Reconstructing Northern Hemisphere temperature over the past millennium

1
Reconstructing Northern Hemisphere temperature
over the past millennium

• Peter Huybers
• WHOI/Harvard

2
The hockey stick (IPCC, 2001)
Millennial Northern Hemisphere (NH) temperature
reconstruction (blue) and instrumental data
(red) from AD 1000 to 1999, adapted from Mann et
al. (1999).
3
Media coverage
The hockey stick has often been trotted out as
the clinching proof of anthropogenic warming.
Skeptics have never liked this chart, which
necessarily relies on statistics to infer
historical temperature from tree rings. ...
McIntyre and McKitrick allege that Dr. Manns
statistics are inaccurate, and that the hockey
stick is a mere statistical artifact.
-The Economist, Feb, 2005.
4
Proxy cross-correlation with instrumental
temperatures
5
Leading principal component using different
normalizations
MBH98
MM05
6
Two normalizations of the Sheep Mountain
bristlecone pine record, CA534
MBH98
?7
MM05
?0.1
7
Variance and normalization conventions
MBH98
x MM05
. This study
8
Leading principal components and the straight
average
MBH98
MM05
This study
9
A simple method to estimate past temperatures

• Normalize all paleo temperature records to zero
  mean and unit variance between 1820 (the shortest
  record) and 1980.
• Weight records according to spatial
  distribution and then compute the average.
• Adjust mean and variance of the average proxy
  record to that of the instrumental record and
  correct for noise bias.
• Uncertainty scales as N-1/2.

10
Re-reconstruction of temperatures
bias corrected average
PCA method (Mann et al, 1998)
11
"It is also likely that, in the Northern
Hemisphere, the 1990s was the warmest decade and
1998 the warmest year" over the last 1,000 years.
---IPCC, 2001 "Likely" is defined as a 60-99
chance.

• More rigorous methods are necessary
• Convert proxy variability into temperature
• Proxy type and location
• b) Seasonal dependence
• c) Non-temperature dependencies
• d) Age-model uncertainties
• 2. Estimate average Northern Hemisphere
  temperatures
• a) Spatial covariability
• b) Spatially dependent noise
• c) Multi-resolution temperature signals

12
Estimating Northern Hemisphere temperature is not
at the cutting edge of mathematics or
statistics, but has important societal
implications. Placing current warming trends in a
longer term perspective requires a more
rigorous statistical approach.
13
Comparison of bias in reconstruction techniques
cross-correlation weighted average
PCA or simple average
Bias corrected average
14
Leading principal components and the straight
average
MBH98
MM05
15
Monte Carlo results using synthetic
records (following MM05)
Non-centered PCA typically gives hockey sticks