I have been given a set of depth points for a waterbody that were taken along the shoreline and along linear transects. What is the best way to use these points to interpolate/extrapolate bathymetric contours for the entire waterbody? Previous attempts result in contours that are "pinched" around the transect points and don't appear to be interpolated fully for an accurate depiction of depth (see images below).

Example 1

Example 2

  • 2
    Can you send your surveyors back and get some North-South transects? I'm no expert in this type of analysis but I thought a more zig-zag transect was usually the way its done? – Hornbydd May 18 '15 at 15:01
  • 2
    You're seeing the "pinching" because the interpolation tools in ArcGIS are naively interpolating based on point locations without any regard for hydrologic processes, sedimentation, etc. For this reason, more specialized bathymetric interpolation tools are probably needed to create a robust result unless you have many transects. This question goes into some of the other options. There's also a lot of literature about bathymetric interpolation. – dmahr May 18 '15 at 15:12
  • 2
    I'm with Hornbydd - this just looks like a lack of sufficient sample points and/or poor distribution of those that are present. I'm sure dmahr's point about methods specific to hydrologic processes is valid, but I don't think that's the cause of what is shown in the images (more it would be the accuracy of any overall interpolation compared to another method). Which interpolation method did you use, and what settings? – Chris W May 18 '15 at 18:49
  • As other's have said, not enough data. Doesn't look like a big lake: hook up a Garmin GPS with a Sonar (fish finder) and a laptop and go for a boat ride. Works surprisingly well once you filter out the noise. Other untested method would be to used a drone with a polarizer filter on the camera lens - this removes glare so you could capture imagery below surface then model terrain from imagery. Probably only doable with shallow lakes. – Jakub Sisak GeoGraphics May 20 '15 at 18:22

You not alone with this problem. I was able to minimize the pinching somewhat by adjusting the density of zero depths represented by the periphery of the lake. More data is always great but I suspect that until you have enough points so that you get equal representation, the interpolation will be influenced by the points of greater density. However, in my case it was sufficient to illustrate the band of greater depth in the center and no one was concerned by the pinching.The overall length in my case was about 400 m and depths were taken at 5 m intervals. enter image description here


You definitely are working with insufficient data and there is no getting around that. The suggestion about lowering the density of your 0 points is good. Personally I don't use points for my shoreline, I use a line... but you can't do that if you are using krigging or something. My "solution" gets to the same issue though. Data along the transect is dense and data between transects is sparse. If you create a TIN from your points and shoreline, and then create a grid of points (density dependent on your waterbody size and data density) and bring your TIN "depths" into your points and then run your interpolation on your original data and the tin based depth points you can get rid of a lot of the scalloping effect. I find this is a balancing act between scalloping effect and TINish looking contours. In the case shown by user18597 if would have likely helped (as did his solution), but given your data distribution... you just don't have very nicely collected/sufficient data.


You have insufficient data. You have two options

  1. Smooth the data with a very large value. 2 divide the body up with polygon cross sections and then clip the data and create contours for each region using the same symbology so the data can be presented uniformly.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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