摘要: |
目的 通过分析机器学习方法与传统方法各自的优势和缺陷,可以融合机器学习方法和传统方法各自的优势,有效克服经验模型的缺陷,同时提高降雨估算精度,增强了降雨预测的可靠性,从而预防自然灾害。方法 文章指出将人工智能等相关算法引入GNSS大气可降水量的反演过程成为了当前研究的重要方向,特别是将深度学习神经网络等模型运用到GNSS大气可降水量的探测实验中不仅能够提高实验运算效率及反演精度,还可以在一定程度上减少工作量。结果 根据系统组合的PPP模式实验数据显示运用北斗系统进行大气可降水量反演具有可行性,反演结果与探空数据有很好的一致性,且将机器学习方法运用到多系统组合的PPP模型的精度更高,能够为天气预测提供技术支撑。结论 通过对GNSS反演大气可降水量发展的分析,归纳了当前利用GNSS反演大气可降水量的研究方法,对GNSS的大气水汽含量反演从理论到技术应用进行了梳理。同时研发基于约束下的GPS系统和我国北斗卫星系统的联合大气水汽反演系统对于进一步推动GNSS技术在气象和农业方面的应用具有重要作用。 |
关键词: GNSS 北斗卫星系统 机器学习 大气可降雨系统 |
DOI:10.7621/cjarrp.1005-9121.20220905 |
分类号:S-1 |
基金项目:2021年宁夏自治区科技创新团队柔性引进人才项目“‘北斗+’土壤水分和植被含水量监测仪器设备研发及应用”(2021RXTDLX14);中央级公益性科研院所基本科研业务费专项项目资助(1610132020014) |
|
PRINCIPLE AND APPLICATION PROGRESS OF GNSS SIGNAL ESTIMATION SYSTEM FOR ATMOSPHERIC RAINFALL |
Sun Yidan1, Guo Zhonghua1, Yang Changzhi1, Mao Kebiao2, Xin Xiaoping2, Wang Yifan1, Wang Ping1
|
1.School of Physics and Electronic-Engineering, Ningxia University, Yinchuan 750021, Ningxia, China;2.Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China
|
Abstract: |
By analyzing the advantages and disadvantages of machine learning methods and traditional methods, the shortcomings of empirical models can be overcome effectively by integrating the advantages of machine learning methods and traditional methods, which could improve the accuracy of rainfall estimation to help prevent natural disasters. The introduction of artificial intelligence and other related algorithms into the inversion process of GNSS atmospheric precipitation had become an important direction of current research. In particular, the application of deep learning neural networks to GNSS atmospheric precipitation detection experiments could not only improve the computational efficiency and inversion accuracy, but also reduced the workload to a certain extent. According to the PPP model experimental data of system combination, it was feasible to use BeiDou system to retrieve atmospheric precipitable water. The retrieval results were in good agreement with radiosonde data, and the application of machine learning method to multi system combination PPP model had higher accuracy, which could provide technical support for weather prediction. In summary, through the analysis of the development of GNSS inversion of atmospheric precipitable, the current research methods of using GNSS to retrieve atmospheric precipitable are summarized, and the inversion of atmospheric water vapor content of GNSS is sorted out from theory to technical application. Meanwhile the research and development of a combined atmospheric water vapor retrieval system based on the constrained GPS system and BeiDou satellite system plays an important role in further promoting the application of GNSS technology in meteorology and agriculture field. |
Key words: GNSS BeiDou satellite system machine learning atmospheric rainfall system |