Sexual dimorphism in adult Little Stints (Calidris minuta) revealed by DNA sexing and discriminant analysis

Species and study area

The Little Stint is a long-distance migrant wader, which breeds in the Eurasian tundra and spends the non-breeding season from the Mediterranean coasts south throughout sub-Saharan Africa to South Africa in the south and southwestern Asia in the east (Del Hoyo, Elliott & Sargatal, 1996; Tulp et al., 2002; Round et al., 2012). A few individuals are also found further east along the East Asian–Australasian Flyway (Tomkovich, Chih-Yuan & Liu, 2009). Little Stint is a numerous species with an estimated world population of 1,500,000 individuals, a Least Concern conservation status and an increasing population trend (Delany & Scott, 2006; Bird Life International, 2016). Most Little Stints leave the non-breeding grounds and attempt breeding at the end of their first year, but a few stay in South Africa over the austral winter (Underhill, 1997; Underhill et al., 1999). Immature Little Stints can be distinguished from adults by the retained chestnut-fringed inner median coverts, which in adults are pale grey-brown tipped whitish (Prater, Marchant & Vuorinen, 1977). After arrival at the non-breeding grounds immature and adult birds undergo a complete moult, including the replacement of remiges. They complete this moult in January–March, after which the immatures become indistinguishable from the adults (Prater, Marchant & Vuorinen, 1977).

We caught Little Stints at Barberspan Bird Sanctuary (26°33′S, 25°36′E; North West Province, South Africa, Fig. 1). This reserve is centered on a shallow alkaline lake whose area varies from 257 ha to 2,000 ha, depending on rainfall (Milstein, 1975; Barnes, 1998). Lake Barberspan is fed by the Harts River. In the austral winter and during droughts it becomes the only permanent waterbody in the surrounding farmland, which makes it an important stopover, moulting and non-breeding destination for waterbirds, including Palearctic migrants (Barnes, 1998; Lipshutz et al., 2011; Remisiewicz & Avni, 2011). Barberspan Bird Sanctuary is a Wetland of International Importance in terms of the Ramsar Convention and is an Important Bird Area according to Bird Life International (Cowan, 1995; Barnes, 1998; Remisiewicz & Avni, 2011).

Data collection

During September–April in 2008–2016 we captured 348 adult Little Stints using walk-in traps (Busse & Meissner, 2015) and mist nets. We ringed and measured each bird. Measurements included: bill length (from the feather line to the bill tip), total head length (from the back of the skull to the bill tip) and tarsus length (from the tarsal joint to the distal end of the tarso-metatarsus), taken with callipers of 0.1 mm accuracy, and wing length (flattened and straightened wing, as in Evans (1986) and Prater, Marchant & Vuorinen (1977)), tarsus-plus-toe (Anderson, 1975) taken to the nearest 1 mm with a ruler, and body mass to 1 g or 0.1 g in different years (weighed with an electronic scale). We used only morphological measurements taken by MR (SAFRING ringing permit 1454), as recommended in Henry et al. (2015) because measurements taken by different ringers can reduce the accuracy of sex discrimination. We also took blood samples from a brachial vein (Owen, 2011) from all ringed Little Stints and preserved the samples in 96% ethanol for DNA sexing. Only trained, experienced team members handled the birds and took the blood samples to ensure safety standards. All the procedures were approved by the management of Barberspan Bird Sanctuary, under permits from SAFRING and the Department of Rural, Environmental and Agricultural Development, North West Provincial Government, South Africa.

Molecular sex identification

We isolated DNA from blood samples using the Blood Mini Kit (A&A Biotechnology, Gdynia, Poland). The next step was PCR with the pair of P2 and P8 primers (Griffiths et al., 1998) using a modified protocol. Total volume of PCR was 20 µl, the reaction mix included: 7.5 µl REDTaq^® ReadyMix™ (Sigma Aldrich, St. Louis, MO, USA), 3.5 µl of water, 1 µl MgCl₂, 1 µl P2 primer (10 µM), 1 µl P8 primer (10 µM) and 2 µl of the DNA sample. For PCR amplifications we used an Eppendorf Mastercycler with this thermic profile: an initialisation at 94 °C for 2 min, 40 cycles of denaturation at 94 °C for 30 s, 40 cycles of annealing at 50 °C for 30 s and 40 cycles of elongation at 72 °C for 2 min, ending with a final elongation at 72 °C for 5 min. This method is based on the amplification of chromo-helicase-DNA-binding (CHD) genes found on avian sex chromosomes. The CHD-Z gene (ca 370 bp) is located on the Z chromosome, therefore it is present in both sexes. The CHD-W gene (ca 390 bp) occurs only on chromosome W, therefore it is unique to females (Griffiths et al., 1998). We separated the products of PCR with electrophoresis on 3.5% agarose gel (75 V, 120 min) stained with ethidium bromide (samples from 2008–2013) and Midori Green Advanced DNA Stain (NIPPON Genetics, samples from 2013–2016). Products were visualised in UV light, one band was visible for males, which indicated ZZ chromosomes, and two bands for females (ZW chromosomes). The method enabled us to sex all birds sampled.

