mintpy on WCape results

[MaryemBl]

Hello everyone,
I'm currently applying mintpy workflow on WCape dataset, link below, (i'm using snap), I've got results but cannot validate them. I need expected results from using mintpy or other techniques (gps data...) applied on this area.
to anyone who have worked with this dataset, how did u validate your results?
https://zenodo.org/record/4318134#.YrWhA3ZBxPY
Thank you

[mosyhey]

Hello,
As you know, the InSAR displacement time-series is relative in both time (i.e. reference date) and space (i.e. reference point). First, you should calibrate the displacement time-series with a CGPS station. Then, you can validate the calibrated displacement time-series with other CGPS stations.

I suggest this process, but I am not sure it's true. I hope @yunjunz, @hfattahi and other friends approve or correct it:

1. In step reference_date, set the reference date at the first date.
2. In step reference_point, set the location of a CGPS station as the reference point. Consider the following conditions in selecting the station:

## 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

3. Set other settings and run MintPy processing. Tropospheric correction is highly recommended.
4. After MintPy processing, calibrate the whole displacement time-series with the reference CGPS station data.
5. Eventually, validate the calibrated displacement time-series with the other CGPS stations.

You can see my primary validation result in our study area here. I tried all sorts of settings, and got R2≈0.85 and RMSE≈8mm at best.

[yunjunz]

Your procedure looks reasonable to me.

[MaryemBl]

Thank u @mosyhey, @yunjunz for your answers, I will try this. In step 4, did u include GNSS data in the InSAR 3D matrix to calibrate the displacement time-series with the reference CGPS station data?

[mosyhey]

Thank you dear @yunjunz.
Yes, dear @MaryemBl , I include the reference GNSS station data in InSAR matrix for calibration, in this procedure:

1. Doing steps 1 to 3 in the previous post. Imagine the resultant InSAR matrix has 3 layers or dates [Date1,Date2,Date3]. Also, imagine we selected CGPS A as the reference point for calibration and CGPS B for validation.
2. Getting data from CGPS stations on those three dates. If you access hourly or minute-by-minute data, get the data for the same time as image acquisition times.
3. GNSS data has three components: North, East, and Up (NEU). For example, component of U in CGPS A is [23,28,32]. Standardize every component of all stations by subtracting the first value from them. In this example, it will be [0,5,9].
4. Converting the CGPS NEU displacement to satellite LOS (Line of Sight) displacement, by this formula. For example, LOS displacement of CGPS A calculated as [0,3,7].
5. Calibrating InSAR matrix [Date1,Date2,Date3] by adding CGPS A LOS displacement [0,3,7] to it (see image below).
6. Extracting calibrated InSAR LOS displacement value at the pixel of CGPS B. In this example, it is [0,5,11] (see image below).
7. Validating it [0,5,11] with CGPS B LOS displacement time-series (calculated in step 4) as here.

Try different settings to get the best correlation. Please share the final R2 and RMSE.