Global climate and weather forecasting use predictive models which divide the Earth's atmosphere into a grid, with each section typically covering tens to hundreds of kilometers. Now a new model, run by a team at the University of Tokyo, has narrowed this grid spacing down to just 220 meters. This higher resolution removed instances of "popcornlike" rain, bursts of heavy rain which appear on simulations but not in real life, providing more accurate predictions of extreme local weather from a global model.
This year has been one of record-breaking weather: Intense storms battered Europe, August was blisteringly hot, and tragic flooding devastated parts of Asia and Africa. Strong and unpredictable storms are expected to become more common due to climate change, so precise climate modeling and weather forecasting, along with early warning systems, are essential to help avert future disasters.
Global storm-resolving models (GSRM) are amongst the most advanced tools we have for such a task. They are used to simulate the global distribution of clouds and storms, and to see how small- and large-scale atmospheric and energy systems interact and influence each other. Looking at the Earth as a whole, even when making local predictions, is important as even distant and small-scale events can have knock-on effects.
Compared to regular global climate models (GCMs), which divide the Earth's atmosphere into a grid with sections spanning tens to hundreds of kilometers, GSRMs use supercomputing power to narrow that resolution down to just 1 kilometer to 10 km per section. Now, a team at the University of Tokyo has created a new simulation which narrows that resolution down even further, to just 220 meters per section, about the length of two football (soccer) pitches.
"This finer resolution allows us to represent the internal structure of convective clouds, including individual updrafts and downdrafts, much more explicitly than before," explained Project Researcher Shuhei Matsugishi from the Atmosphere and Ocean Research Institute at the University of Tokyo. "Performing such simulations globally opens up the possibility of studying not only individual convective clouds, but also how they interact with each other and with the larger-scale atmospheric circulation."
The team has called this the world's first demonstration of a "global large-eddy simulation" ("GLES"), because it can explicitly represent the finer structures of deep convective clouds, such as storm-bringing cumulonimbus, rather than predicting their behavior based on coarser-resolution formulas and parameters.
Thanks to its higher resolution, the model was able to eliminate a long-standing "bias" called popcornlike rain, which occurs in current GSRMs. In this context, a bias refers to a recurring and consistent error in a model. Popcornlike rain is when intense bursts of rain appear on a model which don't actualize in real life. With the GLES, precipitation appeared more realistically, without unrealistically intense, localized events.
The most challenging aspect of this research, according to Matsugishi, was the "computational cost" of running the GLES. The simulation had almost 1 trillion three-dimensional grid points, representing individual data points across the globe. To simulate eight hours (specifically on Aug. 5, 2016), the team had to use more than half of the supercomputer Fugaku simultaneously. That's equivalent to roughly 40 years' worth of electricity consumption for an average household in Japan.
"At present, a global simulation at 220-meter resolution is far too computationally expensive to replace operational weather forecasting systems," said Matsugishi. "For now, these simulations are better suited to research experiments. For example, it can be used to investigate the detailed structure of tropical convection and heavy rainfall, and to provide a high-resolution reference against which coarser climate models can be evaluated and improved."
Also, while the GLES's higher resolution did resolve the issue of popcornlike rain, other biases to do with the distribution of cloud cover remained, highlighting a need for a deeper understanding of the processes involved. With this in mind, the team intends to use this model to next investigate the characteristics and properties of convective clouds in much greater detail.
"We want to better understand how turbulence, cloud microphysics and other unresolved processes should be represented, as global models move from kilometer scale to several-hundred-meter and eventually tens-of-meters resolution," said Matsugishi. "Ultimately, this could better represent extreme weather and reduce uncertainties in future weather and climate predictions."
