Mann et al. 2009, weighed
The western-Mediterranean cloud equation treats one Atlantic series as a cause. That series is the KNMI file Mann_AMO, an AMO reconstruction, cited as Mann, Zhang, Rutherford, Bradley, Hughes, Shindell, Ammann, Faluvegi & Ni, Science 326, 1256–1260 (2009). This page is an audit of that file and of every place this atlas lets it move the sky.
Not a neutral instrument
Mann has a record of not showing the work. The 1998 and 1999 temperature reconstructions did not ship with a complete archive. Getting the data and the code took years of requests. The FTP site later turned up a directory named CENSORED: a run with the bristlecone pines left out. Without those trees the blade shrank. That test was not in the paper. The 2006 National Research Council review said those strip-bark series should not be used as temperature proxies and that the early centuries were far less certain than the picture the public was sold. McIntyre and McKitrick showed the principal-component step, centered on the twentieth century, would mine the same shape out of noise.
The weighing is the method. A small set of proxies that already rose in the twentieth century was allowed to dominate the average, and the result was presented as the climate history. The file used here does the same kind of thing under another name. It is called an Atlantic Multidecadal Oscillation. From 1850 it is principal-component-filtered instrumental data that keeps the global warming trend. A standard AMO subtracts that trend. On the overlap with detrended ERSST the Mann series sits 0.34 °C warmer. The archived 2σ band includes zero in 88% of years, so most of the “oscillation” is not resolved. The name sells a cycle. The numbers sell a modern warm spike.
That is not how an unbiased series is built. It is how an author keeps the version of the history that his public argument needs. Mann has not acted as a neutral archivist. He has been a climate advocate, and he has used journals, institutions, and defamation suits against people who called the result fraudulent. A jury can punish the word. It cannot turn this index into the oscillation the label claims. The appearance is political activism choosing the weights, not a disinterested measurement that happened to fit the slogan.
This lab does not need a motive to measure the effect. The sliders take his definition off the cloud curve and leave the original line in place. What they cannot do is launder the series into an instrument it is not.
See it on the cloud curve
The sliders are the same control as on the atlas and methods pages. They are not a fraud switch. One removes a double count this atlas introduced. One walks the index from Mann’s trend-keeping values to a detrended AMO. The others turn off a hand-drawn shape, the imposed post-1970 drying, or slide the index inside its own uncertainty band. The dashed line stays so the size of each choice is visible.
1990–2022 mean 36.9% · +0.1 points vs the curve as used
of years, 500–2006. The sign of the index is unresolved in a typical year.
in 1,507 years. Lag-1 autocorrelation is 0.986. Annual correlations are not 1,507 degrees of freedom.
Annual, 424 years. Decadal r = 0.50. Same ocean, only a partial match.
Added to detrended ERSST after 2006 so it meets Mann’s warmer 1980–2006 mean.
The series this atlas actually ingested
Decadal means of Mann_AMO, with the archived 2σ envelope. The ribbon is the width shipped with the file, not a bound we invented. A reconstruction whose uncertainty includes zero almost every year cannot, by itself, time Mediterranean cloudiness to the decade.
KNMI Climate Explorer series Mann_AMO (amoall.txt), the file this atlas ingests. KNMI labels 1850–2006 as PC-filtered instrumental observations, not proxies. The parent spatial reconstruction is Mann et al. 2008 (PNAS), which the 2009 Science paper maps onto MCA and LIA dynamics.
Mean of the archived index · sd 0.14 · n = 281
Mean of the archived index · sd 0.15 · n = 451
Mean of the archived index · sd 0.12 · n = 61
Mean of the archived index · sd 0.21 · n = 157
Mean of the archived index · sd 0.23 · n = 27
Mann against an independent tree-ring AMO and against the detrended instrumental index
Each series is a decadal z-score on its own overlap, so a unit difference is one standard deviation, not one degree. Gray et al. 2004 shares some multidecadal phase and misses the centennial swing. ERSST, with the global-mean regression removed, does not follow Mann’s late-20th-century rise.
