Loading the archive…
365-day tracker · 2025–2026
Did you really miss them?
Choose a date to separate a confirmed absence from a late, small, or simply undocumented flight.
The full year
Every night, in context.
Color shows what the public record supports—not what “usually” happens. Select any square for the evidence and its confidence.
Flight observed
Late or small
Explicit no flight
Weather-confounded
No clear record
Descriptive, not predictive
The seasonal shape.
Monthly totals make the winter lull and late/dark edge cases visible. They describe this archived year only; they are not odds for a future visit.
| Month | Flight observed | Late / small | Explicit no flight | Weather-confounded | No clear record |
|---|
Auditable record
Observation log.
The same filters apply here. Each observed row links back to its date-stamped Austin Bat Refuge source; select a date for the full interpretation.
Recovered Wayback evidence · 2011–2025 search
A sparse archive, date by date.
Older Austin Bat Refuge pages were mutable: each new week replaced the last. Internet Archive captures preserve fragments of those tables, yielding 122 dated observations across eight years.
Blank dates and years mean unobserved archive coverage—not “no flight.” This bounded pass recovered evidence from 27 of 72 distinct captures identified in the search. Only dated evidence can receive an observation label; public descriptions are conservative paraphrases.
Look up an older date
Try any date in the research window. A gap is reported as a gap, never converted into a bat conclusion.
Loading recovered observations…
Recovery by year
| Year | Recovered dates | Flight observed | Late / small | Explicit no flight | Weather-confounded | Unclear evidence |
|---|
Recovered observation list
NOAA NEXRAD prototype
A promising signal, not ground truth.
Weather radar can screen for elevated, expanding biological echoes near the bridge. At this distance it cannot prove that an echo began at Congress Avenue Bridge—or that the scatterers were bats.
NOAA candidates are not confirmed bridge observations. None of the seven pilot nights is added to the historical labels above. A missing candidate signal also never means “no flight.”
20held-out labeled nights in 10 adjacent-date pairs
0.80accuracy, balanced accuracy, flight precision, and flight recall
0.584–0.91995% Wilson interval for held-out accuracy
0 / 7unconfounded candidates in the June 2024 pilot
Why KGRK is the primary radar
KGRK is 61.86 km from the bridge. Its lowest sweep is centered about 847 m above the bridge, with a 328–1,365 m half-power vertical span, a 1.04 km azimuthal footprint, and 0.25 km range gates.
That footprint is much larger than the bridge. The defensible observable is an echo appearing aloft nearby and expanding over time—not the exact exit point or emergence instant.
Held-out screening result
Leave-one-pair-out evaluation kept each adjacent flight/no-flight pair together. The 20 held-out predictions produced 8 true no-flights, 8 true flights, and 4 errors.
Accuracy 0.80 · balanced accuracy 0.80 · flight precision 0.80 · flight recall 0.80
The sample is small, seasonally paired, and calibrated against ABR radar interpretations rather than independent field observations. It is a screening result, not biological validation.
Seven-night June 2024 pilot
June 10–16 produced zero unconfounded candidates above the predeclared 0.65 threshold. June 10 had the pilot's highest score, 0.702, but precipitation forced indeterminate_weather. The other six were no_candidate_signal.
Scores are not calibrated probabilities, and neither pilot outcome supplies a flight or no-flight label.
NOAA method & sources
Five Level-II volumes per night were sampled around sunset. The prototype measured weak-to-moderate, low-correlation echoes in a bridge core and plume zone, their change and outward movement, contrast with a control annulus, and precipitation confounding.
Sources: NOAA NCEI Level-II dataset, NEXRAD on AWS, and NOAA radar site database.
Methodology & limits
Built to say “unknown.”
The reconstruction favors a transparent gap over a confident guess. A missing post is never treated as a no-flight night.
Current tracker source & extraction
The complete 365-day tracker comes from Austin Bat Refuge's public WordPress Radar archive. The build script reads the observation date from each title/slug, resolves documented source metadata typos, extracts Congress-specific timing and quantity fields, then emits one ordered row per date. The older sparse archive is kept separate and marked as recovered Wayback evidence; its public descriptions are original status-level paraphrases, with exact captures linked for source review.
Confidence
- High: a Congress-specific flight or no-flight statement.
- Medium: a study-area-wide no-flight statement or an explicitly weather-confounded report.
- None: no clear bridge conclusion or no date-stamped post.
Movement vs. a show
Weather radar can document bats moving even when they emerged after dark or were hard to see from a particular bank. “Reported visible” is used only when ABR explicitly says bats were viewable. Otherwise the page says visibility was not reported. ABR's viewing guide explains why location and darkness matter.
Caveats
- ABR says its quantities are generally maximum-density snapshots unless marked as emergence totals; they are not ground counts.
- Radar can be suppressed or confounded by rain, outflow boundaries, birds, insects, and outages.
- UTC radar times are converted with the America/Chicago time zone. Relative-to-sunset labels are used only when the source states them.
- This is an independent reconstruction and is not affiliated with or endorsed by Austin Bat Refuge.