The strong correlation between magnetic storms and southward interplan
etary magnetic field (IMF) is well known from linear prediction filter
studies using the Dst and IMF data. However, the linear filters chang
e significantly from one storm to another and thus are limited in thei
r predicting ability. Previous studies have indicated nonlinearity in
the magnetospheric response as the ring current decay rate varies with
the Dst value during storms. We present in this letter nonlinear mode
ls for the evolution of the Dst based on the OMNI database for 1964-19
90. When solar wind data are available in advance, the evolution of st
orms can be predicted from the Dst and IMF data. Solar wind data, howe
ver, are not available most of the time or are available typically an
hour or less in advance. Therefore, we have developed nonlinear predic
tive models based on the Dst data alone. In the absence of solar wind
data, these models cannot predict the storm-onset, but can predict the
storm evolution, and may identify intense storms from moderate ones.
The input-output model based on IMF and Dst data, the autonomous model
based on Dst alone, and a combination of the two can be used as forec
asting tools for space. weather.