Considered a grand challenge in weather prediction, the high-resolution forecasting of precipitation up to two hours ahead supports real-world socioeconomic needs of many sectors reliant on weather-dependent decision-making. However, say the researchers, even today's weather predictions driven by powerful numerical weather prediction (NWP) systems that solve physical equations struggle to generate high-resolution predictions for short lead times under two hours.
Now, using generative modeling to make detailed and plausible predictions of future radar based on past radar, the researchers say their approach - called "nowcasting" - fills the performance gap in this crucial time interval, and is essential for sectors like water management, agriculture, aviation, emergency planning, and outdoor events. Using a systematic evaluation by more than 50 expert meteorologists, the researchers' generative model ranked first for its accuracy and usefulness in 89% of cases against two competitive methods.
"Advances in weather sensing have made high-resolution radar data–which measures the amount of precipitation at ground level–available at high frequency (e.g., every 5 mins at 1 km resolution)," say the researchers. "This combination of a crucial area where existing methods struggle and the availability of high-quality data provides the opportunity for machine learning to make its contributions to nowcasting."
The researchers focused on nowcasting rain: predictions up to two hours ahead that capture the amount, timing, and location of rainfall. The use of generative modeling enabled them to make detailed and plausible predictions of future radar based on past radar - conceptually a problem of generating radar movies.
With such methods, say the researchers, they can both accurately capture large-scale events, while also generating many alternative rain scenarios (known as ensemble predictions), allowing rainfall uncertainty to be explored. The researchers were especially interested in the ability of these models to make predictions on medium to heavy-rain events, which are the events that most impact people and the economy, and say they show statistically significant improvements in these regimes compared to competing