Gpsgridder & strain-rate grid butterfly pattern at station locations

Hi all,

I searched the forum to check whether the question has been posted before. I couldn’t see any mention of this, and I hope I am not mistakenly asking it a second time.

Sys specs
GMT version 6.4.0 installed on Mac arm64 Darwin Kernel Version 22.2.0

I calculated the strain-rate field with gpsgridder and grdmath; to do that I studied the scripts in the topex site CGM_V1 directory. When I plot the dilatation field, at the GNSS station coordinates, there are singularities -as if the greens functions are over-fitting at those locations. I checked the velocity grids, and when I zoom in, I cans those singularities in there too. Example dilatation rate screenshot is attached.

What I tried

  • I tried to fudge the solution by both the default value and larger values. But this didn’t solve the doublets of very large and small values.

  • I varied the Poisson’s ratio, but the pattern didn’t improve.

  • I fixed the velocity uncertainties to very large values, to see if the solutions would vary -they didn’t.

  • I played with the number or ratio of eigenvalues to retain, there wasn’t any improvement. The command line below represents my final trial.

    gmt gpsgridder data.lluv -R26/45/36/42 -I2m -Cn95 -W -fg -V -Ff0.01

    gmt grdgradient -Gvxx.grd -A270 -fg -V
    gmt grdgradient -Gvxy.grd -A180 -fg -V
    gmt grdgradient -Gvyx.grd -A270 -fg -V
    gmt grdgradient -Gvyy.grd -A180 -fg -V

    gmt grdmath vxx.grd 1000. DIV 1.e9 MUL = exx.grd
    gmt grdmath vyy.grd 1000. DIV 1.e9 MUL = eyy.grd
    gmt grdmath vxy.grd vyx.grd ADD 2000. DIV 1.e9 MUL = exy.grd

    gmt grdmath exx.grd eyy.grd ADD 2 DIV = dilatation.grd

Does anyone have suggestions to remove those singularities?

Best Regards,

Think your best bet is to post something on the GMTSAR git or wiki sites as those guys are actually doing this. David Sandwell is the man behind the theory on this and he and I wrote a paper on it reference in the docs. But I dont do InSAR so cannot really help.

Thank you very much @pwessel. I read your 2016 paper introducing this function. I would like to do basically the same thing. And actually my data is GPS-derived velocity solutions. And you are right; I should contact David Sandwell.

For those interested in a partial solution: the only time I got smoother results were those when I set a coarser resolution. So, this might have to do with the optimal -I value.

Best wishes