- All mean member combinations are compared with measurements
to find the one with the the minimum RMSE
- Objective is to eliminate members that are not much different from
each other and to reduce ensemble mean bias
- In forecast applications update each analysis cycle with new data
and use the best ensemble for the next forecast
- Without measurement data, reduction techniques can also be used
to compare members to reduce redundancy

hysplit2 |   wildcard base name.??? |
hysplit_datem.006 |   delete 5-member mean from working |


Find Ensembles with Minimum Error

Create 2-member mean (1 and 5)

Edit INFILE


Recompute ensemble statistics



|