Estimating age of harbor seals (Phoca vitulina) with incisor teeth and morphometrics

Age Estimates from Teeth

To extract teeth from live-captured harbor seals, we induced anesthesia by administering isoflourane through a face mask until seals were sufficiently anesthetized to pass an endotracheal tube. Once the seal was intubated, we continued administering isoflourane until general anesthesia and sufficient analgesia were obtained for tooth extraction. The mouth was held open with a mouth gag typically used for dental work in veterinary medicine, and a tooth elevator, also used in veterinary medicine for dental work in dogs, was inserted into the gum along each side of the incisor and used to gently rock the tooth back and forth. Once the tooth was loose it was grasped with a tooth extractor, placing the tips of the extractor at the base of the incisor, and the tooth was gently pulled straight out. Following tooth extraction, an antiseptic solution was used to cleanse the area and direct pressure was applied for approximately one minute to stop blood flow and promote clotting.

In addition to extracting incisor teeth from live-captured seals, we obtained teeth from 450 subsistence-harvested seals acquired via a biosampling program conducted by the Alaska Native Harbor Seal Commission (ANHSC). Skulls from these seals were housed at the University of Alaska Museum (UAM) Frozen Tissue Collection in Fairbanks. Teeth were extracted by UAM personnel and provided to the Alaska Department of Fish and Game (ADF&G) for this study. For subsistence-harvested seals, we obtained a combination of at least two teeth from each individual, including canine, postcanine, and incisor teeth. Canine teeth were extracted for most individuals, the postcanine tooth was taken from only a subset of sampled seals, and 57 seals had two incisors removed to assess variability of age estimates from incisors.

Four of the subsistence-harvested seals and 17 live-captured seals had been previously captured and tagged as pups or yearlings by ADF&G and were thus of known ages when taken during a subsistence hunt or recaptured by ADF&G. An incisor, postcanine, and canine tooth also were obtained from two known-age harbor seals that died from natural causes at the Alaska SeaLife Center (ASLC). We submitted teeth from all seals to a well-established laboratory (Matson's Laboratory LLC, Milltown, MT) with >30 yr experience estimating ages by counting cementum annuli from teeth of other species. All teeth submitted to the laboratory were randomly numbered such that the laboratory had no method of determining which teeth came from which seal. In addition to estimating age for each tooth, the laboratory provides a rating of their confidence in the reliability of age estimates, with a Certainty Code of A being higher than B.

Laboratory Methods

At Matson's Laboratory, teeth were cleaned in a hot water bath (at least 80°C, but never hot enough to boil), followed by firm wiping with nylon mesh to remove dirt. Teeth were treated with acid to remove calcium, leaving behind a collagen matrix. Samples were then rinsed in water to remove acid, treated with isopropyl alcohol to remove water, and toluene to remove alcohol. Next, samples were treated in melted paraffin, allowing paraffin to penetrate into tooth tissue, followed by placement into paraffin embedding molds (Fisher Scientific), which hardens while cooling. Teeth in paraffin blocks were sectioned at 14 microns using a rotary microtome, mounted on glass slides, and dried on warming plates. Slides were treated with toluene to remove paraffin, then hydrated and dried. Giemsa blood stain (Ricca Chemical Company, Arlington, TX; Cat. No. 3250-16) was used to stain the annuli purple and the other tissue blue and pink. A cover slip was affixed to the slide using resin, and annuli were counted at 40×, 60×, 100×, or 160× using a compound microscope with transmitted light. One person with >30 yr experience read all samples. Teeth were aged assuming a 1 June birthday, and that annulus was formed in winter. The first annulus was not always visible at the tip; in most samples the first annulus clearly visible at the tip was the second year annulus, formed in the second winter of life. Further details about methods used by Matson's Laboratory are proprietary.

Age Estimates from Morphometrics

Morphometrics were obtained from 164 live-captured, anesthetized seals from which incisor teeth were also extracted. Body mass was measured to the nearest 0.1 kg, and standard length, curvilinear length, axial girth, maximum girth, and hip girth were measured to the nearest cm. Age was determined from the annuli of an incisor tooth adjusted for calendar date by adding the tooth age plus the percentage of the year from 10 June (assumed birth date for all seals) to the capture date; this assumes that the tooth annuli count accurately reflects seal age.

