At the end of May 2021, 172 runners embarked on an ultramarathon of 100 kilometers (62 miles) in the northwest of China. At noon, as runners made their way through a rugged, high-altitude portion of the course, temperatures plunged, high winds whipped the slopes, and freezing rain and hail battered runners. By the next day, the death toll from the sudden storm had risen to 14.
A new study looks back at the fatal event with the aim of testing how hyper-local modeling can improve the accuracy of predictions for mountain events. The riders got into trouble because the hourly weather forecast for the class underestimated the storm. The steep mountain slopes had very localized effects on wind, precipitation, and temperature on a scale too small for weather predictions of the event, according to the new study published in the journal AGU JGR Atmospheres.
The hourly predictions for the race of 2021 were based on relatively large-scale atmospheric processes, with models operating at three kilometer resolution – enough for most regional forecasts, but too coarse to capture “hyper-nearby” weather like the storm that hit the course, says Haile Xue, a climatologist at the China Earth System Modeling and Prediction Center CMA and lead author of the new study. Even though a wind and cold temperature advisory had been issued the previous night, it lacked the resolution needed to identify danger zones on the course.
“A temperature forecast apparent based on high-resolution simulation can be useful” in addition to general regional forecasts, Xue said. Situations like the 2021 storm are common in mountains at extremely high altitudes, such as Mount Everest and Denali, the document says. Although less common at lower elevations, when such storms do occur they can strike suddenly and lead to injuries and loss of life.
The new study uses topographic data from the course, at tens of meters of resolution rather than kilometers, to model the hyper-local meteorological situations created by the mountains. With a resolution two orders of magnitude greater than the original forecast for this 7-day-stop, as well as detailed considerations of mountainous topography, the model accurately recreated the storm conditions of the race and even offered better preview of what could have happened that day.
The initial forecast included a large-scale cold entrance, which would have resulted in lower temperatures and stronger winds – but not extreme – with only a low level wind advisory issued. The new study found that the apparent temperature could have dropped to -10 degrees Celsius (14 degrees Fahrenheit), approximately 3 degrees Celsius lower than the original models predicted.
The model also generated an “impact forecast”, including including apparent temperature, which could have dropped even further because it takes humidity into account and would ideally include the effect of wet clothing or skin on body temperature. Including them in forecasts, Xue says, could help mitigate the risk of hypothermia.
In addition to the weather, program planning and equipment requirements for runners were discussed after the event. Many endurance events require many layers to protect against heat and rain these were suggested but not required which could have contributed to the loss of life. Accurate weather forecasts and equipment requirements are essential for an event to be safe.