Add DEM Hydro correction pre-processing

[roescob]

While testing the tool I encountered areas within my flood extent that were not drawn. This was due to topographic depressions withing my area of interest hindering flow towards the stream during the HAND map generation. This was fixed with a simple DEM preprocessing (fill depression).

Describe the solution you'd like Using a preprocessed filled DEM as input influence the accuracy of the flood depth results, this is solved by computing (raster calculator) the difference between the filled and unfilled DEMs and adding it to the depth output.

It would be great for readers or testers to have a note of that on the readme. Perhaps, implementing it onto the actual script.

Additional context This issue was experienced during v1.0.1

[cefect]

Thanks Roberto, can you give some more information on the steps you did? show a pre/post filtered DEM and the difference? Which algorithm did you try and which worked the best?

[roescob]

Sure, here is a more detailed answer.

RICorDE requires a filled DEM to hydro-connect each value to the stream throughout its computations, otherwise, the complete flood extent may not be shown correctly. The hydro-correction method may differ depending on the flood scenario, but the Fill Depression Wang and Liu (2006) method was chosen for the Fort Mcmurray 2020 flood because it establishes the outlet from the edge of the DEM dataset and iteratively fills depressions from the outlet to the highest point, similar to how Ice Jam floods inundate.

This is how the initial DEM looks like, lets call it DEM1:

This is how the filled DEM looks like, lets call it DEM2:

The difference from DEM1 and DEM2 is computed following this equation: DEM2 - DEM1

Since fill depression only increases the elevation this will output a positive difference as seen below, let's call it DEM3:

DEM3 might be similarly added onto the RICorDE final flood depth to enhance its validity. Before adding DEM3 to RICorDE output:

After adding DEM3 to RICorDE output:

[roescob]

Answering the question "Which algorithm did you try and which worked the best?"

The algorithms tested were the following:

• Fill Depression Planchon and Darboux (2001)
• Fill Depression Wang and Liu (2006)
• Breach Depression (stream enforcing, drainage enforcing, breach depressions)
• Breach Depression Least Cost
• Fill Single Cell Pits

A single fill depression method performed well as an INPUT among those given. Wang and Liu (2006)'s alternative was chosen because their literature method corresponds to the flood type.

Burning streams or drainage alternatives from the Breach Depression technique were not chosen since it is thought that decreasing real topographic values has a greater impact on other processes than enhancing them.

Fill Single Cell Pits, Breach Depression Least Cost, and Fill Depression Wang and Liu (2006) performed well together (in that sequence), although not as quickly as Fill Depression Wang and Liu (2006) alone.

[cefect]

This is great. Thanks Roberto. Can you mention which implementation of the DEM correction algorithms you used and provide a link? (I assume WBT)

Also, I suggest revising your naming of 'DEM3' to something like 'DIFF' as this grid is no longer a 'digital elevation model' (it has heights not elevations).

[roescob]

Yes, WBT: https://www.whiteboxgeo.com/manual/wbt_book/available_tools/hydrological_analysis.html#FillDepressionsWangAndLiu

Thanks for the suggestion