Dynamical Downscaling of Extra-Tropical Cyclones

In this Chapter the method of dynamically downscaling extra-tropical cyclones is applied to data from the Global Climate Model (GCM) climate simulations. In Chapter 4 the GCM at two resolutions was used to investigate the impact of an increase in resolution, and the effect of a warmer climate, on extra-tropical cyclones in the Northern Hemisphere (NH), with a particular focus on the precipitation field. It was shown that both the increase in resolution and a warmer climate resulted in the identification of more extreme precipitation events. A more regional analysis looking at the UK and Western Europe provided less robust results due to the relatively low number of events identified in the 20-30 year periods of the GCM integrations. This is a typical problem with climate model simulations where a relatively high resolution is required to simulate weather systems correctly and as a consequence only relatively short simulations can be performed with current computational resources, so that when regional analysis of these systems is conducted the sample may not be large enough for meaningful statistical results. It is also difficult to run GCMs at the resolutions that might provide realistic extreme precipitation associated with cyclones, i.e. <4 km. In Chapter 5, a Limited Area Model (LAM) was used to determine whether it was able to predict intense precipitation events by investigating two past intense precipitation events in the UK. The results showed that it was capable of reproducing the location, timing and intensity of precipitation over a river catchment sized area, when compared to raingauge data. The approach in this Chapter is to run the LAM of Chapter 5 using boundary conditions and start dumps from the T213 (60 km) integration of the GCM used in Chapter 4 for particular cyclonic events over the UK which could result in flooding. This event-based downscaling is essential to simulate storms affecting a particular region at the resolutions necessary to obtain realistic precipitation intensities required for flood modelling. The T319 integration could not be used as some of the fields required for the downscaling were unavailable.

The events investigated in this Chapter are NH summer events to be consistent with the seasonal period in which the observed storms examined in Chapter 5 occurred. Flooding during the winter season is generally caused by the cumulative effect of a series of cyclones affecting a region over a longer time period, which is not addressed here, whereas in summer flooding events are generally caused by single cyclonic events (e.g. June and July 2007). Summer flooding is also characterised by mesoscale convective structures, which have a smaller scale than the GCM resolution. The events of interest in this Chapter are events similar to those investigated in the previous Chapter, intense precipitation events that have a large-scale cyclonic feature associated with them, which may have convective cells embedded within the large-scale feature. As shown in the previous Chapter, the resolution of the LAM is an important consideration when predicting the location, timing and intensity of precipitation. The events are identified from those used in Chapter 4 where the impact of an increase in resolution, and the effect of a warmer climate, on such events was highlighted.

In previous work, such as that of Duliere et al. (2011), the importance of downscaling information from GCMs is discussed, to account for the relatively poor representation of local terrain and sub-grid scale processes in the GCM. Previous climatological downscaling studies (e.g. Haylock et al., 2006; Schmidli et al., 2006; Mehrotra and Sharma, 2010) focus on daily or monthly precipitation and downscale a long integration period which tends to limit the achievable resolution due to the available computational resources. In this study a single event approach is taken, producing hourly precipitation data for specific use as input to hydrological models, where a daily temporal resolution is not viable. It is also important to note that the resolution of the GCM used in the downscaling has a resolution of .60 km, which is comparable to many of the RCMs discussed in the previous studies.

The method used in this study, how the events are identified and the generation of the initial and boundary conditions for the LAM, are explained in x3.6. Four events are identified over the UK from each climate and are discussed separately in this Chapter, with the events from the 20C discussed in x6.2 and the 21C events discussed in x6.3. Due to the low number of events identified in each climate, no statistically robust comparisons can be made between the climates. Time limitations, computational resources and problems associated with the reliability of the computational platform limited the number of events that could be downscaled. In the future this will be expanded but for a relatively small region the number of events that can be identified from a GCM will be limited by the GCM integration length. The aim of this Chapter is to assess the viability of the downscaling method to the high resolutions used here and to consider improvements, discussed in x6.4. It also provides a preliminary study of event-based downscaling focussed on producing data for flood models, such as required for the NERC DEMON project (DEMON, 2012).