基于Himawari-8卫星数据的青藏高原大气可降水量反演算法研究  

Precipitable Water Vapor Retrieval Using Himawari-8Satellite Observations over Tibetan Plateau

作　　者：江金昊 刘海磊[1] 王乙竹 张升兰 JIANG Jinhao;LIU Hailei;WANG Yizhu;Zhang Shenglan(College of Electronic Engineering,Chengdu University of Information Technology,Chengdu 610225,China)

机构地区：[1]成都信息工程大学电子工程学院,四川成都610225

出　　处：《成都信息工程大学学报》2022年第5期494-500,共7页Journal of Chengdu University of Information Technology

基　　金：国家自然科学基金资助项目(42030107)

摘　　要：大气可降水量对降水形成、水循环、物质能量交换、天气和气候变化均有重要影响。青藏高原的水汽分布、输送特征是重要的高原气象科学问题。新一代静止气象卫星能够获取更高时间、空间和光谱分辨率的观测数据,为高时空分辨率PWV探测提供良好的机遇。利用Himawari-8卫星的水汽和分裂窗区通道亮温、全球天气预报系统(GFS)的PWV预报场和其他辅助数据(卫星观测角、海拔高度、经纬度和时间信息),构建一种基于神经网络的青藏高原PWV快速反演算法。结果表明,神经网络模型估算的PWV与地基全球定位系统(GPS)PWV的相关系数、均方根误差和偏差分别为0.957、1.33 mm和-0.004 mm。相比之下,未包含GFS PWV模型反演的PWV相关系数、均方根误差和偏差分别为0.943、1.52 mm和0.01 mm,意味着引入GFS PWV预报场可以有效提高模型的反演精度。该方法适用于其他低水汽环境下的PWV反演,也可用于其他静止气象卫星数据。Precipitable water vapor(PWV)plays an important role in precipitation formation,water cycle,energy exchange,weather and climate change.The characteristics of water vapor distribution and transportation over the Tibetan Plateau are important issues in plateau meteorology.The new generation of geostationary meteorological satellites can obtain observation data with higher temporal,spatial and spectral resolution,which provides a good opportunity for PWV retrieval with high temporal and spatial resolution.In this study,a fast neural network based algorithm for PWV retrieval over the Tibetan Plateau was proposed.The inputs of model mainly include the water vapor absorption and split window channels brightness temperature of Advanced Himawari Imagers(AHI)onboard Himawari-8,Global Forecast System(GFS)PWV forecasts and other auxiliary data(satellite observation angle,elevation,longitude and latitude,and time information).The correlation coefficient(R),root mean square error(RMSE)and Bias of PWV estimated by neural network model with GPS(Global Positioning System)PWV are 0.957,1.33 mm and-0.004 mm,respectively.In contrast,the R,RMSE and Bias of the model without GFS PWV are 0.943,1.52 mm and 0.01 mm.This indicates that the introduction of GFS PWV can improve the inversion accuracy.The proposed method is suitable for PWV retrieval in other area with low water vapor content and can be also used for other geostationary meteorological satellite data.

关 键 词：青藏高原 Himawari-8 大气可降水量 神经网络 GFS 

分 类 号：P407[天文地球—大气科学及气象学]