Tropospheric Tomography In Subtropical Regions By Incorporating FY-4A And GNSS Water Vapor Data

Water vapor is an important component of the terrestrial atmosphere and mainly exists in the lower troposphere.Although water vapor accounts for a small proportion of the atmosphere,it varies greatly in space and time.Given that a considerable amount of latent heat is released or absorbed during phase change processes,water vapor plays an essential role in the global energy balance,climate change processes,and the formation and evolution of disastrous weathers.Water vapor retrieval by Global Navigation Satellite System(GNSS)has the advantages of all-weather real-time observation,low cost and high precision,and is an important means of modern water vapor detection.GNSS water vapor tomography technology can effectively obtain the three-dimensional distribution of water vapor with high precision and high spatial and temporal resolution,which has become a research focus of GNSS meteorology.In addition,satellite-based water vapor detection technology can provide a wide range of water vapor images,especially geostationary satellites have the advantages of wide spatial coverage and good temporal continuity,making them an important data source for global water vapor monitoring research.In this study,we explore the tropospheric tomography in subtropical regions by integrating Fengyun-4A(FY-4A)and GNSS water vapor data,and conduct water vapor tomography experiments using the observation data of Continuously Operating Reference Stations(CORS)to obtain the three-dimensional distribution field of atmospheric water vapor density in Hunan Province with high spatial and temporal resolution,and verified the feasibility of this technique.In addition,based on the tomographic water vapor products,the multi-parameter neural network prediction model of rainfall area is constructed in this study,which further explores the application potential of water vapor tomography technology in rainfall prediction.The main work and innovation points of this study are as follows:(1)This study provides a comprehensive assessment of the accuracy of FY-4A PWV products for the period from January 2019 to January 2020using radiosonde,GNSS and the Fifth Generation Reanalysis Dataset(ERA5)from the European Centre for Medium-Range Weather Forecasts water vapor data.The overall characteristics,spatial distribution,monthly variation and daily variation of FY-4A precipitable water vapor(PWV)accuracy are verified in detail,respectively,and the results show that 1)the overall accuracy of FY-4A PWV is better than 4 mm,with severe dry bias at high water vapor content;2)the root mean square error(RMSE)decreases with increasing latitude,and the relative error(R-RMSE)is the opposite;3)in summer,the RMSE is larger than that in winter,and 4)the RMSE is slightly smaller in daytime than in nighttime.The evaluation results validate the capability of FY-4A PWV for further applications in meteorological monitoring and can provide important references for its subsequent wide application.(2)The traditional system of equations of the stratification model with only empirical constraints added has the problem of discomfort.In this study,based on the method of water vapor tomography with additional external constraints,we propose to construct a water vapor tomography model incorporating FY-4A PWV as additional constraints by taking full advantage of the high spatial and temporal resolution of FY-4A geostationary satellite water vapor data.To verify the feasibility of this model in subtropical water vapor monitoring studies,this study conducted a water vapor stratification experiment using Hunan CORS(HNCORS)observation data in June 2020 to obtain the three-dimensional distribution of atmospheric water vapor with 30 min and 15 min resolution,and then analyzed the accuracy of the stratification results using radiosonde and ERA5 data.The results show that the average RMSE of water vapor density of the conventional model at 30 min resolution is 1.63 and 2.77g/m~3(1.77 g/m~3 and 2.70 g/m~3 at 15 min resolution)compared with the radiosonde data and ERA5 data,while the average RMSE of water vapor density of the new model proposed in this study is 1.18 and1.71g/m~3(1.19 g/m~3 and 1.76 g/m~3 at 15 min resolution),with the accuracy improvement of 27.6%and 38.3%(32.77%and 34.81%at 15 min resolution),respectively.In the vertical direction,the new model significantly improves the water vapor density results in the near-surface(0～3 km)and upper layers(above 6 km)compared with the conventional model.(3)Based on the water vapor products with high spatial and temporal resolution obtained by tomography technology,this study analyzes the spatial and temporal variation of rainfall processes and finds that PWV and various meteorological parameters are highly correlated with rainfall,and further constructs a short-time rainfall area prediction model using Back Propagation Neural Network(BPNN)to conduct rainfall prediction experiments for the Hunan province region.The experimental results show that the prediction accuracy of the model is 0.82,and the false alarm rate is 0.42,and the model can accurately predict more than 94%of the moderate or above rainfall events,which verifies the feasibility of the model for large-scale rainfall prediction.