Week 2#
Macropod behaviour and circadian activity#
Dawn, Dusk, and Other Inconvenient Appointment Times#
Recap: what did you predict?#
In your session, which day period did you predict would have the most recorded collisions — dawn, day, dusk, or night? Show of hands: who guessed right?
Your first graph#
| dawn | day | dusk | night |
|---|---|---|---|
| 22 | 17 | 34 | 17 |
Synthetic teaching data, from this week’s session.
Does this graph show when kangaroos are active — or which times get recorded in these rows?
What “crepuscular” actually means#
- most active around twilight — not fixed to a clock time
- a real, evidenced tendency across a population
- not a complete explanation for a specific spike in a specific dataset
Crepuscular is a behavioural pattern, not a booking confirmation for 6:00pm.
Why kangaroos approach roads#
- food and water access on disturbed roadside verges
- habitat fragmentation — the road cuts across an existing range
- established movement routes between shelter and grazing
- seasonal conditions: drought range shifts, joey season
- mobs move together
Animal activity, vehicle exposure and reporting#
- driver visibility drops as light falls; glare is worst near dawn and dusk
- traffic volume peaks at the same hours — commute time
- roadside vegetation can block the sightline early
Observed collision records = animal presence × vehicle exposure × probability of collision × probability of reporting
An inconveniently exact appointment#
- a disproportionate number of records sit at exactly 06:00 or exactly 18:00
- almost all from one
report_source: wildlife carer data_quality_flag: shift-start time recorded, not incident time
Either the kangaroos became remarkably punctual, or a human system has entered the graph.
Pattern versus explanation#
| Observed pattern | Tempting explanation | Alternatives |
|---|---|---|
| more dusk records | more animal activity | traffic, visibility, reporting |
| more records on one road | greater collision risk | traffic volume, reporting coverage |
The revised distribution#
| dawn | day | dusk | night |
|---|---|---|---|
| 18 | 17 | 27 | 17 |
Synthetic teaching data, same scale as the previous graph, with the eleven shift-start-time records removed. Dawn and dusk both drop; day and night don’t move — that’s what a recording artefact looks like once it’s taken out.
The revised conclusion#
The synthetic dataset contains more dusk records, but some timestamps reflect wildlife-carer shift times rather than incident times. The remaining pattern is consistent with several explanations, including animal activity, traffic exposure and visibility, which cannot be separated using these records alone.
A pattern in the spreadsheet may describe kangaroos — or it may describe the person holding the clipboard.
Next week#
Next week we put the problem on a map, where it will immediately look more authoritative.
Bring your revised distribution from this week’s session.