<< Species Index

< prev | next >

Species

Species: Gray Catbird (Dumetella carolinensis)
Prebreeding migration: April 5 to May 31
Postbreeding migration: August 16 to November 29)

Abundance and BirdFlow Migration Traffic Rate (BMTR)

Abundance: Abundance data provided by Cornell Lab of Ornithology | eBird, resampled to 100 km cells. The highest values are truncated down to the 99th percentile of the values to eliminate extreme outliers. BirdFlow models are trained on abundance.

BMTR: BirdFlow Migration Traffic Rate (BMTR) represents the relative number of birds migrating through each part of the landscape in a week given the model parameters and the assumption that the bird follow a straight path for each transition. To calculate BMTR the probability of each movement in the model (excluding stationary birds) is assigned to a line that connects the start and end of the movement. The probability of each line is then added to all the cells within 50 km of the line.

Routes

These stochastic routes are generated from the BirdFlow model (see BirdFlowR::route ) and can be used to assess performance. Each step of the route is created by sampling the probabilities assigned to the bird’s next location given a starting date and location. The initial locations are sampled from the probability distribution for the species at the start of the migration. The stay length is one if the bird moves on immediately and longer when the bird stays in the location for multiple steps.

eBird

The metadata here comes from eBird via ebirdst::ebirdst_runs. Data quality ranges from 0 to 3, with 3 being the highest quality.

Data quality: 3, 3, 3, 3 (prebreeding, breeding, postbreeding, and nonbeeding)
Version year: 2023
Release year: 2025

BirdFlow

Birdflow model ratings range from 1 to 5, with 5 being the highest rating.

Model date: 2025-10-14
Traverse correlation: 0.9954
Mean step correlation: 0.9904
Min step correlation: 0.9034
Mean rating: 4.5
Model Files: Rds , hdf5

Model Tuning

Selected hyper parameters

Selection criteria: Best log likelihood
Distance weight: 0.008658
Entropy weight: 0.001443
Observation weight: 1
Distance power: 0.3

Performance metric trade offs

The candidate and best models plotted against two possible selection criteria: log likelihood and relative distance gain.

Candidate model performance for other metrics

The distribution of each metric across grid search runs where the distribution correlation was above 0.9. The metric values for this model, the best log likelihood, and the best distance gained model are shown as vertical lines. Each x axis represents values for the metric while the y axis is the density of values across the grid search results which represented the training S&T Distributions with high fidelity (min_dist_cor above 0.9).

Learn more

BirdFlow is a joint project between the University of Massachusetts Amherst and the Cornell Lab of Ornithology. It is funded by the US National Science Foundation

License

Bird Flow Models by BirdFlowScience are licensed under CC BY 4.0

Disclaimer and acknowledgments

This material uses data from the eBird Status and Trends Project at the Cornell Lab of Ornithology, eBird.org. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Cornell Lab of Ornithology.

We thank the USGS bird banding lab for the bird banding data collection, Birds Canada for preparing and sharing the data of the Motus Wildlife Tracking System, and the many individual contributors to these projects for collecting and compiling data on bird movement.

Last updated: 2026-04-21