Data for: Genetic relatedness shapes social dynamics in a threatened finch: Implications for population assessment

Tropical granivorous finches often form large flocks around resources. The composition of these flocks, whether they are random groups of individuals or comprise related birds travelling together, is currently unknown. Understanding this distinction would aid in assessing the accuracy of population counts. To bridge this knowledge gap, we combined high-frequency location tracking with comprehensive genetic sequencing to investigate the relationship between pairwise association strength and genetic relatedness in Gouldian finches (Erythrura gouldiae). Our study revealed that birds captured near each other were more inclined to travel together, and their relatedness was significantly linked to the strength of their association. These findings suggest that within-flock associations are influenced by genetic relatedness, contributing to the stability of the flock size. We propose that juvenile kin associations play a pivotal role in this dynamic, potentially enhancing survival rates by forming sibling subgroups. The consistent flock sizes of Gouldian finches during movement have implications for estimating population sizes from waterhole counts, allowing flocks to be considered as distinct units for concurrent counts at multiple waterholes. This approach would offer a reasonably accurate method for estimating local populations, and conducting repeated counts on consecutive days could provide reliable and replicable results.

Methods
Seventy-six Gouldian finches were captured across three capture periods (sessions) in May, July, and October 2019. Birds were trapped at waterholes using mist nets, and each finch was uniquely banded, and a 70 µm blood sample for genetic analyses was extracted. Tracking data were stored and can be accessed via the Motus Wildlife Tracking System (Motus project #241).

Social associations

Raw detection data was downloaded from the Motus platform (www.motus.org) and processed to remove detections within the first 24 hours post-release to allow for a post-tagging acclimation period. Detections with run lengths (number of consecutive uniquely coded detections by a receiver) of < 4 were also removed to exclude observations with a higher probability of false detection. 

We identified associations between tagged conspecifics following a ‘gambit of the group’ approach, with individuals considered to be associating if they were observed co-occurring at fixed receivers simultaneously. We used the data from the first 40 days after the last sampling date to maximize the number of active tags after excluding the 24-hour acclimation period. This left a total of 57 active tags. As individuals could be detected at multiple receivers simultaneously due to the overlapping detection ranges, individuals were assigned to the receiver with the strongest signal strength (sig) every 10 minutes. We then counted the number of co-occurrences per pair of tagged conspecifics in a time window of one hour (i.e., the number of times the strongest sig was detected at the same tower within an hour). We then used the Simple Ratio Index (SRI) to determine the association strength between each pair of tagged conspecifics. Because most individuals were not present throughout all three sampling sessions, we calculated SRI for each tracking session separately.

Pairwise genetic distance

DNA extraction and SNP (single-nucleotide polymorphism) genotyping were carried out by Diversity Arrays Technology (https://www.diversityarrays.com/) using DArTseq™ protocols (Schouten et al., 2012). We filtered the data using the ‘dartR’ package in R version 4.2.2 according to repeatability (repeatability > 0.95), removing monomorphs and custom filtering to remove SNPs with a mean read depth of less than 15 and a mean ratio of sequence depth between alleles of greater than two.

Using the filtered dataset, we calculated the pairwise genetic distance between all pairs of individuals using the individual genetic distance measure of Smouse and Peakall (1999) in the ‘PopGenReport’ package in R.