I need to run a Viewshed analysis (ESRI 10.2 spatial analyst tool) to answer this: what areas will be able to see a proposed 30m high structure.

I have the choice of a DSM or a DTM to use as the input elevation model. I am thinking that the DSM will be better as it will be the tops of trees/buildings etc.) which in the real world will (mostly) obstruct the views of the structure. Is that the best one to use?

If so what is my Offset B ( the height of the observer). Usually I would put this as 1.5m - does this still make sense to do this. Or will this be like putting a person standing tops of the trees?

  • @MappaGnosis covered your first question very well. For the seond question, as far as I know, the offset represents the height of the eyes of the observer. Therefore, 1.5m results in the view of a ~1.6m tall person. Commented Nov 16, 2015 at 8:19
  • I've added some supplementary info for the second part. Commented Nov 16, 2015 at 8:36

3 Answers 3


Your choice of DSM vs DTM will depend on whether there is best-practice guidance or regulations governing the planning process for which you are performing the calculation.

For instance, Viewshed Analysis for planning in the UK, especially for windfarms, requires that you use DTM data (and specifically Ordnance Survey data). Analysis with DSM may be used as a supplementary presentation to demonstrate the screening effect of buildings and trees but a 'bald Earth' calculation is the standard requirement. The reason for this is that screening is rarely total and, in deciduous woodland areas, is also seasonal. On top of that trees can be cut down and buildings demolished.

This is why the phrases 'Zones of Theoretical Visibility' (ZTV) or 'Zones of Visual Influence' (ZVI) are used in many countries' planning systems to describe the output of 'bald Earth' (DTM) viewshed analysis, as it is the worst-case potential impact that is being measured.

You can only answer your question by reading the planning regulations that pertain to your country, state or province. While we are on the subject, the relevant planning regulations will almost certainly also specify the resolution of data required. If you are in doubt, my advice is to use a 10m resolution and present both a worst-case (DTM) and current-reality (DSM) viewsheds with a radius of 15km (for structures up to 50m). Then describe/show the differences in a narrative or 3rd map.

As it is unlikely you are going to use a 0.5m resolution DTM over a radius of 10km plus, a further problem with trying to calculate 'true viewsheds' is that small/thin features such as walls and hedges may well either not be represented at all in the DSM or be over-stated in some areas. Any clever planner/lawyer can instantly nullify the validity of your entire DSM-based calculation by an argument based on this fact and suggest that you are claiming a false level of accuracy. This is another reason why a worst-case analysis is often required because the analysis gives a definite baseline for theoretical visibility within which some screening may be achieved by small structures, eye-height differences etc. In individual contentious locations you are well advised to preform a line of sight calculation based on site-surveyed data of tree and building heights and present that as profile drawings.

Given the resolution of your height data and common radii over which the viewshed analysis must normally be run, it therefore only really makes sense to consider large areas/structures such as entire Woodland blocks and towns or villages in a DSM-based calculation. A common approach for the second part f your question is to perform the calculation normally with a Bald Earth DTM supplemented with DSM data or a simple addition of +15m in woodland areas and +5m for building footprints and then mask these areas in the output (clearly stating that you have done so in notes on the map). The figures of 15m and 5m are deliberately conservative to counter any arguments that your analysis is intentionally disingenuous in its suggestion of screening. These are just commonly used values. Whatever value you use will need to be justified.

Finally, 1.5m is a common value for eye height, but then so is 2m. You can argue the 2m represents a VERY tall person but the logic is again, worst-case scenario with some compensation for terrain inaccuracy. Again, read your planning guidance because there may be a specific height prescribed which you MUST use.

  • This definitely makes sense. Go the conservative route. Great answer to the first part of my question. In my case there is no planning regulation to follow as it's in the preliminary stages so more of a first draft at what may be visible. Commented Nov 19, 2015 at 17:41

The Viewshed tool uses a single DEM as input. @MappaGnosis explains clearly that you must decide whether a DSM, DTM or a combination of the two is best for your input DEM.

By default, the observers eyes are at DEM + 1 (default OFFSET A = 1). So, yes, if you use the DSM as your input DEM and specify nothing else the observers eyes will be one meter above the trees/buildings, if there are trees/buildings at the observer point.

The observer height (eye height) is set as shown

  • If OFFSET A = not specified, SPOT = not specified. Eye height = DEM + 1
  • If OFFSET A = specified, SPOT = not specified. Eye height = DEM + OFFSET A
  • If OFFSET A = specified, SPOT = specified. Eye height = SPOT + OFFSET A

If the SPOT is set, the sum of the SPOT and OFFSET A must be more than the DEM value or the viewshed tool will not work. Also OFFSET A must be >= 0.

Some examples assuming the DEM value at the observation point was 80 m.

  1. OFFSET A = not specified, SPOT = not specified. Eye height = 81 m
  2. OFFSET A = 2, SPOT = not specified. Eye height = 82 m
  3. OFFSET A = not specified, SPOT = 100. Eye height = 101 m
  4. OFFSET A = 2, SPOT = 100. Eye height = 102 m
  5. OFFSET A = 2, SPOT = 72. Eye height = will not work as 74m is less than DEM.

Using a DSM is pretty much required for viewshed analysis. If you use a DTM, then many obstructing objects will have been removed.
However, as you yourself noted, Offset A/B is somewhat tricky, but this can be either completely ignored, or the analysis can be run with different values for Offset A/B, and then combined with a land-use model. This will allow you to differentiate your Offset A/B between open ground (roads, fields etc), where the value should be around 1.5m, and forest areas, where your value should be 0, or potentially negative (if the model can manage that).

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.