Investigating the Hockey Stick Climate Model

1
Investigating the Hockey Stick Climate Model

• EAS 4803- Dr. Wang
• 4/22/08 Robert Binion

2
Problem

• Michael Mann, et al. have proposed the hockey
  stick temperature model, which suggests that
  world temperatures have significantly increased
  during the twentieth century
• I sought to analyze temperature data to determine
  the validity of his assertion that temperatures
  have significantly increased

3
Data

• My data set includes normalized temperature data
  for the years 1391-1991 for both the northern and
  southern hemispheres as compiled by P.D. Jones
  from 17 different proxies
• I used only the most recent 600 years of data
  because of inconsistencies in measuring
  techniques before that time period

4
Methodology

• I performed a principle component regression
  using a singular value decomposition method to
  analyze my temperature data set
• The regression is used to determine whether the
  hockey stick model could be a valid model to
  describe climate changes over the last 1000 years

5
Methodology

• After performing the regression on both the
  northern and southern data, I found the 95
  confidence interval for both slopes and plotted
  the 95 confidence interval of the fitting line

6
Results

• Plot of normalized northern hemisphere
  temperature data with principle component
  regression line

7
Results

• Plot of normalized southern hemisphere
  temperature data with principle component
  regression line

8
Results

• 95 Slope Error estimate for northern data
• low
• -0.0045
• high
• 0.0054
• Actual slope
• 4.6423e-004

9
Results

• 95 Slope error for Southern Data
• low
• -0.0048
• high
• 0.0047
• Actual slope
• -3.9896e-005

10
Conclusions

• Analysis of the data turns out to be inconclusive
• The regression for the northern hemisphere does
  have a positive slope to it, which would
  correspond to the hockey stick model
• However, the southern data dhows a negative
  temperature trend
• Both of the slope error bounds include positive
  and negative slopes, so neither graph is
  particularly conclusive
• We can say that Manns results are not
  statistically vigorous based on the slope error
  bounds I calculated