"Except global temps have not increased in over 8000 years"
Another Smegma lie. In related "this is way over your head" news,
|
Adjusted Global Temperature Data | Open Mind (wordpress.com)
Posted on February 16, 2024 |
Way back in 2011 I co-authored a paper with Stefan Rahmstorf ( Foster & Rahmstorf 2011, hereafter FR11) in which we adjusted global temperature in order to remove (as best we could) the influence of factors we knew were only temporary, and not man-made. Specifically, these are volcanic eruptions (whose aerosols cool the planet), the El Niño Southern Oscillation (ENSO, which warms the world in its positive el Niño phase and cools in its negative la Niña phase), and solar variations (when the sun gets hotter or colder, so does the Earth). These exogenous factors make global temperature fluctuate, but don’t really get anywhere; removing their influence makes the global warming part clearer.
I’ve updated my method for doing so, and extended the time span it covers, so I’d like to share some of the changes to methodology. But before I do I’ll cut right to the chase: doing so removes a lot of the fluctuation in global temperature, so that yearly averages since 1950, which look like this for five prominent global-temperature data sets:
Advertisement
end up looking like this:
There are three main changes in methodology. First is the use of new data sets. For volcanic aerosols, I acquired the data used for CMIP6 model runs, which is much more up-to-date than what we used in FR11. For ENSO, I’m using the Soutern Oscillation Index (SOI) rather than the Multivariate El niño Index (MEI). For solar irradiance, I’m using sunspot numbers as a proxy.
Second, I’m no longer using lagged regression (i.e. regressing temperature against lagged versions of the data for volcanic, el Niño, and solar influence and choosing the lags by least squares), instead I’m exponentially smoothing the data and regressing against the smoothed versions (choosing the smoothing timescales by least squares).
Third, I no longer model the “global warming” part of the signal with a straight line. That was fine for FR11, because during the time span involved (1979 to 2010) it did closely follow a straight line (very closely, in fact). But the new time span starts in 1950, and since then the global warming part of the signal has been quite a bit more complicated than a straight line; now I’m modeling it as a general function of time g(t) which is estimated by a lowess smooth. Hence the model is now
 ,
where g(t) is a generic function of time (mostly man-made global warming), Vj(t) is volcanic aerosol optical depth exponentially smoothed on a time scale of j months, Sk(t) is the Southern Oscillation Index SOI exponentially smoothed on a time scale of k months, Rl(t) is sunspot number exponentially smoothed on a time scale of l months, and the  are a stationary noise process. The function g(t) is determined by a lowess smooth, the coefficients cj, ck, and cl, and the time scales j, k, and l, are determined by least squares. Combining them creates a generalized additive model.
Advertisement
The first iteration of the process is to fit a smooth curve to the data x(t) to approximate the function g(t). This is subtracted from the data, and the volcanic, El niño, and solar factors (and their smoothing time scales) fit to those residuals by least squares. Subtracting that final-step fit from the original data, leaves adjusted data.
For the second iteration, a smooth curve is fit to the adjusted data from iteration 1, the new smooth is subtracted from the original data, and exogenous factors are fit to that. This is the process of backfitting, meant to compensate for the fact that the original smooth removed part of the exogenous influence even before we fit it by least-squares. The second-iteration fit to the exogenous factors is, therefore, a closer match.
And the process can be repeated until it converges to the best approximation (something called projection pursuit). In this case, I found that five iterations (i.e. four iterations of backfitting) converged with adequate precision.
Let’s look in detail at the adjusted data for that from NASA’s Goddard Institute for Space Studies. Here’s the raw data:
The final model of the influence of exogenous factors came out like this:
Advertisements
REPORT THIS AD
The influence of volcanic eruptions, in brown, shows the cooling from the three large volcanic eruptions during this time, especially the massive explosion of Mt. Pinatubo in the 1990s. Solar variations, in purple, have had only a small influence on Earth’s temperature because the solar variations themselves are small. The ENSO, in red, has sometimes warmed and sometimes cooled. The estimated total influence of these temporary factors is:
When we subtract this from the original data, we get the adjusted data for NASA.
The reduced level of fluctuation is evident, and even more so in the graphs of yearly averages. Here’s the raw data:
and here is the adjusted version:
Perhaps the main result of FR11 was that the adjusted data showed no sign of acceleration or deceleration, contradicting ideas that global warming had stopped or even slowed; the evidence didn’t support that. But the new adjusted data do not contradict the recent acceleration which the raw data suggest, in fact they confirm it, mainly because the uncertainty in trend estimates is so greatly reduced that the rate change (acceleration) is easier to demonstrate statistically.
Yes, it appears that the rate of global warming has increased. |
