Home

Awesome

xarray-sentinel

Easily access and explore the SAR data products of the Copernicus Sentinel-1 satellite mission in Python.

This Open Source project is sponsored by B-Open - https://www.bopen.eu.

Features

xarray-sentinel is a Python library and Xarray backend with the following functionalities:

Overall, the software is in the beta phase and the usual caveats apply.

Install

The easiest way to install xarray-sentinel is in a conda environment. The following commands create a new environment, activate it, install the package and its dependencies:

    conda create -n XARRAY-SENTINEL
    conda activate XARRAY-SENTINEL
    conda install -c conda-forge dask "rasterio=>1.3.0" xarray-sentinel

Usage

The SAR data products of the Copernicus Sentinel-1 satellite mission are distributed in the SAFE format, composed of a few raster data files in TIFF and several metadata files in XML. The aim of xarray-sentinel is to provide a developer-friendly Python interface to all data and several metadata elements as Xarray Datasets to enable easy processing of SAR data into value-added products.

Due to the inherent complexity and redundancy of the SAFE format xarray-sentinel maps it to a tree of groups where every group may be opened as a Dataset, but it may also contain subgroups, that are listed in the subgroups attribute.

The following sections show some example of xarray-sentinel usage. In the notebooks folder you can also find notebooks, one for each supported product, that allow you to explore the data in more detail using the xarray-sentinel functions.

The root dataset

For example let's explore the Sentinel-1 SLC Stripmap product in the local folder ./S1A_S3_SLC__1SDV_20210401T152855_20210401T152914_037258_04638E_6001.SAFE. First, we can open the SAR data product by passing the engine="sentinel-1" option to xr.open_dataset and access the root group of the product, also known as /:

>>> import xarray as xr
>>> slc_sm_path = "tests/data/S1A_S3_SLC__1SDV_20210401T152855_20210401T152914_037258_04638E_6001.SAFE"
>>> xr.open_dataset(slc_sm_path, engine="sentinel-1")
<xarray.Dataset>
Dimensions:  ()
Data variables:
    *empty*
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                SM
    swaths:                              ['S3']
    orbit_number:                        37258
    relative_orbit_number:               86
    ...
    start_time:                          2021-04-01T15:28:55.111501
    stop_time:                           2021-04-01T15:29:14.277650
    group:                               /
    subgroups:                           ['S3', 'S3/VH', 'S3/VH/orbit', 'S3/V...
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

The root Dataset does not contain any data variable, but only attributes that provide general information on the product and a description of the tree structure of the data. The group attribute contains the name of the current group and the subgroups attribute shows the names of all available groups below this one.

Measurements datasets

To open the other groups we need to add the keyword group to xr.open_dataset. The measurement can then be read by selecting the desired beam mode and polarization. In this example, the data contains the S3 beam mode and the VH polarization with group="S3/VH" is selected:

>>> slc_s3_vh = xr.open_dataset(slc_sm_path, group="S3/VH", engine="sentinel-1", chunks=2048)
>>> slc_s3_vh
<xarray.Dataset>
Dimensions:           (slant_range_time: 18998, azimuth_time: 36895)
Coordinates:
    pixel             (slant_range_time) int64 ...
    line              (azimuth_time) int64 ...
  * azimuth_time      (azimuth_time) datetime64[ns] ...
  * slant_range_time  (slant_range_time) float64 ...
Data variables:
    measurement       (azimuth_time, slant_range_time) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                SM
    swaths:                              ['S3']
    orbit_number:                        37258
    relative_orbit_number:               86
    ...
    geospatial_lon_min:                  42.772483374347
    geospatial_lon_max:                  43.75770573943618
    group:                               /S3/VH
    subgroups:                           ['orbit', 'attitude', 'azimuth_fm_ra...
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

The measurement variable contains the Single Look Complex measurements as a complex64 and has dimensions slant_range_time and azimuth_time. The azimuth_time is an np.datetime64 coordinate that contains the UTC zero-Doppler time associated with the image line and slant_range_time is an np.float64 coordinate that contains the two-way range time interval in seconds associated with the image pixel.

Since Sentinel-1 IPF version 3.40, a unique identifier for bursts has been added to the SLC product metadata. For these products, the list of the burst ids is stored the burst_ids dataset attribute.

