Coupled model simulations: EC-Earth

Experimental Design:

We applied the same method used in the observational analysis on general circulation model data to decrease the statistical uncertainty at the expense of an increased systematic uncertainty. We used 16 experiments covering 1861–2100 of the EC-Earth 2.3 model (Hazeleger et. al. 2010) using the CMIP5 protocol (Taylor et. al 2012). This model is very similar to the ECMWF seasonal forecasting model. The resolution is T159, ie about 150 km over the UK. Precipitation in this area shows a climatology comparable to ERA-interim (which is made with a very similar model). Extreme winter precipitation is concentrated in October–January, as in the observations.

Temperature:

1) Connection with global warming
We performed the same analysis as in the observations. The value of 2015 was corrected for a small bias in the mean december temperature and a bias in the variability, which is 20% lower in the model than in the observations. Even with these bias corrections, the event falls outside the simulated distribution (as in the Weather@Home ensemble), with a return time of more than a few thousand years. This is higher than the observational analysis. The difference is probably due to a combination of two factors. 1) The tail is better constrained by the high statistics in the model, which would argue that the model value is more reliable. In particular, the normal distribution does not describe the highest values well in the model. 2) The EC-Earth model has a well-known cold bias over the subtropical Atlantic, which is the source region for the mild and moist air being transported to the UK in december 2015. This implies that it has a circulation-dependent bias, with mild weather too cold and cold weather not cold enough. This implies that the model estimate is too high.

Because the observed temperature is very close to the upper bound of the ensemble, it is very hard to determine how much the probability has increased due to the external forcings to the climate, mainly greenhouse gases, but the factor is large.

Figure 10.  A fit of the EC-Earth equivalent of December Central England Temperature to a generalised Pareto Distribution (GPD) shifted by the ensemble average model global mean temperature.

2) Connection to El Niño.
The model has a significant teleconnection of Central England Temperature with the Niño3.4 index. However, at r=0.08 it is very weak. We have not yet performed the extreme value analysis to determine how much El Niño in the model influences the probability of extremes.

Figure 11. Correlation between CET and Niño3.4 in the EC-Earth coupled runs.

Precipitation

1) Connection to global warming
The model simulates an increase in Northern England precipitation in winter, but the trend is not as large as the observed one. For a return time in the current climate similar to the observed one, 200 years, we find an increase in probability of a factor 2.5, with an uncertainty range of 1.2 to 4.5.

Figure 12.  Fit of EC-Earth Northern England precipitation to a normal distribution that scales with the ensemble average global mean temperature.

2) Connection to El Niño.
the model does not simulate any dependence of Northern England precipitation on the state of El Niño, with a correlation coefficient of r=0.01 and no visual indications that extreme events behave differently than the mean.

Figure 13.  Scatterplot of EC-Earth Northern England precipitation as a function of the strength of El Niño in the model. The line represents the best linear fit.