Population Biology of Meroles anchietae and an Elusive Sex Ratio

Study Site

The study site, Helga's Dune, is in the central Namib Desert on the northern margin of the Southern Sand Sea in west-central Namibia near the Desert Ecological Research Unit field station at Gobabeb, Namibia. Helga's Dune is a relatively small, isolated linear dune 2.10 km long with a width ranging from 52–210 m and a maximum height of about 40 m. The dune trends north–south with the long axis at 339° 29’ and lies in an interdunal valley between two much larger dunes. The valley isolated Helga's Dune lizards from those on nearby dunes. Helga's Dune contained 12 slipfaces which comprised about 4% of its area. It harbored a population of shovel-snouted Lizards, M. anchietae, which dive into slipfaces to escape predators and temperature extremes. The dune was sparsely vegetated, with only a few individuals of Nara shrubs (Acanthosicyos horrida) and perennial grass (Stipagrostis sabulicola). Annual rainfall at Gobabeb between 1990 and 2006 was less than 25 mm except for two years of exceptional rainfall: 72.7 mm in 1997 and 99.2 mm in 2006, both of which were survey years. Following occasional heavy rains, the base of the dune is ringed by annual grass (Stipagrostis gonatostachys), which is also scattered across the sand and gravel substrate of the interdune valley. A bi-directional wind regime causes the linear nature of the dune, and a series of slipfaces form leeward to the prevailing wind. The geomorphology and ecology of the dune are described in detail in Besler et al. (2013) and Robinson and Seely (1980).

Survey Protocol

We sampled the study site for M. anchietae multiple times within each of seven surveys between 1997 and 2006. We sampled once in the morning and once in the afternoon during M. anchietae's bimodal activity periods. Poor weather conditions sometimes precluded sampling during morning (fog) or afternoon (wind) activity periods.

Observers walked the dune crest and base looking for active lizards. We captured lizards by hand either by pursuing active lizards until they buried themselves or by tracking lizards that were active earlier and digging where their tracks indicated they had buried themselves in a slipface.

We captured any unmarked individuals we encountered, determined their sex by probing for hemipenes, measured their snout-vent length (SVL) with a clear plastic ruler (SVL, ±1 mm), and mass with a Pesola Spring Balance (± 0.1 g). We marked adults permanently with size 14/0 colored beads (Fisher and Muth, 1989). Juveniles were too small to bead. We toe-clipped (Medica et al., 1971) juveniles, limiting toe clips to the ungual phalanx to avoid damage to toe fringes and temporarily marked juveniles by writing their identification number mid-dorsally with a black Sharpie® marking pen. After marking individuals, we did not physically recapture them during the same field season. Marked lizards were identified from a distance using binoculars.

Adults were assigned to a sex group by probing for hemipenes but because smaller juveniles were too small to probe safely, we used presence of yellow coloration to identify juveniles (Louw and Holm, 1972) and treated them separately from males and females. Because yellow color fades gradually (Louw and Holm, 1972) and differently among individuals, there was some overlap in body size among the largest juveniles and smallest adults.

Body Size Analysis

We analyzed body size meristics to determine whether the relationship between body mass and SVL differed by sex and by annual rainfall categories. We transformed (natural log) meristics and centered them on mean of ln SVL prior to linear regression (Muth et al., 2023).

We analyzed the relationship among sex groups using analysis of variance with three categorical variables (female, male, and juvenile) to test for homogeneity of slopes (H₀: all slopes are equal) and differences among intercepts (H₀: All intercepts are equal). If intercepts differed significantly, we compared residuals by sex group using analysis of variance with Tukey Honestly Significant Difference as a post hoc test.

