Viewshed Analysis adjust for treetop/building roof classifying as visible

Question

I need to depict the areas that a landfill will be visible within a 3-mile radius at the current height.

After pouring over the ARC Visibility toolset page, my thought is that the Viewshed tool is the best way to address this. For the input raster data I used an STRM raster which includes vegetation and building heights and added it to a land surface DEM to get an elevation/vegetation & building hybrid raster. For the input line observer feature I used a line feature class outlining the current footprint of the landfill. Seems simple enough. After running the tool, it looks pretty good.

Here is my concern. My understanding is that all the areas that are visible from the landfill looking out get classified as visible. But the goal here is really the opposite. Most of the time this detail doesn’t seem to matter. I “field checked” some points through google earth streetview and it seemed to match up well. The areas I noticed an issue is where there is very tall tree cover. These locations are being classified as visible, I think because the tree tops are visible from the landfill. But for this analysis, when looking from land surface (where a person would be looking from) those trees actually obscure the view. Is this a valid concern and if so, how could this be adressed?

This is my first time using any 3D analysis tools.

Answer 1

This is likely due to the spatial resolution and accuracy of your STRM data. Ideally, you would have a digital surface model (DSM) with better spatial resolution and, subsequently, better accuracy. I see two options to correct the situation: 1) check if point cloud data is available for your area of interest so you can generate your own DSM at a higher spatial resolution and 2) if no point cloud data is available, consider “burning in” (aka altering) field derived data into your existing DSM.

Answer 2

The issue that you are facing is rather common one but unfortunately it does not have a simple solution. You need to understand that the viewshed function answers the question "What area is visible from a specific point?". The question that you need to answer is in fact opposite - "From what area is a specific point visible?". There are some specific assumptions under which these questions are pretty much equal but your case does not fulfil any of them. What you ideally would need to do is to calculate so called reverse (inverse) viewshed. The problem is that it is not really implemented in common GIS.

Fortunately, there is a way to obtain approximation of the result that you need. It has some limitations but it would provide at least some outcome. You calculate viewshed as you described but you need to set correctly your OFFSETA and OFFSETB. OFFSETB will be the height of observer and OFFSETA is some small elevation above surface. The details are described in this answer. After you obtain the result you need to mask out the areas that the observer cannot really reach (e.g. high trees as you mentioned, or tall buildings etc.). Hopefully, you can do that based on some data - landcover or something similar. That would provide you with roughly the areas from which an observer can see the target area (landfill).

One small note regarding your methodology. Depending on the size the landfill it may not be sufficient to use just the outline. The situation can happen that you do not see the outline but you would be able to see some inner part of the polygon. So you may need to add at least few points inside the polygon to prevent a mistake.