Source: Journal of Geophysical Research: Atmospheres
Thunderstorms and rain showers are difficult to simulate using weather and climate models because they require representation of atmospheric circulations down to the kilometer scale and smaller. This detailed modeling takes a lot of computing power, so models capable of such fine resolution have so far been limited to subcontinental-scale simulations.
However, novel supercomputer designs employing accelerators such as graphics processing units are opening up new opportunities to simulate the details of thunderstorms on a continental scale. In a new study, Leutwyler et al. demonstrate the capabilities of a new version of the Consortium for Small-scale Modeling (COSMO) model, evaluating its ability to simulate storm processes across continental Europe.
At kilometer-scale resolution, COSMO is a convection-resolving model, meaning that it explicitly simulates the deep convection processes in which warm, moist air rises to create precipitation conditions. In contrast, other models use parameterization to simulate deep convection, which is a complex mathematical approach that, despite recent advances, can introduce uncertainties.
To evaluate COSMO’s performance at a continental scale, the research team ran a 10-year simulation of Europe’s climate at a resolution of 2.2 kilometers. They used atmospheric data from the ERA-Interim Reanalysis for the period of 1999 through 2008 to drive the model.
The researchers then compared the COSMO simulation with observational data sets for the same time period. These included data from rain gauge networks and energy balance stations (which measure surface solar radiation and heat transfer) and lightning data sets.
The team found that COSMO successfully simulated Europe-wide precipitation patterns for the 10-year period. The model performed especially well in reproducing summertime precipitation patterns and accurately simulated important aspects of the continent’s annual deep convection cycle.
The scientists also identified areas for improvement. For example, COSMO underestimated or overestimated hourly precipitation to different degrees depending on whether the underlying terrain was flat or mountainous.
This work builds on previous research showing that supercomputers equipped with graphics processing units (GPUs) allow for rapid performance of the latest version of COSMO. The computations were done on the Piz Daint supercomputer of the Swiss National Supercomputing Centre. One day of simulation takes about 18 minutes and uses 144 compute nodes (which is about 3% of the total resources). These current findings serve as a proof of concept that GPU technology allows for continental-scale climate simulation in a reasonable amount of time. (Journal of Geophysical Research: Atmospheres, https://doi.org/10.1002/2016JD026013, 2017)
—Sarah Stanley, Freelance Writer
Citation:
Stanley, S. (2017), New supercomputers allow climate models to capture convection, Eos, 98, https://doi.org/10.1029/2017EO074615. Published on 02 June 2017.
Text © 2017. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.