### Install sentle Python package

Source: https://github.com/cmosig/sentle/blob/main/README.md

Instructions to install the sentle package using pip or by cloning the GitHub repository for local development. The package is tested with Python 3.12.*.

```Shell
pip install sentle
```

```Shell
git clone git@github.com:cmosig/sentle.git
cd sentle
pip install -e .
```

--------------------------------

### API Reference for sentle.process function

Source: https://github.com/cmosig/sentle/blob/main/README.md

Detailed API documentation for the `sentle.process` function, which is the main entry point for retrieving and processing Sentinel data. It outlines all required parameters, their types, and descriptions.

```APIDOC
sentle.process:
  Description: The package contains only one main function for retrieving and processing Sentinel data.
  Parameters:
    target_crs:
      Type: rasterio.crs.CRS or str
      Description: Specifies the target CRS that all data will be reprojected to. You can provide either a rasterio.crs.CRS object or a string (e.g., "EPSG:32633").
    target_resolution:
      Type: float
      Description: Determines the resolution that all data is reprojected to in the target_crs.
    bound_left:
      Type: float
      Description: Left bound of area that is supposed to be covered. Unit is in target_crs.
    bound_bottom:
      Type: float
      Description: Bottom bound of area that is supposed to be covered. Unit is in target_crs.
    bound_right:
      Type: float
      Description: Right bound of area that is supposed to be covered. Unit is in target_crs.
    bound_top:
      Type: float
      Description: Top bound of area that is supposed to be covered. Unit is in target_crs.
    datetime:
      Type: DatetimeLike
      Description: Specifies time range of data to be downloaded. This is forwarded to the respective STAC interface.
    zarr_store:
      Type: str or zarr.storage.Store
      Description: Path of where to create the zarr storage.
```

--------------------------------

### Notes and Constraints for Parameter Usage

Source: https://github.com/cmosig/sentle/blob/main/README.md

Important considerations and dependencies for various parameters, including conditions for applying snow/cloud masks, handling of empty Sentinel-1 asset lists, and requirements for zarr store chunk sizes.

```APIDOC
- If S2_apply_snow_mask is set to True, S2_mask_snow must also be True.
- If S2_apply_cloud_mask is set to True, S2_cloud_classification must also be True.
- If time_composite_freq is set and neither S2_apply_snow_mask nor S2_apply_cloud_mask is set, a warning will be issued as temporal aggregation may yield useless results for Sentinel-2 data.
- When S1_assets is supplied as an empty list, it will be converted to None, meaning no Sentinel-1 data will be downloaded.
- The zarr_store_chunk_size dictionary must contain the keys 'time', 'y', and 'x'.
- When using cloud or snow masking with temporal composites, the masks will be applied before aggregation.
```

--------------------------------

### Visualize Sentinel data cube using lexcube

Source: https://github.com/cmosig/sentle/blob/main/README.md

Illustrates how to visualize a specific band (e.g., B02) of the loaded Sentinel data cube using the lexcube package. This provides an interactive 3D widget for exploring the data, highlighting features like cloud gaps and seasonal coverage.

```Python
import lexcube
lexcube.Cube3DWidget(da.load().sel(band="B02"), vmin=0, vmax=4000)
```

--------------------------------

### Process Sentinel-1 and Sentinel-2 data into a Zarr cube

Source: https://github.com/cmosig/sentle/blob/main/README.md

Demonstrates how to use the `sentle.process` function to download and process Sentinel-1 and Sentinel-2 data for a specified geographic area and time range. It applies cloud/snow masks, selects assets, and performs temporal compositing, saving the results to a Zarr store. The process is parallelized across multiple workers to handle larger-than-memory cubes.

```Python
from sentle import sentle

sentle.process(
    zarr_store="mycube.zarr",
    target_crs="EPSG:32633",
    bound_left=176000,
    bound_bottom=5660000,
    bound_right=216000,
    bound_top=5700000,
    datetime="2022-06-17/2023-06-17",
    target_resolution=10,
    S2_mask_snow=True,
    S2_cloud_classification=True,
    S2_cloud_classification_device="cuda",
    S1_assets=["vh_asc", "vh_desc", "vv_asc", "vv_desc"],
    S2_apply_snow_mask=True,
    S2_apply_cloud_mask=True,
    time_composite_freq="7d",
    num_workers=10,
)
```

--------------------------------

### Scaling Program with num_workers Parameter

Source: https://github.com/cmosig/sentle/blob/main/README.md

Guidance on how to scale the program by adjusting the `num_workers` parameter when calling `sentle.process`. It also specifies memory requirements per worker based on the default spatial chunk size.

```APIDOC
Increase the number of workers using the `num_workers` parameter when calling `sentle.process`. With default spatial chunk size of 4000, specified by `processing_spatial_chunk_size`, you should plan with 2GiB per worker.
```

--------------------------------

### Optional Parameters for Data Processing

Source: https://github.com/cmosig/sentle/blob/main/README.md

This section details the optional parameters available for configuring data processing operations, including settings for spatial chunking, Sentinel-1 and Sentinel-2 data handling, cloud and snow masking, and storage options.

```APIDOC
Parameter: processing_spatial_chunk_size
  Type: int
  Default: 4000
  Description: Size of spatial chunks across which parallelization is performed in pixels.

Parameter: S1_assets
  Type: list[str]
  Default: ["vh_asc", "vh_desc", "vv_asc", "vv_desc"]
  Description: Specify which bands to download for Sentinel-1. Only "vh_asc", "vh_desc", "vv_asc", "vv_desc" are supported. Empty list will be converted to None (no Sentinel-1 data).

Parameter: S2_mask_snow
  Type: bool
  Default: False
  Description: Whether to create a snow mask. Based on https://doi.org/10.1016/j.rse.2011.10.028.

Parameter: S2_cloud_classification
  Type: bool
  Default: False
  Description: Whether to create cloud classification layer, where 0=clear sky, 2=thick cloud, 3=thin cloud, 4=shadow.

Parameter: S2_cloud_classification_device
  Type: str
  Default: "cpu"
  Description: On which device to run cloud classification. Either "cpu" or "cuda".

Parameter: S2_return_cloud_probabilities
  Type: bool
  Default: False
  Description: Whether to return raw cloud probabilities which were used to determine the cloud classes.

Parameter: num_workers
  Type: int
  Default: 1
  Description: Number of cores to scale computation across. Plan 2GiB of RAM per worker. -1 uses all available cores.

Parameter: time_composite_freq
  Type: str
  Default: None
  Description: Rounding interval across which data is averaged.

Parameter: S2_apply_snow_mask
  Type: bool
  Default: False
  Description: Whether to replace snow with NaN.

Parameter: S2_apply_cloud_mask
  Type: bool
  Default: False
  Description: Whether to replace anything that is not clear sky with NaN.

Parameter: overwrite
  Type: bool
  Default: False
  Description: Whether to overwrite existing zarr storage.

Parameter: zarr_store_chunk_size
  Type: dict
  Default: {"time": 50, "x": 100, "y": 100}
  Description: Chunk sizes for zarr storage. Must contain the keys 'time', 'y', and 'x'. Controls the size of data chunks for efficient storage and retrieval.
```

--------------------------------

### Load processed Sentinel data from Zarr with xarray

Source: https://github.com/cmosig/sentle/blob/main/README.md

Shows how to load the processed Sentinel data cube from a Zarr store into an xarray DataArray. This allows for easy access and manipulation of the multi-dimensional data.

```Python
import xarray as xr
da = xr.open_zarr("mycube.zarr").sentle
da
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.