Viewshed and Line of Sight Analysis

Line of sight asks about one pair of points: standing here, can I see there? The computation walks the terrain profile between the two and checks whether anything rises above the straight line joining them.

A viewshed asks the same question for every cell in an area at once, and returns a map: visible or not visible from the observer. It is line of sight repeated in every direction.

STREAM computes viewsheds and can output the result as a raster or as vector polygons, so the visible area can be measured, exported and overlaid on a drawing.

This is the single biggest factor, and the one most often left on the default.

A viewshed on bare-earth terrain sees through buildings and trees, because they are not in the surface. It answers "does the landform block the view" — the right question for a landscape or planning study where the vegetation may be cleared.

A viewshed on a surface model including buildings and canopy answers "is the view blocked today". The visible area will be dramatically smaller, and for anything in a built-up or forested area the bare-earth result is simply optimistic.

Neither is wrong. But a report that does not say which surface was used cannot be checked.

The observer is almost never on the ground. An antenna sits on a mast, a camera on a pole, an eye 1.6 m above the surface. Adding the real offset changes the result far more than people expect, because visibility is decided by grazing angles over intervening ridges.

Target height matters just as much and is skipped more often. Asking "is the ground visible" and asking "is a 30 m turbine visible" give very different maps from the same terrain.

Every viewshed needs a limit. Without one the computation runs to the edge of your data, which produces long thin visible slivers at great distance that are technically correct and practically meaningless.

Set the radius from the physics of what you are studying: radio range, the distance at which an object of that size is actually discernible, or the study boundary you have to report on.

Over short distances the earth is flat enough to ignore. Over tens of kilometres it is not — curvature drops the far terrain away from the sight line, and atmospheric refraction bends the ray slightly back down. The two partly cancel, which is why radio work uses an effective earth radius rather than the true one.

If your radius is a few kilometres, this does not matter. If you are assessing visibility across a valley system or planning a long radio link, it does, and a plain geometric viewshed will be pessimistic about long shots.

Telecommunications: siting masts and checking link paths before anyone climbs anything.

Visual impact assessment: which settlements will see a wind farm, a quarry face or a building, and how much of it.

Security and monitoring: camera placement, and finding the dead ground a fence line cannot see.

Mining and quarrying: designing screening bunds and checking that a working face is hidden from a road or a village.

Forestry and emergency management: lookout tower placement and coverage gaps.

A viewshed is geometry, not optics. It reports whether a straight line is unobstructed — not whether a human eye would notice the object at that distance, in that light, against that background.

It also inherits every flaw in the surface. A DEM that smoothed away a small ridge will report a view that a site visit disproves in seconds. For anything consequential, treat the viewshed as the map that tells you where to go and check, not as the conclusion.