Thursday 09 Feb 2017: Statistical Science Seminar: Downscaling high temperature events on the Greenland ice sheet
Emma Eastoe - University of Lancaster
Rising temperatures on the Greenland ice sheet are of serious concern due to the fact that higher temperatures imply increased ice melt which, in turn, implies increases in global sea levels. Accurate prediction of high temperature events in the future thus plays a key role in prediction of sea level changes. Often these predictions are made using output from regional climate models (RCMs). Whilst such models well represent the mean behaviour of the temperature process, they represent extreme events less well. The reason for this could be due to either the methods by which the RCMs are calibrated, or the fact that extreme events occur at a very localised scale and RCMs provide output on a grid scale (in our case grid cells are 7.5 km^2), or both of the above. In this talk we will look at a statistical downscaling method to 'nudge' the RCM data to produce extreme values that are more consistent with those seen in observational data sets. Using RCM output from the MAR model and observational data from the Greenland Climate network (GC-net), we show that this method produces estimated return levels which are consistent with those estimated using the observational data.