How to set a good reference point / the way of mintpy doing automatically

[Clément Goldmann]

Hey all,

I working on Jakarta zone for subsidence, I have presented some of the results and a good question returned : How i choose my reference point.

In my case I let Mintpy doing by himself, the code about the reference seems to only choose an area with high coherence :

########## 3. reference_point
## Reference all interferograms to one common point in space
## auto - randomly select a pixel with coherence > minCoherence
## however, manually specify using prior knowledge of the study area is highly recommended
##   with the following guideline (section 4.3 in Yunjun et al., 2019):
## 1) located in a coherence area, to minimize the decorrelation effect.
## 2) not affected by strong atmospheric turbulence, i.e. ionospheric streaks
## 3) close to and with similar elevation as the AOI, to minimize the impact of spatially correlated atmospheric delay
mintpy.reference.yx            = auto   #[257,151 / auto]
mintpy.reference.lalo          = auto   #[31.8,130.8 / auto]
mintpy.reference.maskFile      = auto   #[filename / no], auto for maskConnComp.h5
mintpy.reference.coherenceFile = auto   #[filename], auto for avgSpatialCoh.h5
mintpy.reference.minCoherence  = auto   #[0.0-1.0], auto for 0.85, minimum coherence for auto method

In my report a mention about the reference point was asked about the localisation. Because if the reference point is also under subsidence that make difficult to doing on all the zone. Some people said me you can use a gps station as reference if it's stable over time but if the zone you have no gps station / no gps station without movements it can be difficult.

How Mintpy doing this automatically ? Because it seems in the paper of yunjun " Small baseline InSAR time series analysis: Unwrapping error correction and noise reduction The reference pixel should be (i) located in a coherent area; (ii) not affected by strong atmospheric tur- bulence such as ionospheric streaks and (iii) close to and with similar elevation as the area of interest to minimize the impact of the spatially correlated atmospheric".

It seams only put the reference point into high coherence area but not an area which is not affected by subsidence.

Cheers,

Clément

[Eric Fielding]

Hi Clément,

You are correct, the MintPy automatic selection of the reference point is based on choosing a point with high spatial coherence. The automatic selection does not consider the other criteria.

You should manually choose a reference point that you think is not affected by subsidence, relatively close to the area of subsidence, has high coherence, not affected by strong ionospheric streaks, and close to the same elevation as the area you are studying.

Cheers,

   ++Eric

[Clément Goldmann]

Hi Eric,

Really thanks to your reply. 

For the next procedure I will select a GPS from Nevada lab not so far away of my zone with the lower up velocity.

Do you have any papers about the zonation of ionospheric streaks ?

Wish you a great day.

Kinds regards,

Clément

[Eric Fielding]

Hi Clément,

I am not sure I understand what you are asking about "zonation of ionospheric streaks". Strong ionospheric streaks (short spatial scale variations) are generally a problem for L-band InSAR near the equator and near the poles, but for C-band InSAR the short-wavelength variations are less severe. For L-band data, it is important to do an ionospheric correction if possible. The ISCE2 software includes ionospheric corrections in topsApp, stripmapApp, alos2App, topsStack, and alosStack.

See these papers for discussion of ALOS-2 and Sentinel-1 ionospheric corrections:

Liang, C., E. J. Fielding, and M.-H. Huang (2017), Estimating Azimuth Offset With Double-Difference Interferometric Phase: The Effect of Azimuth FM Rate Error in Focusing, IEEE Transactions on Geoscience and Remote Sensing, 55(12), 7018-7031, doi:10.1109/TGRS.2017.2737955.

Liang, C., Z. Liu, E. J. Fielding, and R. Burgmann (2018), InSAR Time Series Analysis of L-Band Wide-Swath SAR Data Acquired by ALOS-2, IEEE Transactions on Geoscience and Remote Sensing, 56(8), 4492-4506, doi:10.1109/tgrs.2018.2821150.

Liang, C., P. Agram, M. Simons, and E. J. Fielding (2019), Ionospheric Correction of InSAR Time Series Analysis of C-band Sentinel-1 TOPS Data, IEEE Transactions on Geoscience and Remote Sensing, 57(9), 6755-6773, doi:10.1109/TGRS.2019.2908494.

All the best,

    ++Eric

[Karthi J]

Hi Mr. Clément & Mr. Eric, I find it as very interesting conversation above.  I have one doubt, i have selected my reference point nearby my study area, but I'm unsure about it stability like it may also goes down. So how do you make sure about it? 

If i just choose a GPS station nearby my study area, and it also has some minimal subsidence, how can i correct it overall. Is there any offset method are you using? Thanks in advance.

[Carolina Hurtado]

Hi Karthi et al.,

I would like to share what I do to answer Karthi's question. If it is not correct or if there is an easy way to correct for the reference point motion I would like to know!

I also take a GNSS station as reference point. I take a station with known motion and with a time series >3 years, and I try to use a station that has not large motions, and should be in a coherent area. As I know the rates in the ENU components, I project the velocity into the LOS direction of the SAR satellite using the incidence and azimuth of InSAR, here is the equation to do that https://groups.google.com/g/mintpy/c/etKnhodR4ZI/m/BcrwZpfZAgAJ?hl=en 

So, let's say that the reference point is moving 1 mm/y in the LOS direction.

Then, to correct the velocities, I do VEL -  1 mm/y (reference point velocity in LOS). I use shapefile format, so I apply this to all the points in the shapefile. This corrects the velocities, but not the time series.

To correct the time series, I take the reference point velocity (e.g., 1 mm/y) and use the per day rate (i.e., 1/365 mm/day). Then remove this from the time series considering the number of days between acquisitions. So, if the time between the first and second image are 50 days, the deformation to remove for the second date is  (1*50/365 mm). 

Hope this is clear (and right).

Carolina