QGIS: Saga Channel network: how to prepare SRTM data

Question

I am trying to use the Saga toolbox (2.2.3) in QGIS (2.18) to create a channel network from SRTM data using this tutorial 1.

It works, but the result I get when using the DEM that comes with the tutorial are much better than what I get when using SRTM DEMs. The channels are mostly not connected to each other, no matter what I try. Is there a way to fix this, or is it because of the SRTM data being not good enough for this analysis?

So far, I am using the following workflow:

1. Catchment area (recursive): Step: 1, Method: Deterministic, Convergence: 1,100000

2: Channel network: Initiation theshold: Greater than 100000. I have also tried Less than 100000 and other values, but that does not help.

Do I need to prepare the SRTM data first before this analysis works?

Answer 1

You might want to try running Fill Sinks. This is available using SAGA via Processing. There may be GRASS versions too. I didn't see any mention of it in the tutorial you linked to, so it's worth a look.

What this does is make tiny adjustments to each cell's height, so that it maintains a minimum downward slope to its neighbours. That way, when you run the algorithms, the virtual water can always find a long uninterrupted path, without getting "trapped" in depressions/pits (these are cells surrounded on all sides by higher neighbours). By the sounds of it, these 'breaks' you see are probably due to these pits.

This tutorial explains things well (it assumes you're using SAGA from the saga_gui, but you can do these equally well using Processing).

I'd suggest looking at the SAGA Fill Sinks (Wang & Liu).

If this is SRTM data, you might also need to check to make sure there are no voids (nodata areas) as these could also need correcting. (To find these I usually set the canvas background to a garish colour like cyan.. if the DEMs are greyscale, voids will stick out like a sore thumb)

Answer 2

You may able to skip the Fill Sinks step by downloading a corrected DEM, recently made available at http://hydro.iis.u-tokyo.ac.jp/~yamadai/MERIT_DEM/index.html. It is a 90 m DEM based on SRTM data and was used to create Hydro rasters, also available to download here.

These data seem to be high-quality. I'm trying to get a stream network out of them, but am not sure of the easiest way to do this.