Improvement of the Ensemble Methods in the Dynamical–Statistical–Analog Ensemble Forecast Model for Landfalling Typhoon Precipitation

Abstract

 The Dynamical–Statistical–Analog Ensemble Forecast model for landfalling typhoon precipitation (the DSAEF_LTP model) identifies tropical cyclones (TCs) from history data that are similar to a target TC, and then assembles the precipitation amounts and distributions of those identified to obtain those of the target TC. Two original ensemble methods in the DSAEF_LTP model, mean and maximum, tend to under- and over-forecast TC precipitation, respectively. In addition, these two methods are unable to forecast precipitation at stations beyond their maxima. To overcome the shortcomings and improve the forecast performance of the DSAEF_LTP model, the following five new ensemble methods are incorporated: optimal percentile, fuse, probability matching mean, equal difference-weighted mean, and TSAI (Tropical cyclone track Similarity Area Index)-weighted mean. Then, model experiments for landfalling TCs over China in 2018 are conducted to evaluate the forecast performance of the DSAEF_LTP model with the new ensemble methods. Results show that the overall performance of the optimal percentile (the 90th percentile) ensemble method is superior, with the false alarm rate lower than that of the original ensemble methods. As compared to five operational numerical weather prediction models, the improved DSAEF_LTP model shows advantages in predicting accumulated rainfall, especially with the rainfall of over 250 mm. When implementing the experiments, above results, however, it is found that the model forecast performance varies, depending on the type of TC tracks. That is, the accumulated rainfall forecast for westbound TCs is significantly better than that of northbound TCs. To address this issue, different schemes are used to forecast the accumulated rainfall of TCs with the two different track types. The precipitation forecast performance for westbound and northbound TCs, using the 90th percentile and the probability-matched ensemble mean ensemble method, respectively, is much better than that using a single ensemble method for all the TCs.