After flow accumulation analysis, what will be the criteria to define the threshold to extract the stream network from DEM? As stream threshold choice will influence the extracted drainage networks and thus the basin morphometric analysis.
I have several papers on researchgate that may be of interest. For channel accuracy, LiDAR is best, especially in areas witth dense forest canopy. The threshold size may depend on intent, full network with ephemeral streams or something else. In gullied terrain, highly erodible, channel network density is greater. Much DEM cleanup may be needed to remove road crossings as culverts and orient streams through DEM fill areas or across relatively flat valleys and coastal plains, where topography is relatively flat. In high rainfall hills to mountains, you might try 5 acres (2 hectares) to start. There is some trial and error. In humid, semitropical coastal plain, I think 29 acres (~12 hectares) did reasonably well. With LiDAR, some added detail may be possible.
An example of the stream network for Indian Creek in SC, USA may be an example of scale and density for a subwatershed that can be plotted and presented.
The most accurate way is to calibrate against real channel head, i.e. beginning of streams you observed. Depending on the scale of your study, you might get these channel heads from a field survey or also from topographic maps. Than, you can extract the flow accumulation values for the locations of the observed channel heads (or better a buffer around these locations) and use these values for getting an idea about the channel initiation threshold. You can see also e.g. the following papers: