Resampling

Resampling in Satpy.

Satpy provides multiple resampling algorithms for resampling geolocated data to uniform projected grids. The easiest way to perform resampling in Satpy is through the Scene object’s resample() method. Additional utility functions are also available to assist in resampling data. Below is more information on resampling with Satpy as well as links to the relevant API documentation for available keyword arguments.

Resampling algorithms

Available Resampling Algorithms
Resampler	Description	Related
nearest	Nearest Neighbor	`KDTreeResampler`
ewa	Elliptical Weighted Averaging	`DaskEWAResampler`
ewa_legacy	Elliptical Weighted Averaging (Legacy)	`LegacyDaskEWAResampler`
native	Native	`NativeResampler`
bilinear	Bilinear	`BilinearResampler`
bucket_avg	Average Bucket Resampling	`BucketAvg`
bucket_sum	Sum Bucket Resampling	`BucketSum`
bucket_count	Count Bucket Resampling	`BucketCount`
bucket_fraction	Fraction Bucket Resampling	`BucketFraction`
gradient_search	Gradient Search Resampling	`GradientSearchResampler`

The resampling algorithm used can be specified with the resampler keyword argument and defaults to nearest:

>>> scn = Scene(...)
>>> euro_scn = scn.resample('euro4', resampler='nearest')

Warning

Some resampling algorithms expect certain forms of data. For example, the EWA resampling expects polar-orbiting swath data and prefers if the data can be broken in to “scan lines”. See the API documentation for a specific algorithm for more information.

Resampling for comparison and composites

While all the resamplers can be used to put datasets of different resolutions on to a common area, the ‘native’ resampler is designed to match datasets to one resolution in the dataset’s original projection. This is extremely useful when generating composites between bands of different resolutions.

>>> new_scn = scn.resample(resampler='native')

By default this resamples to the highest resolution area (smallest footprint per pixel) shared between the loaded datasets. You can easily specify the lowest resolution area:

>>> new_scn = scn.resample(scn.coarsest_area(), resampler='native')

Providing an area that is neither the minimum or maximum resolution area may work, but behavior is currently undefined.

Caching for geostationary data

Satpy will do its best to reuse calculations performed to resample datasets, but it can only do this for the current processing and will lose this information when the process/script ends. Some resampling algorithms, like nearest and bilinear, can benefit by caching intermediate data on disk in the directory specified by cache_dir and using it next time. This is most beneficial with geostationary satellite data where the locations of the source data and the target pixels don’t change over time.

>>> new_scn = scn.resample('euro4', cache_dir='/path/to/cache_dir')

See the documentation for specific algorithms to see availability and limitations of caching for that algorithm.

Create custom area definition

See pyresample.geometry.AreaDefinition for information on creating areas that can be passed to the resample method:

>>> from pyresample.geometry import AreaDefinition
>>> my_area = AreaDefinition(...)
>>> local_scene = scn.resample(my_area)

Create dynamic area definition

See pyresample.geometry.DynamicAreaDefinition for more information.

Examples coming soon…

Store area definitions

Area definitions can be saved to a custom YAML file (see pyresample’s writing to disk) and loaded using pyresample’s utility methods (pyresample’s loading from disk):

>>> from pyresample import load_area
>>> my_area = load_area('my_areas.yaml', 'my_area')

Or using satpy.resample.get_area_def(), which will search through all areas.yaml files in your SATPY_CONFIG_PATH:

>>> from satpy.resample import get_area_def
>>> area_eurol = get_area_def("eurol")

For examples of area definitions, see the file etc/areas.yaml that is included with Satpy and where all the area definitions shipped with Satpy are defined.