Paleoclimate reconstruction lab

The Sheep Mountain dial

Mann, Bradley & Hughes (1998) drew a flat millennium with a sharp twentieth-century blade. McIntyre & McKitrick (2005) showed the blade is what you get when one strip-bark bristlecone site — Sheep Mountain, CA534 — is weighted hundreds of times the rest of the tree-ring network. Drag the bar.

Northern Hemisphere anomaly
Composite-plus-scale, 11-year smooth, baseline 1902–1980
Hard hockey stick2% from Sheep Mountain (CA534)
Your reconstructionHadCRUT5 thermometer1910–1938 and 1976–2026 fitsEqual-weightPublished MBH98

Two thermometer fits, not one century-long smear. 1910–1938 already ran at +1.98. Then the record ducks: 1940–1970 is -0.72. That hole is extra planetary albedo from sulfate lofted into the stratosphere. After the pulse fades, 1976–2026 resumes at +2.95— steep because it starts at the bottom of the hole. The faint line is the 1910–1938 slope continued. Without the volcanic dimming, the century is one moderate, continuous climb, not a pause plus a blade. The stick hides all of this by flattening the shaft and splicing the instrument on at the recovery.

1910–1938
+1.98
°C / century · already climbing
1940–1970 trough
-0.72
°C / century · volcanic albedo dip
1976–2026
+2.95
°C / century · HadCRUT5 through 2025, line to 2026
Eruptions that raised albedo, 1940–1970

Sulfur dioxide oxidizes to sulfate aerosols in the stratosphere. The droplets scatter incoming sunlight — higher albedo, cooler surface — for a year or two per pulse. A cluster of them stacks. Mount Agung (Bali, March 1963, VEI 5) is the climate-scale eruption of the trough: several megatonnes of SO2 and a documented drop through 1964–66. Bezymianny, Taal, Awu, Fernandina and Fuego add further pulses. Together they dent an otherwise continuous 20th-century rise. Take the dent out, and 1976–2026 is not a new physics — it is the 1910–1938 slope recovering from extra reflected sunlight.

  • 1956Bezymianny, KamchatkaVEI 5
  • 1963 · majorAgung, BaliVEI 5
  • 1965Taal, LuzonVEI 4
  • 1966Awu, SulawesiVEI 4
  • 1968Fernandina, GalápagosVEI 4
  • 1974Fuego, GuatemalaVEI 4
Relative weight
1×

Sheep Mountain (CA534) · ITRDB CA534

0×1×50×100×200×390×
Share of the composite
2%
Weight sitting on the target series
Shaft vs blade
-0.34 → -0.14
1400–1850 mean vs 1902–1980 mean, °C
Hockey index
3.26
Blade minus shaft, in shaft standard deviations
Every series, side by side

