Fig. A: Preliminary annual
by the National Weather
Service at San Antonio,
Texas, from 1885 to 2009.
Fig. B: Figure A with a superimposed 10-year running
average that more clearly
reveals patterns and trends
in the data. Fig. C: Figure A
with a superimposed mean
and linear trend (given by
the equation at top) and
a color-coded background.
changed from the starting point to any year.
The regression equation is y = 0.0031x + 63. 1,
where y is the temperature (T) and x is the year.
This formula gives T = 68. 94 in 1885 and 69. 33 in
2009 for an increase of 0.38 degrees.
Illustrations and photograph by Forrest M. Mims III
Figure C is decorated with Excel’s “gradient fill”
option to provide a hot (red)/cold (blue) color-coded background for the chart. This looks good on
the web and in general publications but would be
inappropriate in a formal, peer-reviewed paper in
While it’s common for climate scientists to plot
linear trends of their data, this method misses
significant fluctuations in the data.
For example, the trend lines in Figures B and C
completely miss the warm temperatures of the
1930s and the cool temperatures of the 1970s. This
and the uncertainty of many kinds of experimental
data mean a linear trend line cannot always forecast
changes to come.