Data Access Considerations

The best way to access ICESat-2 data depends on your desired use case. Below, we outline some considerations and recommended data access pathways.

Data Product Select an ICESat-2 data product that aligns with your goals. We recommend starting from a higher level data product and working down if necessary.

Data Volume Select a local download computing environment if you have enough memory to store and proccess the data you or using; otherwise select a cloud streaming environment.

Access Method All access methods support programmatic access, but some also provide GUIs. We recommend using a GUI only for testing small amounts of data and programatically accessing data in your workflows.

Data Use Case Choose if you require subsetting (choosing specific regions or variables to download) or interoperability by downloading additional data from different sources using the same access method.

Access Consideration	SlideRule	Earthaccess	icepyx
Data Product Which ICESat-2 data product do you want?	ATL3, 6, 8, 13, and 24 only	Any product available at NASA Earthdata	Any product available at the NSIDC
Cloud Streaming You need compute beyond your local machine	✔️	✔️1	🔶
Local Download You have enough memory to store and process data locally	❌	✔️	✔️
GUI	SlideRule Web Client	NASA Earthdata Search
Subsetting	Spatial, Temporal, Variables	Spatial, Temporal	Spatial, Temporal
Interoperability	GEDI	Any product available at NASA Earthdata

Reccommendations and Future Work

• Ensure earthaccess has sufficient documentation
• Create documentation for h5coro
• Create issue in xarray to automatically remove s3: from url passed to open_dataset
• DataFrames with attributes (and also unit-aware) - maybe this is called a 1D xarray

Footnotes

1. Using xarray's h5coro backend is much faster than the default h5py backend for opening ICESat-2 granules. We recommend the following workflow when using Earthaccess in the cloud. Note this requires h5coro>=0.0.8.

   import earthaccess
   import xarray as xr
   
   auth = earthaccess.login()
   creds = auth.get_s3_credentials(daac='NSIDC')
   results = earthaccess.search_data(short_name=<short_name>, granule_name=<granule_name>)
   url = results[0].data_links(access="direct")[0][3:] # Remove 's3:' from the beginning of the url
   ds = xr.open_dataset(url, engine='h5coro', group=<group>, credentials=creds)
   

   ↩