STAC Best Practices

1. Do build Collections around distinct sources and methods

  • Do not use spatially or temporally-specific information in Collection name

  • Do use Collection and Item properties to define spatial/temporal information

  • Do place Assets in a new Collection if they come from a different source or use a different method

  • Do thoroughly describe methods or link to description in Collection description.

2. Do use consistent Asset names, particularly across Items in a Collection to create a single layer from tiled data in Navigator

  • Use the same name for Assets throughout multiple Items in a Collection for tiled datasets so that Navigator can mosaic all Assets from a single Collection into a single layer.

3. Do create Items with multiple Assets to show multiple dimensions of a result

  • Do think of this the same way as adding another band to a raster:

    • Does it match in time and space?

    • Does it add related info?

    • If the answer to both is “yes!”, add it. (e.g. asset 1 = softmax, asset 2 = prediction)

  • Do not add assets that were created with various methodologies (e.g. different compositing logic) -> new method, new Collection!

4. Do precisely specify temporal/spatial bounds on Items.

  • Do set start_date and end_date Item properties instead of datetime when applicable.

  • Do not worry about precise temporal/spatial bounds on Collections. It is better to set them “too big” than “too small”, so you can easily add more Items.

5. Do think about how your Collection fits in with other products (if it can be extended, if there’s already a naming convention, …) when setting a name.