资源环境科技发展态势分析平台

In the United States, severe weather poses a threat to society, producing tornadoes and hail that can result in hundreds of casualties and billions of dollars in damages. Fortunately, skillful predictions of severe weather for short lead times of 0-8 days and longer lead times exceeding 1 month have been realized. However, this leaves a forecast gap at subseasonal to seasonal lead times of 2-5 weeks, when early-action decision making by stakeholders is typically made. Here we develop an empirical prediction model that fills this gap during March-June when severe weather is most prevalent across the United States. We demonstrate skillful weekly forecasts of opportunity with lead times of 2-5 weeks of environmental parameters favorable to severe weather, as well as actual tornado and hail activity. To attain this skill, we use as a predictor the current state of active phases of the Madden-Julian Oscillation, known to have physical teleconnections with future weather over the United States. The model has significant skill in regions such as the Plains and the Southeast, providing stakeholders with valuable extended forewarning. Plain Language Summary In the United States, severe thunderstorms produce tornadoes and large hail, responsible for hundreds of deaths and injuries and many billions of dollars in damages on average each year. Because of these devastating impacts, there is a keen interest to accurately forecast when and where severe thunderstorms are likely to occur. While meteorologists and computer models do reasonably well in forecasting severe thunderstorm activity up to a week in advance, their forecasts are less reliable in the 2- to 5-week time frame. In our study, we develop a technique that can accurately forecast severe thunderstorm activity in this time frame by using knowledge of the current state of weather in the tropics. These accurate, extended forecasts offer valuable forewarning to both the general public and stakeholders of when and where potentially deadly severe thunderstorm activity is likely to occur.