Based on these morphometrics, we developed the predictive equation ; we used linear regression to fit the transformed equivalent (Montgomery and Peck 1982) ln  (age) = ln  (β₀) +β*₁x, where ln(age) is the natural logarithm of age, β₀ and β₁ are the regression coefficients, and x is the predictor variable. Initially we fit models with each predictor separately with either a single linear term, or linear and quadratic terms. Next, we used the best of the individual models, based on the bias-corrected version of Akaike's Information Criteria (AICc, Burnham and Anderson 1998), and added the other predictors, again determining the best model based on AICc. Analyses were conducted separately for females and males. We also assessed using a single break point based on curvilinear length for classifying seals as immature or adult. This break point was the curvilinear length that best separated younger seals from adults and was determined graphically.

Accuracy of Age Estimation Methods

We assessed the consistency of age estimates derived from counts of cementum annuli in teeth by examining the distribution of the discrepancies among age estimates from the same seal. In addition, we calculated Pearson correlation coefficients based on the paired age estimates. We examined the accuracy of the age estimates from known-age seals; six based on canines, six from postcanines, 15 from incisors, and 33 from morphometrics (15 morphology-based estimates were from the same animals at different ages).

In previous years researchers in our group had assigned an age class (“field age”) based on visual examination of seals following their capture. To assess our accuracy of subjective estimation of age classifications, we compared our field ages (pup, yearling, subadult, adult [=4 for females, =6 for males]) for a larger data set (n= 544 seals captured from 2003 through 2006) with tooth-based age estimates from incisors of 216 of those seals. To evaluate whether morphometrically derived ages were a better alternative than visual (field) aging, we also compared morphology-based estimates to tooth-based estimates of age for all seals among the 544 that had data for both methods.

Distinguishing Between Immature and Reproductively Mature Seals

We determined how well morphometrically derived age estimates predicted whether seals were mature (i.e., female age < 4 vs. age ≥ 4; male age < 6 vs. age ≥ 6). We estimated misclassification probability as the number of seals assigned to the incorrect age class, divided by the total number of seals classified. These ages were selected as cut-off points based on a study conducted by Lydersen and Kovacs (2005) that determined average age of sexual maturity for harbor seals by measuring estradiol levels in females and testosterone in males. Other studies based on necropsies of harbor seals noted that both sexes typically reach sexual maturity at ages similar to those reported by Lydersen and Kovacs (2005), and average age of first reproduction for female harbor seals throughout their distribution is generally between 4- and 5-yr old (Pitcher 1977, Pitcher and Calkins 1979, Härkönen and Heide-Jørgensen 1990, Bjørge 1992, Lydersen and Kovacs 2005, Hauksson 2006).

Seal Observations After Tooth Extraction

Four seals were recaptured 1 yr after tooth extraction and showed no ill effects as a result of the extraction. An additional 118 individuals were radio tracked following tooth extraction and no mortalities were detected. The average number of days between tooth extraction and last-known location was 443 d (SE = 33, range 4–1,324), although our telemetry results were confounded by indications of transmitter loss or failure, resulting in few relocations for some individuals before radio signals were censored.

Consistency of Age Estimates Among Teeth

Age estimates from canine and postcanine teeth from the same seal generally agreed well, with no systematic bias (i.e., median discrepancy = 0 and symmetric maximum discrepancies) and 90% of the discrepancies were ≤ 1 irrespective of the quality (i.e., Certainty Code provided by Matson's Laboratory) of the age estimates (Table 1). Comparisons involving incisor-based age estimates varied with the Certainty Code of the estimates. If both teeth in a comparison were rated with a Certainty Code of A, the distribution of the discrepancies was similar to comparisons of canine and postcanine estimates (i.e., 0 median discrepancy, 90% of discrepancies ≤ 1; Table 1). Comparisons involving incisor-based estimates were more variable when at least one of the estimates was Certainty Code B; although the median discrepancy was still 0, the 90% range extended beyond 1. In a few instances, the most extreme discrepancies were when age estimates from incisors were much lower than from other teeth (Table 1). Correlations between age estimates from different teeth generally followed the pattern seen with the discrepancies; r= 0.984 for “A” Certainty Codes, with lower Certainty Codes for incisor estimates leading to lower correlations (r= 0.705) than other comparisons (Table 1).

