I need to determine the mudflat area for a group of small islands in the Red Sea. I have read some publications and it seems to be possible using remote sensing, for example Landsat. But I am not exactly sure how to go about it, I know I should try to get images at low tide of my area of interest but I am not sure how to go about selecting the mudflats areas.

The most intuitive way seems to be to select pixels with a specific RGB combination, since the mudflats seem to have a very specific color. I am just interested in mapping the extent of the mudflats.

Someone has suggested to me "multispectral classification using a standard method like the maximum likelihood classifier" and "ratio analysis or decision-tree classifications" but I am not a remote sensing person and these seem to be way too technical.

edit: I have available argic 10.2 full license and might be able to get Imagine 2014

  • What software do you have available? Also, the "multispectral classification [etc.]" suggestion essentially is "selecting pixels with a specific RGB combination", but in more discipline-specific terminology. – Erica Oct 22 '14 at 11:05
  • @Erica That's important I just added that in the question – Herman Toothrot Oct 22 '14 at 11:21
  • A starting point if you are using ArcGIS is to open the help file and search for the topic "What is image classification?". You'll be able to very quickly get up and running with the Image Classification window. – Hornbydd Oct 22 '14 at 11:26

There's a tutorial using ArcGIS here:


(note that probably requires an expensive license).

You basically train the model by selecting some points that are mud, and some that aren't, and then the thing learns what's mud and what's not, and produces a raster of mud/not mud, perhaps with probability that it got it correct.

Or if you want to save a few thousand pounds/dollars/euros then there's a QGIS plugin that might do it:


Note these things are going to get technical, so you might want to read up on some basic RS literature, or at least know what a multi-band raster is.

The easy alternative is to load the image into QGIS and manually digitise the mudflats by drawing round them...

  • ok that seems a potential and intuitive solution. The author also mentions unsupervised classification, what would those be? It might be more accurate perhaps? – Herman Toothrot Oct 22 '14 at 11:26
  • 1
    unsupervised classification is where the algorithm itself works out the classes by looking at the pixels and "clustering" by intensity in the bands. It might not find "mudflats" as a category, or it might find several categories for "mudflats" since smooth mudflats might look different to more rippled mudflats or slightly damp mudflats. In a supervised classification you'd put samples of all those in the "mudflat" training set. – Spacedman Oct 22 '14 at 11:31

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.