Metadata datasets

The measurement group contains several subgroups with metadata associated with the image. Currently, xarray-sentinel supports the following metadata datasets:

For example, the image calibration metadata associated with the S3/VH image can be read using group="S3/VH/calibration":

>>> slc_s3_vh_calibration = xr.open_dataset(slc_sm_path, group="S3/VH/calibration", engine="sentinel-1")
>>> slc_s3_vh_calibration
<xarray.Dataset>
Dimensions:       (line: 22, pixel: 476)
Coordinates:
  * line          (line) int64 0 1925 3850 5775 7700 ... 34649 36574 38499 40424
  * pixel         (pixel) int64 0 40 80 120 160 ... 18880 18920 18960 18997
Data variables:
    azimuth_time  (line) datetime64[ns] ...
    sigmaNought   (line, pixel) float32 ...
    betaNought    (line, pixel) float32 ...
    gamma         (line, pixel) float32 ...
    dn            (line, pixel) float32 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                SM
    swaths:                              ['S3']
    orbit_number:                        37258
    relative_orbit_number:               86
    ...
    stop_time:                           2021-04-01T15:29:14.277650
    group:                               /S3/VH/calibration
    Conventions:                         CF-1.8
    title:                               Calibration coefficients
    comment:                             The dataset contains calibration inf...
    history:                             created by xarray_sentinel-...

Note that in this case, the dimensions are line and pixel with coordinates corresponding to the sub-grid of the original image where the calibration Look Up Table is defined.

The groups present in a typical Sentinel-1 Stripmap product are:

/
└─ S3
   ├─ VH
   │  ├─ orbit
   │  ├─ attitude
   │  ├─ azimuth_fm_rate
   │  ├─ dc_estimate
   │  ├─ gcp
   │  ├─ coordinate_conversion
   │  ├─ calibration
   │  ├─ noise_range
   │  └─ noise_azimuth
   └─ VV
      ├─ orbit
      ├─ attitude
      ├─ azimuth_fm_rate
      ├─ dc_estimate
      ├─ gcp
      ├─ coordinate_conversion
      ├─ calibration
      ├─ noise_range
      └─ noise_azimuth

Advanced usage

TOPS burst datasets

The IW and EW products, that use the Terrain Observation with Progressive Scan (TOPS) acquisition mode, are more complex because they contain several beam modes in the same SAFE package, but also because the measurement array is a collage of sub-images called bursts.

xarray-sentinel provides a helper function that crops a burst out of a measurement dataset for you.

You need to first open the desired measurement dataset, for example, the HH polarisation of the first IW swath of the S1A_IW_SLC__1SDH_20220414T102209_20220414T102236_042768_051AA4_E677.SAFE product, in the current folder:

>>> slc_iw_v340_path = "tests/data/S1A_IW_SLC__1SDH_20220414T102209_20220414T102236_042768_051AA4_E677.SAFE"
>>> slc_iw1_v340_hh = xr.open_dataset(slc_iw_v340_path, group="IW1/HH", engine="sentinel-1")
>>> slc_iw1_v340_hh
<xarray.Dataset>
Dimensions:           (pixel: 21169, line: 13500)
Coordinates:
  * pixel             (pixel) int64 0 1 2 3 4 ... 21164 21165 21166 21167 21168
  * line              (line) int64 0 1 2 3 4 5 ... 13495 13496 13497 13498 13499
    azimuth_time      (line) datetime64[ns] ...
    slant_range_time  (pixel) float64 ...
Data variables:
    measurement       (line, pixel) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        42768
    relative_orbit_number:               171
    ...
    geospatial_lon_min:                  -61.94949110259839
    geospatial_lon_max:                  -60.24826879672774
    group:                               /IW1/HH
    subgroups:                           ['orbit', 'attitude', 'azimuth_fm_ra...
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

Note that the measurement data for IW and EW acquisition modes can not be indexed by physical coordinates because of the collage nature of the image.

