In recent years, we have begun to appreciate that social behaviours might exhibit repeatable among-individual variation. Such behavioural traits may even covary and have critical evolutionary implications. Importantly, some social behaviours such as aggressiveness have been shown to provide fitness benefits, including higher reproductive success and survival. However, fitness consequences of affiliative behaviours, especially between or among sexes, can be more challenging to establish. Using a longitudinal behavioural dataset (2014-2021) collected on eastern water dragons (Intellagama lesueurii), we investigated whether various aspects of affiliative behaviour (1) were repeatable across years, (2) covaried with each other at the among-individual level, and (3) influenced individuals’ fitness. In particular, we considered affiliative behaviours towards opposite-sex and same-sex conspecifics separately. We found that social traits were repeatable and covaried with each other similarly for both sexes. More notably, we found that male reproductive success was positively correlated with the number of female associates and the proportion of time spent with females, whilst females’ reproductive success was not correlated with any of the measured social behaviour metrics.  Overall, these findings suggest that selection may be acting differently on social behaviour of male and female eastern water dragons.

Data collection

We used data collected as part of an ongoing behavioural study that started in 2010 of an urban population of Eastern water dragons inhabiting Roma Street Parkland (RSP), Brisbane, Australia (27°27’46’S, 153°1’11’E). The RSP population has an estimated size of 336 individuals which are highly habituated to human presence. The parkland spans approximately 16 hectares and is composed of various habitats differing in size, complexity and heterogeneity of structure, vegetation and access to water.

Behavioural data

We carried out behavioural surveys twice a day (0730-1030 and 1300-1600), by following a transect which covers approximately 85% of the known population, three days a week from September to April (when dragons are most active) each year. For each individual, we took a photo of their left and/or right profile, along with their GPS location which was later used to calculate spatial proximity and define social associations (see ‘Social traits’ section). An individual’s sex was assigned based on sexual dimorphism and dichromatism. We then recorded the individual’s immediate behaviour at the time of the observation. Behaviours included a spectrum from resting to agonism (e.g., head bobs, tail slaps, arm waves, chasing, and physical confrontations, see for more details).

Profile photographs individual were used to identify individuals using a previously established method for this population. We employed an interactive identification software (I3S Spot, v4.0.2), comparing individual facial patterns with an established photo library. Those without a match were considered new individuals.

Genetic data

Animals were caught during sessions independent from the behavioural surveys. Each season all possible adult individuals were caught using a lassoing technique. A photo was taken of each individual for identification upon capture. Blood samples were collected using the ventral tail caudal venepuncture technique, while tissue samples consisted of collecting the tip of the tail. DNA was extracted using DNeasy extraction kits (Qiagen) and sequenced using restriction-associated digest methods at Diversity Arrays Technology, Canberra, using proprietary DArTcapTM technology. This method applies a selective step after complexity reduction to genotype specific markers from DArTseq representations, which is a reduced representation sequencing approach, similar to RAD-sequencing. Single nucleotide polymorphisms (SNPs) were identified using the DArTsoftS pipeline. This resulted in a total of 6,425 SNPs prior to filtering, including 510 juveniles (which were not included in analyses, but were used in parentage assignment) and 775 male and female adult individuals.

Social traits

Social associations

Social associations were defined as pairs of individuals socially tolerating each other in close spatial proximity (1.85m). Details of these measurements can be found in the Supplementary methods.

For each year, the following aspects of social behaviour were measured: (1) proportion of time an individual spent within 1.85m of conspecifics (i.e., social tendency), (2) number of associates (i.e., degree), and (3) average half-weight association index (hereafter referred to as HWI, see supplementary methods for exact calculations). HWI estimates the strength of social associations using the weighted proportion of time pairs of individuals spend together. Once estimated, each social behaviour was divided into two categories: opposite-sex associations, and same-sex associations. All social behaviours were calculated in individuals’ core home ranges (detailed under ‘Conspecific density’ methods), and only individuals with at least 30 sightings per year were included, as this represents the minimum number of sightings required to estimate stable social measures in this species.

Conspecific density

We calculated the density of conspecifics within each individual’s core home range (HR50). Core home ranges were estimated using the kernel utilization distribution method (UD) in the adehabitatHR package, and a previously optimised smoothing parameter of 7m was used to control for the amount of variation around density estimates. The coordinates of the 50% probability contour were then extracted to obtain HR50. The core home range was chosen as this represents individuals’ primary territory and is where most social associations occur. Density of conspecifics (hereafter density) was estimated for individuals with at least 25 sightings, as this is the minimum number required to calculate robust home range estimates. Once estimated, density of conspecifics was divided into two categories, namely opposite-sex density (i.e., containing overlapping individuals of opposite-sex only) and same-sex density (i.e., containing overlapping individuals of same-sex only).

Fitness

Fitness was defined here as lifetime reproductive success, which was estimated by calculating the number of offspring that survived to adulthood (hereafter referred to as ‘number of offspring'). To estimate the number of offspring, parentage assignments were performed in the sequoia package, using SNP data, within the R statistical environment. Sequoia uses a fast, heuristic hill-climbing algorithm, which has been shown to result in low error and high assignment rates with only a few hundred highly informative, unlinked SNPS. SNP data for use in parentage assignment were obtained by applying the following filters: (1) read depth of at least 5 reads per genotype, (2) individual call rate greater than or equal to 70%, (3) SNPs call rate greater than or equal to 99%, (4) proportion of technical replicate assay pairs for which the marker score was consistently greater than or equal to 99%. Minor allele frequency threshold (MAF) was chosen by undertaking sensitivity analyses, where we generated a set of files in which SNPs were filtered using different MAF threshold values (range: 0.38 – 0.47). We then performed parentage assignments for each dataset and compared the outputs.

Parentage assignments obtained from Sequoia were validated in two ways. First, we compared them to known maternities resulting from field observations of nesting events [40]. Second, genetic relatedness was calculated using the maximum-likelihood dyadic relatedness estimator in Coancestry. We then plotted parentage assignments against relatedness estimates.  SNPs used for the relatedness analysis were obtained from less stringent filtering (individual call rate ≥ 80%, SNP call rate ≥ 95%, MAF ≥ 0.05, proportion of technical replicate assay pairs ≥ 98%), as a higher density of SNPs may increase the accuracy of relatedness estimation. This resulted in a total number of 2100 SNPs to estimate relatedness.

The best MAF value, as indicated by the highest parentage assignment and absence of disagreements with field observations was MAF ≥ 0.43, which resulted in a sample size of 179 SNPs used for parentage assignment. From the 775 adult individuals included in the parentage assignment, 239 dams and 296 sires were assigned surviving adult offspring. When validating assignments against known maternities, 98 from the 106 were correctly assigned (92%), while 6 mother-offspring pairs were not assigned (6%), and 2 were mismatched (2%). These mismatches, however, could be the result of mis-assigned mothers from field observations, due to nest-sharing events (C.D. pers. observation). While it is possible that we may have missed genotyping individuals that would have been assigned to other parents, this likelihood is low (overall genetic sampling rate from the entire population was 80%).

Excel