Warning This Web page has been archived on the Web.

Archived Content

Information identified as archived on the Web is for reference, research or recordkeeping purposes. It has not been altered or updated after the date of archiving. Web pages that are archived on the Web are not subject to the Government of Canada Web Standards. As per the Communications Policy of the Government of Canada. You can request alternate formats on the Contact Us page.

Breeding Bird Survey Statistical Methods

The Breeding Bird Survey (BBS) is an avian survey conducted annually in the United States, Canada, and, starting in 2008, northern Mexico. The survey is designed to monitor trends in relative abundance of North American breeding birds at continental, national, and regional scales. The U. S. Geological Survey (USGS) Patuxent Wildlife Research Center and Environment Canada jointly coordinate the BBS program in Canada and the United States.

Data Collection

BBS data are collected by participants once annually along randomly established roadside routes. The majority of BBS routes are run by skilled volunteer observers; the remainder are run by professional biologists. In Canada, the acceptable dates for running a BBS route is between May 28th and July 7th although observers are encouraged to run their routes after June 1st or for the boreal regions, after June 5th. Routes are located along all-weather secondary roads. The starting point and direction of routes are selected randomly in order to sample a range of habitats representative of the region. Observers are encouraged to run their routes for as many consecutive years as possible in order to reduce the effects of observer variability on data analysis. Because the BBS is designed to monitor long-term changes in bird populations, observers are encouraged to continue to run routes whether or not habitat conditions along the route have changed over time. The route layout is changed only if the road system has been altered, or if traffic has increased to the point that the noise interferes with bird identification and detection or creates conditions under which it is dangerous for a participant to stop a car.

Routes consist of 50 stops spaced 0.8 km apart along a 39.2 km route. Participants record the total number and species of individual birds heard or seen within 0.4 km of each stop during a three minute observation period. Start and finish times for the routes are recorded. Sky and wind conditions are recorded at the beginning of each 10-stop block and at the end of the route. The number of vehicles passing are recorded during each stop. Data are entered by participants via the USGS-Patuxent BBS website or are submitted for data entry to the BBS national offices. Canadian participants submit data forms to the Canadian Wildlife Service, Environment Canada office. All BBS data are stored in the North American BBS database housed at the USGS Patuxent Wildlife Research Center and are available on the USGS BBS website. Original data sheets for Canada are stored at the Canadian Wildlife Service, Environment Canada office.

Data analysis

Several factors, in addition to changes in bird populations, contribute to variation within BBS data; these include changes in weather, date, starting and finishing times and differences among observers. To help reduce the extraneous variability in the data, the data are screened to remove surveys run under unacceptable conditions (weather, date and time). The accepted surveys for a route are then partitioned into blocks run under comparable conditions. The set of matching conditions are that the surveys must have been run by the same observer and all surveys are done with a span of 19 or fewer days.

A particular survey is excluded from the analysis if any of the following conditions are true:

1) survey was outside the acceptable dates (May 28th through July 7th).

2) survey began more than 1 hour before or after the official starting time, the survey finished after 11:00 AM or the survey took more than 7 hours to complete.

3) winds are force 4 (Beaufort scale) at both the start and end of the route or are force 5 at either the start of end of the route.

4) rain is combined with winds equal to force 4 or greater, except on the prairies (where high winds are frequent) where survey are excluded only if winds are force 5 at the beginning and end of the route.

Data analyses were conducted for species and species groups at the Canada-wide, provincial/territorial and Bird Conservation Region (BCR) level where there were sufficient data. The BBS began in the Maritimes and Quebec in 1966, in Ontario and Manitoba in 1967, in Alberta, Saskatchewan and British Columbia in 1968, and in Yukon in the mid 1970s. There are only a handful of routes in Northwest Territories and Nunavut the earliest of which was in 1987. The number of routes run in the initial years of the BBS in Canada is small and the results for the early years of the survey often produce anomalous results. Therefore start dates were adjusted depending on the data available among province/territory and BCR. Species-specific adjustments were made for the Canada-wide analysis. For migration or habitat groups the start date was always 1970.  

