As climate change intensifies the water cycle, communities in mountainous regions face growing threats from flash floods and landslides triggered by sudden, violent rainstorms. An international research team has shown that increasing the resolution of weather forecasting models to the kilometre scale could improve the ability to predict these events—not just in China's Qinghai Province, but in complex terrain worldwide.
The study, published in Advances in Atmospheric Sciences , focused on a catastrophic rainstorm that struck the Hongshui River valley in eastern Qinghai on 13 August 2022. The event caused widespread flooding, damaged crops, and affected nearly 6,000 households. Using the Weather Research and Forecasting (WRF) model, scientists compared simulations at different resolutions: 9 kilometres, 3 kilometres (matching China's operational forecasts), and 1 kilometre.
The results showed clear differences. Only the 1-kilometre simulation accurately captured the storm's intensity, timing, and location.
"The 1-kilometre grid spacing allowed the model to capture the subtle wind patterns within the valley that actually triggered the storm," said Yongling Su, lead author of the study and a forecaster at the Qinghai Meteorological Observatory. "As the sun heated the valley slopes during the day, predictable upslope winds developed. But in the evening, these collided with cooler air draining down the mountainsides, creating narrow lines of forced rising air that ignited the thunderstorm cells. At coarser resolutions, these critical details were simply smoothed out."
The research revealed that the thermodynamic conditions for storms—instability and moisture—were similar across all simulations. The critical difference lay in how well the models represented the low-level valley winds that provide the final trigger for convection.
"For forecasting extreme precipitation in complex mountainous terrain, increasing resolution from 3 kilometres to 1 kilometre can yield noticeable forecast improvements," said Robert Plant, Professor of Meteorology at the University of Reading and corresponding author of the study. "The 1-kilometre grid enables the model to better simulate the intricate flow structures within valleys that govern where and when the most dangerous storms develop. This is relevant not just for Qinghai but for mountain valleys in many parts of the world."
The findings have implications for operational forecasting. While running ultra-high-resolution models across entire continents remains computationally demanding, the researchers propose a more targeted approach.
"Our goal is practical," Su explained. "We want to provide forecasters in Qinghai and similar mountainous regions with more precise tools. By running higher-resolution 'on-demand' forecasts when and where dangerous storms are anticipated—essentially zooming in on high-risk areas within broader operational models—we may be able to issue heavy precipitation warnings earlier and more accurately."
The study also highlighted a limitation of traditional "convective parameterization" schemes—mathematical formulas that approximate storm development. In simulations where these schemes were active, weak rainfall began too early, followed by a delay in the main storm, effectively disrupting the model's timing.
"Using a convection parameterization scheme led to premature removal of early atmospheric instability," Plant noted. "This delayed the real storm and reduced its intensity in the simulation. When we let the model represent convection directly at high resolution, the timing and magnitude aligned more closely with observations."
While the study analysed one event in depth with supporting evidence from a second, the researchers suggest the underlying mechanisms are likely applicable more broadly.
"This is about understanding how valley circulations develop—how air moves up slopes during the day and drains down at night—and how these flows can contribute to storm triggers," Plant added. "Better representation of these wind patterns in models supports better predictions."
This approach, the researchers suggest, could strengthen disaster prevention efforts in mountainous regions globally, from the Andes to the Alps, the Himalayas to the Rockies.
Advances in Atmospheric Sciences
The Benefits of Kilometre-scale Simulations for Extreme Summertime Precipitation in the Eastern Valleys of Qinghai
7-Mar-2026