I've looked at some of the data, and its very true that it is far to
short at time base to model the remote past (thousands of years). It is
even more sloppy to pick an exponential model for the data (since a
linear works much better). I recognize that the exponential model may be
used because of an a priori "feel" for the right form that the data
"must" have. (I've done this myself when looking at chemical reaction
rates as a function of temperature in order to estimate the activation
energy - but never to extrapolate beyond the range of measurements).
If I remember right, the spreading sea floor offers a series of hardened
lava stripes that record the magnetic field orientation and strength over
time. They show not one, but multiple reversals.
I believe that the sun offers and interesting situation of rapidly
reversing magnetic field: in each turn of the sun spot cycle, the spots
are reversed in polarity (this occurs every 11 years). Could someone
help me with this ?
Naval Surface Warfare Center, Phila