 |
 |
 |
 |
|
 |
 |
Training with trend
•
In our test of MBH in science
(2004), the trend in the
calibration period was taken out.
•
When the trend during the
calibration period is used as a
critical factor in the empirical
reconstruction model, then the
contamination of the proxy trend
by non-climatic signals must be
mimicked.
•
Thus, we introduced an error on
the centennial time scale as red
noise.
•
Here: 50% centennial,
75% year-to-year.
•
Again heavy underestimation of
long-term variability.
 |