Assessing the threat to critical functions and infrastructure assets relies on accurate forecasting of climate extremes. Infrastructure is generally built on the assumption that climate is stationary. In a changing climate, this assumption is no longer valid. Decision-makers cannot simply rely on historical observations to predict the statistics of future extreme events. Extracting actionable information from climate model projections, however, is at the cutting edge of climate science.
LLNL climate researchers have been developing the ability to simulate how climate change might change recent historical extreme events (typically referred to as storyline analysis). To demonstrate their approach, they analyzed a recent weather event—the record-breaking rainfall that hit northern China in the summer of 2023. While Beijing is occasionally affected by summertime typhoons, in this instance, a pair of back-to-back typhoons triggered some of the worst flooding in the region’s history, destroying roads, knocking out power, destroying pipes carrying drinking water, and displacing over a million people from their homes. The event was a significant test of the region’s capacity to handle extreme weather.
Beijing and its surrounding cities are bordered by plains to the east and mountain ranges to the north and west. The dramatic variations in regional topography can result in distinctly different rainfall and flood intensities that cannot be captured by the typical resolutions of global climate models. The team worked with the Simple Cloud-Resolving Energy Exascale Earth System Model Atmosphere Model (SCREAM) in two new high-resolution, regionally-refined modeling configurations—one featuring 3 kilometer resolution over the region of interest, and the other an even more highly resolved 800 meter version. Using regionally-refined versions of the model provides much greater computational efficiency than would be possible using a kilometer-scale resolution over the entire globe, making rapid analysis of recent events possible. These high-resolution simulations took less than 1 day of run time on LLNL’s supercomputers.
First, the team simulated the 2023 Beijing flood and demonstrated that their results compared well against satellite and ground-based observations of the event. They then simulated the same event, but with the higher atmospheric and sea surface temperatures and atmospheric water vapor and greenhouse gas levels projected for 2050. The storm is projected to be even worse in a warmer world because the amount of moisture the atmosphere can hold increases nonlinearly as temperature warms, enabling clouds to squeeze out more water as they move across temperature gradients.
The team continues to expand its capabilities and can now do storyline analysis for a wide variety of extreme events—hurricanes, atmospheric rivers, windstorms, etc.—anywhere in the world.
