Here is a link to Nicholas Lewis's paper "Mann's new paper recharacterizing the Atlantic Multidecadal Oscillation", which devastates Michael Mann's paper on the Atlantic Multidecadal Oscillaion.
Michael Mann is famous for his hockey stick climate change work. That work has been shown to be fatally flawed and MM has been shown to have been unethical concerning its flaws.
Here are some excerpts from NL's paper.
--------------------------------------------------------
The conventional view of the AMO NOAA, which provides an AMO index, has a helpful FAQ on the AMO that says:The AMO is an ongoing series of long-duration changes in the sea surface temperature of the North Atlantic Ocean, with cool and warm phases that may last for 20-40 years at a time and a difference of about 1°F between extremes. These changes are natural and have been occurring for at least the last 1,000 years… Since the mid1990s we have been in a warm phase. The AMO has affected air temperatures and rainfall over much of the Northern Hemisphere… It alternately obscures and exaggerates the global increase in temperatures due to human-induced global warming.
----------------------
Did aerosols rather than the AMO drive 20th century Atlantic SST variations? In 2012 a team of scientists at the UK Met Office published a paper claiming that anthropogenic aerosol indirect forcing, rather than natural variability, drove much of the 20th century variability in North Atlantic SST attributed to the AMO. This claim was based on simulations using the HadGEM2-ES climate model. However, in 2013 a team of scientists from GFDL and elsewhere published a counter-paper entitled "Have Aerosols Caused the Observed Atlantic Multidecadal Variability?", which showed major discrepancies between the HadGEM2-ES simulations and observations in the North Atlantic.Mann himself had argued that anthropogenic aerosols rather than the AMO drove variability in tropical Atlantic SST in a short 2006 paper, here . However, he accepted therein that his analysis relied upon the AMO having no influence on GMST, and he also used what is arguably a questionable statistical model. AR5 didn't mention this paper when discussing the AMO (in Section 10.3.1.1.3).
Now, however, Mann has returned to this issue, making the extraordinary claim that trends forced by anthropogenic greenhouse gas and sulphate etc. emissions masqueraded as an apparent oscillation, and that, rather than warming the NH:
"The true AMO signal, instead, appears likely to have been in a cooling phase in recent decades, offsetting some of the anthropogenic warming temporarily."
Mann's other claims
The press release for the paper also says:
According to Mann, the problem with the earlier estimates stems from having defined the AMO as the low frequency component that is left after statistically accounting for the long-term temperature trends, referred to as detrending.
Mann and his colleagues took a different approach in defining the AMO... They compared observed temperature variation with a variety of historic model simulations to create a model for internal variability of the AMO that minimizes the influence of external forcing -- including greenhouse gases and aerosols. They call this the differenced-AMO because the internal variability comes from the difference between observations and the models' estimates of the forced component of North Atlantic temperature change.
They also constructed plausible synthetic Northern Hemispheric mean temperature histories against which to test the differenced-AMO approaches. Because the researchers know the true AMO signal for their synthetic data from the beginning, they could demonstrate that the differenced-AMO approach yielded the correct signal. They also tested the detrended-AMO approach and found that it did not come up with the known internal variability.
While the detrended-AMO approach produces a spurious temperature increase in recent decades, the differenced approach instead shows a warm peak in the 1990s and a steady cooling since.
That is certainly a novel approach. By defining the AMO as the part of the smoothed temperature change simulated by the models that is not observed, the problem of models warming far too fast since ~2000 largely disappears. So does the inconvenient possibility that the fast model-simulated warming in the 1980s and 1990s might have only been matched in the real world due to a significant contribution from the AMO. Mann's differenced-AMO is a high-sensitivity climate modeller's dream. If climate models were perfect apart from not simulating the AMO, then the differenced-AMO approach would make sense. But models are by no means perfect – and if they were then they would simulate the AMO.
Mann's differenced-AMO merely reflects, on a smoothed basis, the extent to which the observed NH temperature outpaces climate model simulated NH temperature, going negative when models simulate an unrealistically high temperature rise. It seems likely that it will represent model failings and unrealistic forcings to a greater extent than unforced multidecadal internal climate system variability. The CMIP5 models typically have very high aerosol forcing, and as aerosol forcing grew fast from 1950 to the mid/late 1970s it seems that their high aerosol forcing typically more than compensated for their high transient sensitivity, so that they partially emulated the effects of the AMO downswing.
