Research On Statistical Forecasting Technology Of Precipitation Dynamics Of Tropical Cyclones Landed In China

Track forecasting is the greatest progress over past decades in numerical weather prediction(NWP)regarding the tropical cyclones(TCs).The landfalling TC precipitation prediction capability of NWP models is still very limited.This article study on combining dynamical and statistical methods of landfalling tropical cyclone precipitation prediction.Summarizes the primary coverage and conclusions as follows to show:(1)A new method named track-similarity-based Landfalling Tropical cyclone Precipitation(LTP)Dynamical-Statistical Ensemble Forecast(LTP_DSEF)model has been developed.And The application of LTP_DSEF model for 21 TCs with accumulated LTP ? 100 mm in South China during 2012-2016 reveals that the new method shows a better performance than the three NWP models(EC,GFS and T639)in predicting LTP.For ?50mm LTP forecasts,the threat score(TS)of LTP_DSEF model's results is superior to those of the three NWP models.Analysis of the values of the LTP_DSEF model's parameters for three best schemes shows that,the initial time being closest to the impact time i.e.12:00 UTC the day before the first day with precipitation on land and the ensemble parameter being the maximum are both beneficial to stable good forecast.(2)Introducing TC intensity similarity factor into LTP_DSEF model and conducting the same experiment in South China.The results reveal that the performance of LTP_DSEF model improved with the factor used and more superior to three NWP models.For three thresholds(?250mm,?100mm and ?50mm)LTP forecasts,the TS of LTP_DSEF model all improved and for ?100mm improved most evident(0.1805 to 0.2221)in modeling sample experiment.At the same time,the highest TS of three NWP models is 0.1681.The independent sample experiment results reveal that the promotion of TS for three thresholds LTP forecast is not significant but still overmatch three NWP models.For ?250mm LTP forecast,the TS of LTP_DSEF model is 0.025,and three NWP models are all 0.Analysis of the values of the LTP_DSEF model's parameters with the TC intensity similarity factor introduced shows that,the initial time and the ensemble parameter are same.(3)Increasing the number of several parameters of LTP_DSEF model and apply it to China for 99 TCs with accumulated LTP ? 100 mm during 2004-2016.According to the experiment result,a set of fixed parameters has been chosen to be the best parameters for China.Analysis of the values of the best parameters shows that,the initial time choose 12:00 UTC of the same day better than the day before and the change of ensemble parameter don not show promotion temporarily.The TS for three thresholds LTP forecasts of LTP_DSEF model perform stability in both modeling sample experiment and independent sample experiment.