-
Abstract:
The landfall of tropical cyclones in the eastern part of China falls in the category of small probability events. Constructing a step function with intervals adequately divided can help reflect the non-linear distribution of conditional probability for a landfall event. For the prediction of landfall event probability, factors applying the step function in transformation are superior to the standardized factors that are linearly related. The prediction scheme discussed in the work uses transformation factors of step function to formulate prediction models for tropical cyclones making landfalls in eastern China, through screening with non-linear correlative ratios and REEP analysis. Classified models for statistic-synoptics, statistic -climatology and statistic-dynamics have been constructed using initial field data and numerical prediction output. Forecasting skills have been improved due to ensemble of predictions using these classified models. As shown in forecasting evaluations and experiments, the scheme is capable of predicting tropical cyclones that make landfalls in eastern China.
-
-