Apples and oranges

Tree-ring series are standardized growth indexes — ring width, not thermometers. They mix temperature, rainfall, drought, CO2 fertilization, strip-bark growth, and the late-20th-century divergence problem. HadCRUT5 is a mercury-and-satellite temperature field. Plotting them on one axis is the original splice. The cards below keep the units apart: trees as z-scores of growth, the instrument as °C.
HadCRUT5 Northern Hemisphere
Actual instrumental temperature · 1850–2025 · not a tree
°C
140019102000
Sheep Mountain390×
CA534 · Great Basin bristlecone pine · 10001990
growth z
Campito Mountain
CA535 · Great Basin bristlecone pine · 10001979
growth z
Cirque Peak
CA533 · Great Basin bristlecone pine · 10001983
growth z
Flower Lake
CA528 · Foxtail pine · 10001987
growth z
Green Mountain
AZ520 · Ponderosa pine · 14601986
growth z
Hill 10842
NV514 · Great Basin bristlecone pine · 10001985
growth z
Indian Garden
NV519 · Limber pine · 10002000
growth z
Mammoth Creek
UT509 · Great Basin bristlecone pine · 10001989
growth z
Mount Moriah
NV518 · Western juniper · 10001998
growth z
Mount Washington
NV513 · Great Basin bristlecone pine · 10001983
growth z
Pearl Peak
NV510 · Great Basin bristlecone pine · 10001984
growth z
Pearl Peak
NV512 · Great Basin bristlecone pine · 10001985
growth z
San Francisco Peaks
AZ510 · Rocky Mountain bristlecone pine · 10001983
growth z
Spring Mountains
NV511 · Limber pine · 13001980
growth z
Spruce Mountain
NV516 · Great Basin bristlecone pine · 10001984
growth z
Timber Gap Lower
CA530 · Foxtail pine · 10001987
growth z
Timber Gap Upper
CA529 · Foxtail pine · 10001987
growth z
Ward Mountain
NV517 · Great Basin bristlecone pine · 10001984
growth z
Wild Horse Ridge
UT508 · Great Basin bristlecone pine · 10001985
growth z
Chisana Air
AK076 · White spruce · 14151994
growth z
Coastal Alaska
AK096 · Tree ring · 10002002
growth z
Copper River Delta
AK169 · Tree ring · 10002001
growth z
Copper River Delta
AK171 · Tree ring · 10002001
growth z
Cordova Eyak Mountain
AK020 · Tree ring · 13651992
growth z
Exshaw Tunnel Banff
CANA022 · Douglas-fir · 14601965
growth z
Fennoscandia tree rings
FENNO · tree · 10001980
growth z
France tree rings (fran010)
FRANCE · tree · 10001974
growth z
Granite Mountain
WA056 · Tree ring · 12591979
growth z
Herring Alpine
AK001 · Western hemlock · 14221972
growth z
Island Lake
CO511 · Limber pine · 11691989
growth z
Junction of HWYS 51 and 97
OR052 · Ponderosa pine · 14191995
growth z
Kamloops Fraser River Basin
CANA012 · Douglas-fir · 14201965
growth z
Kobuk/Noatak
AK046 · White spruce · 10001992
growth z
Lemon Creek
AK131 · Tree ring · 14502010
growth z
McGinnis Trail
AK102 · Tree ring · 13801999
growth z
Miners Well
AK021 · Tree ring · 14281995
growth z
Mt. Eddy
CA523 · Jeffrey pine · 13881981
growth z
Nabesna Mine
AK074 · White spruce · 14711997
growth z
Naramata Penticton
CANA018 · Ponderosa pine · 14511965
growth z
Nellie Juan
AK120 · Tree ring · 10331999
growth z
Northern Patagonia tree rings
PATAGONIA · tree · 10001983
growth z
Paddy-Go-Easy Pass
WA057 · Tree ring · 14301980
growth z
Pavilion Lake
CANA011 · Douglas-fir · 14601965
growth z
Pleasant Island
AK119 · Tree ring · 13242009
growth z
Polar Urals tree rings
URALS · tree · 10001990
growth z
Powerhouse
CANA020 · Douglas-fir · 14101965
growth z
Prince William Sound
AK094 · Tree ring · 10001991
growth z
Quelccaya ice δ¹⁸O
QUELCCAYA · ice · 10001984
ice
Rock Tor
AK091 · Tree ring · 14062002
growth z
Sandpass
ID010 · Tree ring · 10001991
growth z
Sheenjek River and Flats
AK033 · White spruce · 12961979
growth z
Southern California (Network 41)
CA511 · Tree ring · 14581966
growth z
Spruce Canyon
CO509 · Douglas-fir · 13731978
growth z
St. Anne River
CANA036 · Northern white-cedar · 14041982
growth z
Tasmania tree rings
TASMANIA · tree · 10001989
growth z
Tebenkof Glacier
AK167 · Tree ring · 10001991
growth z
Tebenkof Glacier
AK025 · Tree ring · 14501991
growth z
Uhl Hill
WY001 · Limber pine · 14001971
growth z
Upper Nanatuk
AK093 · Tree ring · 14721999
growth z
Upper Sandpass
ID011 · Tree ring · 12101992
growth z
West Greenland ice δ¹⁸O
GREENLAND · ice · 10001982
ice
Wittier Canal
AK028 · Tree ring · 14121991
growth z
Wolverine Glacier
AK030 · Tree ring · 12471991
growth z
Wynoochee Falls
WA058 · Tree ring · 14171980
growth z
Yakutat Moraines
AK121 · Tree ring · 14031995
growth z
Yolla Bolly
CA555 · Foxtail pine · 11771988
growth z

Each tree card is independently z-scored, so a tall wiggle is that site’s own variance — not degrees Celsius. Bristlecones with a 20th-century growth surge (Sheep Mountain, Hill 10842, Timber Gap) are the series the stick was built from. Most of the others are not hockey sticks.

Overlay desk

Satellites, stations, ocean — pick any mix

Rebased to 1981–2010 so the shapes can sit on one axis. They are not the same thermometer. Toggle whatever you want; nothing here is a tree.

HadCRUT5 NHCRUTEM5 land NHHadSST4 NHUAH TLT NHRSS TLT globe
Homogenized surface
Land stations
Ocean (no UHI)
Satellites (TLT)
Nikolov & Zeller (CERES → GSAT)
Derived
SeriesLayer1979–2024 °C/century2001–2023What it is
HadCRUT5 NH2 m air + SST mix+2.91+2.92NH · land+ocean mix
CRUTEM5 land NH2 m air, weather stations+3.54+3.23NH land · station UHI possible
HadSST4 NHsea-surface temperature+2.06+2.39NH ocean · no urban heat island
UAH TLT NHMSU/AMSU lower troposphere+1.81+1.95NH · no urban heat island
RSS TLT globeMSU/AMSU lower troposphere+2.28+2.2370S–82.5N · no urban heat island
Homogenization is not a thermometer

Ground records are adjusted for time-of-observation, station moves, and instrument changes (CRS to MMTS). Pairwise homogenization can also spread an urban breakpoint into a neighboring rural site. The land series you overlay on satellites have already been through that mill. Ocean and MSU/AMSU have not.

