摘要
为提高地基微波辐射计反演大气温/湿度廓线的精度,提出了一种直接利用高垂直分辨率探空资料与地基微波辐射计观测亮温训练反演温/湿度廓线BP神经网络方案.基于地基微波辐射计的观测特点,提出了一种基于微波辐射计地面观测资料和探空资料的观测亮温综合质量控制方案,利用质量控制后的观测亮温训练BP神经网络(OBS-BP),并与基于MonoRTM辐射传输模式模拟亮温训练BP神经网络(SIM-BP)的方法进行了对比.结果表明,OBS-BP反演温度廓线的均方根误差随高度逐渐增大,范围为0.62~2.81 K,偏差范围为-0.67~0.43 K,相关系数随高度的升高逐渐减小,变化范围为0.92~0.99;相对湿度廓线的均方根误差在0~4.75 km随高度升高而增大,在4.75 km以上随高度升高而减小,范围为8.21%~24.37%,偏差范围为-3.87%~4.54%,相关系数随高度升高逐渐减小,变化范围为0.13~0.94.将OBS-BP和SIM-BP反演高时间频次的温/湿度廓线的效果进行了对比,得出OBS-BP的反演结果能更好地反映对流层内大气温/湿度演变过程,相对于利用SIM-BP的反演结果,OBS-BP反演温/湿廓线在各个高度层上均优于SIM-BP,与探空资料具有更好的一致性,更适用于实际观测中地基微波辐射计温/湿度廓线的反演.
In order to improve the accuracy of ground-based microwave radiometer on temperature and relative humidity retrieval,a training method of BP neural network(OBS-BP)to retrieve temperature andrelative humidity profiles was presented in this study.The observed brightness temperature,quality control by observed microwave and radiosonde data were used to train OBS-BP,and the retrieval results were compared with SIM-BP using simulated brightness temperature calculated by the monochromatic radiative transfer model.The results showed that the root mean square error(RMS)of temperature profiles retrieved by OBS-BP increased with the altitude and range of 0.62-2.81 K,the bias(B)of temperature profiles range of-0.67-0.43 K;the correlation coefficient(R)of temperature decreased with the altitude and range of 0.92-0.99;the RMS of relative humidity profiles increased with the altitude in 0-4.75 km and decreased with the altitude above 4.75 km,and the RMS range of 8.21%-24.37%,B range of-3.87%-4.54%;the R of humidity decreased with the altitude and range of 0.13-0.94;the retrieval results of OBS-BP were superior to SIM-BP at all levels,and the high-frequency temperature and humidity profiles retrieved by OBS-BP exhibited a better consistency with the troposphere temperature and humidity evolution process.A comparison between the retrieval results of the two training methods showed that OBS-BP had a better consistency with the radiosonde data,which was more suitable for the actual observation of the ground-based microwave radiometer temperature and relative humidity profiles retrieval.
作者
樊旭
吴肖燕
曲宗希
张北斗
张文煜
Fan Xu;Wu Xiao-yan;Qu Zong-xi;Zhang Bei-dou;Zhang Wen-yu(Key Laboratory for Semi-Arid Climate Change with the Ministry of Education,College of Atmospheric Sciences,Lanzhou University,Lanzhou,730000,China;Key Laboratory for Cloud Physics,Chinese Academy of Meteorological Sciences,Beijing 100081,China)
出处
《兰州大学学报（自然科学版）》
CAS
CSCD
北大核心
2019年第5期587-596,共10页
Journal of Lanzhou University(Natural Sciences)
基金
国家自然科学基金项目(41875085,41630421)
中国气象科学研究院“西北区域人工影响天气能力建设”项目(ZQC-R18208)
中央高校基本科研业务费专项资金项目(LZUJBKY-2018-k03).
关键词
地基微波辐射计
BP神经网络
质量控制
温/湿廓线反演
ground-based microwave radiometer
BP neural network
quality control
retrieval of temperature and humidity profile
作者简介
通信联系人:张文煜(1964-),男,甘肃陇南人,教授,博士,博士研究生导师,e-mail:zhangwy@lzu.edu.cn,研究方向为大气探测和大气物理