Statistical analysis

For each morphometric measurement we used Storer’s dimorphism index (SDI) to assess the degree of sexual size dimorphism in Little Stints (Storer, 1966), using the formula: ${\displaystyle \mathrm{SDI}=\frac{\text{mean}♀-\text{mean}♂}{\left(\text{mean}♂+\text{mean}♀\right)\ast 0.5}\ast 100.}$

We compared all the morphometrics we had measured of males and females using the two-sample t-test. Then we divided the birds into two groups: those before and those after primary moult, and compared the measurements of birds from these two groups. Birds caught in active moult were classified as “before moult” because they moult their outermost primary last. However, we did not measure wing length of any bird whose outermost primary was heavily damaged or if it was growing. For each group we used discriminant function analysis to determine the best set of measurements for sexing Little Stints with a two-fold approach. First, we used wing length alone as a discriminant factor. The second approach used a stepwise method including other measurements after conducting pairwise correlation of all the measurements. We used only one of a pair of correlated measurements at a time in the stepwise procedure to avoid multicollinearity. We did not include the body mass of Little Stints, because it changes during the non-breeding season during pre-migratory fuelling. The aim of producing two different discriminant functions for each group was to make these functions applicable for different data sets, because wing length is the most commonly taken measurement, in contrast to tarsus and total head lengths. We present the discriminant functions developed using different sets of morphometrics in the Supplemental Information. All the assumptions of discriminant function analysis were met (Tabachnick & Fidell, 2001), including the homogeneity of covariances (Box’s M test), the homogeneity of variance (Levene’s test), and the normal distributions of the measurements for males and females separately in each of the two groups. We confirmed no multicollinearity of the selected measurements (r < 0.50 for all pairwise correlations). We computed prior classification probabilities from the group sizes because of the unequal number of males and females in our sample (Tabachnick & Fidell, 2001). To validate our models we used a jack-knife procedure to assess the percentage of correctly sexed individuals by discriminant function analysis (Dechaume-Moncharmont, Monceau & Cézilly, 2011). This cross-validation technique predicts the sex of each individual using a discriminant function calculated for all the birds except the individual being classified (Hair Jr et al., 1995). We had unequal samples of males and females, so we assessed the effectiveness of our proposed functions by calculating Cohen’s kappa statistic (Titus & Mosher, 1984), which estimates the improvement made by the results of discriminant analysis over random chance: 0 = no improvement over chance, 1 = full compliance (Titus & Mosher, 1984). The optimal cutting score was calculated as a weighted average of the group centroids (Hair Jr et al., 1995). Statistical analyses were performed in IBM SPSS Statistics for Windows, version 22.0 (IBM Corp., Armonk, N.Y., USA). All tests were two-tailed and the accepted level of significance was P < 0.05.

Morphological differences between the sexes

We identified 185 males and 163 females using DNA sexing. Analysis of morphometrics and Storer’s dimorphism index (SDI) also revealed sexual differences. On average females were bigger than males in all morphological measurements (Table 1). The most dimorphic features were respectively bill, wing and total head lengths (Table 1). In all our measurements birds before and after wing moult differed only in wing length (t₃₁₁ = 7.69, P < 0.001), which was on average 2.3 mm longer in those with fresh primaries after moult than in those before moult with worn primaries. We therefore conducted the discriminant analyses separately for these two groups.

Some morphological measurements were correlated (correlation coefficients for males and females between tarsus and tarsus-plus-toe were r = 0.73 and r = 0.79 and for culmen and total head length r = 0.81 and r = 0.83). We chose wing length, total head length (because it is less prone to errors than bill length measured to the feather line, which might be worn difficult to determine Prater, Marchant & Vuorinen, 1977), and tarsus length as the best factors for discriminant analysis.

Discriminant functions for adult Little Stints before primary moult

Using measurements of 156 adult Little Stints (70 females and 86 males) taken before they had moulted their primaries, and using only wing length as a discriminant factor, we obtained the equation: ${\displaystyle D_1=-47.496+0.484\left(\text{wing}\right),}$ which allowed us to correctly classify 78.8% of the birds. A jack-knife cross-validation procedure yielded the same success rate, and our random chance-corrected procedure showed that our proposed classification was 56.8% better than chance (kappa = 0.568 ±0.082 SE, P < 0.001). If D₁ > 0.17, the bird was classified as a female, and if D₁ < 0.17 as a male (Fig. 2).

In the stepwise procedure including three selected measurements (wing, tarsus and total head length), only the combination of wing and tarsus length was a significant discriminant factor. The best discriminant function we obtained was ${\displaystyle D_2=-50.428+0.421\left(\text{wing}\right)+0.420\left(\text{tarsus}\right).}$ This function correctly classified 82.7% of the birds and was more accurate than the previous equation. The cross-validation procedure correctly classified 80.8% of individuals and the classification was 65.1% better than chance (kappa = 0.651 ±0.081 SE, P < 0.001). If D₂ > 0.18 then the individual was a female and if D₂ < 0.18 it was a male (Fig. 3).

Discriminant functions for adult Little Stints after primary moult

The second group we analysed comprised 159 adult Little Stints (82 females and 77 males) with all new primaries after their complete post-breeding moult. The function obtained using only wing length was ${\displaystyle D_3=-52.184+0.520\left(\text{wing}\right).}$ This function correctly classified 83.4% of the individual birds, which was confirmed by the cross-validation procedure. The proposed classification was 66.9% better than chance (kappa = 0.669 ± 0.080 SE, P < 0.001). When D₃ >  − 0.038 the individual was classified as a female, when D₃ <  − 0.038 as a male (Fig. 2).

The stepwise analysis revealed that the best discriminating combination of measurements was wing and total head length (THL), according to the equation ${\displaystyle D_4=-59.310+0.445\left(\text{wing}\right)+0.377\left(\text{THL}\right).}$ This function correctly classified 84.7% of the individuals in the sample, which was the highest proportion of all the equations we present. The cross-validation procedure showed that the proposed equation was 84.1% correct and 69.4% better than chance (kappa = 0.694 ±0.080 SE, P < 0.001). If D₄ >  − 0.041 a Little Stint was classified as a female and if D₄ <  − 0.041 as a male (Fig. 3).