Help needed: How should I interpret the PS processing output in SAR geometry?

[Shima-shoki]

Hi, I was able to run the dolphin by using the coregistered SLC files by using ISCE2. I believe that my files are in SAR geometry and there is no geometric transforms in *.vrt files. I chose Jakarta area as an example of InSAR processing. Some intermediate files such as the amp_mean.tif (Fig. 1) and ps_pixels (Fig. 2) looked okay as shown below. The question is, the final outcome of the PS processing, such as the velocity, has a strange unit. Based on a previous discussion #503 , the unit of the dolphin output is in meters. However, the output shows the unit in -20 ~ 0 based on the plot, which is unlikely in a usual subsidence events. The velocity output also has a range from -11 ~ 10.

My guess is that if a SAR geometry data is supplied to the software, the output is calculated in the unit of radians, but I am still wondering how to correctly interpret the output in this case. I also really appreciate to know whether dolphin can make a geocoded output as a final product. Any kind help is very much appreciated!

The log & config file: dolphin_log-config.zip

Figure 1. "amp_mean.tif"

Figure 2. "velocity.tif"

Figure 3. A visualization of displacement following the "basic walkthrough" example. The "masked" parameter was changed from "True" to "False", since the visualization was not shown if it is set to "True".

[scottstanie]

Hi I haven't checked the logs, but right now there's no attempt to automatically recognize the sensor of input data; therefore the output would be in radians if you don't add "wavelength" to the config.
(OPERA sentinel 1 does have a switch now, since it's been the main test cases. Considering adding more later, but unsure now)

And if the "masked" argument for loading shows nothing, I believe that means it is a nodata value... do you mean the time series is masked? Or it's masked based on the connected component labels?

[Shima-shoki]

@scottstanie
Thank you very much for the quick reply!

there's no attempt to automatically recognize the sensor of input data; therefore the output would be in radians if you don't add "wavelength" to the config.

Well noted about the output of the PS processing. I rechecked the documents, and there is a "wavelength" option in the code reference as per the advice. However, I could not find it when I called "dolphin config -h". Is it an intentional behavior of the software ...? (sorry if I missed it in the help!)

And if the "masked" argument for loading shows nothing, I believe that means it is a nodata value... do you mean the time series is masked? Or it's masked based on the connected component labels?

Sorry my question was vague where the "masked" option was changed. This is the code I used to visualize the final output as per the "basic walkthrough" instruction. I believe that there should be some kind of output values in the file, but my understanding of the Dolphin is yet to be quite limited...

fig, axes = plt.subplots(ncols=2, figsize=(20, 6))
ax = axes[0]

last_displacement = load_gdal(file_list[-1], masked=True)#The visualization in axes[0] is empty (axes[1] is showing a time-series)

axim = ax.imshow(
    last_displacement - last_displacement[0].mean(), cmap="RdBu", vmax=0, vmin=-20
)
fig.colorbar(axim, label="LOS [radians]")

ax = axes[1]
row, col = np.unravel_index(np.argmin(last_displacement), last_displacement.shape)
dates = [get_dates(f)[1] for f in file_list]
timeseries_values = [
    load_gdal(f, rows=slice(row, row + 1), cols=slice(col, col + 1)) for f in file_list
]

ax.plot(dates, np.array(timeseries_values).squeeze(), ".-")

I also tested the visualization of the "velocity.tif" with masked=True option, then the output looks okay.

velocity=load_gdal('timeseries/velocity.tif', masked=True)
plt.imshow(velocity)

[scottstanie]

Ah I didn't add a clear enough comment- that subtracting of the mean of the first row was just an arbitrary way to spatially reference the image. Your image looks like it has nans in the first row, so I think you're subtracting nan

[Shima-shoki]

@scottstanie
>that subtracting of the mean of the first row was just an arbitrary way to spatially reference the image
I see! Thank you for your elaboration on the procedure. I now understood how to interpret the outcome of my processing :)

If possible, would you mind me to give a guidance on processing the final output map (i.e., velocity) to generate a georeferenced image in Dolphin?

[scottstanie]

There's no dolphin utility now to do this with images in radar coordinates, but I know that the UNAVCO tutorial introduction to GDAL demonstrate how to use GDAL along with the lat and lon files to do so (search for geocodeUsingGdalWarp at the bottom)

[Shima-shoki]

@scottstanie

I know that the UNAVCO tutorial introduction to GDAL demonstrate how to use GDAL along with the lat and lon files to do so (search for geocodeUsingGdalWarp at the bottom)

Thank you for kindly sharing the information about the UNAVCO tutorial! I'll have a look into the material to make geocoded stacked SLC files to be supplied to Dolphin🐬