Figures
Recreated panels from Diodato, Seftigen & Bellocchi (2025), then the graphs a millennial climate paper of this kind almost always needs and this one mostly skipped: attribution, running correlations, spectra, volcanic superposed epochs, and a radiation translation.
From the paper
Climate engine
Schematic of the water cycle, Atlantic forcing, soil moisture and cloud radiative effects that the reconstruction compresses into Eq. 2.
Western Mediterranean domain
CM-SAF-style geography: clearer south, cloudier north. Toggle the era on the Methods page for the millennial shift of the same pattern.
Calibration
Scatter of modeled versus analog observed TCC (1935–1980) and the residual histogram. The original also draws 90% and 95% prediction bands.
Spatial correlation of predictors (schematic)
Summer storms build cloud. Strongest where Atlantic fronts still reach the peninsula.
Wet soils recycle vapour. The land memory is why a dry year stays clearer.
Warm Atlantic, fewer western-Med clouds — if the index is a real AMO. Mann’s late-century values keep global warming a detrended AMO removes, so this sign is partly a definition. The atlas buttons show the size of that.
Paper panels: (A) ERA5 SRI vs CRU TCC — positive, stormiest in the north-west; (B) summer scPDSI vs TCC — positive over land; (C) Mann AMO vs TCC — negative across the basin. Signs, not the native grids, are what the model uses.
Millennial forcing and reconstruction
Paper: (A–C) water-balance maps for MCA, LIA, MWE from Cook PDSI; (D) solar; (E–F) AMO and temperature; (G) Crowley volcanoes; (H) reconstructed TCCfG with MCA/LIA/MWE 10th and 90th percentile thresholds, observed TCC 1935–2022, and eruptions with sulfate > 12 kg km⁻².
The maps in A–C are 95th-percentile PDSI fields, i.e. the wet extremes, not the mean drought atlas. They are a climate moisture balance. They are not a map of rivers after dams, inter-basin transfers, and groundwater pumping rearranged where the water sits.
Predictor–cloud scatter, 1935–1980
(A) WCI vs TCC, two points outside 95%. (B) AMO vs TCC, four points outside 95%. ANOVA p < 0.05 in the paper for the AMO relationship.
Graphs the paper needed
None of these contradict the authors; they make the same reconstruction speak more clearly, and they flag where the original analysis is thin.
Standardised attribution
Stacked contributions of WCI, AMO, GTA and volcanic seeding after z-scoring on 1935–1980. Companion-paper weights put AMO and PDSI nearly equal and ahead of temperature. The late-20th-century drop is AMO+ plus GTA+, with WCI (drying) in the same direction.
31-year running correlations
Stationarity test the paper does not show. PDSI×TCC stays positive. AMO×TCC stays negative but is not constant — weaker in parts of the LIA, stronger in the industrial era. A single millennial coefficient is an average, not a law.
Low-frequency spectrum
Periodogram of the 21-yr TCC. Power at multi-decadal (AMO-like 50–80 yr) and near-centennial bands. The paper’s Eddy-cycle (~1000 yr) claim is a two-point MCA–MWE rhyme more than a resolved spectral peak in a 1054-year series.
Interannual spectrum
After removing the 21-yr mean: weather-scale power, no sharp clock. Summer storms are noisy. Do not overfit a 11-year solar cycle in annual western-Med cloud.
Volcanic superposed epoch
Mean TCC anomaly for eruptions ≥ 12 kg km⁻², lags −5 to +8 years. A short-lived cloud increase is the aerosol-seeding / cooling signature. Tambora is special because it sits at the hinge of a longer AMO+/solar+ decline, not because one eruption rewrote the millennium.
Era distributions
Median and IQR of annual TCC. LIA is shifted up; MWE has the long clear-sky tail that the 10th-percentile line in Fig. 6H is catching.
Surface shortwave anomaly
Order-of-magnitude translation: 0.01 of TCC ≈ 0.8 W m⁻² extra (or missing) downwelling SW at the surface, mid-latitude annual. The modern clear-sky trend is a several W m⁻² brightening — comparable to a non-trivial slice of greenhouse forcing, and a reason the basin is a climate hotspot.
Not a full radiative-kernel calculation. Sign and scale are the point.
Change points
Mean TCC in a window before versus after each hinge the paper names.
| Year | Hinge | Before | After | Δ |
|---|---|---|---|---|
| 1250 | MCA → LIA | 43.8% | 46.5% | +0.027 |
| 1600 | LIA cloud peak | 50.5% | 52.5% | +0.020 |
| 1815 | Tambora | 48.8% | 49.9% | +0.011 |
| 1850 | LIA → MWE | 49.2% | 46.7% | -0.025 |
| 1975 | Late-20th-century drop | 42.5% | 38.4% | -0.041 |