Province/territory/BCR analysis

The default first year used for the analysis for different geographic regions is shown in the following table. Data were too few to calculate results for Newfoundland, the Northwest Territories or Nunavut.

The default first year used for the analysis for different geographic regions is as follows:

  • 1970: Nova Scotia, New Brunswick, Quebec, Ontario, BCR 8, BCR 12, BCR 13, BCR 14
  • 1973: Manitoba, Saskatchewan, Alberta, British Columbia, BCR 5, BCR 6, BCR 9, BCR 10, BCR 11
  • 1986: Yukon, BCR 4

Data were too few to calculate results for Newfoundland, the Northwest Territories or Nunavut.

Canada-wide analysis

Because starting years vary across province/territory and BCRs, there were consequences for the Canada-wide analysis of western species. If the start year was after 1970 predominately western species would be represented only by routes from peripheral (Eastern) regions during the early years of the survey. Thus the results may not have provided a correct representation of the population trends on a Canada-wide basis. It was decided to suppress the analysis for early years for these predominately western species in the Canada-wide analysis.

To identify predominately western species, population estimates were made for each province by BCR combination. If more than two-thirds of the population occurred within BCR 4 (where the analysis started in 1986) then the long-term,Canada-wide analysis was started in 1986 (2 species only). If more than two-thirds of the population occurred within BCR 4, 5, 6, 9 10 and 11 combined then the long-term, Canada-wide analysis was started in 1973 (76 species). Otherwise, the long-term, Canada-wide analysis started in 1970. 

For shorter time periods that include the early years, the start dates were adjusted accordingly (e.g. 1970-1989 was changed to 1973-1989). In the latter half of the BBS, data were more consistently available across the country and there are no differences in start dates (i.e. 1989-2009 and 1999-2009).

Analytical method

The BBS is analysed using a model with terms for year, route and block. The observed counts are assumed to have a Poisson distribution and the model is fitted using a weighted maximum likelihood. The logarithm of expected value of the count is assumed to be a linear model in the factors.

BBS routes are established within degree blocks (one degree of latitude and longitude) which are considered the basic sampling unit for the BBS. Although effort is made to provide an even distribution of routes across each province, active routes tend to be concentrated close to population centers and in the southern portion of the province. In order to accommodate this uneven distribution in the analysis, the individual routes are given an area weighting based on the area of the degree block divided by the number of useable routes in the degree block.

The maximum likelihood model cannot handle routes or route-blocks where the species was not recorded. Once the model has been fitted, the annual index predicted by the model is scaled to compensate for these two types of non-inclusion. The objective of adjusting the model-based annual indices is to estimate the number of individuals which would be recorded on a route by an average observer and to capture the local extirpations or colonisations which might not be adequately captured by the model. To compensate for non-inclusion of route-blocks, the annual index for each year is multiplied by the proportion of useable routes that have a route-block that is useable in that year. To compensate for non-inclusion of entire routes, the annual index is then multiplied by the proportion of useable routes. The former adjustment can affect the year-to- year trends in the data while the latter adjustment has no influence on trends. Once the annual indices have been estimated, a linear trend is fitted through the logarithm on the annual indices. This provides a summary of the annual percentage change in the population.

The standard error (SE) of the annual indices and trend slope are estimated through a jackknife procedure. The variability among the set of estimates produced by discarding one route at a time provides an estimate of the among- route variability in the annual indices and slopes. This provides a data-based estimate of variability rather than relying on an assumed distributional model for among-route variability.

Credibility measures

To quantify further the adequacy of the BBS results in describing overall population trends, three additional measures of the credibility of the BBS results are calculated: precision, coverage and non-linearity of the trend.

Precision was calculated as the power of the survey to detect a slope of -0.0346. This corresponds to a 50% decline over a period of 20 years, a baseline that has often been used as a criterion for a trend with biological significance (see for example Bart et al., 2004). Power was calculated based on a 2-sided t-test with p=0.10.This integrates the observed standard error of the trend and the sample size (number of routes where the species was seen) to assess whether the available data are sufficient to detect a trend of biological significance.

