Hi, allthough I have not tried it, I can imagine you could work with some EDGE DETECTION algorithms, a number of which are available out there. To implement them yourself could be quite a  challenge, but there are several GIS / Remote Sensing programms that deal with it already available. Even though this could identify the bank edges, you would still have to discriminate them from other types of edges, with something like proximmity to river network and so on.

A more simple approach could start by taking the "slope of the slope" (second derivative from elevation for those mathematicians out there). With a proper threschold (maybe using some local statistic)  you can then identify places where the slope changes abruptly as in river banks. Hope it gives you some clue on how to proceed.   

I am not a big fan Dave Rosgen's, "Applied River Morphology", but if you have not looked at it you might check Chapters 2 and 5 on bankfull determination.

The River bathymetry toolkit can be set to provide the bankfull limits. It can be downloaded at http://www.fs.fed.us/rm/boise/AWAE/projects/river_bathymetry_toolkit.shtml

Daniele,

very intresting tool. My idea was to implement a gis software extension able to compute the bankfull limits.

Could you please address to a resource where the approach used is explained?

Thanks

Thank you for this question

For European flood directive, we developped in France semi-automatic 1D models for extreme event. For this type of event (more than 1000 year return period), the reach and the knowledge of limits between main channel and floodplain don't need to be very accurate. The tool developped is Cartino and available in French with theses links (http://wikhydro.developpement-durable.gouv.fr/index.php/CARTINO_PC, download http://www.cete-mediterranee.developpement-durable.gouv.fr/outils-pour-la-phase-cartographie-r55.html)

Now, we are looking to work with frequent events before overtopping on dykes, levees or floodplain.

The first tests are to launch a lot of calculation increasing the discharge to estimate some limits. The procedure can work but the return period is not the same in each rivers. So I try to do some tools near yours and combine geometric analyse with 1D hydraulic models (see pdf File).

My main difficulties are on rivers with the main channel higher than the floodplain, mainly in mediterranean areas with a lot of morphology due to sedimentation and shape of cones.

With yours information, I will try to intregrate your promising developments in addition to 1D hydraulic model.

To compare my results with your developments , I try to download extension but I cannot find on qgis extension the tool described on this paper http://2014.ogrs-community.org/2014_papers/DeRosa_OGRS2014.pdf and I hope you can help me.

Best regards

Dear Frédéric

The plugin is still in development but you can find the version in development here:

https://github.com/pierluigiderosa/BankFullDetection

Try and let me know

Regards

This paper might be helpful

Johansen K, Tiede D,Blaschke T, Arroyo LA,Phinn S. 2011. Automatic geographic object based mapping of streambed and riparian zone extent from LiDAR data in a temperate rural urban environment, Australia. Remote Sensing 3: 1139–1156.