tooth cementum were counted (Fay, 1982; Garlich-Miller, Stewart, Stewart, & Hiltz, 1993) and their positions marked on the images to denote the positions of the first five years of life on the laser ablation transect. Growth layer groups consist of paired bands of cementum, which accrete onto the tooth annually (Fay, 1982). The terminology used to describe growth layer groups has not been standardized and differences in the appearance of growth layers under reflected and transmitted light may cause confusion. In this study, the term “light layer” refers to the opaque, hypercalcified layer that accretes during periods of faster growth. This layer represents the period from approximately mid-April to mid-December (Clark, Horstmann, & Misarti, 2020). Under reflected light, this layer appears white, but it is dark under transmitted light. The term “dark layer” refers to the more translucent, hypocalcified layer that is built during periods of slower growth. The dark layer represents the period from around mid-December to mid-April (Clark et al., 2020). This layer appears dark under reflected light, because it allows light to pass through, thus it appears white when using transmitted light. Individually, each light or dark layer is referred to as a growth layer, whereas a pair of light and dark layers, representing one year of growth, is referred to as a growth layer group (Laws, 1952). Growth layers were identified and marked collaboratively by three observers (C.T.C., L.H., and N.M.), and the positions of these growth layers were revisited on at least two additional days to confirm their positions on the laser ablation transect. Estimates of overall age (age at death) were also generated for each animal using the methods described in Clark et al. (2020), and median age estimates were used to calculate approximate birth year for all animals examined in this study (Table S1).

Analysis of Sr and Ba data for age-at-weaning estimation was restricted to the first five years of life. Prior to analysis, data were smoothed using a Savitzky-Golay filter from the R package prospectr (Stevens & Ramirez-Lopez, 2014) with a window of 15 data points. This approach was chosen because it reduces noise in the data, allowing underlying patterns to be seen more clearly, but results in very little data loss and uses a centered method, thus does not shift the data left or right like some other smoothing methods. Weaning is a process, rather than a single event, beginning with the first intake of non-milk food items and ending when consumption of mother’s milk has ceased completely. For the purposes of this study, a walrus was considered to have weaned at the point where the early life signal in its cementum Sr or Ba time series, associated with consumption of milk, transitioned to relatively stable values that persisted throughout the animal’s adult life. The weaning age estimates generated here thus represent the end of the weaning process and the transition to an entirely non-milk diet. Data were first analyzed visually, with a single observer (C.T.C.) identifying patterns in Sr and Ba indicative of weaning (e.g., abrupt decreases/increases in concentration, changes in slope, etc.) based on expected patterns from the published literature (Humphrey, Dean, et al., 2008; Humphrey, Dirks, et al., 2008; Austin et al., 2013; Humphrey, 2014; Tsutaya & Yoneda, 2015; Smith et al., 2017). Because Pacific walrus tooth cementum typically begins growing in the second or third month after birth (Fay, 1982), a signal associated with the onset of nursing (low and stable Sr concentrations/rapidly increasing Ba concentrations) was not expected. Once a suspected weaning signal was selected, a plot of the data was overlaid on an image of the individual tooth from which they were collected, with the x-axis aligned to the laser ablation scar (Fig. 1). In this way, the location of the estimated weaning signal could be directly assigned to the growth layer in which it occurred. After visual analysis, weaning estimates were made using a mathematical approach, in which the data were subjected to an iterative method of change point detection using the R package segmented (V.R.M. Muggeo, 2008). This tool detects underlying changes in slope within a time series and, for the purposes of this analysis, was restricted to assigning a single change point within the first five years of an animal’s life. Because the mathematical approach was susceptible to both positive and negative changes in slope, and was restricted to assigning only a single change point to the elemental time series, periods of increasing or stable element concentrations at the beginning of an animal’s life (relatively rare, and possibly associated with the onset of and/or a period of sustained nursing) were omitted from the Sr and Ba data when applying the mathematical weaning estimation approach. While these variations in the patterns of element accumulation in the first year of life may contain important information, they were excluded for their tendency to interfere with the ability of the segmented regressions to assign a change point associated with the end of weaning and the attainment of an entirely non-milk diet.

            Weaning age data were analyzed qualitatively, by comparing estimates of age-at-weaning generated by the visual and mathematical estimation methods for Sr and Ba to walrus weaning age estimates from the published literature. The performance of these two different estimation methods was compared by examining the average differences in the predictions generated by the two approaches. Chi-squared tests were used to examine differences in the weaning age predictions generated for female and male walruses. Significance was assessed using an alpha of 0.05. Weaning age estimates were plotted by approximate year of birth for each individual, and visually examined to determine whether any patterns in estimated age-at-weaning existed across the ~100 year span represented by the walrus teeth in this study. Regional differences in weaning age estimates were not assessed, as only location of harvest/collection was available for the animals in this study. The Pacific walrus population is large and panmictic, with little evidence for internal structure (Beatty et al., 2020). Individual walruses move widely over the species’ range (Jay, Udevitz, Kwok, Fischbach, & Douglas, 2010; Jay, Fischbach, & Kochnev, 2012; Beatty et al., 2016), thus location of collection is unlikely to provide useful information about the regions in which an individual walrus spent its life or to reflect association with a distinct group within the larger population.