Body Size Analysis and Rainfall

We compared SVL and mass between rainfall categories that reflected the annual rainfall variation at Gobabeb, “typical/normal” (<25 mm) and “wet” (>25 mm) years, analyzing sex groups independently. Prior to this analysis, we had determined that body size differed among females, males, and juveniles. Hence, we computed regression equations for normal and wet years and tested for homogeneity of slope and equal intercepts as noted earlier. Lizard growth rates and body mass are influenced by variable rainfall and seasonal availability of resources (Dunham, 1978; Muth et al., 2023).

Sex Ratio and Encounter Probability

We used Pearson χ² analysis to determine whether there was a significant difference in: (1) number of individual males and females captured and (2) number of total captures (initial plus recaptures) of sexes for each survey year. We used Pearson χ² analysis for all years combined to determine whether there was a significant difference in cumulative number of captures of individuals of each sex and cumulative number of total captures (initial plus recaptures) of the sexes.

To explore possible behavioral differences between sexes, we analyzed capture probability between adult males and females using the Cormack-Jolly-Seber method (Cormack, 1964; Jolly, 1965; Seber, 1965) implemented in Program MARK (https://sites.warnercnr.colostate.edu/gwhite/program-mark). This step involved analyzing two interconnected probabilities: survival and encounter. We analyzed survey years independently to examine differences between sexes during the short duration (approximately 2 weeks) of each survey. We hypothesized that the population was effectively a closed population during a given survey and that survival probabilities between sample days likely approached 1.0 because inter-sample interval was relatively short. To increase sample size, we pooled morning and afternoon captures into daily captures, using only adults.

Within a survey year, we first modeled survival differently among samples and between the two sex groups and capture probability differently among samples and between sexes. The full model, written in notation suggested by Lebreton et al. (1992), is {ϕ_g*t p_g*t} where ϕ is survival probability, p is capture probability, and the subscript g*t signifies that the probabilities differ by sex group (g) and by survey (time, t). We used the Fletcher-ĉ (Fletcher 2012), calculated by Program MARK, as a measure of goodness-of-fit where a perfect fit is 1.00. A difference of 0.01 required adjusting ĉ, which likewise adjusted the Akaike Information Criterion (AIC_C; Akaike, 1973; Sugiura, 1978; Hurvich and Tsai, 1989) to obtain quasi-AIC_C (QAIC_C); Program MARK uses AIC_C and QAIC_C to rank and weight models. Both survival (ϕ) and encounter probability (p) can differ between sex groups (g), among sample days (t), both (g*t), or neither (.). The result was 15 potential models in addition to the full model for each year. We ran all 16 models so there were equal numbers with and without a group factor allowing comparison of summed AIC_C weights; the “all-combinations model strategy” with model averaging recommended by Doherty et al. (2012). Thus, the sum of AIC_C weights for models with and without sex groups can be compared to look for sex differences in capture probabilities (i.e., behavioral differences between sexes). Burnham and Anderson (2004) and Cooch and White (2019) suggested that summing support (weight) over models is superior to inferences made based only on the best model. The full models for all years revealed Fletcher-ĉ near 1.00 (mean 1.008, range 0.917–1.092) but some required adjustment. We used the following sixteen models each year: {ϕ_g*t p_g*t}, {ϕ_g p_g*t}, {ϕ_g p_g}, {ϕ_g p_t}, {ϕ_g p_.}, {ϕ_t p_g*t}, {ϕ_t p_g}, {ϕ_t p_t}, {ϕ_t p_.}, {ϕ_. p_g*t}, {ϕ_. p_g}, {ϕ_. p_t}, {ϕ_. p_.}, {ϕ_g*t p_g }, {ϕ_g*t p_t}, and {ϕ_g*t p.}.

