High resolution base map for time series exploration & manual point input

[alex-s-gardner]

I wonder if adding a base map option would be valuable for time series exploration. We developed a serverless (Java Script) Zarr exploration app for time series exploration of a large (>50TB) Zarr archive. The Zarr data is optimally chunked for time series exploration, providing the users with a highly responsive exploration tool. The user is able to select and move points on a map where they want to extract the time series. I believe that this access pattern is common in Earth science and likely other domains as well. For example, if one wants to plot a time series of a given variable for New York it is currently not very easy to do.. or if someone wanted to plot a time series for where they live it would also be challenging. If one could select a location on a base map then such visualizations would become trivial... this functionality would also better support Zarr data that is chunked for time-series analysis that does not load easily load as a map. It would also be use full if the user could manually enter/modify point coordinates for where time series should be extracted.

[TheJeran]

Are you saying that instead of viewing the raw data, they can browse a map like in your example. And it will plot the timeseries of that data?

So like instead of seeing that map, they would see a high resolution map. And when they click it will plot the variable seen in said map?

Also now just realized I never put the timeseries selection option in the 2D map. I will have to do that.

There are plans eventually to allow streaming of maps/data for the 2D view option so that people can overlay their own data and basemaps.

EDIT: and instead of a timeseries of just the available data it's a timeseries of the entire time extent along optimally chunked data?

[alex-s-gardner]

Instead of seeing a map generated from the data the user would see a reference high-resolution base map (e.g. ESRI base map: https://openlayers.org/en/latest/examples/xyz-esri.html). And when the user clicked on the map it will plot the variable in a reference Zarr (which often don't provide detailed spatial context.. e.g. your above figure.. if you want a time series over Lake Erie it would be hard to do).

This approach would prevent a Zarr that is not efficiently chunked for map representation to be loaded into RAM as the user could just access specific time-series without needed to generate a map from the raw data, if that makes sense? (i.e. we chunk our data in units of 10x10x20,000 so would need to load 20,000 times more data than needed for a 2d map)

[TheJeran]

Okay gotcha, I understand. I will add this to feature requests #380.