the main difference is that ASTER is based on optical data (kind of a stereoscopy) and SRTM is based on radar interferometry. This has consequences on the surfaces described by theses approaches.

ASTER can be noisy in areas of strong topography whereas SRTM can be noisy in areas of low coherence (vegetation cover). To a certain degree, this noice can be reduced by filtering. But that, however, destroys fine-scaled structures.

I would have a look at both (hillshading is a good way to get a visual expression but make sure the data isn't projected to WGS84 as this produces faulty hillshading), maybe calculate a delta image (difference between both) to see where errors might occur. Then you can estimate which of both is more suitable for your area of interest.

Two interesting comparisons:

• http://www.digital-geography.com/dem-comparison-srtm-3-vs-aster-gdem-v2/
• http://www.opentopography.org/blog/comparison-aster-gdem-srtm

If none of both is of sufficient quality, you can also try the ALOS World DEM which is also available for free at a spatial resolution of 30 meters: http://www.eorc.jaxa.jp/ALOS/en/aw3d30/index.htm

The reason why it might be better (true to my experience) is because it was originally generated at a resolution of 5m (not for free) and was then upsampled to 30 meters. This might have preserved structures which are not present in ASTER and STRM as their original resolution is around 30 at best.

Article SRTM vs ASTER elevation products. Comparison for two regions...

I include a few co-authored papers on my researchgate suggesting benefits of LiDAR coverage detail is a substantial benefit (2007, 2015).  When working with 30 m resolution, expect more mistakes, especially in flat or heavily wooded terrain without well defined topographic features.  Many DEM coverages are based on topographic contour coverages, developed by cartographers, and there is some mistakes to be expected in photo interpretation with manual photographic methods.  For critical research grade work, the detail provided by LiDAR for defining hydrologic boundaries and stream detail and georeferencing, will almost always be worth the effort, especially in low gradient terrain like coastal marine terraces, even though the surface data detail and need for cleanup of flow diversions or barriers like roads, logs, dikes, beaver dams can be substantial effort.  In areas with well defined terrain, the 30 m DEMS are more likely to be suitable enough for hydrologic boundaries with some checking of poorly defined local areas.  The 30 m DEMs often have poor georeferencing of streams in heavily wooded areas and low gradient valley systems.  

Thank you all,  giving valuable information.

In my experience, 1"SRTM is much better in delineation of watersheds most of the time. Also, in last time I use Global Mapper (v.17 and above) because of very fast and accurate delineation (plus you can merge smaller catchments into larger one or define that within the options). Accuracy depends not only of DEM, but of used software and algorithm also (I'm very satisfied with GM for now, and I export the data in Arc or other software). 

Thank you very much sir.

you can see this link http://www.digital-geography.com/dem-comparison-srtm-3-vs-aster-gdem-v2/.

SRTM (1-arc) is better to extract drainage network as well as watershe boundary as compare to ASTER data with same resolution

I agree with Javed Ikbal.

I agree with Javed Ikbal