Discussion and Conclusions

As we have seen in the two previous cases, ensemble forecasts can help in the assessment of large changes from one operational run to the next. Typically, these changes indicate a high degree of uncertainty in the forecast. This is confirmed by ensemble forecasts when we find:

• Large differences among ensemble members within an ensemble run (large ensemble "spread" in ensemble mean/spread diagrams, "spaghetti" diagrams that indeed look like spaghetti),
• A low relative measure of predictability (RMOP), and
• A low probability of the ensemble mean verifying, based on the previous 30 days of ensemble runs.

In such cases, the current operational model run usually falls within the (wide) range of solutions offered in the immediately previous ensemble runs. The disparity between consecutive operational runs therefore reflects forecast uncertainty rather than a radical change in the forecast resulting from new data.

Cases where the ensemble mean forecast shows a significant change should be taken more seriously. Comparisons to the ensemble spread, spaghetti diagrams, and RMOP graphics can help in the assessment of the significance of the change in the forecast.