Population Size Estimates

We estimated population size, N̂, for each year using Huggins Closed Capture models (Huggins, 1989, 1991) using Program MARK. Huggins Closed Capture uses probability of capture (initial, p, and recapture, c) in terms of sex group and time and then removes the number of animals in the population that are never captured. The Huggins Closed Capture technique allows estimates of N̂ (the estimated population size) by sex group where groups can have differing capture probabilities (both p and c). We pooled capture data from both morning and afternoon activity periods of each survey day and estimated N̂ for each year independently because we could only attest to the closed population assumption within a survey year but not during intervening years. Summed AIC_C weights for those models that modeled sex independently were compared with those models that lacked the sex variable, again using the “all-combinations model strategy” (Doherty et al., 2012).

Activity Frequency and Duration

Herein we distinguish between “activity period” and “emergence period.” Activity period refers to bimodal morning times, afternoon times, or both, when lizards emerge from sand and are on the surface. Emergence period refers to amount of time—duration—that the lizard spends on the surface.

We used focal observations to generalize frequency and duration of emergence and activity of adult males and females for two years: 2002 (2–8 November) and 2009 (1–4 December). December 2009 was later in the summer and was consistently warmer than November 2002. Time of day that activity occurred and duration of surface activity are expected to differ based solely on environmental temperatures. Therefore, we did not combine observational data for 2002 and 2009. However, the observations were focused on differences in frequency and duration of emergence periods between males and females and not on environmental temperatures or time of day that an observation occurred. The activity and emergence period observations were not conducted in conjunction with mark-recapture surveys.

In 2002, we focused on duration of emergence periods: amount of time elapsed from a lizard's (n = 30) emergence from sand to the time it buried (submerged), as well as on time budgets: proportion of time emerged individuals (n = 25) spent moving (active) versus stationary. Observers arrived at a slipface prior to morning or afternoon emergence and seated themselves about 20 m from the base of the slipface to monitor marked individuals. Observers were stationary and moved as little as possible to not disturb lizards, and previously marked individuals appeared oblivious to our presence. As a marked individual emerged, we recorded time of emergence as well as time of initiation and termination of movement by that individual until it submerged. We calculated average length of emergence periods for each sex and for both sexes combined and compared means for each sex using a two-sample t-test. We also calculated proportion of time spent moving versus stationary for each sex and compared them using the Mann-Whitney U test.

In 2009 (1 December to 4 December), we focused on duration of activity of individuals (n = 14) inhabiting a small, isolated portion of dune containing a slipface. During the first day, we marked all active adults on the dune. On subsequent days and activity periods, we arrived prior to activity and recorded when each individual lizard emerged and submerged and verified that they still emerged throughout the activity period by periodically accounting for each individual lizard.

Energy Expenditure

We used doubly-labeled water techniques (Nagy, 1989) to determine FMR of 12 females and 22 males in 2001. If females spend less time emerged on the surface than males, as postulated in our previous surveys, then they should expend less energy (lower FMRs) than males when scaled by body mass. We estimated that the biological half-life of tritiated water in Meroles is about 9 days. We used a tritium dose of 1.0 μCi/kg or about 0.0045 kilograms per lizard. Blood samples were taken after 9 days and were transported to California for analysis in the laboratory of K. Nagy at the University of California, Los Angeles.

Statistical Analyses

We used SYSTAT® 10.0 (IBM, 2000) for all computational and statistical procedures. To determine the significance of a statistical test, we set α = 0.05.

Survey Results

Survey dates and duration of field time varied from year to year (Table 1). Number of sample days per survey varied from 7 to 17 days for a total of 76 sample days and 133 activity period samples.

Table 1.Survey dates and number of samples.

View Fullscreen

Body Size Analysis

The largest lizards were 57 mm SVL and 7.0 g for males and 50 mm SVL and 5.5 g for females. Descriptive statistics by survey year and all years combined are given in Table 2.

Table 2.Descriptive statistics, mean, and SD of snout-vent length (mm) and body mass (g) of male, female, and juvenile lizards by year and all years combined. (X ≁ N) indicates that the distribution of a variable was nonnormal.

