-
Write forecast model script
-
What set of conditioning variables are needed to produce better representations?
- Adding X variables in time
- Adding X variables in space
- Value of main X variable + residuals in time and space
-
What machine learning model produces the best estimates of the full pdf, especially the mean and variability?
-
How should random numbers be initialized and updated to preserve the continuity of the unresolved state?
-
What GAN configuration produces the greatest diversity of samples?
-
Three main experiments
- Weather forecast problem
- Constant climate problem
- Climate change problem
-
Forecast metrics
- RMSE of ensemble mean vs. ensemble spread
- Reliability of different ensemble forecasts
- RPSS
- Extreme scores
- Sharpness
- Hellinger distance (pdf of observed state to forecast state)
-
Climate Metrics
- Quantile-quantile plot
- Autocorrelation function in time for given X
- Spatial autocorrelation averaged over all X locations
- PDFs of X values and comparison with truth PDF
- Hellinger distance on climatological PDF of forecast vs truth