I've been using the STRM 90m DEM for determining hydrological characteristics for projects that we are working on in Africa. To date I've found very few anomalies within the DEM. The only drawback with the SRTM 90m DEM is that the the river reaches (flowpaths) are not always well defined which affects the hydrological characteristics. I've recently downloaded the HydroSHEDS updated version of the SRTM. I'm currently busy with a project and decided to use the HydroSHEDS DEM to derive the hydrological characteristics and found that the DEM has anomalies\artifacts (i.e. stripping) that run through the DEM that looks likes squares most probably introduced during the merging of tiles.
Is there anyway of removing these systematic errors from the DEM without grossly altering the DEM? I've come across some users using Fast Fourier Transforms filter to remove systemtic errors from DEM. I'm not sure how to apply the following within ArcGIS. Any suggestions are welcome.
I've attached an example image of the systematic errors.
If you want to apply something like a Fast Fourier Transformation you will probably have to set up and run the appropritate Spatail Analyst Math tools to create an FFT algorithim, a difficult task in ArcGIS but easily performed in ERDAS. In looking at your example though, I don't think you are dealing with an image anamoly, but rather a processsing anaomly created during a merge as you suggest, as such an FFT wouldn't necessarily be appropriate. I would try using the Majority Filter tool (essentially a high-pass filter) first, changing the input paramters for neighbors and definition unitl you produce a reasonable surface.