OSPREY Migration Stopover Analysis

Automated detection of stopover sites used by ospreys (Pandion haliaetus) during fall and spring migration, reanalyzing a legacy Movebank satellite telemetry dataset (Martell & Douglas, 1995–2002).

---

Research Questions

1. Can automated analysis detect discrete stopovers from coarse-resolution (date-only) satellite telemetry?
2. Do ospreys from different breeding populations share stopover sites during migration?

---

Dataset

Source: Martell & Douglas (1995–2002) via Movebank
Tracking records: 86,383 Argos Doppler-shift location fixes across 131 deployments
Files:

• Osprey in North and South America 1995-2002 (Martell).csv — raw tracking data (19 MB, 47 columns)
• Osprey in North and South America 1995-2002 (Martell)-reference-data.csv — deployment metadata (131 records)

Data Caveats

• Argos positional error ranges from ~150 m (Class 3) to several km (Class A/B/Z). Class Z fixes have no valid location and are excluded from analysis.
• Date-only timestamps prevent fine-grained step-speed filtering; analysis relies on NSD curves instead.
• Tag redeployment issue: 41 of 80 tags were redeployed on multiple birds in the original study. Analyses using tag ID alone blend trajectories across different individuals. All scripts from v2 onward use a composite (tag_id, animal_id) key to correctly separate deployments.

---

Methods

Stopover Detection

Pipeline steps:

1. Filter to visible=TRUE records only
2. Compute Net Squared Displacement (NSD) from each bird's breeding site over time
3. Identify flat segments in the NSD curve as stopover candidates (stationary periods)
4. Cluster candidate fixes spatially using DBSCAN (eps = 75 km, min_samples = 3, Haversine metric)
5. Filter out pre-departure staging at breeding sites
6. Flag stopover sites shared by 2+ birds across deployments

---

Results

Metric	Value
Total stopovers detected	78
Birds with at least one stopover	58 of 126 deployments
Shared stopovers (2+ birds)	24 (31%)
Highest-use site	Central Cuba — 4 birds, 3 years
Second-highest site	Northern Colombian coast — 4 birds, 4 years

Stopover duration increases markedly at continental barriers: median ~4 days inland vs. ~12 days at coastal barriers.

---

Dependencies

pandas
numpy
scikit-learn   # DBSCAN

Install with:

pip install pandas numpy scikit-learn

---

Usage

Run stopover detection

python detect_stopovers.py

Citation

Raw data: Martell M, Douglas D. 2017. Osprey in North and South America 1995–2002. Movebank Data Repository.