Atmospheric science: Supercomputer improves tropical climate forecasts

The use of supercomputers and high-resolution climate models together can lead to month-long tropical climate forecasts, according to a paper published in Nature Communications. This finding has potentially important implications for the future prediction of monsoon and cyclone dynamics in the tropics.

The Madden-Julien Oscillation (MJO) is a travelling band of thunderstorms, characterised by a churning mass of clouds. It moves eastwards over the tropical Indian, Pacific and sometimes Atlantic oceans, within a 30-60 day window, causing anomalous bouts of intense rainfall that strongly influence monsoon dynamics and tropical cyclone development. Understanding and predicting the MJO is difficult due to the high-resolution cloud microphysics involved. Complex cloud-resolving numerical models have been developed, but detailed analysis is limited due to a lack of computational power.

Tomoki Miyakawa and colleagues use the enhanced power of the recently developed Japanese ‘K’ supercomputer to conduct a series of 40-day MJO simulations using a highly complex cloud-resolving model, known as NICAM. The immense computational power (10 peta-flops) available, allows the team to statistically evaluate 54 high-resolution simulations and to subsequently show that NICAM can be used to provide a valid prediction of MJO up to one month in advance.

While the results show that increases in computational power are promising for complex climate modelling, the rarity of these high-powered computer systems, and the intense competition for time utilizing those that currently exist, may mean that it will be some time before this prediction method becomes widely available.