Long-term and large-scale multispecies dataset tracking population changes of common European breeding birds

National breeding bird monitoring schemes

Fieldworkers record all, or a fixed, pre-defined set, of bird species heard or seen in the main breeding season in 28 European countries on an annual basis (Fig. 1). All observations are recorded following a scheme-specific standardized protocol based on established field methods for counting birds: point count transect, line transect, or territory or spot mapping^10,33,34. Here, we provide a short description of the field methods used, as each scheme provides its fieldworkers with specific fieldwork instructions and training.

1. 1.

   Point counts: A fieldworker counts all detected birds at census points, often placed along a transect (typically >200 meters apart) during a fixed time period to sample birds in a defined study area. Each point is usually visited twice a year.

2. 2.

   Line transects: A fieldworker moves along a transect and records all detected birds along the predefined path to sample birds in a defined study area. Each transect is usually visited twice a year.

3. 3.

   Territory or spot mapping: A fieldworker records all birds showing territorial behaviour in a defined study area and marks their positions and their territorial behaviour on a map. The study area is visited multiple times a year (usually 5–12) to map breeding bird territories based on the individual species-specific behaviour recorded. The species counts reflect the number of present territories.

National scheme coordinators provide all fieldworkers with instructions with the prescribed number and timing of survey visits, and information on how to record observations in terms of sampling effort, time of day, seasonality and weather conditions. This ensures the temporal and spatial consistency of data quality within individual national schemes³⁵. The standardization of conditions during counting then enables unbiased comparison of results between years and individual study sites within each country.

For the selection of sampling plots, national monitoring schemes use either random, stratified random, systematic selection, or allow a free choice by fieldworkers^8,34. Sampling plots are selected randomly within the study boundaries using a random selection method or randomly within the stratum under the stratified random method. Under these methods, study plot selection is conducted by random generators (by computer programs) and stratum is predefined as a region with similar attributes; these might be proportions of habitat types, altitude bands, bird abundance, accessibility of survey sites, or fieldworker density, depending on the local circumstances. Systematic selection predefines a spatial grid for sampling plot selection while free choice enables fieldworkers to select their study areas without restrictions³⁴. The use of a free choice, or stratified random selection of sampling plots may result in a biased sampling of specific habitat types (typically species-rich habitats) and regions (remote areas poorly covered), but post-hoc stratification and weighting procedures are generally used to correct for unequal sampling and reduce sampling bias as long as the number of plots per stratum is sufficient³⁶. Moreover, national coordinators provide fieldworkers with recommendations or oversee the study plot selection to prevent oversampling of specific habitat types and regions. Detailed information on scheme-specific counting protocols, study plot selection and breeding period specification can be found for each national monitoring scheme⁸.

National species indices

A species annual index reflects population size change relative to the population size in the reference year. On an annual basis, coordinators of the national monitoring schemes produce species indices for recorded species using a tailor-made implementation of loglinear regression models (TRIM models – Trends and Indices for Monitoring data) from time series of recorded species counts at the study plots^37,38. Species counts from a study plot reflect mean (or maximum) of individuals recorded during visits at the study plot when using point counts or line transects. For some species, only the number of individuals recorded on the second visit is used because the period of the first visit coincides with the migratory period and consequently the mean number of recorded individuals might not reflect the number of breeding individuals. The method to estimate the species counts in a plot is constant within a national scheme.

Missing data occur in the species counts at specific sites in individual years for various reasons, such as severe weather conditions during the counting period, abandonment of the study site, restricted access, or where counts are repeated in multi-year intervals. The TRIM model imputes missing data using species counts either from surveyed sites with similar environmental characteristics (stratified imputing) or all other sites with available data^37,39. This process is based on the assumption that changes in populations at non-counted sites are similar to those at counted sites within the same stratum. To derive expected between-year changes in species population sizes, the program fits a log-linear regression model assuming Poisson distribution to time series from counted plots. Finally, we use this model to calculate missing species-specific counts for individual years^37,39. The resulting time series of species counts with imputed missing values cover the whole period of counts in the national monitoring scheme. These imputed data are then used to estimate annual population sizes from all study plots and to derive population size indices for species¹¹.

European species indices and trends

The individual national indices for a given species are combined to create the European species indices. Subsequently, long-term population size changes (trends) are calculated as the multiplicative linear slopes from species indices and represent an average between-year relative population size change over a predefined period.

The European combination process is very similar to the production of national scheme species indices, but with three differences⁴⁰. Firstly, the indices are calculated using national TRIM output data, consisting of imputed species counts, standard errors per year and covariance matrices. Secondly, species counts are weighted by the most recent species population size estimates (updated every three years) in a given country derived from national bird atlases, official data reports and national experts (http://datazone.birdlife.org/) to account for the country-specific population sizes and thus the unequal contribution of national indices on the European index. Thirdly, missing national time totals due to different start years of the schemes⁸ are imputed using species counts from a set of countries from the same geographical region^6,11. For this purpose, we divided all national schemes into seven geographic regions – Central & East Europe, East Mediterranean, North Europe, South Europe, Southeast Europe, West Balkan and West Europe⁸. We then use a set of national indices from a given region to impute missing national indices. Therefore, the earliest periods of population size changes are based on data from a reduced number of study plots and schemes.

The species trends are then imputed from species indices for four periods: 1980 onwards, 1990 onwards, 2000 onwards and using only the last ten years of data if the data are available. Despite higher uncertainty of the earliest estimates, we do provide the population index estimates for this period as no alternative and continuous measures of bird population size changes exist for this period.

The uncertainty estimates of indices and trends are presented by the standard error^11,37 allowing a calculation of 95% confidence limits (±1.96 × standard error). The magnitude of the trend estimates together with their 95% confidence intervals are then used for trend classification into six classes facilitating communication and interpretation of the outputs³⁷ (Table 1).

Finally, European species indices and trends are presented only for a group of common and widespread bird species (hereafter ‘common bird species’) meeting two criteria:

1. 1.

   The estimated breeding population (http://datazone.birdlife.org/) is at least 50 000 pairs in PECBMS Europe (EU countries, Norway, Switzerland and the United Kingdom; Fig. 1). Additionally, Red-billed Chough (Pyrrhocorax pyrrhocorax) and Spotted Redshank (Tringa erythropus) with population sizes below 50 000 pairs are included, as large parts of their breeding populations are covered in the PECBMS Europe.

2. 2.

   The estimated breeding population in PECBMS countries providing data for a given species⁸ covers at least 50% of the whole PECBMS Europe breeding population (http://datazone.birdlife.org/).

The resulting datasets of European population size indices and trends consist of relative population changes for 170 common bird species.

Updates

We aim to maintain the PECBMS database with annual updates. The annual updates will be available through the PECBMS database deposited at the Zenodo repository⁸ to ensure long-term public availability of the data.

Source: Ecology - nature.com