How to get full resolution result of geo_velocity.h5?

[gjustin40]

The starting point for this question is here. Miaplpy Example

I am trying to Visualize the result of land displacement with tif data, How can i get the resolution that I prefer?

From above link, I found it that the key of resolution of result is mintpy.geocode.laloStep
I tried several cases with command smallbaselineApp.py factory.txt --dir ./miaplpy2/network_single_reference/ --start geocode by:

mintpy.geocode.laloStep     = 0.000092593,0.000092593 (which is 10m)
mintpy.geocode.laloStep     = 0.000555556,0.000555556 (which is 50m)
mintpy.geocode.laloStep     = 0.00092593,0.00092593 (which is 100m)

and visualized geo_velocity.h5 with save_kmz.py and upload to google earth. the results are:

(100m, 50m, 10m)

But when I use save_gdal.py function to result of each options, the resolution of tif data are same on QGIS
Furthermore, the resolution is different from what I expected.

I want to generate the tif data for visualize the resolut of land displacement with the resolution the i expected,
How to do that?

Thanks

[yunjunz]

Good question. By default, mintpy/geocode.py uses this approach to determine the output pixel size, which uses the following criteria:

1. the output pixel should have the same area as the input pixel
2. the output pixel lat/lon ratio matches the height/width of the bounding box of the radar pixel on the ground, as shown here (https://math.stackexchange.com/questions/4001034/calculate-the-dimensions-of-a-rotated-rectangle-inside-a-bounding-box).

Therefore, for a full-resolution product in geo-coordiantes, I think the default output grid (without manually setting the mintpy.geocode.laloStep option) is your best choice. And note that, the output grid in geo-coordinates will not be exactly 5 x 20 m, but should be close to that. Another point is that mintpy uses the nearest resampling only during the geocoding, thus, setting a mintpy.geocode.laloStep value smaller than the actual pixel size, as in the case of 10 m, will only duplicate some values in nearby pixels.

Alternatively, if you only need to show the data in KML, using the radar-coded velocity file, as in save_kmz.py velocity.h5 would avoid the resampling procedure, thus, no issue of choosing an output pixel size.

[gjustin40]

Thanks for your informations.

I want to generate TIFF data with a resolution as close as possible to 20x5m, although it doesn't need to be an exact match.
However, as you can see in the image above, around 24x40 seems to have too much difference.

What should I do to create a TIFF as close as possible using save_gdal.py?

Thanks

[yunjunz]

Make sure your AZIMUTH/RANGE_PIXEL_SIZE metadata is correct in your velocith.h5 file (they should be close to 2.3 and 14.1), by running info.py velocity.h5 to check.

1. Unset mintpy.geocode.laloStep option in smallbaselinApp.cfg; run smalllbaselineApp.py or geocode.py velocity.h5 -o geo/geo_velocity.h5
2. Run save_gdal.py geo/geo_velocity.h5

[falkamelung]

We just use the radar-coded files and then use the lat/long information for each pixel to plot them on a map (aka geocoding). We have code for plotting but it is not in a good shape. I can share though if you are interested.

[gjustin40]

Thanks for your support. @falkamelung

If possible, I would like to receive the code for this.
Thanks

[falkamelung]

Here the two scripts. The first plots velocity or DemError on a map. The second also plots the time series. But they are messy and the python needs re-factoring. You are of course welcome to do this. The second one will be renamed to view_full_res_timeseries.py or so.

https://github.com/geodesymiami/rsmas_insar/blob/master/minsar/view_scatterplot.py
https://github.com/geodesymiami/rsmas_insar/blob/master/minsar/view_ts.py