ARCO - Evapotranspiration datasets?

[sfoks]

Hi @amsnyder @alamotte-usgs I was wondering if we have any cloud optimized evapotranspiration (AET) datasets on OSN or NHGF s3? The eval team is looking specifically for CONUS-wide coverage, daily to monthly timestep.

We are hoping for: SSEBop daily (1-km MODIS or 30-m LandSat if available), Gleam v3.7, and potentially month water balance model (mwbm) ET. We may modify this list with time but wanted to work with something to get started. Thanks!

[amsnyder]

First check open cloud data catalogs:

• AWS Open Data Catalog
• Microsoft Planetary Computer Data Catalog

The only data that NHGF has processed so far is PRISM.

[amsnyder]

You can also check the GDP to see if you can access it from an OPeNDAP endoint (not ARCO, but another access point):

• slightly more human readable version of GDP catalog: https://labs.waterdata.usgs.gov/gdp_web/
• the THREDDS server where you would actually go for the url to access the data: https://nhgf.wma.chs.usgs.gov/thredds/catalog/catalog.html

If you want to know if a dataset has been turned into a zarr store, @gzt5142 is probably your best point of contact.

[sfoks]

Monthly and annual in the NHGF Catalog & GDP, but daily is not. We can put daily SSEBop MODIS 1-km on OSN and chunk if needed later (if taking a long time to load/work with). Unchunked, it's ~85 gb for 2000-2021.

@rsignell-usgs @gzt5142 can I sit sometime as a fly on the wall again watching "a rechunk in progress" (in case we need to rechunk SSEBop)?

[amsnyder]

Or better yet - another use case for the chunking tutorial...

We probably have an old recording from January when Rich did the rechunking demo if you ever need it.

[rsignell-usgs]

This recent amazon blog post is basically taking the same rechunking approach that we do -- doing the rechunking one time chunk at a time. The only difference is that we use kerchunk instead of open_mfdataset and set up an empty zarr dataset and insert the chunks instead of appending: https://aws.amazon.com/blogs/publicsector/decrease-geospatial-query-latency-minutes-seconds-using-zarr-amazon-s3/