Research On Sea Surface High Wind Speed Inversion Model Based On CYGNSS

Typhoon is closely related to the production and life of human society.As a rainfall system,it can provide sufficient rainwater for human use.However,typhoon is surprisingly destructive and easy to cause huge property losses.As a natural disaster,it is also very sudden.Typhoons with huge energy reserves can produce wind speeds of more than 17m/s,and even hurricane winds of 60m/s.The typhoon on the sea is more ferocious than on the ground.Its huge power can produce tens of meters of huge waves,which always threatens the life safety of operators such as marine operation,fishing,aquaculture,exploration and oil production.Therefore,it is an urgent task for us to measure the sea surface wind speed simply,quickly and accurately.Cyclone global navigation satellite system(CYGNSS)studies the characteristics of ocean surface to promote the prediction and tracking of tropical cyclones and improve the prediction of extreme weather.CYGNSS can receive the Global Navigation Satellite System(GNSS)signals reflected from the earth’s surface,and use these reflected signals to inverse the characteristics of the earth’s surface.The L1 level spaceborne data of CYGNSS can inverse the sea surface wind speed.The 10 meter referenced sea surface wind speed data(10)of ECMWF and NCEP are taken as the true wind speed,and the processed satellite data of CYGNSS are mapped to the true wind speed.In order to improve the accuracy of sea surface wind speed inversion in typhoon weather,two methods of wind speed inversion are proposed,that is,geophysical model function(GMF)established by delay-doppler map(DDM)and machine learning method to establish sea surface high wind speed inversion model.The wind speed inversion is carried out using the data during the typhoon,and finally the feasibility of the two CYGNSS sea surface high wind speed inversion methods is verified,thus,this paper makes up for the blank of the research on the inversion of sea surface wind speed greater than 20 m / s using CYGNSS data.The following are the main research contents of this paper:1)Aiming at the idea of the method proposed in this paper,the current research status of sea surface wind speed inversion of global satellite navigation system reflection measurement(GNSS-R)at home and abroad is introduced,which provides a theoretical basis for the experiment.2)By extracting the normalized bistatic radar cross section(NBRCS)from the L1 level data of CYGNSS satellite and matching the wind speed data,the GMF wind speed inversion model is established by DDM.In particular,different GMF models are established under the typhoon sea state in the wind speed range of 0-70 m / s,including the general GMF model without sample size control and the high wind speed GMF model established after randomly sampling samples to control the sample size of each wind speed range.The performance of high wind speed GMF model is better than that of ordinary GMF model,with bias increased by 96.97%,root mean square error(RMSE)increased by 46.17%,mean absolute error(MAE)increased by 49.74%,and standard deviation(STD)increased by 38.52%.It is verified that high wind speed GMF model is more suitable for wind speed inversion in typhoon weather.3)Three machine learning methods: the Support Vector Regression(SVR),the combination of Principal Component Analysis(PCA)and SVR(PCA-SVR),and Convolutional Neural Networks(CNN)are used to retrieve the wind speed of CYGNSS data with wind speed above 20m/s.Because the unbalanced sample size of each wind speed interval will lead to the deviation of the training model,the samples are randomly selected by the under sampling method to control the sample size of each wind speed interval and ensure the generalization of the training model.On the whole,CNN model has the best effect on inversing sea surface high wind speed,PCA-SVR is the second,and SVR is the worst.The MAE of CNN is 2.71 m/s and the RMSE is 3.8 m/s.compared with the SVR model,the MAE is increased by 33.90% and the RMSE is increased by 30.66%.