Code and data for: Familiarity breeds success: pairs that meet earlier experience increased breeding performance in a wild bird population

This is a Data package that contains three separate datasets and the analysis code for a manuscript 'Familiarity breeds success: pairs that meet earlier experience increased breeding performance in a wild bird population '. The first Dataset (pairs_data_2007_10) and the second dataset (pairs_data_2011_14) contain the data used to analyse the influence of meeting time of a pair of Great tits (i.e. month in the first dataset, week in the second dataset, when a pair was first detected in a flock together) on different components of breeding success of a pair. The third dataset (meeting_and_divorce) was used to analyse whether meeting time of a pair (in winter prior to the breeding season t) influence the probability of a pair to stay together or separate (divorce) to the following breeding season (t+1).

Methods
Study system

The great tit is a small, short-lived (adult survival rates 25-65%), cavity-nesting passerine. Pairs are socially monogamous and breed on distinct territories, laying on average 8.5 eggs, and fledging on average 7 chicks. After the breeding season, birds form fission-fusion flocks and breeding pairs generally form between birds co-occurring together in the same winter flocks. Most of the birds recruit into the breeding population in their first year.

The data come from a wild population of great tits in Wytham Woods, Oxford (51° 46’ N, 1° 19’ W) that breeds almost entirely exclusively in nest boxes, between April and July, and pairs usually (>98%) have only one breeding attempt per season. Breeding data are collected using standard protocols to record parental identities and breeding parameters of a pair (date of first egg, maximal clutch size, number of hatched chicks, and number of fledged young). The data on winter flocks were collected between the winter beginning in 2007 until the end of the winter in 2014.

Winter data collection set-up

Data on winter flocks were collected using radio-frequency identification (RFID) technology: birds marked with passive integrated transporders (PIT-tags) were detected on feeders equipped with RFID-antennae. An estimated 82% of the primary population is marked with PIT-tags. The winter data collection protocol (spatial positions and opening times of feeders) changed half way through the study. In the first three winters (2007/08 to 2009/10) data were collected between August and March when 16 (out of 67 feeders) were always opened, and these were rotated every four days. In the second three winters (2011/12 to 2013/14) data were collected between early September and early March when all of the 65 feeders were opened during each weekend. The feeder-availability protocol minimised the possibility that flocks would get attracted to the constant food sources and we assume that data gathered at feeders represent snapshots of the social composition of different flocks at the time of recording. Further, feeders do not influence reproductive success.

Meeting time and social structure

We applied the GMMEvent (Gaussian mixture model for event streams) method to the spatio-temporal datastream collected from the loggers to determine the flocking events. This method is robust for inferring flocking events and preferable to other methods such as using arbitary time intervals. Further, the flocking events extracted using this method are known to be non-random in social composition, related to individuals social associations in other contexts and important to various social processes such as information spread and mating. We defined pair meeting time as the month (the first 3 winters) or the week (the last 3 winters) when the members of the future breeding pair were first identified in the same flock (i.e. gambit of the group approach). Time-frames for these were chosen based on the collection set-up in each set of winters.

A typical individual in this dataset experiences approximately five flock mates. Thus, it is reasonable to assume that future breeding partners have indeed met in the foraging flock when first observed together. Further, we know that birds of new pairs tend to spend the prior winter in the same flocks together, and previous work on the same dataset has shown that calculating the time when a pair first met in this way is good approximation of the beginning of pair bonding and relates to future pairing behaviour. Thus, we can be confident that majority of pairs not only meet, but also begin bonding in the time period (month, weekend) in which they were first detected in the same flock. Nevertheless, we also carried out additional sensitivity analysis ensure that our investigations are robust to these assumptions (see Supplementary methods and Statistical analysis section).

Social networks were constructed based on flocking events using Simple Ratio Index (SRI), that describes the affinity of two individuals to co-occur in the same flock, and can range from 0 (never observed together) to 1 (always observed together). SRI is calculated as the number of of the times (within a certain period: winter, month, weekend in our case) that two individuals were observed in the same flocking event together, divided by the sum of times when they were observed at all (in the same flock, in different flocks at the same time, and when only one of the individuals was observed).

Dataset construction

To be included in our primary analysis a pair had to meet several requirements. First, it had to be newly formed in the winter prior to the breeding season. This excludes pairs that were familiar as they have previously bred together. Second, both pair members were detected at winter feeders at least once. Third, both members were tagged either prior to the winter of interest (as nestlings or adults), or had been newly tagged in the current winter as immigrants at least two sampling periods before the pair meeting time. This eliminates pairs that had been associating but not detected as such purely because one or both members were not tagged at the time. Finally, we excluded six pairs (out of overall close to 400 pairs, i.e. <2%), where both individuals were known to be tagged prior to the winter, but were detected for the first time in the late autumn/winter (i.e. after October) as already paired. These pairs likely formed outside of the main woods, and returned already paired. These requirements were met by 169 pairs in 2008-2010 breeding seasons, and by 252 pairs of the 2012-2014 breeding seasons. Amongst these, members of 29, and 9 pairs, respectively, were never detected in the same flock. In all of these pairs at least one of the birds was either detected at feeders very rarely (in the lowest 20th percentile of the distribution of the number of times each bird was detected in a given winter), or had a larger number of detections but in one month/weekend only. Thus, the final sample size for the analysis included 140 pairs breeding in 2008-2010, and 243 pairs breeding in 2012-2014 spring.

Meeting time and future mating decisions

To tested whether new pairs (first breeding together in spring t) that met earlier in winter (preceding t) had a higher probability of breeding together in breeding season t+1 compared to pairs that started to associate later. We used data on only those pairs where both members had survived to breed in t+1.