Now the 9th burst out of 9 can be cropped from the swath data using burst_index=8, via:

>>> import xarray_sentinel
>>> xarray_sentinel.crop_burst_dataset(slc_iw1_v340_hh, burst_index=8)
<xarray.Dataset>
Dimensions:           (slant_range_time: 21169, azimuth_time: 1500)
Coordinates:
    pixel             (slant_range_time) int64 0 1 2 3 ... 21166 21167 21168
    line              (azimuth_time) int64 12000 12001 12002 ... 13498 13499
  * azimuth_time      (azimuth_time) datetime64[ns] 2022-04-14T10:22:33.80763...
  * slant_range_time  (slant_range_time) float64 0.005348 0.005349 ... 0.005677
Data variables:
    measurement       (azimuth_time, slant_range_time) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        42768
    relative_orbit_number:               171
    ...
    group:                               /IW1/HH
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...
    azimuth_anx_time:                    2136.774327
    burst_index:                         8
    burst_id:                            365923

If IPF processor version is 3.40 or higher, it is also possible to select the burst to be cropped using the burst_id key:

>>> xarray_sentinel.crop_burst_dataset(slc_iw1_v340_hh, burst_id=365923)
<xarray.Dataset>
Dimensions:           (slant_range_time: 21169, azimuth_time: 1500)
Coordinates:
    pixel             (slant_range_time) int64 0 1 2 3 ... 21166 21167 21168
    line              (azimuth_time) int64 12000 12001 12002 ... 13498 13499
  * azimuth_time      (azimuth_time) datetime64[ns] 2022-04-14T10:22:33.80763...
  * slant_range_time  (slant_range_time) float64 0.005348 0.005349 ... 0.005677
Data variables:
    measurement       (azimuth_time, slant_range_time) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        42768
    relative_orbit_number:               171
    ...
    group:                               /IW1/HH
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...
    azimuth_anx_time:                    2136.774327
    burst_index:                         8
    burst_id:                            365923

Note that the helper function also performs additional changes, such as swapping the dimensions to the physical coordinates and adding burst attributes.

As a quick way to access burst data, you can add the burst_index to the group specification on open, for example, group="IW1/VH/8". The burst groups are not listed in the subgroup attribute because they are not structural.

>>> slc_iw_v330_path = "tests/data/S1B_IW_SLC__1SDV_20210401T052622_20210401T052650_026269_032297_EFA4.SAFE"
>>> xr.open_dataset(slc_iw_v330_path, group="IW1/VH/8", engine="sentinel-1")
<xarray.Dataset>
Dimensions:           (slant_range_time: 21632, azimuth_time: 1501)
Coordinates:
    pixel             (slant_range_time) int64 ...
    line              (azimuth_time) int64 ...
  * azimuth_time      (azimuth_time) datetime64[ns] 2021-04-01T05:26:46.27227...
  * slant_range_time  (slant_range_time) float64 0.005343 0.005343 ... 0.005679
Data variables:
    measurement       (azimuth_time, slant_range_time) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              B
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        26269
    relative_orbit_number:               168
    ...
    geospatial_lon_max:                  12.093126130070317
    group:                               /IW1/VH
    azimuth_anx_time:                    2210.634453
    burst_index:                         8
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

Calibration

xarray-sentinel provides helper functions to calibrate the data using the calibration metadata. You can compute the gamma intensity for part of the Stripmap image above with:

>>> xarray_sentinel.calibrate_intensity(slc_s3_vh.measurement[:2048, :2048], slc_s3_vh_calibration.gamma)
<xarray.DataArray (azimuth_time: 2048, slant_range_time: 2048)>
dask.array<pow, shape=(2048, 2048), dtype=float32, chunksize=(2048, 2048), chunktype=numpy.ndarray>
Coordinates:
    pixel             (slant_range_time) int64 dask.array<chunksize=(2048,), meta=np.ndarray>
    line              (azimuth_time) int64 dask.array<chunksize=(2048,), meta=np.ndarray>
  * azimuth_time      (azimuth_time) datetime64[ns] 2021-04-01T15:28:55.11150...
  * slant_range_time  (slant_range_time) float64 0.005273 0.005273 ... 0.005303
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              A
    mode:                                SM
    swaths:                              ['S3']
    orbit_number:                        37258
    relative_orbit_number:               86
    ...
    geospatial_lat_min:                  -12.17883496921861
    geospatial_lat_max:                  -10.85986742252814
    geospatial_lon_min:                  42.772483374347
    geospatial_lon_max:                  43.75770573943618
    units:                               m2 m-2
    long_name:                           gamma

Advanced data access via fsspec

You need the unreleased rasterio >= 1.3.0 for fsspec to work on measurement data

xarray-sentinel can read data from a variety of data stores including local file systems, network file systems, cloud object stores and compressed file formats, like Zip. This is done by passing fsspec compatible URLs to xr.open_dataset and optionally the storage_options keyword argument.

