The BTO/JNCC/RSPB Breeding Bird Survey (BBS) was launched in 1994, following two years of extensive pilot work and earlier desk-based studies. The introduction of the BBS was a move designed to overcome the limitations of the Common Birds Census (CBC), which had monitored bird populations since 1962. In particular, it improves the geographical spread of UK bird monitoring, thus boosting coverage of species and of habitats.
The BBS uses line transects rather than the more intensive territory-mapping method that had been used by the CBC. The average time observers spend per visit on counting birds is only around 90 minutes and, even with travel and data-input time, this survey is relatively quick to undertake and is therefore accessible to a large number of volunteers. Sampling units are the 1x1-km squares of the Ordnance Survey national grid, of which there are some 254,000 in the UK. From these we make random selections for inclusion in the scheme (see Square selection, below). The BBS requires a relatively large sample of survey squares, and the initial aim was to achieve coverage of about 2,500 squares (1%). This total is now well exceeded.
An important aspect of BBS is its coordination through a network of volunteer BBS Regional Organisers. The Regional Organisers find and encourage willing volunteers for their squares and provide paper forms as required. Since 2003, when online submission of BBS data was introduced, most data have been returned online – see the BBS pages of the main BTO website for details.
Fieldwork involves up to three visits to each survey square each year. The first is to record details of habitat and to establish or re-check the survey route, while the second and third (termed 'early' and 'late') are to count birds. A survey route is composed of two roughly parallel lines, each 1 km in length, although for practical reasons routes typically deviate somewhat from the ideal. Each of these lines is divided into five sections, making a total of ten 200-m sections, and birds and habitats are recorded within these ten units. The two bird-count visits are made about four weeks apart (ideally in early May and early June), ensuring that late-arriving migrants are recorded. Volunteers record all the birds they see or hear as they walk along their transect routes. Birds are noted in three distance categories (within 25 m, 25–100 m, or more than 100 m on either side of the line, measured at right angles to the transect line), or as in flight. Recording birds within distance bands provides a measure of bird detectability in different habitats and thus allows population densities to be estimated more accurately. The total numbers of each species, excluding juveniles, are recorded in each 200-m transect section and distance category, as well as the timing of the survey and weather conditions. In 2014, the optional recording of the method of detection was included in BBS for the first time, and observers can now record whether they detect each individual bird by sight, by song or by call. This information is not currently used to calculate trends, but it is anticipated that it will help further refine the calculation of population densities for some species.
By 1998, more than 2,300 BBS squares were being surveyed annually, close to the original target of 2,500. Only around a quarter of these plots were covered in 2001, owing to Foot & Mouth Disease access restrictions, but (thanks to our keen observers) the sample recovered immediately to over 2,205 in 2002 and had increased further to 2,328 squares in 2003, 2,533 in 2004, 2,893 in 2005 and 3,313 in 2006. The sample soared to 3,759 in 2007 and ran marginally below that level over the next few years during and just after the 2007–11 Bird Atlas, before reaching a new high of 3,837 squares in 2016 (Harris et al. 2017). Squares are distributed throughout the UK and cover a broad range of habitats, including uplands and urban areas. There are now 111 species that are present on 40 or more BBS squares annually and so can be monitored with good precision at the UK scale (Joys et al. 2003, Harris et al. 2017), although a few present special difficulties because of their colonial or flocking habit or their wide-ranging behaviour. For most of these species, BBS can also assess annual population changes within England alone, using data from 30 or more squares, and for about half the species also within Scotland and Wales as separate units. Sample sizes in Northern Ireland already allow more than 30 species to be indexed annually.
Survey squares are chosen randomly using a stratified random sampling approach from within 83 sampling regions, which in most cases are the standard BTO regions. Survey squares are chosen at random within each region, to a density that varies with the number of BTO members resident there. Regions with larger numbers of potential volunteers are thereby allotted a larger number of squares, enabling more birdwatchers to become involved in these areas. This does not introduce bias into the results because the analysis takes the regional differences in sampling density into account.
Change measures between years are assessed using a log–linear model with Poisson error terms. For each species and square, counts are summed across all sections and distance bands for each visit ('early' and 'late') and the higher value is used in the model (or the single count if the square was visited only once). Counts are modelled as a function of square and year effects. Each observation is weighted by the number of 1-km squares in each region divided by the number of squares counted there, to correct for the differences in sampling density between regions. The upper and lower confidence limits of the changes indicate the certainty that can be attached to each change measure. When the limits are both positive or both negative, we can be 85% confident that a real change has taken place (see here for details).
Trends are presented as graphs in which annual population indices are shown alongside a smoothed trend and its 85% confidence limits. A caveat, 'small sample', is provided against the trends for England, Northern Ireland, Wales and Scotland where the mean sample size is between 30 and 40 plots per year.