Rainfall Forecast And Typhoon Movement Analysis In Taiwan Based On GNSS-PWV

GNSS Precipitable Water Vapor(GNSS-PWV)can provide high-precision and high-temporal-resolution water vapor information for weather monitoring.The spatial and temporal variation characterization of GNSS-PWV is an important means for rainfall prediction and typhoon analysis.This paper presents a more comprehensive analysis and discussion of the spatial and temporal variation characteristics of PWV in Taiwan using PWV data inferred from GNSS stations in Taiwan,related meteorological data,topographic features of Taiwan,and monsoon systems.Based on GNSS-PWV information and joint conventional meteorological parameters,a new method of rainfall forecasting is proposed using Adaptive kernel learning based relevance vector machine(aRVM).Based on the spatio-temporal variation characteristics of GNSS-PWV and the spatial location relationship of GNSS stations during the 2014 Fung-wong typhoon,a new method for estimating the core parameters of typhoon motion was constructed.The experimental results show that:(1)By analyzing the spatial and temporal variation characteristics of PWV in Taiwan,it is found that the water vapor in different regions of Taiwan has significant annual cycle variation characteristics during different seasons,and the elevation,temperature,monsoon and latitude have obvious effects on the water vapor;Through the correlation analysis between PWV and rainfall,it is found that rainfall and PWV changes are closely related.(2)In GNSS-PWV for rainfall prediction:A new method of rainfall prediction is proposed by introducing a variety of conventional meteorological parameters to address the shortcomings of current conventional methods.The experiments show that the average correct rate of rainfall prediction is around 93%and the error rate is 25%.Compared with current conventional methods,the correct rate is improved by about 10-30%,and the false forecast is reduced by about 45-55%.And through the comparative analysis with BP-NN algorithm and SVM algorithm,we found that the rainfall forecasting method proposed in this paper still has obvious advantages.(3)In GNSS-PWV for typhoon analysis:By analyzing the spatio-temporal variation characteristics of PWV during the 2014 Fung-wong typhoon,it is found that there is a significant time-delayed relationship between the PWV peaks of different GNSS stations,which is highly consistent with the spatio-temporal characteristics of typhoon passage.A new method for typhoon moving speed and azimuth estimation is constructed based on this feature,and the estimated typhoon moving speed and azimuth in Taiwan are v=26.53±1.43km/h and ?=16.76±10.55°,which are highly similar to the typhoon moving speed 25.73 km/h and azimuth 21.5° calculated from the data given by Central Meteorological Observatory-Typhoon Network,which confirms the effectiveness of the new method.