Skip to contents

Changelog

Source: NEWS.md

bbsBayes2 1.1.3.3 Released January 29, 2026 - Bug fix for custom stratifications.

  • Fixed bug in stratify() (bug introduced in v.1.1.3) that removed the area weights for custom stratifcations. Bug did not affect built-in stratifications or subsets of those stratifications.

bbsBayes2 1.1.3.2

  • Model fitting options and optimization.

  1. Changed model files to take advantage of the sum_to_zero_vector constrained parameter type in Stan. This parameter type is now used for all of the random effects and spatial random effects that previously used a soft sum-to-zero constraint. This parameter type improves sampling efficiency for all models, but has no effect on the parameter estimates.

  2. Added an optional argument to run_model(compiler_optimization = "O1"). The default "O1" is the same compiler settings used in v.1.1.3.1. These compiler optmization options can improve sampling efficiency for some models.

bbsBayes2 1.1.3.1

  • New Stratifications.

  1. stratify(by = "bcr",...): Includes the 2025 version of the North American Bird Conservation Regions (BCRs). The changes are primarily in Northern Canada, where the largest northern BCRs have been subdivided into regions that are more similar in scale to the BCRs in the rest of the continent. These new BCRs have replaced the previous BCR stratification (i.e., stratify(by = "bcr",...) will now use this updated BCR map). For reproducibility, the previous BCR map is still available by calling stratify(by = "bcr_old",...).

  2. stratify(by = "bbs",...): Intersection of the 2025 updated BCRs with states, provinces, and territories in Canada and the United States. This stratification is conceptually the same as stratify(by = "bbs_usgs",...) but uses the updated BCRs.

  1. stratify(..., use_map = TRUE), the default is TRUE, where the BBS routes are stratified based on the spatial overlay (sf::st_join()) of the strata polygons and the starting locations of each BBS route. This spatial overlay approach has always been used for custom stratifications, and now is also the default for the standard built-in stratifications. In previous versions, the BCR, province, state, or lat_long information for each BBS route was taken from the columns in the BBS database, so not an explictly spatial process. By setting use_map = FALSE this previous approach is still available for the following stratifications "bcr_old", "bbs_cws", "bbs_usgs", "latlong", "prov_state". The new stratifications "bbs" and "bcr" do not allow use_map = FALSE (will be changed to TRUE with a message), because the full list of new BCRs does not yet exist in the BBS database.

  2. stratify(..., distance_to_strata), optional numerical value indicating the distance in meters within which routes will be joined to the nearest stratum polygon, if the route start point does not intersect any strata. This argument is particularly useful to ensure that routes with starting locations on or near the coast or shoreline of large lakes are included. These coastal routes may be otherwise excluded due to errors in the spatial information (either the starting point coordinate or the strata map). For example, the map associated with the standard bbs stratification excludes 3,877 surveys on 72 routes when this argument is NULL. All of these 72 routes have starting locations on the coasts. Setting this argument to 4000 (any route within 4 km of at least one polygon) ensures all of these coastal routes are included in the nearest stratum. Users should be cautious of using this argument if the strata map does not represent the full landmass of Canada and the United States. For example using a subset of an existing strata map, such as all of the bbs strata within one country will treat routes within the set distance of the national border (e.g., some routes in the US as if they are in Canada) the same way it treats routes that happen to fall just off the coast of the supplied strata map.

bbsBayes2 1.1.3

  • 2025 release. Includes access to the 2025 release of the BBS database (observations from 1966 through to 2024).

Some of the key fixes and improvements:

  1. first-difference models (hierarchical and spatial) now better handle the missing data from 2020. The models estimate a single value of change (i.e., the difference between 2021 and 2019) and then use random draws from a beta distribution (beta(1,1)) for each posterior draw to randomly partition the difference between the first interval (2019-2020) and the second (2020-2021). This has the result of forcing the mean trajectory to follow a straight line between 2019 and 2021, while increasing the uncertainty of the estimated index for 2020.

  2. the fetch_bbs_data() function now allows the user to retain the BBS observations that do not conform to the survey design. The documentation includes warnings about how this is probably a bad idea.

