Today, weather and climate forecasts are based largely on computer models. These models need huge computing power to be able to cope with the complexity of the global atmosphere, because of which they use the very latest and biggest computers possible. The faster computers become, the easier it is for weather and climate models to describe the atmosphere in more detail. As a result, the behaviour of clouds and atmospheric turbulence in particular can be calculated better and fewer assumptions are necessary when describing these processes. This paves the way for better weather forecasts and more accurate climate scenarios.
The results of this research make it possible to give more accurate weather and climate forecasts in the future
In the research that we did with the Royal Netherlands Meteorological Insitute (KNMI), we looked at possibilities to use a new technology to produce weather and climate calculations. When doing this, we drew on the very latest computer-related developments: calculations are carried out on the graphics card. This chip, which is now a standard part of all computers, has a large number of small calculation units instead of just one strong one. As such, it needs to be controlled slightly differently.
You can compare it with the way in which tasks are allocated in a team: although coordination may be far easier if just one person takes on all of the various tasks himself, the work will get done faster if tasks are allocated efficiently among the members of a team. Given the extra computing power offered by the graphics card, we are able to produce calculations with a far higher resolution than has been possible in the past: for example, we have been able to run a simulation of the weather throughout the Netherlands at a resolution of 100 metres. The above means that we are able to establish how, where and when clouds will develop far more accurately and, by doing this, provide better weather and climate forecasts in the future.
Dr. Jerôme Schalkwijk was a PhD candidate at Delft University until his doctoral defense (cum laude) in April 2015. His PhD research was about the studying and forecasting of atmospheric phenomena using High Performance Computing and about the development and use of a Large-Eddy Simulation program that employs GPU (CUDA) technology to utilize the Graphical Processing Unit (GPU, video card).
Clouds, atmosphere, air quality, forecast, climate