Geoscience Reference
In-Depth Information
When the IPCC specifies a confidence interval, it is generally 90% (
very
likely
)
1
. The sum of the probabilities to exceed the given upper or lower
limits is therefore 10%, without them necessarily being equally shared at 5%
each. It should not be assumed either that the most likely value is situated in
the middle of the range.
Regarding an assertion
(not quantified), the degree of confidence results
from both the reliability of the indicators (type, quantity, quality and
consistency), codified as
limited, average
or
robust
,
and from the degree of
agreement amongst the experts in their collective judgment:
weak, average,
high
. Hence, skillfully combining these two factors, gives a scale of overall
confidence:
Very low
10 %
Low
20 %
50 %
2
Medium
High
80 %
Very high
90 %
Since giving a level of confidence is an assertion, it is generally subject
itself to an assessment of confidence (see section 5.2), to which it in turn
could be subject…but let's stop there.
The probabilities indicated (in %) may result from effective statistical
calculations, especially the ranges of confidence, but not necessarily.
Conversely, it is clear that a level of agreement between the members of a
group of experts has nothing to do with a formal statistical calculation based
on observed data, even though providing evaluations in % in the context of a
scientific report may strongly suggest that it results from statistical
calculations.
Moreover, some climatic parameters arise from a combination of other
parameters (e.g.:
). In most cases, even using fuzzy logic,
it will be impossible to accurately combine the evaluations of the
λ
=
log(
2
)
α
1
/
S
G
clim
1 Statisticians often prefer 95% (for confidence ranges of +/- two standard deviations in the
Gaussian case).
2 When the IPCC gives the label of
medium confidence
to some catastrophe caused by climate
change, this allows them to announce it without taking the risk of an explicit assertion.
Search WWH ::
Custom Search