Skip to content

Starting Point of DePSI

The DePSI package is designed to work with coregistered interferogram stacks or reconstructed SLC stacks. It expects this stack data to be in the form of xarray.Dataset, with specific attributes and dimensions. Below is a minimum example interferogram stack in xarray.Dataset that DePSI can work with:

<xarray.Dataset> Size: 2GB
Dimensions:    (azimuth: 2000, range: 4000, time: 10)
Coordinates:
  * azimuth    (azimuth) int64 16kB 0 1 2 3 4 5 ... 1995 1996 1997 1998 1999
  * range      (range) int64 32kB 0 1 2 3 4 5 ... 3994 3995 3996 3997 3998 3999
  * time       (time) datetime64[us] 80B 2018-09-20 2018-09-26 ... 2018-11-13
Data variables:
    complex    (azimuth, range, time) complex64 640MB dask.array<chunksize=(500, 500, 1), meta=np.ndarray>
    amplitude  (azimuth, range, time) float32 320MB dask.array<chunksize=(500, 500, 1), meta=np.ndarray>
    phase      (azimuth, range, time) float32 320MB dask.array<chunksize=(500, 500, 1), meta=np.ndarray>

This is an corregistered Sentinel-1 interferogram stack with 10 interferograms, each having dimensions of 2000 (azimuth) x 4000 (range) pixels. DePSI expects the xr.Dataset to match the following criteria: - It should contain dimensions for azimuth, range, and time (number of interferograms or SLCs). - It should include data variables for complex, amplitude, and phase corresponding to each interferogram or SLC.

These criteria are the bare minimum requirements for DePSI to function correctly. Additional variables are need for certain functionalities in DePSI, such as latitude (lat), longitude (lon), will be needed for some distance-related calculations. The Height-to-phase conversion factor (h2ph) is required for deformation phase to displacement conversion.

Loading a binary interferogram stack with sarxarray

We recommend using the sarxarray package to read binary interferogram stacks into xarray.Dataset. This package provides efficient reading and handling of large SAR datasets using dask for lazy loading. For exmaple, the xr.Dataset above is created using the following code: 

import sarxarray

# Path to the interferogram dataset
path = Path('nl_amsterdam_s1_asc_t088')

# Make a list of interferograms to read
list_ifgs = [p for p in path.rglob('*_cint_srd.raw')]
list_ifgs.sort()

shape = (2000, 4000) # Assume (azimuth, range) shape known beforehand
reading_chunks = (500, 500) # Reading chunks for Dask lazy loading
stack = sarxarray.from_binary(list_ifgs, shape, dtype=np.complex64, chunks=reading_chunks)

The file structure of the dataset is as follows: 

nl_amsterdam_s1_asc_t088/
├── 20180920_cint_srd.raw
├── 20180926_cint_srd.raw
├── 20181002_cint_srd.raw
├── 20181008_cint_srd.raw
├── 20181014_cint_srd.raw
├── 20181020_cint_srd.raw
├── 20181026_cint_srd.raw
├── 20181101_cint_srd.raw
├── 20181107_cint_srd.raw
├── 20181113_cint_srd.raw
where each interferogram binary file is named with its acquisition date followed by _cint_srd.raw. We assumen the metadata such as the shape of the interferograms and data type are known beforehand.