M. Cortini et Cc. Barton, CHAOS IN GEOMAGNETIC REVERSAL RECORDS - A COMPARISON BETWEEN EARTHS MAGNETIC-FIELD DATA AND MODEL DISK DYNAMO DATA, J GEO R-SOL, 99(B9), 1994, pp. 18021-18033
The Earth's geomagnetic field reverses its polarity at irregular time
intervals. However, it is not clear whether a reversal is a determinis
tic (low-dimensional) or a random (high-dimensional) process; the dura
tion-frequency distribution of the polarity time intervals resembles t
hose generated by random processes, but many models suggest that a geo
magnetic field reversal can be the outcome of a deterministic dynamics
, that of the convection in the Earth's outer core. The latter, in tur
n, is only a part of an extremely complex system, made up of both terr
estrial and extraterrestrial subsystems nonlinearly interacting with e
ach other over a wide range of time scales. We studied the geomagnetic
field reversal patterns by means of several techniques of nonlinear d
ynamics and compared the results obtained on actual geomagnetic revers
al data with synthetic reversal sequences generated by the Rikitake an
d Chillingworth-Holmes models of the Earth's magnetic field. We analyz
ed both the geomagnetic and the synthetic reversal scales by nonlinear
forecasting and found that we cannot predict the geomagnetic reversal
sequence with nonlinear forecasting. Predictability of the synthetic
data varies widely depending on the model parameters. Phase portraits
of data obtained from the magnetic field models show fractal structure
s similar to those associated with the Lorenz attractor. We measured t
he correlation dimension Dc of the synthetic and geomagnetic data by m
eans of the Grassberger-Procaccia method and found that Dc always has
a value of about one for the synthetic data. The correlation integrals
for the geomagnetic reversal sequence behave very differently from th
ose of randomized reversal sequences and suggest that the Earth's geom
agnetic field reversal dynamics is not random. However, the limited si
ze of the magnetic reversal data set (282 points) and the poor converg
ence of the correlation integrals make a quantitative assessment of lo
w-dimensional chaos impossible. Our analysis sets a lower limit to the
correlation dimension of the geomagnetic reversal dynamics: D-C > 3.