Comparison of Tooth Ages with Known Age of Seals

Although the sample size was small, age estimates from teeth of known-aged seals (Table 2) were highly correlated with their actual ages (canine r= 0.987; postcanine r= 0.996; incisor r= 0.992). These correlations were based on use of all estimates of ages provided for all teeth submitted regardless of Certainty Code. Discrepancies between estimated and known ages were generally larger for older seals, especially for the 29-yr-old captive seal. Although often considered to be the most reliable, canine-based age estimates were not always the most accurate (Table 2). Discrepancies between estimated and known ages were most often negative, indicating that annuli were missed and ages underestimated. A partial explanation for the observed discrepancies likely was because tooth-based estimates, unlike known ages, were not adjusted for the time of year when the sample was acquired.

Morphometric-based Estimates

The morphometrics we obtained all were highly correlated (r > 0.81), so only one to three predictors (curvilinear length, mass, and axial girth) were needed in the regression models. Curvilinear length likely was a better predictor than standard length because, while curvilinear length involves running a tape measure the length of the spine during which any curvature in the spine can be matched and measured, standard length involves holding the tape on a parallel plane to (above) the seal, with an observer at either end sighting down the tape to the tip of the nose and the tip of the tail. The latter measurement introduces three possible sources of measurement error: measuring plane not parallel; straight-line distance may not incorporate curvature in the spine of the seal; observer sighting-angle from tape to end-point of the seal is angled rather than perpendicular.

For each sex there were two models with very similar AICc values (Table 3); each model differing from its pair by an additional variable. To assess the accuracy of predictions based on these morphometric equations, we used the model for each sex with the smaller median residual. The prediction accuracy of these models was moderate for older animals where the predictors changed little for a wide range of ages, once a seal reached adult (maximum) body size.

For our known-age animals, most of which were younger, correlation between morphology-based ages vs. known ages and incisor ages was r > 0.90 (Table 2). Morphology-based age estimates were within 1 yr of the known age for seals ≤ 3-yr old (Table 2), however, older pups and yearlings, especially males, could not be reliably classified into the correct age group (Table 4). Seals > 3-yr old had larger discrepancies between morphology-based age and known age (Table 2).

Identifying Adults

We have three methods that can be used to separate younger seals from adults (i.e., female age < 4 vs. age ≥ 4; male age < 6 vs. age ≥ 6; Lydersen and Kovacs 2005) in a field situation: predictive morphometric-based equations, length-based cut points, and subjective “field” classification. In predicting whether seals were mature, morphometric-based age estimates produced reasonable predictions of incisor-based ages for females with a misclassification probability of 0.051 using the equation (Table 4) or a misclassification probability of 0.045 for a cut point of 142 cm (Fig. 1). For males, a single cut point with a curvilinear length of 162 cm (Fig. 2) was superior (misclassification probability of 0.033) to the predictive equation (misclassification probability of 0.180), which had difficulty correctly assigning older subadults (Table 4); adult males were correctly classified by both methods.

Utility of Subjective “Field” Age Classes

Subjective “field” classification of age was only moderately successful, often classifying seals as one age class older than indicated by the incisor-based age estimate (Table 4a, b). This pattern was evident for both sexes, although slightly more pronounced for males. It should be kept in mind that age estimates from teeth tend to be slightly low, so some of the difference between field and tooth-based age estimates could be due to errors in the tooth estimates, which would contribute to the observed pattern. Morphology-based estimates are somewhat more accurate than field classifications for females (Table 4c), but still have high misclassification probabilities for yearling and subadult males (Table 4d).