First page Back Continue Last page Overview Graphics
Cross Correlation Analysis
Can identify which of 2 variables changes first
- Temperature and CO2
- Temperature and CH4
- CO2 and CH4
Smoothing is required to normalize variability
- Smoothing does not mask cause and effect
- Smoothing makes short term dependence apparent
Notes:
The cross correlation analysis must be performed with a smoothing interval of at least 1500 years for the Vostok ice core data. Otherwise, the analysis will become distorted by more rapid changes in temperature, which aren't reflected in the CO2 and CH4 data. The DomeC data analysis can be performed with a shorter smoothing interval, although care must still be exercised to avoid incorrect correlation across very short period temperature variability.
Smoothing doesn't mask cause and effect, even when the delay is small. Smoothing only minimizes short term periodic effects shorter than the smoothing interval. Only by cancelling these short term periodic effects can short term causality be ascertained. In fact, smoothing accentuates time delayed, causal dependencies.