School of Earth and Environment

A Seamless Approach to Assessing Model Uncertainties in Climate Projections of Severe European Windstorms (SEAMSEW)

Funding: AXA Research Fund

Project period: Oct. 2010 to Sept. 2013

Total grant: €450k

PI: Peter Knippertz

Co-Is: John Marsham, Doug Parker, Alan Haywood, Piers Forster

Project scientists: Tomek Trzeciak, Jenny Owen, Steven Pickering

Summary

Despite the enormous advances made in climate change research, the recently published UK Climate Projections (2009) conclude that "robust projections of the latitude and strength of the Atlantic stormtrack near the British Isles are not yet possible, as differences between individual models, and also between single-model ensemble members with respect to the current climate are too large." This uncertainty about the evolution of tracks, frequency and intensity of damaging windstorms over the North Atlantic in the decades to come bears enormous risks to European societies and the (re-)insurance industry. This project will identify and quantify the sources of error in climate model simulations, making use of a "seamless" approach of simulating severe storms in "climate" and "weather" versions of the same model. The results will enable a much more robust prediction of future regional climate impacts across Europe.

Left: Meteosat satellite image of windstorm Klaus over the Bay of Biscay on 23 January 2009 at 2230 UTC. The red line shows the subsequent track and the red shading areas with heavy damages as illustrated by toppled eucalyptus trees in a cultivated forest in Galicia (Spain) shown in the right panel (courtesy of Ángel Manso, La Voz de Galicia).

Left: Meteosat satellite image of windstorm Klaus over the Bay of Biscay on 23 January 2009 at 2230 UTC. The red line shows the subsequent track and the red shading areas with heavy damages as illustrated by toppled eucalyptus trees in a cultivated forest in Galicia (Spain) shown in the right panel (courtesy of Ángel Manso, La Voz de Galicia).

Previous studies have addressed the problem of climate model uncertainty through statistical comparisons of simulations of the current climate with (re-) analysis data and found that many models struggle to realistically reproduce the observed climatology of intense cyclones. The usage of different models, resolutions and approaches to measure storminess make it difficult, if not impossible, to identify and separate single causes of errors. To illustrate this, two extreme scenarios of the influence of a climate model's basic state on storminess are given: (1) The model is in principle capable of reproducing severe storms given the right initial conditions, but slow processes in the climate system such as ocean circulations lead to an incorrect occurrence frequency of storm-prone initial conditions. (2) The model has a mean state and variability close to those observed, but does not capture the crucial dynamics of extreme cyclone intensification due to over-simplistic model physics or insufficient horizontal resolution. In a statistical evaluation the two scenarios will look very similar, but the interpretation of climate predictions should be fundamentally different. Compensating effects might even conceal errors and suggest higher reliability than there really is.

Flowchart

This project takes a completely different approach to the problem. Three state-of-the-art climate models used for the fourth Assessment Report of the IPCC will be evaluated according to their performance in weather prediction for 15–20 carefully selected historical severe storms (single and serial) over Europe. Such a “seamless prediction” concept is increasingly followed by operational centres today. The main objectives are (I) to clearly separate and quantify different sources of uncertainties in the potential evolution of damaging storms over Europe and (II) to understand the physical mechanisms that lead to disagreement between different models or model configurations.
The main deliverables resulting from this work will be:

  • a thorough evaluation of the reliability of existing climate runs
  • a calibration of projected changes in storm climate based on this evaluation
  • recommendations for optimal model configuration/future model design.

To achieve this we will conduct simulations with predefined initial conditions on time-scales of several days, which separates uncertainties associated with an incorrect basic state and internal variability of the atmosphere from those associated with fast processes. This cost-effective case-study approach allows us to perform carefully designed experiments to test the relative importance of model formulation, horizontal resolution and variations to the initial conditions using a common metric to assess forecast performance. The lessons learned from this exercise will be translated into implications for long-term climate integrations through comparisons of common biases and the development of physically meaningful calibration techniques for existing climate simulations. A secondary objective is to examine sensitivities with respect to initial conditions and to assess implications for storm clustering. These investigations will be based on ensemble techniques developed for estimating uncertainties in operational weather prediction and will allow the identification of robust storm-prone large-scale settings to be used for downscaling and pattern recognition analysis as a further deliverable.

The project will make a strong contribution to activities of the World Weather Research Programme THORPEX (The Observing System Research and Predictability Experiment) and build a bridge between scientists in climate-change impact research, atmospheric modelling, climate research and dynamical meteorology for the sake of a much more robust prediction of future windstorm impacts across Europe.

The project has close links with other storm-research and model evaluation activities in the UK and with individual storm researchers across Europe: