University of Cambridge > > Isaac Newton Institute Seminar Series > Biases and uncertainty in multi-model climate projections

Biases and uncertainty in multi-model climate projections

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Mustapha Amrani.

Mathematical and Statistical Approaches to Climate Modelling and Prediction

The ensemble approach has originally been derived in probabilistic medium-range weather forecasting, and is now broadly used in numerical weather prediction, seasonal forecasting and climate research on a wide range of time scales. Applications geared towards climate projections are usually based on a heterogeneous ensemble with typically a mere handful of ensemble members, stemming from different models in an only partly coordinated framework.

An important feature of ensemble approaches in climate research is the inability to rigorously quantify climate model biases. While biases of climate models are monitored for the control period, the lack of long-term comprehensive observations (on the centennial time-scales considered) implies that it is difficult to decide how the model biases will change with the climate state. In contrast to other studies, we look not only at 20 or 30 year averages, but also at the interannual variability. This allows us to consider additive and multiplicative biases. In the talk, I will discuss two plausible assumptions about the extrapolation of additive biases, referred to as the ``constant bias’’ and ``constant relation’’ assumptions. The former is used implicitly in most studies of climate change. The latter asserts that over-/underestimation of the interannual variability in the control period leads also to over-/underestimation of climate change, and this assumption is closely related to the statistical post-processing of seasonal climate predictions. In addition we explicitly allow the additive and multiplicative model biases to change between control and scenario periods, resolving the resulting lack of identifiability by the use of informative priors.

An analysis of of GCM /RCM simulations from the ENSEMBLES project shows that bias assumptions critically affect the results for several regions and seasons.

This talk is part of the Isaac Newton Institute Seminar Series series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2021, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity