![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Bias adjustment for climate predictions: an overview |
VerfasserIn |
Jens Grieger, Henning W. Rust, Uwe Ulbrich |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250150744
|
Publikation (Nr.) |
EGU/EGU2017-15237.pdf |
|
|
|
Zusammenfassung |
While general circulation models are designed to be able to reproduce observed climatologies,
deviations are found in various aspects of statistics, both of individual parameters and
of their relationships. A first order approach to describe the model-observation
discrepancies is the difference in mean states (or expectations) for individual parameters:
the bias B. Quantification and adjustment of bias is important if model output is
compared to observations or is used for further processing, e.g. to drive impact
models.
Whenever a climate model is initialized with a climate state distant from its preferred one
(e.g. from observations), the model tends to drift towards its own climatology during the
numerical integration. Thus, the model bias depends on i) forecast lead time τ (model drift:
B(τ)) and ii) the distance between initialization and model preferred states. In the context of
climate predictions, these effects have different importance and need to be taken care
of.
In this contribution, an overview of different methods for bias and drift adjustment for
climate prediction is given, with a focus on decadal time scales. Starting from the
recommendation of the Decadal Climate Prediction Project (DCPP), approaches are
discussed for a bias adjustment depending on the distance between the state for initialization
and the model’s climatology. One popular way in decadal predictions is to consider
initialization time as a proxy for this distance. Advantages and disadvantages of approaches
with different complexity are discussed. |
|
|
|
|
|