I would like to be able to identify saddle points in my grid.
Does anyone know of a method to do this? Search has led me in the way of Numpy. Would that be the right direction to go or can this be achieved with Spatial Analyst?
I geographical terms, I think of the saddle point as a mountain pass like in the second picture in the link. From that point, the surface curves up in one direction and down in another.
In more mathematical terms, this means that the point will be the minimum value in its row and maximal in its column. This, I believe, is only valid for a smaller neighborhood. If you consider a larger area like most DEM's, you could potentially have a min value in a row in a sink, i.e. the surface curves upwards in all directions around it. At the same time you can have a "true" saddlepoint in the same row somewhere else in the DEM.
This means I am not sure how I can adapt the solution from Sebastians suggestion to take a larger area into account. Or I am thinking in the wrong direction here?
I guess that you could fit some helmets to your geographical features and obtain your saddle point by analyzing the deterministic surface you would obtain from that fitting. Another suggestion I would explore is passing a low pass filter through your raster so your geography gets smoother.
I geographical terms, I think of the saddle point as a mountain pass like in the second picture in the link. From that point, the surface curves up in one direction and down in another.
You may want to look at the Curvature tool. Comparing that value with some other measures (slope for example) may help you isolate the area you're interested in.
Stefan, I am trying to find/locate saddle points in a DEM as well. Did you ever find a solution? If so, can you point me to it or describe it to me? You can email me at jed.flint -at- yahoo.com
