Breeding timed to match optimal resource abundance is vital for the successful reproduction of species, and breeding is therefore sensitive to environmental cues. As the timing of breeding shifts with a changing climate, this may not only affect the onset of breeding, but also its termination, and thus the length of the breeding period. We use an extensive dataset of over 820K nesting records of 73 bird species across the boreal region in Finland to probe for changes in the beginning, end, and duration of the breeding period over four decades (1975-2017). We uncover a general advance of breeding with a strong phylogenetic signal, but no systematic variation over space. Additionally, 31% of species contracted their breeding period in at least one bioclimatic zone, as the end of the breeding period advanced more than the beginning. We did not detect a statistical difference in phenological responses of species with combinations of different migratory strategy or number of broods. Nonetheless, we find systematic differences in species responses, as the contraction in the breeding period was found almost exclusively in resident and short-distance migrating species, which generally breed early in the season. Overall, changes in the timing and duration of reproduction may potentially lead to more broods co-occurring in the early breeding season – a critical time for species’ reproductive success. Our findings highlight the importance of quantifying phenological change across species and over the entire season, to reveal shifts in the community-level distribution of bird reproduction.

1. Description of methods used for collection/generation of data:

Raw original data: The Bird Nest Ringing Database is coordinated and curated by the Finnish Museum of Natural History.

This database contains records for ringing events of individual bird nestlings (marked by unique Ring IDs) in Finland. These data are a subset of the data in the “Database of birds ringed in Finland and all reported encounters of the birds” (http://tun.fi/HR.48), which is available from the Finnish Biodiversity Information Facility (FinBIF) at https://species.fi. The primary database was queried by J. Valkama and M. Piha in summer 2018 to include only records of ringed nestlings. At the point of data extraction, the database contained over 3.6 million ringing records of nestlings for 166 species (data available from the Finnish Biodiversity Information Facility (FinBIF) at https://species.fi), which represents c. 66% of the 250 bird species known to breed in Finland. Ringing is done by experienced volunteer ornithologists who have received training in the procedure (Lokki H. (ed.) 2017: Rengastajan käsikirja. Version 1.0. Finnish Museum of Natural History, University of Helsinki. (In Finnish). Crucially, nestlings can only be ringed when of a certain size (which is species-specific), and thus the timing of ringing is tightly linked to the timing of egg-hatching (Møller et al. 2010: Climate change affects the duration of the reproductive season in birds. Journal of Animal Ecology https:/doi.org/10.1111/j.1365-2656.2010.01677.x; Lehikoinen 2011: Advanced Autumn Migration of Sparrowhawk Has Increased the Predation Risk of Long-Distance Migrants in Finland in PloS One, (2011) https:/doi.org/10.1371/journal.pone.0020001.). We therefore use the timing of ringing as an indicator for breeding time, since, within species, the possible timing of ringing is relatively constant in relation to egg-hatching, and has been identified as a functional phenological indicator (Møller et al. 2010). Additionally, ringing data has been shown to be a reliable estimate for breeding phenology (Eeva et al 2012: Breeding time trends of the Crested Tit (Lophophanes cristatus) in southern Finland: comparison of data sources. J Ornithol 153, 653–661).

2. Methods for processing the data:

To quantify changes in breeding period timing and duration of bird species in Finland, we use the Bird Nest Ringing Database coordinated and curated by the Finnish Museum of Natural History (see 1 above).

In order to identify single ringing events per nest and day-of-year for each species, we took a number of processing steps with the raw data. First, we aggregated individual ringing events likely done in the same nest. We used a combination of unique Ringer ID, Species, Location (unique longitude and latitude) and Date (unique day, month and year) of ringing to identify unique ringing events per nest (RSLD combinations); this yielded 584 354 events which correspond to a ringing event occurring in separate nests by a given ringer on a given day of year for each species in the dataset. However, nestlings in several nests may have been ringed in the same locations (when nests are located very close to each other) and/or more than one ringer may have ringed nestlings cooperatively in the same nest, causing the same nest to occur multiple times in the database. Therefore, we further divided these records into three subsets (A, B and C), based on a brood number (i.e. each ringer’s personal chronological sequence of ringed broods) and an abundance value (i.e. number of nestlings observed per brood), when available. Specifically, a unique brood identifier was created based on the combination of Brood and the variables described above (RSLD+B=RSLDB). For each RSLDB identifier, we tallied the number of ringed nestlings (i.e. number of unique Ring IDs) and compared this to the reported abundance value. If these values matched, each of these unique broods was considered a unique nest and given a unique ID (subset A; 392 421 unique nests). In cases where the tallied number of ringed nestlings and the reported abundance value did not match (either because the ringer had not ringed all nestlings in the brood, or because either the brood number or abundance value was incorrectly reported - which of the two was the case we cannot separate) and/or abundance values were missing, we used the tallied number of ringed nestlings per unique RSLD and disregarded the reported abundance value by the ringer, which might or might not be correct. Next, the tallied number of nestlings was compared to the maximum brood size for each species (literature-based estimates; see traits below). When these values matched, the records were assigned a unique nest ID (subset B; 247 399 unique nests). For the remaining records, we calculated the potential number of nests for each RSLD, by dividing the number of ringed nestlings by the maximum brood size for each species. This step was used to calculate the required number of nests to “accommodate” the recorded number of nestling ringed, which were given a unique ID (subset C; 294 284 unique nests). Subsets A, B, and C were then combined, yielding 934 104 records of ringing events in unique nests. Finally, we only included records after 1975, since records before this year were sparse and only represented a small subset of the bird species in the dataset. This processing resulted in 919 713 records of bird ringing events for 166 species, spanning 43 years and four bioclimatic zones in Finland (166_species.csv).

For the purposes of our analyses we further filtered the above data set (166_species.csv) through additional criteria. As some species had very sparse records, either within a specific bioclimatic zone or over time, we included only species-zone-year combinations with a minimum of 30 records. Next, we included only those species-zone combinations that had records from at least 10 years over our study period, and with at least one year with observations in the first and one in the last ten years of the study period. These relatively conservative inclusion criteria meant that many ringing events, and thus several species and species-by-zone combinations, were not included in the analysis. These steps were intended to ensure that: 1) there were sufficient records to estimate phenological metrics for each species in each zone and per year (i.e. enough records to represent the day-of-year distribution of nesting events); 2) there was a sufficient number of years sampled for each species and zone to reliably estimate shifts in the breeding period (i.e. at least 10 data points over time); and 3) the species analyzed were present in the beginning and end of our study period. We finally excluded one species (Columba oenas) as it breeds irregularly and several times throughout the season, and thus has an intermittent and varying breeding phenology that is not well-captured by ringing data. Finally, we removed apparent outliers in the data, i.e. ringing events clearly outside the general breeding season of birds in Finland, which most probably represent data entry errors (by inspecting the distribution of the nesting events over time for each species-zone combination). The final dataset that we used for our analyses contained 821 413 records of ringing events in unique nests for 73 species (73_species.csv). These data represent a large variety of species including birds of prey, waders, passerines, and owls that are characterised as Resident ( R) or Short distance migrants (S) versus Long distance migrants (L) and single-brooded (1) versus multi-brooded (2) species (Traits_73_species.csv).

See attached README_RingingDataBorealBirds.txt file