Noseleaf, grounded in real geography

Three ways of combining the bat prototype with real map data. Same game underneath, same real place, three different answers to "where does the data show up?"

Variant 1
Ground Truth
The world is one real hillside instead of an invented one. Nothing about the hunt changes - the ground just has the valley's real fall, and the title card tells you where you are standing.
Variant 2
The Survey
Ground Truth, plus a field notebook. Your own catches are pinned on the real map next to 1,570 real bat records, at three zooms. It shows what your instincts and the published survey have in common - and is careful about what they do not.
Variant 3
Read the Landscape
The map comes first. You pick tonight's hunting ground from an 11x11 grid of the real valley - reading relief, drainage, cover and where bats have actually been recorded - and the world is built from the cell you chose.

The place is Woodchester Park, Gloucestershire. It holds maternity roosts of greater and lesser horseshoe bats - the bat you fly - and Roger Ransome has studied that population since 1959, the longest continuous study of a wild mammal by one researcher.

The data is real and open. Ground height from AWS terrarium terrain tiles at z14, 129x129 samples over 1200 m. Bats from the GBIF occurrence API, filtered to CC-BY-4.0 and CC0 only, because roughly a third of GB bat records are non-commercial and this is a commercial venture. Baked once, offline, into static files - no API keys, no login, nothing phones home.

On resolution. The elevation source resolves to about 30 m and most bat records are published rounded to two decimal places, which means a cell roughly 690 m by 1113 m. The playfield is 100 m across. Every screen says so. The 3D world is built from the data; the map is where correlation is shown; no pin claims a bat was recorded at a particular tree.

Prototype builds for review, not finished levels. Variant 3 is the least complete of the three.