Extracting GRIB data for 3D Tiles.

I’m trying to visualize cloud data stored in a GRIB-2 file in Cesium. Currently, my solution involves manually extracting the relevant raster bands, using Cesium’s primitive to individually generate tiles for each cloud and then displaying them. While it works, it is simply not performant enough and despite having attempted two optimization passes, I am unable to make it run smoothly.

The Cesium 3D Tiles system seems to be the way to go, but I’m a little lost on how to prepare data for it. I’ve read the specs on GitHub for how the tiles need to be defined, but I was wondering if an automated workflow for extracting raster bands from GRIB-2 files and converting them to a format compatible with Cesium 3D Tiles is already available.

I believe a simple point cloud system would work, but even 3D cubes would suffice. Any advice on how to go about accomplishing this task would be appreciated.

Just to add some more specifics, if I have .XYZ file ready to go - (where X is Latitude, Y is Longitude, Z is the value at the location), is there a generator to convert it to a format that 3D tiles can digest?

Hey Sanjeet,

We have been talking internally about the idea of just tiling files with X/Y/Z. I opened an issue with some more details about this here: https://github.com/AnalyticalGraphicsInc/cesium-ion-community/issues/179

Feel free to chime in there. In the mean time, do you have any examples/screenshots of you expect your tiled GRIB-2 files to look in Cesium? I’d be curious if treating them as point clouds and tiling would be a good solution here. I assume you might need some per-point metadata in order to color certain types of points differently, or perhaps scale them by different amounts, or are they all uniformly the same type?