Assessment And Improvement Of GRAPES Ensemble Forecast Perturbation Method On Landing Typhoon

Typhoon occurs on tropical oceans and has the characteristics of high strength,the sudden,there are a lot of uncertainty on typhoon forecasting.Forecast of heavy rainfall brought by landing typhoon is one of the hardest forecasting problems all over the world.The forecast uncertainty is better provided by the Ensemble forecast,but there need further study on landing typhoon of ensemble forecast perturbation technology.This study revolves around the core issue of landing typhoon Ensemble forecast skill-initial and model perturbation method,This study use the data of GRAPES-REPS,includes July to September 2015 and 2016 which make use of ensemble statistics diagnostic analysis method and synoptic analysis method,Compare the effect of Ensemble Transform Kalman Filter(ETKF),Multi-Scale Blending(MSB)and Stochastic Perturbation of Physics Tendency(SPPT)to assess the forecast performance of the track,landing time,location,strength and rainfall of landing typhoon.On the basis,improving the ETKF observation data vertical density experiment and SPPT stochastic perturbation spacial scale sensitivity experiment to forecast No.1510 Linfa.Thus providing meaningful basis for landing typhoon ensemble forecast theoretical method.The mean results are:(1)Results of GRAPES region ensemble forecast based on ETKF show that ensemble forecast average effect of typhoon track is better than CRTL.And GRAPES-REPS forecast on typhoon track dispersion is small before landing,and the predictability is high.The average error of the 24 h and 48 h ensemble mean landing location is closer to observation.Landing location distribution show that there's no obvious error on GRAPES-REPS forecast of landing location.In strength forecasting,typhoon is universally early weak and late strong at the landfall time,the lowest pressure and highest speed of wind are all in the ensemble member forecast,which means that ensemble forecast includes the truth value and discrete ensemble member also shows the uncertainty of forecast.Different orders of rainfall AROC grade is between 0.56 and 0.76,which means the grade of the rainfall in landing area is significant and have forecastreference value.(2)Based on the GRAPES-REP forecast system of MSB initial perturbation and SPPT mode perturbation show the 0 to 12 h track ensemble discrete of landing typhoon is relatively concentrated,reliability is high,and landing location forecasting effect is good.Observation landing are falling in the collection members landing area.In addition,except NO.1604 Nida set the average prediction error is greater than CTRL,the set of the remaining three typhoon is obviously better than CTRL,but the differences between different typhoon landing site of the error distribution.The forecast error characteristics of the intensity set are similar to that of the initial value disturbance of ETKF,which also has the tendency of early weak and late strong at the landfall time.Different orders of rainfall AROC grade is 0.58 to 0.81.(3)The forecast results set typhoon moving path contrast ETKF initial perturbation method and MSB initial perturbation and SPPT mode perturbation method,the typhoon track set dispersion and evolution of ensemble average method is more reasonable than the ETKF,48 h landing location prediction error decreases nearly 13km;two methods of actual landing time forecasting than landing time,and MSB the initial perturbation and SPPT perturbation method forecast model is closer to the observation,the landing time accuracy increased 36%;AROC and BS grading precipitation score showed that two kinds of disturbance precipitation forecast method of rainstorm precipitation technique,the other is MSB initial perturbation and SPPT perturbation method slightly better forecast skill.(4)Through the improvement of ETKF initial perturbation gain matrix calculation program,increase the vertical density observation results show that the 850 hPa height field,U component and V component of wind wind and temperature field sequential probability score(CRPS)report rate and probability prediction of leakage(Outlier)improved significantly;Not only can reduce the root mean square error value of each element,but also improve the collection of dispersion.Experimental results of ensemble prediction of Typhoon No.1510 "Linfa" show that increasing the density of observation data can significantly improve the ensemble mean path prediction error,especially during the period from 12 to 48 h.In precipitation forecast,the technique of heavy precipitation forecast is improved.(5)Based on SPPT random perturbation type of four spatial scales(750km,500 k,300km and 250km)showed that the sensitivity of the test,the typhoon path set forecast degree on SPPT random disturbance type four spatial scales have a certain sensitivity,which 300 km the spatial scale of the SPPT type stochastic disturbances of No.1510 typhoon "Linfa" set minimum average error.For precipitation forecast,scale of 300 km test improved best effect,can significantly improve the precipitation forecast center,precipitation from the main sub grid scale precipitation,large scale grid scale precipitation variation.