I'm working on producing a graph that realistically represents the visible landscape in a 360º view from a single point. I'd appreciate all ideas on how I could accomplish this. I've tried many things without success and I'm really a newbie at GIS.
I'm currently working with the Zonal Statistics as Table tool to have a table of values that relate DEM elevation with one degree classes. My main difficulties with this are:
1. Mountain tops that are farthest away appear huge in my graph, which is not consistent with what is actually seen (and the viewshed results, for that matter, which do take into account this factor).
2. I would need all elevation values within a degree to have a sensible depiction of the landscape: the table tool only provides the maximum value every one degree which at the moment is giving me mostly hilltops. An interesting mountain pass that could be seen from the observing point has disappeared completely in the resulting graph.
I hope I've provided enough information and thank you all.
Now what is the distance threshold from you observation point that you are interested?
You indicate that viewshed is not good, but you can control the vertical view angles so can can capture that interesting valley.
I picture something along the line of sampling the raster cells within a distance bound .... going back to your original idea... say between 900 to 1000 m for example. The average distance being 950 as a centerline. you can then average the raster values normal to this line, then unravel the circle to a straight line... this effectively gives you a circular 'profile' of whatever you are sampling. You need not be worried as much about the angular sampling. It can be less dense closer to the observation and more dense the further away. You sampling is going to be controlled by cell size in any event. Think about it, in the extreme... if you have 10m cell sizes, the only angles you can sample at a distance of 10m from the observation point are the 8 cardinal cells surrounding the observation point (cell). As you move substantially away from the observation point, your angular sampling will actually increase.
so the question should have some focus on what it is you intend to do with this elevation data? (or any other values for that matter) You will have a circular profile that can be unravelled to a straight line. Each profile will be averaged normal to the line taking care of variance in that direction. The only remaining dimension to analyse is that which is along the circle ... hence line. Focus less on the sampling strategy and elaborate more on the end goal.
Thank you Dan, I want to read carefully what you're suggesting to make sure I understand it properly and test it out.
My project is in archaeoastronomy, specifically analysing the cultural landscape at an archaeological site. The first part is to determine the visible landscape within a 60km radius. The terrain is quite flat where the observing point is situated, but rather elevated, and the adjacent mountain chain can reach 4,000m mark. My viewshed analysis actually renders other hilltops farther away as visible (even with the earth-curvature box checked) but I've not taken them into account because this is a coastal site and the fog often obstructs visibility. Situating the landscape in degree-frame is needed for the study: for instance, the mountain pass I mentioned is observed at 90º and there are other distinctive features observed at specific angles. The next parts of the analysis, where the astronomical features and the architectural remains come into the picture, also would require this degree-frame to be set.
then that makes sense... you don't really need to sample at fixed degree increments... you just need the ability to sample by degrees at some orientation... they are very different things. My initial suggestion is for creating the sampling profile from the 'circle', the next step would be to sample that information at a particular orientation and/or distance from an observation point. This as you are probably aware, has been dabbled with and is simple to implement... /blogs/dan_patterson/2017/01/04/circles-sectors-rings-buffers-and-the-n-gons . Keep us posted, particularly if you intend to linearize the circlular sample of the elevation data