Cary Olsen about Costello:

[Peter Wloch]

Name: Cary
Date: Tue, 23 Jul 2002 02:03:31 -0600
The Costello predictions use a fairly complex neural network technique to predict Kp behavior from solar wind data. The model
has been studied and validated in quite considerable detail. It is in general the best modelled predictor available. But it does have some
weaknesses. For example, the higher end of the geomagnetic spectrum (Kp values > 6) suffer from a lack of samples to train the
network reliably to, while Kp values of between 1 and 4 are plentiful and therefore are modelled more accurately. What we have
observed over the last few days has been a bit odd and Costello hasn't handled it very well. We had to switch over to an alternate
model several times during the last disturbance in order to generate more accurate Kp predictions than what Costello was producing.
But these are exceptions. Usually the model handles conditions rather well.

[Ulrich Rieth]

Der Mann heißt OLER, kommt aus Kanada und ist NICHT Mitglied der Olsen-Bande *o.T.*