I'm trying to make a 3D map with multi-beam bathymetry (65536 features, xyz), but it seems that the interpolation methods are too rigid. I need to smooth the data so it looks more natural and realistic.

Which parameters do I have to vary so I can get this?

  • Im using ARCGIS 9.2 – Romina Jul 26 '13 at 14:26
  • +1 Would also be curious if Lidar processing techniques would be applicable. – Kirk Kuykendall Jul 26 '13 at 16:45
  • 7
    I'd be very suspicious about the data source because 65536 is exactly the row limit of an excel spreadsheet prior to v2010 :-) – WolfOdrade Jul 26 '13 at 16:53

This is likely not entirely an issue with the interpolation model. Bathymetric data can exhibit considerable noise. Because of an equal weight associated with each TIN facet and outlier effect, A TIN base interpolation can extenuate this noise and is not recommended. I would apply a Topogrid (Topo to raster tool) Spline interpolation and then apply a smoothing filter to the result. I commonly use a Gaussian weighted filter with a sigma of 2, but in ArcGIS you could just use a focal mean. The size of the window will depend on the resolution of the interpolated surface and an error criteria. You do not want to oversmooth the data so, assessing the Root Mean Squared Error (RMSE) of the observed vs. predicted is essential. Find a window size that exhibits an acceptable balance between smoothness and error.


Try FFT (Fast Fourrier Transform) in ENVI or other Image Processing Tool after creating your raster. You could also apply it directly on your raster using IDL.

  • very good suggestion! – Jeffrey Evans Jul 27 '13 at 20:22
  • 1
    The FFT by itself does not solve this problem--it is just a way of re-expressing the same data. What exactly do you propose to do with the FFT in order to smooth the data? – whuber Jul 28 '13 at 22:10
  • You have to apply a filter to some frequencies (I think to remove high frequencies) on your FFT image and then do a reverse FFT – Below the Radar Jul 29 '13 at 11:48

Try making a TIN from the bathymetry. A TIN will do interpolation of the points (lines), creating a smoother surface than a grid. You could also run a hillshade on the interpolated grid, this may produce something you like visually.


You can try the Topo to Raster interpolation just in one step, trying different values for its smoothing parameters: the discretization error factor (1.5, 2 or higher), tolerance #1 (try 2-3) and tolerance #2 (about 100). Drainage enforcement should be off for bathymetry (no enforce), and data type "spot". Maybe you manage to get the desired smoothness without filters.


I hate the unnatural contours you get from most models. Here is my workflow:

  1. Create shoreline (try to use newer lidar/imagery)
  2. Create TIN based on shoreline and point sounding data
  3. Convert Tin to raster (I use 1m resolution typically)
  4. Use Multi Values to Points to extract raster values to sounding points
  5. Calculate a "Difference" field in your sounding data and manually inspect areas where the difference is > 0.5 metres. Make necessary deletions.
  6. Create TIN based on corrected soundings and shoreline
  7. Convert TIN to raster (I use 5m resolution typically)
  8. Convert raster to point.
  9. Make selection from fill point data set with sounding point data set (I use within 10 metres) and delete these points form the fill points data set
  10. Make random selections to reduce the density of the fill data set (I just query for FID/some number = CEILING(FID/some number) after making initial deletions and saving.
  11. Use the randomized fill points, the sounding points, the shoreline, and an extent polygon (from the shoreline) in the Topo to Raster tool.
  12. Create contours.

This gives you smoothed contours, but preserves the measured values for your sounding data. It is no better, but I think it looks much better.


In terms of simplicity a TIN can provide a very reasonable return.

I see no reason why noise precludes the use of a TIN per se. They will exactly model your data points if you set the parameters up that way or fit more smoothly to the surface. They also have the distinct advantage of being scale and grid orientation independent, unlike any moving window based method.

I would suggest :

  1. Check WolfOdrade's point first!
  2. If the bathymetry data is arranged in rows or some other irregular geometry (e.g. from boat soundings) then topogrid these these to a raster surface. If it is points in a regular grid or randomly scattered, make it into a TIN that fits every point.
  3. Visualise in something simple, like ArcScene.
  4. Then Post a picture of it somewhere so we can actually see the data, showing your points of concern.

The right smoothing algorithm is very dependent on the type of terrain, Guassian filter might be good for smoother DEM. FFT would be worth the effort for varied terrain and could handle sharper angles, but iteratively weeding the TIN nodes could be a sufficient and simpler option that minimises simulating data.

  • 2
    "Exact modeling" of noise only propagates error. When there is noise, you ought to prefer methods that smooth the data rather than honor them. That suggests looking at alternatives to TINs. Your suggestion near the end of iteratively weeding TIN nodes is in this spirit, but unfortunately once you're done you are still pinning the TIN to the remaining nodes, none of which are correct (due to the noise). Applying a statistical smoother sounds like a better idea. – whuber Jul 28 '13 at 22:09
  • My point as regards TIN is that they are just a data structure, with their own smoothing methods. But whether in TIN or GRID, filtering leaves raw data with errors. Smoothing requires good point distribution. Both remove maxima and minima which might be real and necessary. All grid methods are vulnerable to scale and orientation. Without seeing the data and know the end use we don't know what methods it needs or can sustain. If it is just for visualisation then a bit of limited random error might actually make it look more realistic thus filter/weed rather than smooth. – AnserGIS Jul 29 '13 at 10:18

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.