Analysis and Reduction of Climate Model Systematic Errors through a Unified Modelling StrategyPrimary Author: Martin, Gill
Analysis and Reduction of Climate Model Systematic Errors through a Unified Modelling Strategy
Gill Martin & co-authors
Met Office, FitzRoy Road, Exeter EX1 3PB, UK
Various techniques are used in understanding and reducing climate model systematic errors. These include the use of idealised models (single column model, aquaplanet,
dynamical core), sensitivity tests designed to shed light on processes and investigate teleconnections, and "spin-up" tests which allow us to determine whether a systematic
bias is the result of long-term feedbacks, e.g. through model drift, or an immediate movement of the model away from the initial observed state and from which it does not
recover. In the latter case, it may be possible to attribute the source of the error to a particular parametrisation scheme, thus making its solution potentially rather easier. Spin-up tests using climate models also provide a parallel with numerical weather prediction models, which themselves have the benefit of assimilated observations as well as (usually) higher resolution.
At the Met Office, the same model is used for both daily forecasting and climate prediction, allowing direct comparison between the two systems. Each system brings a unique perspective to model development. The use of state of-the-art variational data assimilation in the forecast model minimises the errors in the large-scale synoptic flow and can allow parametrisation errors to be identified more easily. In addition, comparisons can be made against detailed observational datasets (ARM, TOGA-COARE, GCSS) to investigate individual physical processes. Long climate runs show how a parametrisation behaves in equilibrium and in combination with the rest of the model physics and dynamics. Coupled climate modelling represents a stringent test of the model physics as systematic errors in the atmosphere can feed back on the ocean. Although it is not always the case that model systematic errors are shared between the daily forecast model and the climate model, when this is the case we find that a unified approach to model development is beneficial to both systems.
Examples of the use of our joint approach to analysing and reducing systematic errors in the Met Office Unified Model are described, with particular reference to the representation of ENSO and the Asian summer monsoon in HadGEM1. Further examples are given in a contribution by Milton et al..
|
