SGER: Real Time Multi-model Superensemble Forecasts for Atlantic Hurricanes of 1999

点击翻译按钮获取中文摘要

The award is being made as a Small Grant for Exploratory Research. Dr. Krishnamurti will extend some intriguing preliminary work on the use of multi-model superensemble forecasts for the prediction of hurricane track and intensity. The method relates the past performance of the multiple models to an observed "analysis" state using a simple regression procedure. All of the multi-model results are regressed with respect to this analysis for a control period and then the statistics are applied to a forecast period and a weighted combination is produced which is the superensemble. In the initial work, the superensemble has always outperformed the individual model forecasts. Under this award, Dr. Krishnamurti will carry out the superensemble forecasts in real-time as a further test of the method. After the 1999 operational hurricane season is over, he and his colleagues will analyze and evaluate the 1999 superensemble forecasts and explore the efficacy of using statistical methods other than simple regression in their superensemble forecasts.