View Fullscreen

Our analyses of the length-mass relationship resulted in the following regression equations: for males, y = 2.9042x + 1.4506; n = 219, R² = 0.584, P < 0.0001 (1 outlier omitted); for females, y = 3.1808x + 1.1804; n = 159, R² = 0.711, P < 0.0001; and for juveniles, y = 3.1968x + 0.2449; n = 183, R² = 0.881, P < 0.0001. Slopes did not differ significantly among males, females, and juveniles (F_{2, 555} = 2.356, P = 0.0957). However, intercepts

differed significantly among regression lines (F_{2, 557} = 5.86, P = 0.0030). A Tukey Honestly Significant Difference post hoc test revealed that differences between pairs of sex groups were only significant between females and juveniles (P = 0.0147) and between females and males (P = 0.0042). The difference between males and juveniles was not significant (P = 0.9531). For further analysis, we treated females, males, and juveniles independently.

Body Size Analysis and Rainfall

We compared the length-mass relationship during wet versus ‘average’ rainfall years. For females, the resulting regression equations (centered at the mean ln SVL) were: for wet years, y = 3.2919x + 1.2381; n = 46, R² = 0.948, P < 0.0001; and for average years, y = 3.1226x + 0.8996; n = 113, R² = 0.928, P < 0.0001. Slopes were homogeneous (F_{1, 154} = 1.650, P = 0.2008) but intercepts differed significantly (F_{1, 155} = 9.223, P = 0.0028). Females were heavier per unit length during wet years. We fit the mean ln SVL of all females, 3.765, into each regression equation by first subtracting from it the mean ln SVL from each rain group to center the data. At the maximum body length (50 mm), the back-transformed mass differed by 0.78 g; females weighed more during wet years. For males, neither slopes nor intercepts differed significantly between rainfall categories (F_{1, 215} = 0.054, P = 0.8159; F_{1, 216} = 0.514, P = 0.4743; slopes and intercepts, respectively); thus, we omitted regression equations for the rain groups. For juveniles, slopes differed significantly between rainfall categories (F_{1, 178} = 6.038, P = 0.0150) making a comparison moot between intercepts.

Sex Ratio

Results of the Pearson χ² analysis (Table 3) varied by year and indicate that sex ratio of adults varied depending on if the calculation was based on cumulative number of individuals captured or on cumulative number of total captures (initial plus recaptures) of the sexes. For individuals, the sex ratio was male-biased for 2 years, but for 5 years was 1:1. However, the sex ratio of individuals for all survey years combined was male-biased. During each survey year the sex ratio of total captures was strongly male-biased.

Table 3.Number of individuals captured and the total number of captures (initial plus recaptures) by sex and adult sex ratios. Differences between male and female frequencies were compared using Pearson χ² where H₀ was a 50:50 sex ratio. H₀ was rejected (bold font) when χ² is greater than the critical value of 3.841 at 1 DF and P > 0.05.

View Fullscreen

Survival and Detection Probabilities

Probability estimates for apparent survival (ϕ) and capture (p) with 95% confidence limits are given in Table 4. As expected, survival probability during each short survey period approached 1.00. Estimates of male capture probability were higher than for females. However, 95% confidence intervals overlapped in three of the seven years: 2001, 2005 and 2006. Likewise, the summed AIC_C weights for all models within a year that modeled sex group independently in capture probability, p(g) or p(g*t), showed strong support for those models in all but two years: 2001 and 2005 (Table 4).

Table 4. Apparent capture (p) and survival (Φ) probabilities estimated using Program MARK and model averaging. Summed AIC_c weights where sex group was modeled separately for capture probability indicated high support for modeling sexes independently with the exception of years 2001 and 2005; apparent probability of capture for males differed from that of females. This sex group difference did not extend to apparent survival probability, which, for most years, approached 1.00.

