Here is a message on behalf of project scientist Dr Peter Watson:
We are in the final stages of developing a new pair of global atmospheric models for use in CPDN. These have a big resolution upgrade compared to the existing global model used by Weather@Home, meaning that the spacing between grid points is being reduced, making use of the larger memory and processor power available in modern computers. The purpose of this is to address the fact that the high-impact weather events that CPDN is often used to study (like extreme heatwaves and heavy rainfall) usually involve extreme "dynamical" phenomena, like storms and blocking anticyclones. So it is very important to be able to simulate these events faithfully, and increasing the model resolution helps with this.
The resolution sets how much fine detail the model can resolve through calculations of the fundamental fluid dynamics of the atmosphere, and increasing it reduces the reliance on more approximate parameterisations of sub-grid processes. These models will have resolutions of 0.83x0.5 degrees (~60km in middle latitudes) and 1.25x0.83 degrees (~90km). The global atmospheric model that is currently used has a resolution of only 1.875x1.25 degrees (~150km). The number of vertical levels will also be doubled. Studies that have been done over the past few years have found that reducing the grid spacing below 100km starts making substantial improvements to the quality of simulated weather phenomena such as those previously mentioned.
The model with the higher resolution has a grid spacing close to that of the 50km regional model, but simulates features at this scale for the whole globe, effectively simulating the results for all of those regional models simultaneously. Whilst the regional models that are currently used help to capture fine details, they are dependent on the winds simulated by the 150km-resolution global model at large scales. The new higher resolution models will also simulate these large-scale winds more faithfully, giving further improvements in the results for small regions. The regional models may still be more appropriate for some projects, though, and are likely to still be used.
Our initial tests have found that the quality of weather simulations from these models appears similar to that in the current state-of-the-art model used at the Met Office. So with the ability to produce multi-thousand member ensembles with these models, there will be many opportunities to do cutting-edge science.
For now, these models are only going to be run on Linux systems. Some of you will already have received work units using the model with ~90km resolution. This work is pushing at the boundaries of what it is possible to achieve in a project like this, so please bear with us as there may be quite high failure rates at first as issues are ironed out.