Pangeo climpred presentation
- 1. Riley X. Brady (University of Colorado Boulder) & Aaron Spring (MPI) http://climpred.readthedocs.io/ riley.brady@colorado.edu rileyxbrady
- 2. Decadal climate prediction lies at the intersection of an initial conditions problem and an external forcing/boundary condition problem. IPCC AR5 Chapter 11
- 3. Earth System Model (ESM) predictions use the same principles as numerical weather prediction: initialize an ESM with a reanalysis or model solution, and then integrate it forward. IPCC AR5 Chapter 11
- 4. This experimental configuration leads to high-dimensional output, but we can anticipate a few in particular.
- 5. This experimental configuration leads to high-dimensional output, but we can anticipate a few in particular. ~200 billion data points for 3D ocean output. >1TB at single precision for one variable
- 6. This experimental configuration leads to high-dimensional output, but we can anticipate a few in particular. anticipate dimension names: handle large datasets:
- 9. Future Direction Support for sub-annual predictions (subseasonal to decadal) xarray accessors Advanced predictability analysis
- 10. Give climpred a try! Plenty of tutorial/sample data to work with Looking for users, fresh ideas, contributions http://climpred.readthedocs.io/ riley.brady@colorado.edu rileyxbrady
Editor's Notes
- #3: Mention initialization vs. external forcing and how we need to get both right.
- #4: Mention that this is hindcast prediction. Perfect model prediction initializes off a control run to keep the prediction study self-contained and purely theoretical. We also are dependent on external forcing. You see the increase in SST.
- #7: Can nod to other existing packages