Abstract
This study considers the application of the Ignorance Score (also known as
the Logarithmic Score) in the context of ensemble verification. In particular,
we consider the case where an ensemble forecast is transformed to a Normal
forecast distribution, and this distribution is evaluated by the Ignorance
Score. It is shown that the Ignorance Score is biased with respect to the
ensemble size, such that larger ensembles yield systematically better scores. A
new estimator of the Ignorance score is derived which is unbiased with respect
to the ensemble size, and thus allows for a fair comparison of ensembles of
different sizes. In an application to seasonal climate predictions it is shown
that the biased Ignorance score can assign better scores to large ensembles
with poor quality than to small but skillful ensembles. By contrast, the new
bias-corrected Ignorance score correctly ranks these ensembles according to
their actual skill, independent of the number of members. It is shown that the
unbiased estimator has smaller estimator variance and error than the biased
estimator, and that it is a fair verification score, which is optimized if and
only if the ensemble members are statistically consistent with the
observations. A broader discussion is provided as to when a correction of the
finite-ensemble bias of verification scores is actually desirable, and when it
is not.
Users
Please
log in to take part in the discussion (add own reviews or comments).