Sentinel-1¶

Overview¶

Sentinel-1 is a C-band synthetic aperture radar (SAR) mission supporting Copernicus land and marine monitoring services. It provides all-weather, day-and-night radar imagery with a spatial resolution ranging from 5m to 40m and a revisit time of 6 to 12 days, depending on the location and acquisition mode.

In GeeFetch, Sentinel-1 data is accessed via the Google Earth Engine collection COPERNICUS/S1_GRD (Ground Range Detected).

Processing¶

Filtering by date range and area of interest
Filtering by orbit availability (see configuration options)
Resampling to target resolution
Mosaicking of overlapping acquisitions
Scaling to maximize precision within the requested data type. Pixels outside of the range \((-30, 0)\) saturate. For instance, if the requested datatype is uint8, the image is scaled by \(x \mapsto (x + 30) \cdot 255/30\).

Available Bands for download¶

Band	Description	Native resolution	Download by default
VV	Vertical transmit, vertical receive (dB)	10m	yes
VH	Vertical transmit, horizontal receive (dB)	10m	yes
HH	Horizontal transmit, horizontal receive (dB)	10m	no
HV	Horizontal transmit, vertical receive (dB)	10m	no

Refer to COPERNICUS/S1_GRD for more details.

Configuration Options¶

See common configuration options for non Sentinel-1-specific configuration. Additionally, you may provide the following options.

Attributes:

Name	Type	Description
`orbit`	`S1Orbit`	Orbit direction to filter Sentinel-1 acquisitions. Can be ASCENDING, DESCENDING, BOTH, or AS_BANDS to download ascending and descending composites as separate bands. Defaults to BOTH.
`speckle_filter`	`SpeckleFilterConfig \| Literal['default'] \| None`	Configuration dataclass for speckle filtering, or None for no speckle filtering. Can also be "default" to use baseline speckle filtering parameters. Defaults to None.
`terrain_normalization`	`TerrainNormalizationConfig \| Literal['default'] \| None`	Configuration dataclass for terrain normalization, or None for no terrain normalization. Defaults to "default" which uses baseline terrain normalization parameters.

Example Usage¶

Command Line¶

Write the following config.yaml

data_dir: ~/satellite_data
satellite_default:
  aoi:
    spatial:
      left: 2.2
      bottom: 48.7
      right: 2.5
      top: 49
      epsg: 4326
    temporal:
      start_date: "2023-06-01"
      end_date: "2023-06-30"
  gee:
    ee_project_ids: ["your-gee-id"]
  tile_size: 2000
  resolution: 10
s1:
  orbit: ASCENDING
  speckle_filter:
    framework: "MONO"
    filter_name: "BOXCAR"
    kernel_size: 3
    nr_of_images: 10
  terrain_normalization:
    flattening_model: "DIRECT"
    layover_shadow_buffer: 0
    dem: "USGS/SRTMGL1_003"

then download with

geefetch s1 -c config.yaml

Terrain Normalization¶

S1 images can be preprocessed with terrain flattening algorithms. See the table below for the possible configurations.

Parameter	Type	Accepted Values	Description
dem	string	The GEE snippet of any DEM dataset	Digital elevation Model used for terrain corrections.
flattening_model	string	'VOLUME', 'DIRECT'	The flattening model to be used.
layover_shadow_buffer	integer	\(i\in\mathbb{N}\)	Layover and shadow buffer distance.

By default, terrain normalization will be implemented with the following configuration:

terrain_normalization:
  model: "VOLUME"
  layover_shadow_buffer: 3
  dem: "USGS/SRTMGL1_003"

which is equivalent to passing terrain_normalization: default.

If you do not want any terrain normalization done you can explicitly deactivate it by setting the following:

s1:
  terrain_normalization: null

More detail on the functions and conventions here.

Speckle Filters¶

S1 images can be preprocessed with speckle filtering. See the table bellow for the possible configurations.

Parameter	Type	Accepted Values	Description
framework	string	'MONO', 'MULTI'	Whether the speckle filter will be applied on a single image or to a temporal stack of images neighboring each image in the filtered collection
nr_of_images	integer	\(i\in\mathbb{N}\)	The number of images to be used by the multi temporal speckle filter framework
filter_name	string	'BOXCAR', 'LEE', 'REFINED LEE', 'LEE SIGMA', 'GAMMA MAP'	The name of the speckle filter to use
kernel_size	integer	\(i\in\mathbb{N}\), \(i\) odd	Size of the kernel that will be used to convolve the images.

By default, speckle filtering isn't applied. You can get baseline speckle filtering by passing speckle_filter: default, which is equivalent to

s1:
  speckle_filter:
    framework: MONO
    filter_name: BOXCAR
    kernel_size: 3
    nr_of_images: 10

More details on the filters and conventions here.

Python API¶

Though this is not geefetch main intended use, you can bypass the configuration and directly download Sentinel-1 data with the geefetch.data.get.download_s1 function. For instance, the CLI command above is roughly equivalent to

import logging
from pathlib import Path

from geobbox import GeoBoundingBox

from geefetch.data.get import download_s1
from geefetch.utils.enums import S1Orbit
from geefetch.utils.gee import auth
from geefetch.utils.log import setup

setup(level=logging.INFO)

data_dir = Path("geefetch_data/")
data_dir.mkdir(exist_ok=True)

auth("your-gee-id")

download_s1(
    data_dir,
    bbox=GeoBoundingBox(2.2, 48.7, 2.5, 49),
    start_date="2023-06-01",
    end_date="2023-06-30",
    tile_shape=2000,
    orbit=S1Orbit.ASCENDING
)

See the API reference of geefetch.data.get.download_s1 for more details.

Relevant library code¶

S1
geefetch.data.get.download_s1