Temperature And Humidity Profiles Retrieval Based On Ground-based Microwave Radiometer With Multi-methods

Accurate regional numerical forecasting and meteorological disaster monitoring in megacities require atmospheric profiles with high accuracy and frequency.The atmospheric temperature and humidity profiles are the most basic atmospheric parameters,which have important significance in disaster warning,numerical prediction and other fields.They provide strong support for improving the prediction accuracy.Ground-based meteorological observation networks provide a variety of remote sensing observation equipment.How to give full play to the advantages of these equipment for atmospheric detection is one of the problems worth studying.In this paper,the ground-based MWR(Microwave Radiometer)in the meteorological observation network is the main instrument for retrieving atmospheric temperature and humidity profiles.With the existing problems,the paper combines a variety of forecast data,reanalysis data and different sensors as supplements to carry out the corresponding research and propose the regarding solutions.Aiming at obtaining atmospheric temperature and humidity profiles,retrieving algorithm at MWR sites without adjacent radiosondes,one-dimensional variational(1D＿Var)retrieving algorithm based on MWR,1D＿Var algorithm based on MWR with VPR-CFMCW(Vertical Pointing Radar with C-band Modulation Continuous Wave)were established.Using sounding data and reanalysis data from ECMWF(European Centre for Medium-Range Weather Forecasts)as true values,validation of temperature and humidity profiles was carried out,and the conclusions are followed:Retrieving models of temperature and relative humidity profiles for ground-based microwave radiometer without radiosondes co-located were established based on machine learning algorithm,which employing ERA5 reanalysis data after quality control and simulated brightness temperature as training samples.The validation shows that the quality control for reanalysis data can improve the retrieval accuracy of atmospheric temperature profile above500 m and relative humidity profile of 0-10 km.The proposed algorithm can better reflect the local weather characteristics of MWR sites,so as to improve the retrieval accuracy of the nearsurface temperature and upper atmosphere humidity profiles.The retrieval results are also superior to the ground-based MWR Level2 products.A one-dimensional variational atmospheric temperature and humidity profile retrieval algorithm was established at Nanjiao site in Beijing,using FNL as background field and groundbased microwave radiometer as observation vector.The observation deviation correction algorithm adopts a third-order Taylor expansion method considering cloud influence,which can improve the temperature and humidity detection accuracy under cloud conditions.Compared with the numerical model data,the temperature error of 0-2km and the humidity error of the middle and upper troposphere can be reduced by adding the ground-based MWR data.Comparing the 1D＿Var retrieval results with the neural network retrieval results,the temperature errors of the whole layer are reduced,especially in the middle and upper troposphere.Relative humidity error decreases at 1-8km.Compared with Level2 product of ground-based microwave radiometer,the root mean square error of temperature and relative humidity is reduced by 54.5% and 38.79% respectively.At Longmen Station in Guangzhou,the observations from ground-based microwave radiometer and VPR-CFMCW were combined as the observation vector.Using FNL 6-hour forecast data as the background field,the active and passive combined detection retrieval experiments were carried out.Compared with only using ground-based MWR as observation,the improvement of atmospheric temperature profile is small,while the root-mean-square error of atmospheric relative humidity profile is reduced by 14.68%,and the improvement of relative humidity is mainly reflected below 4Km.