For example you can open a product directly from a zip file with:

>>> slc_iw_zip_path = "tests/data/S1B_IW_SLC__1SDV_20210401T052622_20210401T052650_026269_032297_EFA4.zip"
>>> xr.open_dataset(f"zip://*/manifest.safe::{slc_iw_zip_path}", group="IW1/VH", engine="sentinel-1")  # doctest: +SKIP
<xarray.Dataset>
Dimensions:           (pixel: 21632, line: 13509)
Coordinates:
  * pixel             (pixel) int64 0 1 2 3 4 ... 21627 21628 21629 21630 21631
  * line              (line) int64 0 1 2 3 4 5 ... 13504 13505 13506 13507 13508
    azimuth_time      (line) datetime64[ns] ...
    slant_range_time  (pixel) float64 ...
Data variables:
    measurement       (line, pixel) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              B
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        26269
    relative_orbit_number:               168
    ...
    number_of_bursts:                    9
    lines_per_burst:                     1501
    group:                               /IW1/VH
    subgroups:                           ['orbit', 'attitude', 'azimuth_fm_ra...
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

As an example of remote access, you can open a product directly from a GitHub repo with:

>>> xr.open_dataset(f"github://bopen:xarray-sentinel@/{slc_iw_path}", group="IW1/VH", engine="sentinel-1")  # doctest: +SKIP
<xarray.Dataset>
Dimensions:           (pixel: 21632, line: 13509)
Coordinates:
  * pixel             (pixel) int64 0 1 2 3 4 ... 21627 21628 21629 21630 21631
  * line              (line) int64 0 1 2 3 4 5 ... 13504 13505 13506 13507 13508
    azimuth_time      (line) datetime64[ns] ...
    slant_range_time  (pixel) float64 ...
Data variables:
    measurement       (line, pixel) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              B
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        26269
    relative_orbit_number:               168
    ...
    number_of_bursts:                    9
    lines_per_burst:                     1501
    group:                               /IW1/VH
    subgroups:                           ['orbit', 'attitude', 'azimuth_fm_ra...
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

fsspec is very powerful and supports caching and chaining, for example you can open a zip file off a GitHub repo and cache the file locally with:

>>> xr.open_dataset(
...     f"zip://*/manifest.safe::simplecache::github://bopen:xarray-sentinel@/{slc_iw_zip_path}",
...     engine="sentinel-1",
...     group="IW1/VH",
...     storage_options={
...         "simplecache": {"cache_storage": "/tmp/zipfiles/"},
...     },
... )  # doctest: +SKIP
<xarray.Dataset>
Dimensions:           (pixel: 21632, line: 13509)
Coordinates:
  * pixel             (pixel) int64 0 1 2 3 4 ... 21627 21628 21629 21630 21631
  * line              (line) int64 0 1 2 3 4 5 ... 13504 13505 13506 13507 13508
    azimuth_time      (line) datetime64[ns] ...
    slant_range_time  (pixel) float64 ...
Data variables:
    measurement       (line, pixel) complex64 ...
Attributes: ...
    family_name:                         SENTINEL-1
    number:                              B
    mode:                                IW
    swaths:                              ['IW1', 'IW2', 'IW3']
    orbit_number:                        26269
    relative_orbit_number:               168
    ...
    number_of_bursts:                    9
    lines_per_burst:                     1501
    group:                               /IW1/VH
    subgroups:                           ['orbit', 'attitude', 'azimuth_fm_ra...
    Conventions:                         CF-1.8
    history:                             created by xarray_sentinel-...

Reference documentation

This is the list of the reference documents:

Design decisions

Project badges

on-push codecov

Contributing

The main repository is hosted on GitHub. Testing, bug reports and contributions are highly welcomed and appreciated:

https://github.com/bopen/xarray-sentinel

Lead developers:

Main contributors:

See also the list of contributors who participated in this project.

Sponsoring

B-Open commits to maintain the project long term and we are happy to accept sponsorships to develop new features.

We wish to express our gratitude to the project sponsors:

License

Copyright 2021-2022, B-Open Solutions srl and the xarray-sentinel authors.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.