Retrieval Of Near Sea Surface Air Temperature Based On Multi Source Microwave Radiometer Data

Microwave radiation has the advantages of wear through clouds,all day long and allweather.It has been widely used in atmospheric,marine and terrestrial microwave remote sensing and other fields.Near sea surface air temperature is an important parameter on description of the marine environment,and is also one of the projects in the basic observation.There is no inversion algorithm and products on studying the near sea surface air temperature.So,based on the multi-source microwave radiometer(AMSR2,TMI and SSM/I)data,buoy measured data and ECMWF reanalysis data,using multiple regression algorithm and BP neural network method,we got the near sea surface air temperature.The main contents and conclusions are as follows:(1)Systematically expounds the transmission principle of Microwave Radiation transfer equation,and briefly introduced the basic knowledge of microwave radiation.There are many types of data used in this paper,radiometer(AMSR2,SSM/I and TMI)data,buoy data(Tao,Rama and Pirata)and ECMWF reanalysis data.Data sources and pre-processing method are given in detail.(2)Inversion of near sea surface air temperature.Based on the multivariate statistical regression method,using AMSR2,SSM/I and TMI data,we got the global near sea surface temperature.With buoy data inspection,the average deviation of AMSR2 inversion near sea surface air temperature is 0.025?,root mean square error is 1.22?,correlation coefficient is 0.99;The average deviation of SSM/I inversion of near surface air temperatures is 0.24?,root mean square error is 1.19?,correlation coefficient is 0.99;The average deviation of TMI inversion of near surface air temperature is-0.0054?,root mean square error is 0.70?,correlation coefficient is 0.94.Based on BP neural network algorithm,get the global near sea surface temperature inversion.Checked by buoy data,the average deviation of AMSR2 inversion near sea surface air temperature is 0.17 ?,root mean square error is 1.19?,correlation coefficients is 0.99;The average deviation of SSM/I retrieve the global near surface temperature is 0.12?,root mean square error is 1.06?,correlation coefficient is 0.99;The average deviation of TMI inversion global near surface temperature is 0.38?,root mean square error is 0.85?,correlation coefficient is 0.92.The results of BP neural network algorithm is superior to the linear regression method,linear regression method is not accurately at low temperature.(3)Fusion of near sea surface air temperature.Near surface air temperature inversion results are fused by optimal interpolation algorithm,improve the spatial coverage.The fusion results of near surface air temperature is tested by buoy data,the deviation is-0.05?,the root mean square error is 0.39?.The correlation coefficient is 0.95.The RMS error of every month is about 0.4 degrees Celsius,the correlation coefficient is about 0.9.