Wingbeats

Wingbeats analyzes eBird checklist data for Knox County, TN to estimate typical migration timing for each species. The analysis covers the 10 most recent complete calendar years (2016–2025), drawn from 84,965 checklists across all location types. Partial years are excluded to prevent incomplete seasonal data from skewing frequency estimates. The Birding Guide uses a different dataset scoped to eBird hotspot checklists only, so its statistics will differ.

For each species, a weekly detection frequency is calculated — the proportion of checklists that recorded the species in a given week. This is then smoothed with a circular 3-week moving average that wraps around the calendar year, so winter visitors whose presence spans December–January are handled correctly.

Migration status is determined by analyzing the smoothed frequency curve for sustained absences. The pipeline first identifies candidate gaps using a threshold (10% of peak frequency or 0.1%, whichever is greater) and a minimum gap length, then applies a two-tier verification to avoid false positives driven by seasonal detectability. The verification requires a deep enough gap (for example, mean frequency < 3% of peak), or an extended gap where many weeks have near-zero raw detections. This also supports multiple migration windows for species with genuinely separate passages (e.g., spring and fall), while keeping summer residents that merely go quiet mid-season as a single window rather than splitting them into false spring/fall periods. Species that pass these checks are classified as migratory.

All migration classifications reflect detection patterns within Knox County, TN only and should not be interpreted as representing a species' behavior across its full range. Many species have both migratory and year-round resident populations depending on geography — a species classified here as migratory may be sedentary in other parts of its range, and vice versa. Additionally, species with sparse local records may occasionally be classified as migratory due to factors other than true seasonal movement, such as juvenile dispersal, rather than a consistent annual migration pattern.

Arrival and departure weeks are set to the boundaries immediately surrounding the absence gap — arrival is the first week after the gap ends, and departure is the last week before it begins. For species with multiple migration windows the site reports arrival/departure and peak separately for each window; peak migration is the week with the highest smoothed detection frequency within that window. Windows that never rise clearly above the detection floor are treated as noise and dropped, so a few stray off-season records don't carve out a spurious window.

Confidence is scored on a 0–99% scale combining year coverage (60% weight) and checklist volume (40% weight, log-scaled). Species observed in fewer than 4 of the 10 years or with confidence below 65% are flagged as Rare, as are a few species that are locally rare but reported most years (sometimes through misidentification), which the automatic score would otherwise miss. These rarely reported species are excluded from migration alerts and the main timeline despite being almost exclusively migratory — with so few records, their apparent migration windows are unreliable.

Citation: eBird Basic Dataset. Version: EBD_relMay-2026. Cornell Lab of Ornithology, Ithaca, New York. May 2026. Data provided by eBird. Taxonomy from the eBird/Clements Checklist.