How to use TerraSAR data to creat DEM?

maybe some of these tutorials help you in general on how to deal with SAR data:

Input to Google Earth DEM elevation is relative elevation, right? How can I determine its absolute elevation,otherwise, how do I know the accuracy of the generated DEM.Thanks

did you perform Phase to Elevation?
Which steps were applied?

yes,I did perform phase to elevation>Range-Doppler Terrain correction,then export to geogle earth.

after Phase to Elevation you should be getting absolute heights.
The Google Export is not necessary, I would rather choose to include the DEM in the TC module as well by checking the respective box so you can compare your results to the SRTM data.

When I do terrain cottection step,show the dialog ,What is the role of “DEM” in the dialog box? Do i need to check it?

It will add a DEM-band, in your example SRTM 3sec (Auto download).

OK,Why are these two elevation values have such a big difference?Did I understand the wrong or the process is wrong?

you probably did nothing wrong but repeat pass interferometry (scenes from different dates) cannot produce high quality DEMs in most cases. How many days are your scenes apart?

Oh,this is the data I use.
TSX1_SAR__SSC______SM_S_SRA_20140610T225348_20140610T225356.xml TSX1_SAR__SSC______SM_S_SRA_20140702T225349_20140702T225357.xml

can you open both data sets and select Menu > Radar > Interferometric > InSAR Stack Overview
Add Opened > Overview

and post a screenshot of it?

Thanks,this is screenshot,

The front ofthe screenshot is my subset of the image,the full image is

as you see, they are 22 days apart. Bperp means the distance between the satellite positions at the two different dates:

High coherence is needed for good results. Did you compute a coherence layer with your interferogram? Only locations with a coherence of > 0.4 can be used/trusted for the DEM generation.

X-band is comparably short and looses much coherence with longer time. So it’s most likely temporal decorrelation which lowers your InSAR quality .

Please also see here:

Oh,maybe it’s temporal decorrelation.This is coherence of interferogram.

yes, probably. You can also have a look at the coherence raster where in your image your result is of good quality.

You can then mask your interferogram according to these areas:
Layer properties > valid pixel expression, for example coh_HH_11Feb2008_04Mar2008 > 0.6

Even if coherence was > 0.9 the atmospheric artefacts would still cause large DEM-errors in the repeat-pass case.

maybe I could try use a image(the time is 20140621)to interferorgam with image which I have ,:pensive:

you can compare coherence of different image pairs and see which is best.
Maybe also atmospheric influences are different for other pairs.

OK,thanks,I will try.