  3. all models can now generate prior predictions using prepare_model(..., use_likelihood = FALSE). This should be particularly useful for applications with a custom model.

  4. the function plot_map(..., col_ebird = TRUE, alternate_column = "percent_change") will plot the estimated percent change values using the same categories and colour scheme used by the eBird status and trend team.

  5. the function generate_trends(..., export_full_posterior = TRUE) will now add a tibble to the output list that includes the full posterior distribution of every trend estimate. Allows for formal tests/estimates of the differences between two trends (e.g., between two time-periods or two regions).

bbsBayes2 1.1.2.1

  • minor fixes to bugs in 1.1.2

bbsBayes2 1.1.2

  • 2024 release. Includes access to the 2024 release of the BBS database (observations from 1966 through to 2023).

  1. adds options to generate spaghetti plots of population trajectories in plot_indices. Spaghetti plots show a selection of posterior draws of the population trajectories as many distinct lines. These lines provide a different way to summarise the uncertainty of the population trajectories that separates the uncertainty of the trends (shape of the population trajectory) from the uncertainty in the mean abundance (vertical placement on the graphs).

  2. adds options to estimate trends from any model using a gam-based smooth. The gam_smooths argument in the generate_indices(gam_smooths = TRUE) function creates an array of posterior draws of gam smooths through the population trajectory. These posterior draws can be used in the generate_trends(gam = TRUE) function to estimate trends and their uncertainties based on these smooths, for any custom range of dates. These smooth trends provide an alternative to the end-point trends that are the default in bbsBayes2. If the gam argument is set to TRUE, all of the trend and population change calculations are based on the posterior array of smooths, rather than the posterior array of annual indices. Calculating the smooths in generate_indices requires significantly more time and so the gam_smooths option is set to FALSE by default.

  3. adds informative error messages to: a) identify challenges with interpreting the observed annual means relative to estimate trajectories for composite regions (all except stratum); b) identify species and time-periods where analyses may be complicated by changes in species taxonomy (see bbsBayes2::species_notes); and c) identify when the strata_custom optional sf object passed to stratify does not have the correct crs and strata_name column type.

  4. prepare_data now exports columns including latitude and longitude of each BBS route start-location in the raw_data component of the output list.

  5. fix a number of errors and issues identified in version 1.1.1.

bbsBayes2 1.1.1

1.1.1 represents a patch to correct some failing checks and tests in 1.1.0, and includes all of the functionality of the 1.1.0 release.

bbsBayes2 1.1.0

  • Second release. Includes access to the 2023 release of the BBS database. Also includes improvements to the management of the csv files created by Stan, increased ability to map values from generate_trends(), tweaks to the first_diff models to better handle the missing data in 2020 (when no BBS surveys were conducted), and options to use highest posterior density intervals (hpdi) to describe the posterior distribution in generate_indices() and generate_trends().

More specifically, to support version 1.1.0, the following changes were made:

  1. adds some tweaks to the first_difference, spatial model to force 0 spatial variance in 2020 - the year when the BBS was cancelled and so there are no data and therefore no information with which to meaningfully estimate the variance in annual differences among strata.

  2. allows the plot_map() function to plot any of the numerical values in the output from generate_trends(). e.g., to display trend uncertainty by plotting the credible limits or width of the credible intervals of the trend estimates, or to display the mean relative abundance among strata during the trend period. Example added to the Advanced Vignette.

  3. package can access the newest database release to include field observations from 1966 - 2022.

  4. allows for the use of highest posterior density intervals in the generate_indices() and generate_trends() functions. HPDI often provide a better description of the posterior distribution, particularly for skewed distributions, such as those from a log-transformed predicted count (i.e., the indices of annual relative abundance). Example added to the Advanced Vignette.

  5. run_model() function by default now cleans up the .csv files created by Stan, once the model has finished and has been saved to .rds file (also done by default). Saves local disk space, as once the model output is saved in the rds file, the csv files are redundant.

bbsBayes2 1.0.0

  • First release of bbsBayes2, the successor to bbsBayes!