New AI model brings breakthroughs in five-day regional weather forecasting
Peer-Reviewed Publication
Updates every hour. Last Updated: 31-Jul-2025 04:11 ET (31-Jul-2025 08:11 GMT/UTC)
A novel deep learning–based framework that dramatically improves the accuracy of forecasts, even when data are limited. The task involved using past atmospheric variables to predict five key surface weather indicators—including temperature, wind, and precipitation—every 6 hours for the next 5 days.
As the climate becomes warmer on average, it makes intuitive sense that we will see more hot days and we've had predictions of this for some time. However, the duration of heatwaves — how many days in a row exceed a temperature that is unusually hot for a given region — can be very important for impacts on humans, livestock and ecosystems. Predicting how these durations will change under a long-term warming trend is more challenging because day-to-day temperatures are correlated — tomorrow's temperatures have a dependence on today's temperature. This study takes this effect into account, along with the warming seen in current and historical observations and projected for the future by climate models for a wide range of land regions. Not only do the heatwave durations increase, but each additional increment of warming causes a larger increase in the typical length of long heat waves. In other words, if the next decade brings as much large-scale warming as a previous decade, the additional increase in heatwave durations would be even larger than we've experienced so far.
Low carbon fuel policies are intended to reduce greenhouse gas (GHG) emissions from transportation. However, rigid carbon intensity (CI) accounting procedures in current policies may limit CI responsiveness across candidate sites and facilities. This study examines how low carbon fuel programs capture or overlook spatial variability and net electricity production in biofuel carbon intensity, influencing crediting outcomes and fuel selling prices.