Improved Blocking at 25km Resolution?

Tuesday, 16 December 2014
Reinhard Schiemann1, Marie-Estelle Demory1, Matthew Mizielinski2, Malcolm Roberts2, Len Shaffrey3, Jane Strachan4, Pier Luigi Vidale1 and Mio Matsueda5, (1)NCAS Climate, Reading, United Kingdom, (2)Met Office Hadley center for Climate Change, Exeter, United Kingdom, (3)University of Reading, Reading, United Kingdom, (4)UK Met Office, Exeter, United Kingdom, (5)University of Tsukuba, Tsukuba, Japan
It has been suggested that relatively coarse resolution of atmospheric general circulation models (AGCMs) limits their ability to represent mid-latitude blocking. Assessing the role of model resolution for blocking is computationally expensive, as multi-decadal simulations at the desired resolution are necessary for a robust estimation of blocking statistics. Here, we use an ensemble of three atmosphere-only global models for which simulations that fulfil this requirement are available at resolutions of roughly 25km horizontal grid spacing in the mid-latitudes. This corresponds to about a fourfold increase in resolution over the highest-resolution CMIP5 (Coupled Model Intercomparison Project, Phase 5) models. The three models are (i) the ECMWF model (IFS) as used in the project Athena, (ii) the MRI-AGCM 3.2, and (iii) our own HadGEM3-GA3 simulations obtained in the UPSCALE project (UK on PrACE - weather-resolving Simulations of Climate for globAL Environmental risk).

We use a two-dimensional blocking index to assess the representation of blocking in these simulations and in three reanalyses (ERA-Interim, ERA-40, MERRA). We evaluate the spatial distribution of climatological blocking frequency, the interannual variability of blocking occurrence as well as the persistence of blocking events. Furthermore, the degree to which blocking biases are associated with mean-state biases is quantified in the different models.

We find that the representation of blocking remains very sensitive to atmospheric resolution as the grid spacing is reduced to about 25km. The simulated blocking frequency increases with resolution, mostly so as to reduce the model bias, yet there is considerable variation between the results obtained for different models, seasons, and for the Atlantic and Pacific regions.