首页 >  , Vol. , Issue () : -

摘要

全文摘要次数: 132 全文下载次数: 151
引用本文:

DOI:

10.11834/jrs.20210110

收稿日期:

2020-04-20

修改日期:

2020-09-29

PDF Free   EndNote   BibTeX
基于神经网络的太赫兹冰云探测反演算法研究
陈柯1, 张兰1, 张幼明2, 董杉彬1, 刘艳1, 吴琼3, 商建3
1.华中科技大学;2.武汉船舶通信研究所;3.国家卫星气象中心
摘要:

太赫兹频段在冰云探测上具有独特优势,但是目前的太赫兹冰云反演算法将不同种类冰相粒子(主要是冰和霰)视为冰粒子统一计算。本文根据冰云太赫兹辐射特性实现了一种预分类的神经网络算法,能够从太赫兹亮温中分别反演得到冰、霰两种粒子的统计参数和廓线分布。首先基于WRF数值模式和ATMS载荷真实观测的冰云霰数据构建了包含冰、霰粒子密度廓线的混合冰云数据库,然后使用DOTLRT辐射传输模式模拟183GHz-874GHz七个频段的星载太赫兹冰云探测亮温,最后开展冰云参数探测仿真试验,验证反演算法性能。仿真试验中反演得到的冰和霰的路径总量均方根误差分别为8.97 g/m2和10.9 g/m2,等效粒径均方根误差分别为7.54μm和25.38 μm,反演的冰、霰密度廓线也具有较高的精度。研究结果表明本文算法能够以较好的精度从多频太赫兹冰云探测亮温数据中分别反演得到冰、霰两种粒子的路径总量、等效粒径、等效云高和密度廓线,突破现有研究仅仅计算单一冰粒子的局限,更加符合冰云真实情况。

Research on Retrieval Algorithm of Terahertz Ice Cloud Sounding based on Neural Network
Abstract:

Terahertz band has unique advantages in ice cloud sounding application, but current terahertz ice cloud sounding research regard various ice particles (mainly ice and graupel) as the same kind of particles. In this paper, a pre-classified neural network algorithm based on terahertz radiation characteristics of ice cloud is proposed,which can separately retrieved the statistical parameters and density content profiles of ice particles and graupel particles from the multi-frequency terahertz brightness temperature data. Firstly, a hybrid ice cloud database including ice and graupel particle parameters is built based on WRF model and graupel profile data observed by ATMS. Then DOTLRT radiative transfer model is used to simulate the brightness temperature data of spaceborne terahertz ice cloud sounding covered seven frequency bands of 183GHz-874GHz. Finally, simulation experiments of ice cloud sounding are conducted to evaluate the performance of the proposed algorithm. In the simulation experiments, the average root mean square errors of the retrieved IWP and GWP are 8.97g/m2 and 10.9 g/m2 respectively, and the average root mean square errors of the retrieved I_Dme and G_Dme are 7.54μm and 25.38μm respectively, and the retrieved density profiles of ice and graupel particles also have high accuracy. The results indicate that the proposed algorithm can retrieve the total path amount, equivalent ice particle size, equivalent ice cloud height and density profile of ice and graupel particles respectively with high accuracy. It is more consist with the real condition of ice cloud than the current retrieval algorithm which only treats a single ice particle.

本文暂时没有被引用!

欢迎关注学报微信

遥感学报交流群 分享按钮