View Fullscreen

Population Size Estimate

The number of individuals captured each year and the estimated population size (N̂) are given in Table 5. The summed AIC_C weights for those models that modeled sex independently ranged from 0.98 to 1.00, thus showing very strong support. For the Huggins (1989, 1991) Closed Captures method, initial capture probability, p, and recapture probability, c, can be modeled as being independent or as being equivalent. Summed AIC_C weights when p = c was 1.00 for 5 years, was 0.75 for the year 2003, and was 0.00 for the year 2006.

Table 5.Number of individuals identified of each sex (n) compared with estimates (N̂) and 95% confidence intervals (CI). Population size estimates for the two sex groups were similar, and CIs overlapped in all years except 1999.

View Fullscreen

Activity Frequency and Duration

Emergence-time, the time an individual spent from first emerging from sand to burying at the end of its emergence period, was compared for 30 individuals. There was no significant difference between mean emergence duration for morning (80.9 min., standard deviation (SD) = 39.1) and afternoon (58.1 min., SD = 33.1) activity periods for a total of 20 morning and 10 afternoon values (t = 1.63, df = 1, P = 0.109). Surface times for 23 males and 7 females showed that males spent significantly more time on the surface per emergence period than did females (t = 3.197, df = 1, P = 0.004). Males spent a mean of 80.9 minutes on the surface whereas females spent a mean of only 48.4 minutes on the surface. The mean time from when the first individual emerged to when the last individual buried during an emergence period was 185.5 min (SD = 19.1) for the two morning emergence periods that lacked fog and 156.5 min (SD = 0.71) for the two evening emergence periods.

In 2002, there were 13 observation periods: 7 in the morning and 6 in the afternoon. We were able to follow a total of eight individuals, four males and four females, through the entirety of an emergence period. From these data we calculated the average duration to be 49.4 min (n = 8), ranging from 4 to 150 min. Males averaged longer emergence periods than females: the average for males was 75.5 min. (n = 4, range 22–150 min) and for females 23.2 min. (n = 4, range 4–50 min). But emerged females spent a significantly greater proportion of their time moving than males (Mann-Whitney U statistic = 118, P = 0.0034). Males (n = 17) were actively moving (walking, feeding, displaying, etc.) 50.7% of the time they were on the surface, whereas females (n = 8) spent 89.4% of their time on the surface actively moving. Thus, even though females emerged for a shorter period of time, they spent more of that time “active” than males who were inactive about 50% of the time.

In 2009, we conducted five observation periods: three in the morning and two in the afternoon, using a total of fourteen individuals, eight males and six females that were previously captured and marked. Cumulative observation frequencies ranged from zero to five during the five emergence periods during which we made focal observations (Table 6). On average, 8.0 marked individuals (5.8 marked males and 2.2 marked females) were seen per activity period. We counted an individual only once in each activity period and over the course of the five activity periods, the fourteen individuals accounted for 40 sightings. Frequency of sightings of males (n = 29) and of females (n = 11) differed slightly but significantly (χ² = 3.97, df = 1, P = 0.048). Males were active on the surface more frequently and longer than females and thus were sighted more often than females in mark-recapture surveys.

Table 6. Initial capture data and subsequent resightings (×) during five activity periods. Lizards were marked on 1 December 2009.

View Fullscreen

We were able to follow only two individuals through their entire emergence period. One male emerged for a 136 min emergence period of which it spent 25 min active and 111 min stationary. One female emerged for 13 min (7 min active and 6 min stationary). As generalized in 2002, the male was active on the surface (stationary and active) much longer than the female.

Energy Expenditure

Contrary to our expectation, there was no significant difference in FMR between males and females (Table 7). Mass-specific FMR of females was greater than that of males.

Table 7.FMR of 12 adult female and 22 adult male Meroles anchietae where FMR_p is predicted field metabolic rate. Elapsed time is number of days between injection of doubly-labeled water and blood sampling. There was no difference in mean FMR of females and males (t-test). * = P < .05, ** = P < .01, *** = P < .005, NS = not significant (P > .05).

View Fullscreen