The sky of the last thousand years
What was actually ingested
Real series: Mann et al. 2009 AMO, with an ERSST splice after 2006. That index is not a neutral Atlantic oscillation. Mann’s earlier reconstructions withheld the full archive, and a censored run without the bristlecones was left off the paper. This file keeps a warming trend a real AMO removes, which is the same habit: weigh the series so it sells his history. The audit is on the Mann 2009 page. Also ingested, and not his: Steinhilber TSI, Gray 2004 AMO, SILSO sunspots, Crowley and Unterman 2013 sulfate for the eruption dots, and HadCRUT5, which is not forced into the equation.
Modeled from the paper’s constraints, because the authors did not publish a downloadable annual CSV: western-Mediterranean PDSI, storm rainfall, the temperature tail, and total cloud cover itself. Year-to-year wiggles are a methodological recreation, not a digitised Fig. 6H.
The sliders show how much one of those choices moves the sky. They do not replace the recreation, and they do not turn the modeled years into the authors’ file. Both the real series and the lab series can be downloaded.
First millennial reconstruction of annual total cloud cover in the western Mediterranean, 969–2022 CE. Cloudiness rose into the Little Ice Age, peaked near 1600, then fell — and recent decades sit at the lowest cover in the record, coupled to a warm Atlantic, a bright Sun, and a climate-moisture index that looks very dry. That dryness is not the same thing as the managed landscape: the ranking against the last 500 years compares catchments before and after people re-routed the water. The Roman Warm Period is earlier than the Atlantic index, so it is not given a cloud fraction; solar irradiance for those centuries is on the Before 969 page.
Biases are shown with sliders, not by rewriting the study. The curve as published in this recreation stays on the chart as a dashed line. Each slider turns one methodological choice up or down — a double-counted Atlantic index, a trend left inside the AMO, a hand-drawn millennial shape, an imposed drying after 1970, or the index walked across its own uncertainty band — so you can see how many points of cloud cover that choice is worth. The original series is not replaced, refit, or saved over. Moving a slider only draws the difference.
Dotted terms open a textbook definition and a note on how this lab uses them. Hover, focus, or tap.
Research (Washington, D.C.) 8:0606 · DOI 10.34133/research.0606
Reconstructed cloud fraction
21-year mean with millennial 10th / 90th percentiles (dashed). Dots are Crowley and Unterman 2013 Northern Hemisphere sulfate of at least 12 kg km⁻² — the real ice-core file, not the short list baked into the curve. That volcanic term was not recomputed. Dark line: instrumental-period analog after 1935, itself modeled, not CRU. The sliders move the solid curve by one methodological choice at a time. The dashed gray line is the curve as used.
Each slider is one methodological choice, not a new climate. The dashed line stays put. Moving several at once adds their cloud effects, including the cross-term when a detrended index is also counted only once. Nothing here refits the paper’s coefficients.
1990–2022 mean 36.9% · +0.1 points vs the curve as used
AMO (Mann et al. 2009), ERSST AMO, Steinhilber TSI, Gray AMO and SILSO sunspots are the published observational/proxy series. PDSI, SRI and the TCC reconstruction apply the paper's equation and millennial constraints because the authors' exact predictor files are not bundled here. Treat year-to-year wiggles as a faithful methodological recreation, not a digitised copy of Fig. 6H.
Three climate eras
The millennium is not one story. Clearer medieval skies and a clearer modern sky bookend a cloudy Little Ice Age, but the bookends are not the same climate.
- Medieval Climate Anomaly43.2%
969–1249 · Warm, dry, clearer skies
- Little Ice Age49.1%
1250–1849 · Cool, wet, persistently cloudy
- Modern Warming Era43.4%
1850–2022 · Warming, drying, declining cloud cover
What actually moves the clouds
In the published model, soil moisture (PDSI × storm rainfall) raises cloud fraction; a positive AMO and a warm global temperature tail lower it. The Mann series is not a clean version of that Atlantic term: it keeps late-20th-century warming that a detrended AMO removes, and this atlas also copies it into the moisture and temperature predictors. Use the sliders on the chart to see those choices taken out.
WCI = α · PDSI · (β + SRI^σ)
TCCf = B + A·WCI + γ·AMO + η′·GTA25
α=0.30 · β=2.00 · σ=0.50 · γ=-3.00
The climate engine
Paper Fig. 1. Water is lifted from ocean and land, condensed into cloud, and returned as rain. Atlantic SST gradients steer the storm track; soil moisture decides how much of that water is recycled locally. Fewer clouds lower the albedo and let in more sunlight — a feedback the western Mediterranean is now running in the clear, dry direction.
CRU, Mann AMO, OWDA PDSI, TSI, volcanoes, documentary checks.
The published equation, calibration, and what was fitted.
The paper’s panels, plus the checks it did not show.
Radiation, drought, AMO phase, and what the model cannot say.