Impact and partners
GEMS scientists have teamed with scientists at National Snow and Ice Data Center (NSIDC) to generalize their existing discrete global gridding system, EASE Grid 2.0, to make it more useful for agricultural problems. Our new GEMS Grid has grid sizes similar to existing satellite length scales and focuses on ag-relevant portions of Earth.
Process and problem solved
Agriculture is inherently spatial. To solve analytical problems in ag, researchers spend an inordinate amount of time reprojecting spatial maps and changing resolutions to match other data sets in the computation. The GEMS Grid reduces this effort by projecting all data onto a common grid, and by providing functions to change resolution via the API.
GEMS Services Used
Making spatial ag data accessible
Discrete Global Gridding Systems have a long history in Earth Sciences Observations. But very few of them besides Uber’s H3 grid are widely used. In selecting a grid for Agricultural data, we explored H3 and others. H3 is fit for Uber's purposes, but it did not preserve counts upon successive rounds of aggregation and disaggregation. Similarly other grids we explored failed our necessary criteria (e.g., equal area cells, grid sizes relevant to agriculture). So we worked with NSIDC scientists to adapt their EASE Grid, and created the GEMS Grid – a unique grid designed for agriculture – published and peer reviewed.
The GEMS Grid allows researchers to combine many data streams on the same projection and resolution. So when they are looking at weather and soil variables, pest prevalence, crop production and other agricultural variables within the same spatial cells at the same time range, they know they are comparable.
- The GEMS Grid is a peer-reviewed standard for interoperating spatial ag data.
- Values within common grid cells are de facto comparable.
- The Grid has nested resolutions of 36 km, 9 km, 3 km, 1 km, 100 m, 10 m, and 1 m.
- API functions are provided for grabbing neighbor, parent and child grid cells.
- You can specify how cells should be aggregated to the next level (e.g., min, max, mean, sum).
- Smart disaggregation functions are coming soon.