Global Climate Change - Taking the
temperature of the earth.
You have seen a number of places the graph showing global temperature
fluctuation over the last century. What is methodology behind
producing the necessary numbers? Our body temperature is relatively
uniform and so is relatively easy to measure. But temperature
on the earth varies considerably.
Several basic challenges confront us in trying to determine
- How do you average readings from around the globe and reduce
it to a single number that represents some meaningful average?
- In the fluctuating and complex time series that results how
do you separate out long term versus short term signals versus
noise? Think of the stock market as an example of where people
would love to recognize the different signals.
- There is a focus on the near surface, lower atmosphere temperatures,
but there is a lot of vertical structure to the atmosphere. How
should this be taken into account?
If there was only a way to measure the thermal calorie load
of the entire atmosphere!
Change over what time frame?
- seasonal hemispheric warming?
- warming in yearly averages?
- warming in decadal averages?
- smoothing out the curve!
- Signal to noise ratio.
Discussion question: what are the biases in ground T record
that must be overcome?
How do you accommodate biases?
- comparisons and corrections: e.g. rural vs. urban T in an
area -> correction factor.
- weighting, and functions: e.g. temperature with elevation
(not a simple function). Not mentioned - but proportionally by
- throw the bad data out.
- nearest neighbor comparisons for consistency
Natural indices of T change?
- glacier inventories.
- others we will focus on, but must avoid circular reasoning.
What is Jones & Wigley's interpretation of the T record:
- there is a half degree warming in last century.
- .2 to.3 could be noise (not a warming trend). Make important
point - .2 to .3 either way.
- volcanic eruptions show up in data.
- low end of computer models.
I'm struck at how important early baseline records are
in this endeavor. Mere cataloging can be a very worthy endeavor,
and hence an intrinsic love of data by many scientists.
Discussion question for next time: computer models and prediction.
How might you test the potential accuracy of a given computer
model for a time span of several years, and for a time span of