One valley, seven ways of seeing it.

Relief, measured.

The elevation grid as a hidden-line mesh. Contours are traced by marching squares; every fifth is an index contour. Hover the terrain to read a height and trace its contour in red.

Contour levels drawn

–

Contour length

–

Under the cursor

hover the terrain

Hillshade and slope, one gradient.

Hillshade and slope come from the elevation gradient at every pixel, drawn as line work: a line screen for light, cross-hatching for steepness, or illuminated contours.

What a mast can see.

Click the terrain to place a mast. Rays from the antenna to every edge cell test line of sight across the grid, with Earth curvature and standard refraction (k = 0.13), for a receiver 2 m above the ground.

Visible area

–

Share of the tile

–

Farthest visible cell

–

Mast ground elevation

–

Mast position

–

One pass, one swath, line by line.

A sensor on a sun-synchronous orbit sweeps its swath down the valley. Optical shows real HLS imagery of 21 August 2025 as a line screen. SAR shows backscatter modelled from the terrain geometry, with layover and shadow.

Ground-track heading

–

Orbital period, 700 km

–

Tile inside the swath

–

Revisit at 51.4° N, about

–

SAR layover · shadow

–

GeoAI as a method, not an oracle.

Candidate slopes are computed from the elevation grid. The detection boxes and their scores show how a model’s output is filtered by confidence and reviewed. A method demonstration on real terrain, not a hazard assessment.

Candidate area

–

Candidate regions

–

Detections at or above threshold

–

Every station, counted in its cell.

Open-network station positions are aggregated, in your browser, into the geodesic cells of a Goldberg polyhedron, the same idea as H3 hexagons. Filter by domain; extrude by count. Drag the globe to turn it.

Stations aggregated

–

Occupied cells

–

Most in one cell

–

Cell holding this valley

–

Cell under the cursor

hover the globe