FIELD NOTES / IMPERFECT DETECTION
Quietmarsh
A quiet reed bed is not necessarily empty.
Which habitat holds more occupied sites?
Spend up to 48 visits across two habitats. A bird call confirms presence; a quiet visit might mean an absent bird or a missed detection. Revisit sites to separate these possibilities. Each site permits three visits. A sweep surveys every unfinished site in one habitat once and costs one visit per site.
Submit the habitat with higher occupancy when its model support reaches 80%, after at least 24 visits and four revisited sites in each habitat. The instruments infer both occupancy and detectability from your observations.
Open marsh
Reed marsh
● = bird detected · ○ = quiet · — = not surveyed. Tap a site for one visit.
Start by covering both habitats, then return to quiet sites.
Reed marsh estimate
Write the field conclusion
How this inference works · source & license
The likelihood of a detected history is occupancy × detection-history probability. An all-zero history also includes the possibility of absence. This is a JavaScript port of unmarked’s occupancy likelihood, integrated over a 19 × 19 parameter grid with uniform priors. Charts show relative support for occupancy; they are not confidence intervals. Hidden scenarios are synthetic fixed histories, not field data. Real surveys require stronger assumptions and uncertainty analysis.
GPL-3.0-or-later. Complete editable source: model.js, game.js, style.css, source/build notes. No compilation or dependencies.
GPL license · unmarked source