In fact my eyeball would say that a somewhat steeper straight line would be
an even better fit to the data.
Not exact matches
However,
even in the ocean the
fit to the
data is not that
good in many regions — particular the southern oceans and Antarctica, but also in the Northern mid-latitudes.
Craig, I you
even read my posts, you will know that I have said that a linear trend is not the
best fit to the
data.
Truncating down
to 1950 has yet another benefit: it shows that if we ignore the temperature
data beyond 1970 (since we're using 1950 - 1970 temperature
data to avoid end effects) and find the
best fit using only HadCRUT3 up
to 1970, we predict the next four decades of temperature remarkably
well,
even predicting the relatively flat temperature for 2000 - 2010, which the model shows is entirely attributable
to SOL and has nothing
to do with a cessation of long - term global warming.
The
fit is
even better for the latest years (which are not as significant due
to the modified and asymmetric filtering) but for most of the full period the difference between the filtered
data and filtered model oscillates more with a period of about 32 years and full amplitude of about 8 mK.
It would certainly make an interesting study
to determine the fractal dimension of your saw and then «zoom out»
to the next level and re-apply your method — it seems
to me that this may
even provide a
better fit to the available
data (inc long term proxies) than the single level you have already calculated — certainly the change in the projections would be most interesting!
Can't find a recent item on arctic sea ice but hoped it might be worth pointing out Peng et al Sensitivity Analysis of Arctic Sea Ice Extent Trends and Statistical Projections Using Satellite
Data http://www.mdpi.com/2072-4292/10/2/230/htm «The most persistently probable curve -
fit model from all the methods examined appears
to be Gompertz,
even if it is not the
best of the subset for all analyzed periods.
I'm merely pointing out that the physical model of greenhouse gas induced warming over the last 165 years is an excellent
fit to the
data, one that is
even better when one adds an purely empirical «natural» variation on top of it.
This suggests that, when the adjustments implied by historical changes in measurement techniques and
data sources are properly applied, the past SST record will end up cooler on the whole relative
to current temperatures, and will also provide an
even better fit to climate models.