Mann vs ERSST, 1880–2006, annual r = 0.36 (n = 127). In 1980–2006 Mann averages 0.23 °C and the detrended ERSST AMO averages -0.11 °C.
What the weighting actually does
The usual AMO (van Oldenborgh; Enfield) is North Atlantic SST after the global mean has been removed, so a globally warm decade is not automatically a positive AMO. Mann’s 1980–2006 mean sits 0.34 °C warmer than that definition. This atlas then adds that offset to ERSST from 2007 onward. The modern “positive AMO, clear sky” tail is partly an index choice: warmth that other AMO definitions assign to the forced trend is left inside the predictor the cloud equation treats as Atlantic variability.
The series is not annual
Year-to-year changes have a standard deviation of 0.0134 °C before 1000 and 0.0313 °C after 1800, against a full-series standard deviation of 0.174 °C. KNMI marks 1850–2006 as PC-filtered instrumental data. Early centuries are smoother still. A regression that counts every year from 969 to 2006 is counting the same swing many times. The median archived half-width is 0.274 °C; in 88% of years the value is smaller than that half-width.
Inside this atlas the same index moves the sky three ways
Cloud-fraction contribution, in percentage points, from turning each Mann pathway off. Direct: the published γ · AMO term. Hydro: the −1.15 × AMO term written into the synthetic western-Mediterranean PDSI, which then enters the water-cycle index. Temperature: the 0.35 × AMO term written into the synthetic global temperature, which then enters GTA₂₅. These traces stop before the hand-placed centennial spline.
The three pathways nearly add. From the wet LIA core (1550–1710) to 1990–2022 they make the sky clearer by about 3.5 percentage points. They are not three independent lines of evidence.
Share of the reconstructed clearing
The atlas’s own LIA mean is 49.1% and the 1990–2022 mean is 36.8%, a drop of 12.4 percentage points. The three Mann pathways account for about 3.5 points of that drop — roughly 28%. The rest is the prescribed dry-MCA / wet-LIA / post-1970 soil-moisture envelope, the anthropogenic temperature ramp, volcanic seeding, and a centennial spline of about ±1.5 points that was added so the curve would match the paper’s shape. Mann’s index is a real lever. It is not the whole result, and it is not an independent one.
What is a bias, and what is not a finding
Shown in this file
Retained late-20th-century warming relative to a detrended AMO. Heavy smoothing sold as an annual index. An uncertainty band that includes zero in most years. Moderate agreement with Gray’s tree-ring AMO. In this atlas, the same series is also baked into soil moisture and temperature, so a regression of cloud on “AMO, PDSI and GTA” counts it more than once.
Disputed in the parent reconstruction
Mann et al. 2008 (PNAS) is the spatial engine; the 2009 Science paper is the dynamical reading KNMI cites for this index. McIntyre & McKitrick’s 2009 PNAS comment flagged calibration-period screening that can raise hockey-stick shapes from red noise, CPS and EIV intervals that exclude each other around 750–1100, and the Tiljander sediments used upside down despite agricultural contamination. Mann and colleagues replied that the hemispheric result held when those series were withheld. This AMO file does not publish the loadings, so Tiljander’s weight on the Atlantic index specifically is not measurable here.
Not established
The archived numbers are a public series, not a series we caught being invented. The 1998–99 short-centered principal components — the procedure the Wegman report and the National Research Council both faulted — are a different method in a different paper. They are not the algorithm in amoall.txt. Misconduct inquiries did not establish that these values were fabricated. “Biased weighting” fits the definition mismatch and the screening dispute. “Fraud” does not follow from what is in this file.
The later paper that cuts against this use
Mann, Steinman & Miller (2014) argued that the apparent AMO is largely a response to volcanic and aerosol forcing, not a free internal oscillation. Diodato and colleagues, and this atlas, put that index in the same equation as volcanic sulfate and global temperature. If Mann’s later reading is right, the 2009 index is partly those other terms under another name. Either way the cloud model is not identifying an Atlantic cause that is cleanly separate from the forcing.