-
-
Notifications
You must be signed in to change notification settings - Fork 6
Update Satellite Zarr reader for new format #313
Comments
Will start on this once the big redesign #307 is merged! |
@peterdudfield @JackKelly For this, what do you guys think about just adding a second Sattelite source to the Batch? Then we give one the HRV Zarr and one the normal Zarr, and nothing else needs changing? |
Whats the advantage of this? There is something nice, about just having it in one, and perhaps the satellite xr.Dataset loading both of them? Perhaps they are singificant different to suggest they should be in a different 'DataSource' |
The advantage is minimal code changes, the HRV DataArray we can't add to the non-HRV DataArray because of the different resolutions. And to cover the same area, the HRV cutout has to be 3x the size of the non-HRV image. Other than that, they would be the same, but the different size/shape of an HRV example vs non-HRV example suggests to me it might make sense to keep them as separate entries? Also, unlike the separate Optical Flow Data source, it wouldn't be loading the same data multiple times, if they are different data sources. Keeping them as the same one would then involve a decent amount of checks for if one, or both of the non-hrv and hrv data was loaded. So makes the code a bit messier. |
Yeah, I'd go with whatever is simplest to implement :) |
Cool, I'll do the HRV channel as a separate data source then! |
Sounds good Although, that said, I do agree with @peterdudfield that it feels kinda weird having two SatelliteDataSources... 🙂 but let's give it a go and see how it feels! |
Detailed Description
Satellite data is now stored in the zarr in unprojected format, so the original geo projection. This does make it slightly more difficult to save the OSGB coordinates in the zarr, as each pixel has a unique OSGB x and y coordinate. Currently, they are stored in the metadata field, under 'osgb_x' and 'osgb_y'. There might be a better way to store it, which we could also do, but for now, we probably need to read in the metadata of one example each for the HRV channel and all the other channels, to get the mapping from image coordinates to OSGB coordinates.
Context
Possible Implementation
The text was updated successfully, but these errors were encountered: