Data from: Using high frequency GPS data to assess wintering goose proximity to commercial poultry facilities on the Delmarva Peninsula for avian influenza risk management and surveillance.
Data files
Jul 29, 2026 version files 1.02 MB
-
PLOSONE_data.zip
1.01 MB
-
README.md
10.55 KB
Abstract
Supporting data underlying analyses of wintering Canada Goose and Greater Snow Goose proximity to commercial poultry facilities on the Delmarva Peninsula. The archive includes the values underlying reported summary statistics (e.g., model deviance explained and predictor importance) and the numerical data used to generate partial dependence plots from boosted regression tree models for both species.
This dataset (PLOSONE_data.zip) contains supporting data underlying the reported summary statistics and partial-response figures from boosted regression tree (BRT) models evaluating Canada Goose (cang) and Greater Snow Goose (sngo) proximity to commercial poultry facilities on the Delmarva Peninsula.
The archive includes:
- Values behind reported means, standard deviations, standard errors, and related model-performance summaries.
- Predictor-influence values from individual BRT models and their across-model summaries.
- Numerical values used to generate the partial-response graphs.
- Individual model partial-response values used to calculate the plotted mean responses and uncertainty intervals.
The raw GPS telemetry data are not included in this archive.
SPECIES CODES: cang = Canada Goose sngo = Greater Snow Goose
FOLDER STRUCTURE: Each species folder contains model summaries, predictor influence values, and a graph_data subfolder containing the numerical values used to generate every partial-response figure.
Directory: PLOSONE_data/cang/
File: cang_deviance_values_by_model.csv
Description: model-specific values used to calculate the reported mean, standard deviation, and standard error of model performance
Variables
- species
- model_number
- cv_correlation
- deviance_explained
- deviance_explained_percent
File: cang_deviance_summary.csv
Description: summarize model performance across all valid BRT submodels
for each species
Variables
- species
- n_models
- mean_deviance_explained
- sd_deviance_explained
- se_deviance_explained
- mean_deviance_explained_percent
- sd_deviance_explained_percent
- se_deviance_explained_percent
File: cang_predictor_influence_by_model.csv
Description: relative influence of each predictor for every BRT submodel
Variables
- var: Predictor variable included in the boosted regression tree (BRT) model. See definitions below.
- rel.inf: Relative influence (importance) of the predictor within an individual BRT model. Larger values indicate greater contribution to model performance.
- model_number: Unique identifier for an individual BRT submodel.
- species: Goose species represented in the model.
cang= Canada Goose;sngo= Greater Snow Goose
Definitions of 'var' values
- week: Week of the year corresponding to the observation (integer 1–53)
- avg_temp: Mean daily air temperature (°C)
- dsewage: Distance to the nearest wastewater treatment facility or sewage lagoon (meters)
- dcoast: Distance to the nearest coastline (m)
- pasture_hay_2km: Proportion of land within a 2-km radius classified as pasture or hay
- dnwr: Distance to the nearest National Wildlife Refuge (m)
- day_night: Time-of-day category (
day,night) - woody_wetlands_2k: Proportion of land within a 2-km radius classified as woody wetlands
- dwaterbody: Distance to the nearest mapped waterbody (e.g., lake, pond, river, canal, or wetland) (m)
- forest_2km: Proportion of land within a 2-km radius classified as forest
- open_water_2km: Proportion of land within a 2-km radius classified as open water
- crops_2km: Proportion of land within a 2-km radius classified as cultivated crops
- emergent_wetlands_2km: Proportion of land within a 2-km radius classified as emergent herbaceous wetlands
- grassland_2km: Proportion of land within a 2-km radius classified as grassland/herbaceous vegetation
- developed_2km: Proportion of land within a 2-km radius classified as developed land
- barren_2km: Proportion of land within a 2-km radius classified as barren land
- adjacent_farms: Number of commercial poultry facilities within 2 km of the focal poultry facility
- farm_pond_binary: Indicates whether a farm pond was present within the focal poultry facility (
0= No,1= Yes) - shrub_scrub_2km: Proportion of land within a 2-km radius classified as shrub/scrub vegetation
File: cang_predictor_influence_summary.csv
Description: summarize predictor importance across BRT submodels
Variables:
- species
- rank
- predictor
- mean_relative_influence
- sd_relative_influence
- se_relative_influence
- n_models
Directory: PLOSONE_data/sngo/
File: sngo_deviance_values_by_model.csv
Description: model-specific values used to calculate the reported mean, standard deviation, and standard error of model performance
Variables
- species
- model_number
- cv_correlation
- deviance_explained
- deviance_explained_percent
File: sngo_deviance_summary.csv
Description: summarize model performance across all valid BRT submodels
for each species
Variables
- species
- n_models
- mean_deviance_explained
- sd_deviance_explained
- se_deviance_explained -
- mean_deviance_explained_percent
- sd_deviance_explained_percent
- se_deviance_explained_percent
File: sngo_predictor_influence_by_model.csv
Description: relative influence of each predictor for every BRT submodel
Variables
- varPredictor variable included in the boosted regression tree (BRT) model. See definitions below.