Coverage is the estimated proportion of the population in the degree blocks where the BBS was run. For each province by BCR stratum, an estimate of the average number of individuals on a BBS route was calculated using BBS results from 1997-2007. In areas where the BBS was not run, comparable counts were developed from other monitoring programs. The coverage for a BCR is calculated as follows: the denominator is the total area of the BCR in each province multiplied by the average count in the province and totalled across provinces, while the numerator is done in parallel but only including those degree blocks where the BBS was run. In this scheme, if there is one route within a degree block then the degree block is considered covered. This measure assesses whether a sufficiently large proportion of the population has been monitored to justify extrapolation of the survey results to the population.

Non-linearity of the trend was assessed through autocorrelation of the residuals between the predicted trend line and the annual indices. This measure assesses whether summarising the annual indices as a single trend slope is sufficient to characterise the population change. A high value for autocorrelation indicates that the annual indices exhibit strings of values above and below the overall trend line suggesting that there may be shorter term trends in the annual indices and that the annual percent change value may not be representative of the change in population. A small positive value for autocorrelation indicates that the data fluctuate above and below the trend line and there is no indication of a departure from a linear trend, suggesting that the annual percent change value is a good summary of population change. A large negative value for the linearity could indicate a substantial departure from the model used for the analysis. This would also raise a concern over the validity of summarising the trend as an annual percentage change. The statistical significance of autocorrelation is not quantified. For large values of autocorrelation the user should inspect a graph of the annual indices to assess the non-linearity.

Changes to the analysis of 2009 data

The current analysis presented here differs from that done in previous years in three ways: i) start dates were modified and ii) the Canada-wide analysis was better adapted to the species’ range for those species that occur predominately in Western Canada and iii) data from Canadian Wildlife Service’s Grassland Bird Monitoring program were included.

The initial years of the BBS survey was a time of expansion. Volunteers were gradually recruited, trained and assigned to routes. Previously the initial year of data analysis was set at the first year the survey was run in a geographic region. However, examination of the annual indices indicated that these indices were highly variable in the early years for many species. This was due to the small sample size that may not have been able to provide a representative sample for some species in the early years. As described above, the use of new start dates for provincial/territorial and BCR analyses and species-specific start dates for the Canada-wide analysis should reduce the effect of these small sample sizes in the early years.

The Grassland Bird Monitoring (GBM) project was started in 1996 as a pilot project to monitor the grassland species in BCR 11 more intensely (see Dale et al., 2005). The GBM routes were set up and run using the BBS protocol except that the routes were allowed to be set up on tertiary roads and were selected within those degree blocks that included prairie grassland habitat. The standard analysis treats the degree block as a stage in sampling which ensures that more intensive sampling in a selected degree blocks does not introduce a bias towards these particular habitat. However, it increases the number of individuals recorded for grassland species and helps to provide more stable annual indices. Inclusion of GBM data affects the analysis in BCR 11, Alberta, Saskatchewan and the Canada-wide analysis.

For more information

Estimating an annual index for the breeding bird survey (PDF; 134 KB) 

Differences between Breeding Bird Survey (BBS) trends produced by the U.S. Geological Survey and Environment Canada

Literature Cited

Bart, J., K.P. Burnham, E.H. Dunn, C.M. Francis and C.J. Ralph. 2004. Goals and strategies for estimating trends in landbird abundance. J. Wildl. Manage. 68: 611–626.

Dale, B., M. Norton, C. Downes and B. Collins. 2005. Monitoring as a means to focus research and conservation - the grassland bird monitoring example. In Bird Conservation and Implementation in the Americas: Proceedings of the Third International Partners in Flight Conference, vol. 2; C. J. Ralph and T. D. Rich, eds. USDA For. Ser. Gen. Tech. Rep. PSW-GTR-191. Albany, CA. http://www.fs.fed.us/psw/publications/documents/psw_gtr191/psw_gtr191_0485-0495_dale.pdf