How we calculate a viewshed

A viewshed is a map of the ground a camera can see. It is a geometric statement about the land, built from public elevation data and the facts you give us about the camera. This page sets out what we assume, where those assumptions stop holding, and what the result is reasonably used for.

Last updated: September 2026.

What the shaded area means

Shaded ground is ground with an unobstructed line of sight to the camera, within the direction it faces and the distance we analysed. It answers one question: could a straight line reach this spot without the land getting in the way?

It is not a statement about what is visible on any particular day, in any particular weather, or at any useful size in the picture. Those are different questions, and the sections below say what we do about each.

What we build it from

Ground elevations come from public federal survey data — the USGS 3D Elevation Program where it covers the site, and NASA SRTM elsewhere. Water, roads, place names and protected areas come from federal and openly licensed mapping sources. Every sheet names the sources it actually used; where a sheet and this page disagree, the sheet is right, because it was produced from the data of the day.

Elevation data describes the bare ground. Trees, buildings and structures are not part of it. That matters enough to have its own section below.

We account for the curvature of the Earth and for atmospheric refraction, which bends light slightly around that curve and puts the true horizon a little further out than pure geometry would. We use the standard refraction coefficient used by mainstream GIS viewshed tools, so our horizon distance is comparable to theirs.

How far we analyse, and why

This is the assumption most worth understanding, because it is the one that most often surprises people. A camera's reach is limited by three separate things, and we take the smallest of them.

The land

Whether the curve of the Earth, a ridge, or a nearer hill gets in the way. On its own this can be a very long distance: from a mountain overlook, a high ridge 200 km away can be in line of sight.

The air

Distance is not the same as visibility. Haze, humidity and particulates set a practical limit that is measured routinely — the National Park Service reports that Shenandoah National Park averages about 64 miles of visual range, and about 32 miles on its haziest days. Terrain beyond the prevailing visual range is geometrically in view and practically invisible. Visibility is also strongly regional: eastern sites are typically far more limited than western ones.

The camera

A camera has a finite number of pixels spread across its field of view, so distant ground occupies less and less of the picture. At 60 miles, a single pixel of a typical wide webcam covers well over 100 feet of ground — an entire ridge becomes a few pixels of haze. A narrow lens reaches usefully further than a wide one from the same spot, so this limit is specific to your camera rather than a general rule.

We analyse the ground that satisfies all three. Where a camera could geometrically see something far beyond that, we may tell you so separately — but we do not extend the map to cover it, and you should not read the edge of the analysis as the edge of what exists.

What we assume about your camera

A viewshed is only as good as the description of the camera it starts from. We use the position, the height above the ground it stands on, the direction it faces, the width of its view, and its tilt.

Two of those deserve particular attention. Position is the one nothing downstream can recover: because nearby features shift more than distant ones when the viewpoint moves, an error of a few tens of metres can change which ridge blocks the view. And a field of view taken from a specification sheet, or estimated, is not the same as one measured against the landscape — we record which of those we used, and it is shown alongside the camera.

Where a camera has been aligned against its own photograph, the alignment replaces the entered figures. Where it has not, the entered figures are used as given.

What we do not model

The honest list, because a viewshed that looks precise invites more trust than it has earned.

• The weather on any given day. Fog, rain, snow, smoke and glare are not in the model. The analysis describes a clear day.
• Foliage and season. Elevation data is bare ground. Where vegetation matters we model it as screening you can review and adjust, but a viewshed does not change between summer and winter on its own.
• Most structures. Buildings, towers and fences are not in the elevation data. Near-field obstructions can be added deliberately; nothing is assumed automatically.
• The lens and the sensor. We assume an undistorted view. Very wide lenses bend straight lines, and that effect grows toward the edges of the frame.
• Anything that moves. Vehicles, cranes, growing trees, a camera nudged by wind or by a person.
• Light. A viewshed says nothing about whether a place is lit, in shadow, or facing away from you.

How precise the edges are

Elevation data is sampled on a grid of tens of metres, and a viewshed computed from it inherits that. A boundary on one of our maps is therefore a good approximation of where visibility changes, not a surveyed line. Narrow features — a single tree, a wall, a gap between two buildings — can fall between samples.

Edges are also least reliable where a sight line grazes the ground almost tangentially, because a small error in elevation moves the boundary a long way. In practice this means long, flat sight lines are softer than short, steep ones.

What it is reasonably used for

Good uses — the ones the product is built for:

• Telling visitors what they are looking at, and what lies beyond the frame.
• Comparing candidate camera positions before committing to one, and understanding what a mount height buys you.
• Explaining coverage to a colleague, a board, a permitting authority or a funder — with the sources named.
• Planning where a camera cannot help, which is often the more useful half.

Uses it is not suitable for, without independent verification:

• Anything safety-critical. Do not rely on a viewshed to establish that a hazard, a person or a vehicle would be seen.
• Legal, boundary or privacy determinations. Whether a camera can see onto a particular property is a question with legal consequences and metre-scale stakes; this is not a survey.
• Detection or recognition claims. That ground is in line of sight says nothing about whether anything on it could be identified.
• Regulatory submissions that require surveyed accuracy, unless the relevant authority accepts modelled elevation data.

What would make a result wrong

If a map does not match what you see, these are the usual reasons, in the order they are usually at fault: the camera's position; its height above the ground; the direction and width of its view; trees or a structure that the bare-ground data does not contain; and last, and much more rarely, the elevation data itself.

All of those are correctable, and a corrected camera can be reanalysed. If something looks wrong, it is worth telling us — the first four are facts about your installation that we can only get from you.

When results change

A viewshed is recomputed when you change the camera's description or ask for it. It does not update on its own, and it does not track changes on the ground. Every sheet carries the date it was generated; that date is the honest answer to “how current is this?”