- rel.inf: Relative influence (importance) of the predictor within an individual BRT model. Larger values indicate greater contribution to model performance.
- model_number: Unique identifier for an individual BRT submodel.
- species: Goose species represented in the model.
cang= Canada Goose;sngo= Greater Snow Goose
Definitions of 'var' values
- week: Week of the year corresponding to the observation (integer 1–53)
- avg_temp: Mean daily air temperature (°C)
- dsewage: Distance to the nearest wastewater treatment facility or sewage lagoon (meters)
- dcoast: Distance to the nearest coastline (m)
- pasture_hay_2km: Proportion of land within a 2-km radius classified as pasture or hay
- dnwr: Distance to the nearest National Wildlife Refuge (m)
- day_night: Time-of-day category (
day,night) - woody_wetlands_2k: Proportion of land within a 2-km radius classified as woody wetlands
- dwaterbody: Distance to the nearest mapped waterbody (e.g., lake, pond, river, canal, or wetland) (m)
- forest_2km: Proportion of land within a 2-km radius classified as forest
- open_water_2km: Proportion of land within a 2-km radius classified as open water
- crops_2km: Proportion of land within a 2-km radius classified as cultivated crops
- emergent_wetlands_2km: Proportion of land within a 2-km radius classified as emergent herbaceous wetlands
- grassland_2km: Proportion of land within a 2-km radius classified as grassland/herbaceous vegetation
- developed_2km: Proportion of land within a 2-km radius classified as developed land
- barren_2km: Proportion of land within a 2-km radius classified as barren land
- adjacent_farms: Number of commercial poultry facilities within 2 km of the focal poultry facility
- farm_pond_binary: Indicates whether a farm pond was present within the focal poultry facility (
0= No,1= Yes) - shrub_scrub_2km: Proportion of land within a 2-km radius classified as shrub/scrub vegetation
File: sngo_predictor_influence_summary.csv
Description: summarize predictor importance across BRT submodels.
Variables
- species
- rank
- predictor
- mean_relative_influence
- sd_relative_influence
- se_relative_influence
- n_models
PLOSONE_data/sngo/graph_data/ & PLOSONE_data/cang/graph_data/
The graph_data folder contains the numerical data underlying all partial-response figures presented in the manuscript. Separate files are provided for each predictor and each species (Canada Goose, cang; Greater Snow Goose, sngo).
For every predictor, two complementary files are provided:
*_graph_data.csv
These files contain the averaged partial-response values calculated across all retained boosted regression tree (BRT) submodels. These data were used directly to generate the plotted response curves and associated uncertainty intervals presented in the manuscript.
Each file corresponds to a single predictor for a single species (e.g., week_graph_data.csv, avg_temp_graph_data.csv, dcoast_graph_data.csv).
For continuous predictors, columns include:
- species
- predictor
- predictor_value
- mean_log_response
- sd_log_response
- se_log_response
- mean_response
- lower_95_CI
- upper_95_CI
- n_models
For categorical predictors, predictor_level replaces predictor_value.
*_values_by_model.csv
These files contain the partial-response values from every individual retained BRT submodel before averaging.
Columns include:
- species
- predictor
- model_number
- predictor_value (or
predictor_levelfor categorical predictors) - partial_response_log_scale
- partial_response
Each *_values_by_model.csv file corresponds directly to a matching *_graph_data.csv file for the same predictor and species. These files provide the underlying model-specific values used to calculate the mean response, standard deviation, standard error, and approximate 95% confidence intervals reported in the corresponding *_graph_data.csv file.
Together, the paired files provide all numerical values required to reproduce the partial-response figures and associated summary statistics presented in the manuscript.
UNCERTAINTY CALCULATION
- Standard error was calculated as: SE = SD / sqrt(n)
- Approximate 95% confidence intervals were calculated as: mean ± 1.96 × SE
- The values were then exponentiated to produce the plotted response-scale
values. - SOFTWARE
Partial-response values were generated from fitted BRT models using the gbm package in R and exported as comma-separated value (CSV) files. - The supporting data were exported from fitted boosted regression tree models in R using the gbm, data.table, abind, and plyr packages.
DATA USE
These files provide the numerical values underlying the reported model summaries and figures. They do not include the complete raw GPS telemetry dataset or exact bird movement locations.
MISSING VALUES
Blank cells indicate missing or unavailable values.
CONTACT
Questions regarding the dataset should be directed to the corresponding author listed in the associated manuscript.