You are viewing an old version of this page. View the current version.

Compare with Current View Page History

Version 1 Next »

A few tentative observations can be made regarding AIFS and IFS and differences in output:

  • AIFS-Single, AIFS Control (and AIFS-ENS) have distinctly different training methods.
    • AIFS-Single has RMSE training strategy, stochastic physics (SPP) not active.
    • AIFS-Control (and AIFS-ENS) has regression training strategy, SPP not active. AIFS-Control also has perturbations, somewhat analogous to SPP).
  • IFS Control (and IFS ENS) uses SPP.


Generally, using same initial conditions:

  • AIFS-Single is similar to IFS-Control but has a smoother look.
  • AIFS-ENS sample member x and IFS-ENS sample member x are very different (possibly result of no SPP in AIFS).
  • AIFS-Control similar to a sample AIFS-ENS member (more similar at short leads as expected).
  • AIFS-Control similar to a sample IFS ENS member but has more noteworthy peaks (resolution differences).
  • IFS-ENS sample member x has much larger peaks of values than IFS-Control sample member x (strong impact of SPP).


  • No labels