I am trying to virtually put a blanket over a DEM so to say, in more technical terms, a convex hull over 3d point data. To get some insight in what I'm doing exactly, I will first explain my entire plan so it becomes clear what the purpose of the convex hull is.

I am trying to detect sand bars close to the shore of the Netherlands. I have point measurements of the depth and may plan is as follows: first I remove the slope of the coast, secondly I invert the DEM and lastly I use ArcGIS' hydrology toolset to find the spots in the landscape where water will accumulate, these areas are my sand bars.

Here is a sample of my point data. It's transects from a part from the coast. And here is what I would like to to visually: enter image description here

The order of the three images is arbitrary. I realised after making this image, it's also possible to first remove the angle and than invert it.

My question right now is that in order to remove the slope, my plan is to create a convex hull over the DEM, which would look like putting a blanket over the 3d landscape (enclosed on the bottom, but this is not relevant for me). After this I will subtract the point measurements form this straight slope, resulting in a flat landscape with indentations in it. Does anybody know of a python, arcgis or qgis library or tool that could do this? I have been looking at ArcGIs' minimum bounding geometry, but this only works for 2d data. R's Convhulln might work, but there might work, but I'd rather find something in Python.


2 Answers 2


I don't recommend the approach you've outlined. As an example based on the simple profile you've provided, the "hydrology toolset" approach would require a "fill" operation to remove all depressions (I.e. all the areas you're interested in). As a result, the whole region containing both inverted dunes would be flooded and the associated flow accumulation would run through the centre of the flat.

I recommend either trying to run the topographic position index (TPI), Or calculate elevation residuals. In either approach you can play with the window size of the operation to best suite your needs.


Have you tried one of the following: 1) to calculate a curvature raster (either planform or profile), then thresholding it in order to highlight the features you are interested in ?; 2) to calculate the average elevation from the DEM, and then subtracting it from the original DEM, in order to highlight what is convex and what is concave? Both strategies (easily to put in practice in ArcGis) require a trial-and-error approach.

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.