After defining the differenced-AMO, Mann purports to show – using synthetic temperature histories containing a known AMO signal – that his differenced-AMO approach yields the correct signal, whereas the detrended-AMO approach does not. So how does Mann achieve this impressive feat?
-----------
The flaws in Mann's case
The first part of what Mann writes is obviously true, but are his conclusions warranted? It is true
that the detrended-AMO signals diagnosed from the noisy EBM simulations are indeed largely all
in phase with, and very similar in amplitude to, the detrended-AMO signal diagnosed from
observations (the black line). But their real such relationship is to the detrended-AMO signal
diagnosed from the noise-free EBM simulation. Although that signal is not shown in Mann's
published Figure 3b, it is actually plotted by his code, and is shown by the blue dashed line in
Figure 6b (same as Figure 3 thick blue line). As can be seen both there and in Figure 3 above, the
low-passed detrended-AMO signal diagnosed from observations and the low-passed detrendedAMO signal diagnosed from the noise-free EBM simulation are almost identical, reflecting the
success of Mann's fitting of his EBM simulation to the smoothed observations. Therefore, the
detrended-AMO signals diagnosed from the noisy EBM simulations appear also to be related to the
detrended-AMO signal diagnosed from observations. But that apparent relationship is purely an
artefact of the similarity of the detrended-AMO signal diagnosed from observations and the
detrended-AMO signal diagnosed from the noise-free EBM simulation.
Mann's random red-noise series have low-passed components typically only a quarter as large as the
smoothed signal from applying his detrended-AMO approach to the EBM forced simulation
(compare coloured lines in Fig 6a with blue dashed line in Fig 6b, noting different scales). So it is
unsurprising that one recovers something close to that signal (as in Fig 6b) – and hence close to the
nearly identical detrended-AMO from observed temperatures (black line in Fig 6b) – when applying
the detrended-AMO approach to the EBM forced simulation with the random red-noise added,
whatever realisation of noise is used.
So Figure 6 does not prove Mann's claim. The detrended-AMO signals are in reality largely all in
phase with, and of similar amplitude to, the detrended-AMO signal diagnosed from the noise-free
EBM simulation, not (as Mann claims) with that signal derived from observations. One would
expect to end up with something close to a smoothed version of the signal when adding a noise
component with a small low-frequency amplitude to a signal with a ~4 times larger low-frequency
amplitude and low-pass smoothing their sum, where there are only two cycles of signal in the pass
band.
Figure S7.b3 in Mann's Supplementary Information, reproduced as Figure 7, very much supports
my conclusion. It shows the results when an alternative volcanic forcing series (Crowley) is used.
When that is done, the application of the detrended-AMO approach to Mann's EBM simulation
gives a significantly different signal (blue line – present, but not identified, in Mann's SI graphs)
from when it is applied to the actual temperature observations (black line), and the coloured lines
cluster closely around the blue line rather than the black line.
-----------
Conclusions
I have shown that the evidence Mann claims disproves the detrended-AMO, and supports his
differenced-AMO, is illusory. I have also shown that his code produces different results from those
shown in his accepted paper. I have pointed out that graph lines produced by his code that would
have made it much easier to spot the flaws in Mann's evidence, although appearing in the figures in
his Supplementary Information, were omitted from the figures in his main paper.
A differenced-AMO approach has attractions in principle, but only makes sense if climate models
are near-perfect, which is far from the case. The ease with which a simple EBM model can have its
parameters adjusted to produce a nearly flat differenced-AMO shows the very low number of
degrees of freedom involved, with only two full AMO cycles during the instrumental period. The
very heavy, 50-year low-pass, smoothing applied by Mann arguably exacerbates this problem.
The detrended-AMO approach is not perfect, but the pattern exhibited by NOAA's standard
detrended AMO index based on North Atlantic SST appears to be supported by much more
sophisticated approaches. The stadium wave theory, if it holds up, offers physical insight into the
mechanisms underlying the AMO and may lead to more reliable estimation of its state and
influence on surface temperatures and other climate variables.
No comments:
Post a Comment