Bird Migration Loop Signatures

ISEF Category: Mathematics

Subcategory: Geometry and Topology  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A bird track can look messy on a map, but math can pull out a hidden shape. Some migrations form loops, detours, and repeated routes that simple distance measures miss. Your project can ask whether those shapes are different enough to separate species. That turns raw GPS points into a real topological test.

What Is It?

This project studies bird migration as a shape problem. Instead of asking only where a bird flew, you ask what kind of path it made. Did the route keep bending back on itself? Did it form repeated loops around stopover sites? Did some species trace the same broad geometry year after year?

Persistent homology is a branch of topological data analysis. Topology studies features that stay after you stretch or bend a shape, like holes and loops. Persistent homology tracks which loops appear and survive across many spatial scales. For bird GPS data, that means you can turn a migration path into a summary of its looping structure. A loop signature is your own chosen summary rule for those persistent loops, such as their number, size, lifetime, or pattern across scales.

Why This Is a Good Topic

This is a strong science fair topic because the data are public, the question is real, and the math is novel enough to feel original. You can test whether different species, seasons, or route types produce different loop patterns, then use statistics to see if the separation holds up. You will learn data cleaning, geometric thinking, topological summaries, and model comparison, all in one project.

Research Questions

• How does the number of persistent loops differ between bird species in public migration GPS tracks?
• What is the effect of stopover clustering on the lifetime of loops in a migration path?
• Does a loop signature separate spring migration from fall migration for the same species?
• To what extent do route smoothing choices change the persistent homology output?
• Which distance threshold settings produce the clearest species separation in topological features?
• How does path sampling rate affect the stability of the loop signature?

Basic Materials

• Laptop with enough storage for GPS datasets.
• Free Movebank account access to public bird tracking data.
• Spreadsheet software for cleaning and labeling tracks.
• Python installed on your computer.
• Jupyter Notebook for running analysis and saving notes.
• Free plotting software such as Matplotlib or Plotly.
• Public species and route metadata from Movebank or related museum records.

Advanced Materials

• Access to a university or school research computer with more memory.
• Python libraries for topological data analysis, such as Ripser, GUDHI, or Giotto-tda.
• Geospatial tools for coordinate handling and map projection.
• Statistical analysis software, such as R or Python SciPy.
• Version control software, such as Git.
• High-quality external storage for large track files.
• Access to a GIS package for route visualization and spatial comparison.

Software & Tools

• Python: Cleans GPS tracks, computes distances, and runs topological summaries.
• Jupyter Notebook: Lets you document each analysis step and keep code with notes.
• R: Runs statistical tests and makes comparison plots for species or seasons.
• ImageJ: Not needed for the math itself, but useful only if you build custom figure measurements from exported maps.
• GitHub Desktop: Helps you track code versions and avoid losing analysis changes.

Experiment Steps

1. Define one biological comparison, such as species, season, or age class, and keep it narrow.
2. Select a track-cleaning rule that removes bad GPS points without flattening real route structure.
3. Convert each migration path into a mathematical object that your topology software can read.
4. Decide how you will define and measure your loop signature from persistent homology output.
5. Build a null comparison, such as random walks, shuffled tracks, or resampled paths, to test whether your results matter.
6. Plan the statistics that will compare groups and show whether the separation survives repeated trials.

Common Pitfalls

• Mixing species with very different sampling rates, which makes one group look more complex just because it has more points.
• Projecting latitude and longitude badly, which distorts distances and changes the apparent loop shape.
• Choosing a cleaning rule that removes real stopover loops along with GPS noise.
• Comparing tracks with different lengths without normalizing for path duration or point count.
• Treating a pretty map as proof, when the loop signature needs a statistical test against a null model.

What Makes This Competitive

A strong version of this project will do more than make attractive maps. You can define a clear loop signature, test whether it stays stable under resampling, and compare it against a null model that mimics random movement. You can also ask whether topology adds information beyond standard measures like path length, curvature, or home range size. That kind of layered analysis shows real mathematical judgment.

Project Variations

• Compare loop signatures across two migration corridors for the same species and test whether route geometry changes by region.
• Build a species classifier that uses persistent homology features instead of only speed, distance, or stopover count.
• Study how missing GPS points change the persistence diagram, then design a correction method and test its effect.

Learn More

• Movebank: Search the public data portal for bird tracking datasets and metadata.
• NIH PubMed: Search review articles on persistent homology, topological data analysis, and movement ecology.
• NOAA Open Data: Use as a model for working with large public environmental datasets and coordinate data.
• MIT OpenCourseWare: Search linear algebra, differential geometry, and data analysis course notes for math background.
• Topological Data Analysis for Genomics and Evolution: Search for peer-reviewed review articles to see how TDA is applied to biological data.