My discipline is geology. I have studied all of the natural sciences including meteorology. My method is to study cause and effect. My creed is that nature is to be deciphered if we are to understand her. I see none of that approach in the cogitations of the mathematical thinkers. They do not like to consider natural processes, perhaps because they have no founding. Hence the whole of their science is based on unwarranted assumptions.

]]>They didn’t break the curves into three linear parts. They represented each of T and N (their H) as the sum of three exponetially-relaxing terms, with common time constants for the T and N expressions but differing amplitudes for T and N. Doing so enables remarkably accurte reproduction of the N vs T curves. But it is not really possible accurately to determine the time constant for the fastest response term nor, if you want to estiamte ERF for F2x, can one separately estiamte an amplitude for N for the fastest resonse term. ]]>

Firstly I would think that an objective breakpoint analysis would be required to find the statistically significant breakpoints in the series.

Breakpoints in a series with abrupt changes would typically require showing how these changes could occur and particularly so if a linear relationship between variables is strictly imposed. Alternatively the linear breakpoint segments could be considered a tractable means of working with a non linear series which poses the question of why a non linear relationship is not used and the abrupt changes of breakpoints and required explanation is avoided.

In the SI I get the feeling that the choice of using 3 segments instead of 2 was based on the residuals between the GCM simulations and Bayesian fits for temperature and TOA energy flux, but that criteria does not rule out over fitting of the model.

Another point to consider is that the early time points in the abrupt 4X CO2 experiment are few in number and have a heavy weight on the requirement for a segmentation with 2 or 3 slopes. If those early points have a greater uncertainty there could be some misdirection resulting.

Projecting a line to equilibrium (TOA flux = 0) from a cluster of points at the end of the 4XCO2 experiment could be problematic. When viewing all the CMIP5 model plots in the SI for TOA flux versus temperature the need for line segmentation becomes less apparent.

When I have time I would like to do my own analysis of these plots.

]]>