owned this note
owned this note
Published
Linked with GitHub
# xdggs development meetings
Meeting link: https://numfocus-org.zoom.us/j/81977735338?pwd=pVM3UvnSAJORc2p4Oad39TESPvBzV5.1
## May 6, 2024
### Attendees
- Justus Magin / @keewis
- Daniel Loos / @danlooo
- Wai Tik Chan / @tik65536
- Alexander Kmoch / @allixender
### 60 second updates
- Justus:
- upstream-dev CI: can't build healpy in upstream-dev, thus disabled
- grid info objects: basically done for healpix, h3 to be done, still needs a good test strategy
- looking into nth-order neighbours with healpix
- Alex:
-
### Agenda
- meeting room (host privileges)
- Multi dimensional Index
- e.g. DGGRID PROJTRI (face, x, y) or [H3 IJK](https://h3geo.org/docs/core-library/coordsystems/#ijk-coordinates)
- some DGGS of aperture 4 allow to store the data in just 10 matrices without gaps ![image](https://hackmd.io/_uploads/Hyq1yLLzR.png)
- 2D index suited for 2D convolutions on high res data at global scale [Convolutions at Uber](https://www.youtube.com/watch?v=z3PaGIQTFSE)
- 2D lon/lat Bounding boxes can be descibed easier with a multidimensional index, no need to list ids of all cells individually
- However, we currently use PandasIndex which must be 1D.
- can be replaced with a suitable multi-dimensional index by the DGGS index implementation (DGGS indexes are nested indexes)
- Index choice depends on use case
- 1D Index: small datasets, few points, data frame, GIS
- 2/3D Index: big datasets, all points, raster, convolutions
## April 22, 2024
### Attendees
- Justus Magin / @keewis
- Alexander Kmoch / @allixender
- Wai Tik Chan / @tik65536
- Tina Odaka / @tinaok
### 60-second updates
- Justus:
- presentation about xdggs at the EGU last week
- `decode` function
- Tik:
- DGGRID initial prototype for DGGSIndex:
- ![image](https://hackmd.io/_uploads/HkOXrAQ-A.png)
- ![image](https://hackmd.io/_uploads/r1OIHR7b0.png)
- tina:
- grib2 extension for healpix should be able to recognised by xdggs
- information from Kluft, Lukas(mpg)
```
Section 3 in a GRIB file describes the grid and its defining parameters. The important information for you is the following:
gridDefinitionTemplateNumber = 150 [150 means HEALPix]
Nside = 1024 [The nside parameter defining the resolution of the HEALPix grid]
ordering = 1 [A flag denoting the ordering of the grid, 1 = nested]
```
### Agenda
- workflow:
- separate resources for regridding (out-of-scope for xdggs)
- demonstrate the use with xdggs
- DGGS resolution:
- small number, starting with 0 / 1 (depending on the specific dggs)
- paper: deadline in two weeks
## April 8, 2024
### Attendees
- Justus Magin / @keewis
- Anne Fouilloux / @annefou
- Benoît Bovy / @benbovy
- Alexander Kmoch / @allixender
- Wai Tik Chan /
- Tina Odaka / @tinaok
### 60-second updates
- Justus: current status of xdggs
- index implementations:
- h3: h3ronpy
- healpix: healpy (replace with cdshealpix?)
- initialization, conversion to/from cell centers, selection using cell centers
- PR: [grid objects](https://github.com/xarray-contrib/xdggs/pull/39)
- Alex/Tik;
- DGGRID (ISEA7H), now would start with dggrid4py test plugin
- OGC towards DGG RS registry -> how to metadata define DGGS type and its properties (think proj string or WKT CRS)
- hope to have Python bindings eventually
- saw a guy do a DGGRID WASM, could we have a wasmtime?
### Agenda
- high-level overview of usage workflow
- data production, regridding (out of scope for the library): xesmf, binning, ...
- examples of producing data in a separate repository
- calculations:
- transforms: cell centers, cell outlines / boundaries (coordinates, shapely/spherely polygons), ...
- indexing: selection, neighbor search, ...
- repeating values?
- geometric operations: overlap, intersection, ... (using xvec?)
- hierarchical operations: parent cells, siblings, up / downscaling (binning, interpolation), ...
- storing to disk: conventions?
- → OGC DGG RS?
- custom decoder?