Drainage basins ignoring lake

Question

This is my first time writing in this forum and I hope this question offers some quality of information easing a solution. Here goes.

I am new in analyzing hydrology with DEMs and I have followed several steps to identify watersheds and drainage basins. (I am using QGIS for all my operations)

1. Obtained the DEMs and combined them using "Build Virtual Raster"
2. Filled the holes using "Fill sinks (wang & liu)" providing a filled DEM and Flow Directions
3. With the filled DEM I've calculated the Strahler order (which I'm not familiar with) and identified rivers with this tool.
4. A new raster was calculated with the Strahler order's values >= 8 resulting in an acceptable river raster layer (validated through a basemap).
5. Then I've done a Channel network and drainage basins with a threshold 8 (to have the same acceptable result as my Strahler order) and saved the channels and drainage basins. (I suspect my error might be in this step)
6. Then the result that I get gives me a drainage basin going straight through a lake, which makes no sense I find. Shouldn't the lake it'self be a receiver of a whole network of watershed?

Here is a screenshot of the result I'm getting with the DEM in the background.

The drainage basins are the red limits. I figured the limits of the drainage basins would either be where I inserted the arrows 1 or 2 or both, as we can clearly see the lake/reservoir in the DEM. I figure I might have missed a step to tell my procedure the limit of every water plan in this process as it doesn't seem to take them into account.

Answer 1

Through discussions and research, I have figure out that the processes I have used simply do not identify the lake as being a lake and "simulate" a stream flowing through what the DEM identifies as being a flow direction.

My main issue with this lake was that I know for a fact that water does not flow out from the south, it's actually a reservoir and infrastructures were probably built to prevent the water from flowing out. It's natural that the DEM would not support such a situation and that it would try to identify the logical flow of water. Some parameter in r.watershed allows you to define the locations of "real depressions" and in this parameter I have vectorised the reservoir, given it it's altitude in a new field and rasterized it with pixel value being it's altitude. This seems to have partially fixed my problem but still some zones remain untrue to what they should be. Some water basins that should flow in the reservoir show as the opposite and I will need to fix this.

Here are a few links to the related topics and questions I have had:

Slecting point to obtain water basin leading to it

Lowering density of GRASS r.stream.extract result?

drainage basin for whole lake QGIS 2.18.7

Get drainage basin of a polygon (with QGIS, GRASS or other FOSS)