ERA5再分析数据在微波辐射计数据反演中的初步应用

为了研究ERA5再分析资料在微波辐射计数据反演中的应用效果，利用探空和ERA5作为训练样本分别训练得到两个神经网络，对微波辐射计观测数据进行反演。并以探空资料作为参考分别统计基于探空和ERA5再分析资料反演得到的数据的相关系数、绝对偏差和均方根误差，用于评估ERA5再分析资料在微波辐射计数据反演中的应用效果。结果表明：基于ERA5和探空反演的温度廓线与探空的相关系数分别为0.988、0.99，绝对偏差分别为2.402 ℃、2.607 ℃。反演水汽与探空的相关系数分别为0.975、0.979，绝对偏差分别为0.412 g/m3、0.369 g/m3。反演的相对湿度相关系数分别为0.663、0.696，绝对偏差分别为15.587%、13.976%。反演的积分水汽总量相关系数分别为0.959、0.968，绝对偏差分别为0.258 cm、0.227 cm。从结果来看，基于ERA5资料反演的温度、水汽资料与利用探空反演的结果相差较小，温度廓线反演效果优于相对湿度廓线。

Preliminary Application of ERA5 Reanalysis Data to Retrieval Method of Microwave Radiometer

The ground based microwave radiometer is employing passive remote sensing techniques to achieve real time retrieval of atmospheric temperature, humidity profiles and related parameters. It possesses advantages of high spatiotemporal resolution, unmanned operation, and continuous functionality, gradually emerging as a potent supplementary method for remote sensing of atmospheric temperature, humidity, and cloud liquid water profiles both domestically and internationally. However, the inversion methods of microwave radiometer data rely on local historical radiosonde data and are restricted by factors such as the distribution of radiosonde stations. Consequently, its utilisation in areas lacking radiosonde stations is limited. To investigate the application effectiveness of ERA5 reanalysis data in microwave radiometer data retrieval, two neural networks are trained separately using radiosonde and ERA5 data as training samples. These networks are used to invert the microwave radiometer observation data. Subsequently, utilising radiosonde data as a reference, the correlation coefficient, absolute deviation, and root mean square error of the data inverted based on radiosonde and ERA5 reanalysis data are computed to assess the application effectiveness of ERA5 reanalysis data in microwave radiometer data retrieval. The results are indicating that the temperature profile inversions based on ERA5 and radiosonde data are exhibiting correlation coefficients of 0.988 and 0.99, with absolute deviations of 2.402 ℃ and 2.607 ℃, respectively. The water vapour density inversions are showing correlation coefficients of 0.975 and 0.979, with absolute deviations of 0.412 g/m3 and 0.369 g/m3. The relative humidity inversions are exhibiting correlation coefficients of 0.663 and 0.696, with absolute deviations of 15.587% and 13.976%. Furthermore, the integrated water vapour total inversions are presenting correlation coefficients of 0.959 and 0.968, with absolute deviations of 0.258 cm and 0.227 cm. Comparison of the microwave radiometer inversions based on radiosonde and reanalysis data is revealing closely aligned temperature profiles and integrated water vapour statistics; however, noticeable overestimations are observed in relative humidity. Analysis of error distribution with altitude is indicating relatively minor changes in root mean square errors for both radiosonde and ERA5 data. Temperature and humidity profile errors from both radiosonde and ERA5 retrievaled are performing better at lower altitudes, gradually increasing with altitude. The temperature profile error derived from radiosonde retrievals is notably smaller compared to ERA5 based inversions, exhibiting a distinct difference in the variation of relative humidity errors with altitude. This study is currently focusing solely on the application effectiveness of ERA5 data near the Changchun radiosonde station. To comprehensively validate the utility of ERA5 in microwave radiometer data retrieval, further research is essential in other radiosonde stations, including regions devoid of radiosonde data. Future research is necessary to investigate the application and correction methodologies of ERA5 data in microwave radiometer data retrieval in these areas.