Tag Explorer

Methodology — sort modes

Count and Signal Strength and Speed are the raw detection count, the strongest signal reading, and the fastest observed signal change for each ID, respectively.

Nocturnal and Diurnal split detections at true sunrise/sunset for the monitoring site's coordinates (computed with the standard NOAA solar-position equations), so the two percentages are complementary and sum to 100%.

Matutinal, Vespertine, and Crepuscular instead use a 90-minute twilight window centred on sunrise and sunset (45 minutes either side): Matutinal ranks by the dawn-window share, Vespertine by the dusk-window share, and Crepuscular by the two combined.

Cathemeral scores how evenly an ID's detections spread across the 24 hours of the day, using normalised Shannon entropy of its hourly histogram. 100% is perfectly even; 0% means every detection landed in a single hour.

Full Moon / New Moon split on lunar illumination (more vs. less than half-lit) and sum to 100%; Waxing / Waning split on which half of the lunar cycle the detection fell in and are likewise complementary. See Detections by Moon Phase for the underlying phase calculation.

First Appearance ranks IDs by their earliest detection, oldest first — like an arrival log. Last Appearance ranks by most recent detection, newest first — "who's been around most recently." Both are the only two sort modes that don't rank by a 0–100% value: the sidebar shows a compact MM-DD date instead, and the bar length is scaled between the earliest and latest date actually present in the list, not from zero (raw timestamps are all huge, close-together numbers, so scaling from zero would make every bar look nearly full).

Walleye / Bass / Muskellunge (prob.) are a heuristic behavioural classifier, not a measured identification — each ID's timing signature is scored against a hand-tuned profile of that species' known chronobiology, and the three scores are normalised to sum to 100%:

• Walleye — photophobic; rewards a high Nocturnal share plus concentrated (low-Cathemeral) activity.
• Bass — diurnal sight-feeder; rewards a high Diurnal share plus evenly spread (high-Cathemeral) activity.
• Muskellunge — crepuscular ambush predator; rewards a high Crepuscular share plus fast burst movement (high Max Speed).

Treat species probabilities as a prior, not evidence — the weights reflect published behaviour, not a fit to this data. See Methods & References for the sources behind each species profile.

A further Smallmouth / Largemouth split exists but isn't in this dropdown either — like Buddies, it's not a sort mode. It only appears in the Single-ID detail panel, and only when Bass is already that ID's top species guess. See that section for the reasoning.

One more sort mode, Buddies, exists but isn't in this dropdown — it only appears once triggered from the Single-ID detail panel's "Find my buddies" button.

Methodology

Day/night shading uses the standard NOAA solar-position equations (Julian day → solar mean anomaly → ecliptic longitude → declination → hour angle) for the site's coordinates, with the horizon taken at −0.833° to allow for atmospheric refraction and the sun's radius. It's recalculated per calendar date, so seasonal day-length changes are reflected.

Moon-phase icons — a dark circle (New Moon), right-half-lit (First Quarter), fully lit (Full Moon), and left-half-lit (Last Quarter) — mark the exact moment each phase occurs, computed from a standard synodic-month approximation (29.530589 days) anchored to a known reference new moon. This doesn't need the site's coordinates, since lunar phase (unlike sunrise/sunset) doesn't depend on location. Only phases falling inside whatever range is currently zoomed/panned are drawn, and they redraw as you navigate.

Offline periods come from offline.csv, a separately-computed log of recording gaps derived from AudioMoth filenames.

Methodology — Find my buddies

For every detection of the selected ID, each other ID gets credit if it has at least one detection of its own within ±30 minutes of that timestamp — at most one credit per target detection, no matter how many nearby detections that other ID actually has. The score is that count divided by the selected ID's total detection count, so 100% means "detected within 30 minutes of every single one of this fish's detections," and the target ID itself always shows 100%. IDs with no co-occurrences at all are left off the list rather than shown at 0%.

This puts the sidebar into a Buddies sort mode that isn't in the regular Sort by dropdown — it only exists once "Find my buddies" has been clicked, is scoped to whichever ID it was clicked from, and reverts to a normal sort the next time a different sort mode is chosen.

Methodology — Smallmouth vs. largemouth

This is a sub-split of the Bass guess, not a fourth peer species, so it only shows up once Bass is already the highest of the three species scores for this ID — otherwise the breakdown wouldn't mean anything. Smallmouth bass tend to be more active, roaming chase-feeders that stay busy through daylight, so that score rewards a high Diurnal share and a high Max Speed. Largemouth bass are slower, more sedentary ambush predators that lean nocturnal, especially in warm water, so that score rewards a high Nocturnal share and a low Max Speed. Diurnal/Nocturnal are already complementary, and so is speed vs. "not fast," so with matching weights on both sides the two percentages always sum to exactly 100%. Source: Methods & References.

Methodology

Each detection is matched to the nearest-timestamp temperature reading, restricted to dates that have at least one detection, then binned to the nearest 0.5° (i.e. ±0.25°).

Methodology

Lunar phase is computed from a standard synodic-month approximation (29.530589 days) anchored to a known reference new moon — no site coordinates are needed, since the phase itself doesn't depend on location. The cycle is split into four equal quarters, each centred on its named phase: New Moon spans the 12.5% of the cycle around phase 0, and so on for First Quarter (~25%), Full Moon (~50%), and Last Quarter (~75%).

This is the same phase model behind the Full Moon / New Moon / Waxing / Waning sort modes in the sidebar and the moon icons on the main timeline, so all three stay consistent with each other. It's a linear approximation, not full orbital mechanics, so exact phase moments can be off by roughly an hour up to about half a day depending on the time of year — closest for full moons.

Methodology

Speed is |change in signal| divided by the minutes between two successive detections of the same ID, reported in m/s after dividing the raw value by 100. Pairs more than 30 minutes apart, or with a zero reading at either end, are ignored — this rule is shared by both modes.

Max attributes each pair to the hour its first detection falls in, and a cell shows the largest speed among pairs starting in that hour.

Avg instead uses a sliding 2-hour window centred on each hour column — 30 minutes before it starts through 30 minutes after it ends — and averages every pair whose first detection falls in that window. Because neighbouring hours' windows overlap by an hour on each side, the same pair can contribute to two adjacent columns; this is intentional smoothing, not double-counting a single "true" value. Two points (one pair) is already enough to fill a cell — there's no larger minimum sample size.