They do not live in the same place

A weather station is a 2-metre box, clustered in the US, Europe, and coasts. RSS TLT is a microwave brightness temperature of the bulk lower troposphere, 70°S–82.5°N, ~100 km footprints. UAH covers the globe but still not the surface. HadSST is buckets, buoys, and ships. Overlaying them is useful. Pretending they are one number is the original splice, again.

Rural is not downtown

Cities run warmer than the countryside, especially at night, and the gap grows with asphalt and population — the urban heat island. Historical rural series were farms, villages, treelines. They are cooler and usually less trended than airport/city records. Ocean has no UHI. Satellites do not sit on a tarmac. CRUTEM (land stations) climbing faster than HadSST (ocean) and UAH (troposphere) is that contrast on one chart. The land–ocean residual is not “rural,” but it is the part of the surface record UHI can leak into.

CERES · Nikolov & Zeller

Albedo dropped. Absorbed sunlight rose.

NASA CERES at the top of the atmosphere, 2000–2023, from the Nikolov & Zeller 2024 supplement. Their GSAT model (Eq. 16) is a toggle on the overlay desk above — albedo plus a sliver of TSI, not a greenhouse coefficient.

CERES ASW W/m²CERES EEI W/m²CERES OLR W/m²Albedo %
CERES / N&Z2001–2023 / centuryRead it as
CERES ASW+7.81 W/m²CERES absorbed shortwave anomaly
CERES EEI+5.10 W/m²CERES Earth energy imbalance anomaly
CERES OLR+3.03 W/m²CERES outgoing longwave anomaly
CERES albedo-2.30 %CERES Bond albedo anomaly
N&Z GSAT+2.41 °CCERES albedo + TSI → GSAT (Eq. 16)
N&Z albedo+2.35 °CCERES Bond albedo → GSAT (Eq. 15a)
N&Z TSI+0.06 °CTSI → GSAT (Eq. 8)
What CERES actually measures

Not 2-metre air temperature. Reflected sunlight, outgoing longwave, and the residual Earth energy imbalance at the top of the atmosphere. Albedo here is Bond albedo: reflected SW / incoming SW. It fell about 0.44 percentage points by 2023 versus the CERES climatology — more sunlight staying in the system, mostly from fewer low clouds over ocean.

What Nikolov & Zeller did with it

Their Eq. 16 maps absorbed solar flux and TSI onto GSAT with a planetary sensitivity of 0.297 K per W/m² absorbed — pressure and sunlight, not a CO2 coefficient. In their supplement, albedo carries almost the whole 2000–2023 rise; TSI is a rounding error. Turn on N&Z GSAT in the overlay above to sit that curve on HadCRUT and UAH. It is their model of CERES, not another thermometer.

Controls

Include ice cores
Quelccaya + West Greenland from the MBH99 archive

Short-centering subtracts the 1902–1980 mean, so any series that rises in the thermometer era is thrown far below zero in the earlier centuries. Principal components then hunt those series.

MBH98 never published that each North American tree-ring series was centered on the 1902–1980 mean, not the full record, before principal components were taken. A chronology that explodes in the calibration era then looks like a huge negative anomaly for five hundred years. PCA rewards that variance. The first North American PC becomes a portrait of the Graybill strip-bark bristlecones.

Sheep Mountain (ITRDB CA534), a Graybill Pinus longaeva site at 3,475 m in California’s White Mountains, was the most heavily loaded series — over 390 times Mayberry Slough, whose hockey-stick index is near zero. Fifteen bristlecone and foxtail sites accounted for 93% of that PC1. Many of those trees are strip-bark; Graybill & Idso (1993) already argued their twentieth-century surge is CO2 fertilization, not temperature.

Glue a thermometer onto that flattened shaft and the 20th century becomes a corner. Fit the instrument in two pieces — 1910–1938 and 1976–present — and the early climb is already the same class of slope as “today.” Between them sits a volcanic-albedo trough (Agung 1963 and a cluster of VEI 4–5 eruptions). Without that dimming the record is one moderate, continuous rise, not a pause plus a blade. And the splice itself is apples to oranges: ring-width growth indexes are not air temperature.

Data, sources, methods

Take the files with you

Everything plotted here is public. Buttons save the lab’s copies. Links go to NOAA, the Met Office, and the papers.

Original archives
Papers
Models in this lab
  • This lab’s model: composite-plus-scale (CPS). Z-score each proxy (short-center 1902–1980 or full period), weighted mean, OLS calibrate to HadCRUT5. Not Mann’s 1998 climate-field reconstruction.
  • OLS thermometer trajectories. Ordinary least squares on HadCRUT5 NH for 1910–1938 and 1976–present. No GCM.
  • No climate model. HadCRUT5 is an instrumental analysis. MBH98’s climate-field reconstruction is not re-run; the dial shows the weighting effect with CPS.

Sheep Mountain original file: ca534-crn-noaa.txt · HadCRUT5 CSV: Met Office · MBH98 NH: mannnhem.dat