Font Size: a A A

Research On Ground-based GPS Water Vapor Retrieval And Shortterm Rainfall Prediction

Posted on:2021-10-25Degree:MasterType:Thesis
Country:ChinaCandidate:K GaoFull Text:PDF
GTID:2480306473482674Subject:Surveying and Mapping project
Abstract/Summary:PDF Full Text Request
As an important part of the atmosphere,water vapor has a complicated spatial distribution,which plays an important role in the formation and evolution of extreme weather such as rainfall,thunderstorms,and typhoons.At present,due to the characteristics of all-weather,real-time,high accuracy,low operating cost,and not affected by extreme weather,the ground-based GPS water vapor retrieval has become an effective means of detecting water vapor content.Therefore,the use of ground-based GPS water vapor retrieval technology to accurately monitor the spatial and temporal changes of water vapor and establish a high-precision rainfall prediction model are of great significance for meteorological analysis and forecasting in various regions.In this thesis,the GPS observation data from Hong Kong CORS Network is used to water vapor inversion,conduct relevant research on issues such as atmospheric weighted mean temperature modeling,GPS-derived PWV spatial interpolation,and establishment of SVM-based rainfall prediction model.The main research contents are as follows:1.The atmospheric weighted mean temperature at the location of the Hong Kong sounding station was calculated by ERA-Interim reanalysis data and GPT2w model respectively,and the accuracy of the two methods was verified.The results shows that T8)calculated using the ERA-Interim reanalysis data is more accurate than the result obtained by the GPT2w model,and its accuracy is very stable throughout the year,while T8)obtained by GPT2w model has obvious seasonal effect.2.Based on the ERA-Interim reanalysis data,a localized model of the atmospheric weighted mean temperature in Hong Kong was established.After accuracy verification,it was found that the RMS of the model was 1.732K.At the same time,the results shows that the localized model of atmospheric weighted average temperature can support the long-term use of this region.3.The trend surface,inverse distance weighted,ordinary kriging,and Co-Kriging method were used to interpolate the GPS/PWV of stations in different weather conditions.It is found that the trend surface method has the highest interpolation accuracy whether it is rainy or rainless.The elevation factor should be considered when performing regional PWV interpolation.When the spatial variation degree of PWV is higher,the interpolation accuracy is lower.The regional PWV distribution map can better reflect the spatial and temporal changes of water vapor,and can intuitively analyze the movement and transportation of water vapor during rainfall,providing a certain reference for regional weather analysis and rainfall forecast.4.The changes of PWV,ZTD,temperature,and relative humidity during rainfall are analyzed.The result shows that PWV will increase rapidly before rain,and the probability of rainfall before and after the maximum value of PWV is large,however,the rainfall does not necessarily occur at the peak of PWV,so the comprehensive analysis should be carried out in combination with the PWV change rate.The change of ZTD time series before and after rainfall is similar to that of PWV,and it can also indicate the rainfall.Before the rain,there will be a continuous drop in temperature,and the moment of maximum rainfall also corresponds to the moment of minimum temperature.The relative humidity will rise rapidly before the rain.When the rain occurs,the relative humidity can reach more than 90%.5.A rainfall prediction model based on SVM was established,which was trained by using data sets of different years and combinations of different predictors.It was found that increasing the training samples can improve the prediction accuracy.The prediction model established using PWV,temperature,and relative humidity has the highest accuracy,the experimental evaluation using a two-year database shows a true detection rate of 85.12%,a false alarm rate of 18.39%,and an overall accuracy of 81.58%.The accuracy of the ZTD prediction model is equivalent to that of the PWV prediction model.When the meteorological elements cannot be obtained,the ZTD prediction model can be used instead of the PWV prediction model.
Keywords/Search Tags:ERA-Interim reanalysis data, weighted mean temperature, GPS/PWV interpolation, Support Vector Machine, rainfall prediction model
PDF Full Text Request
Related items