###
DOI:
本文二维码信息
人工智能技术在重庆临近预报业务中的初步研究与应用
顾建峰1, 周国兵2, 刘伯骏2, 聂磊3, 张亚萍2, 张勇2, 吴胜刚2
(1.重庆市气象局;2.重庆市气象台;3.百度智能云事业部)
Study on Artificial Intelligence Technology and its Application to Chongqing Operational Nowcasting
gu jian feng1, zhou guo bing2, liu bo jun2, nie lei3, zhang ya ping2, zhang yong2, wu sheng gang2
(1.Chongqing Meteorological Bureau;2.Chongqing Meteorological Observatory;3.Department of AI and Cloud Technology,Baidu Inc.)
摘要
相似文献
本文已被:浏览 32次   下载 86
投稿时间:2019-11-22    修订日期:2020-07-17
中文摘要: 从大量研究工作中可知,人工智能(AI)技术可以用于改进强对流天气临近预报。重庆市气象局与百度公司合作,基于2008-2018年的天气雷达数据,采用TrajGRU技术,初步建立了三维雷达回波智能临近预报模型。结合地面观测资料和U-net技术,对雷暴大风和冰雹进行智能识别。将以上技术用于降水、雷暴大风和冰雹的临近预报业务中。与SWAN产品相比,AI产品在回波演变预报方面有明显优势。雷达回波智能预报能力随时间降低的速度低于传统外推方法,对长时效雷达回波预报明显提升。降水智能预报较SWAN降水预报有一定程度提高。在2019年业务试运行过程中,雷暴大风、冰雹智能预报模型提前量分别为77 min和37 min。AI技术在重庆的强对流天气临近预警业务中的研究与应用已初见成效。
Abstract:Previous researches showed that the application of artificial intelligence (AI) techniques could substantially improve severe convective weather nowcasting capabilities. Cooperating with Baidu Inc., Chongqing Meteorological Bureau preliminarily implemented the intelligent nowcasting model of three-dimensional radar echo using TrajGRU technique based on the radar data of 2008-2018. Together with the ground observation data and the U-net technique, the intelligent identification model of thunderstorm gale and hail were established. AI techniques above were used to operational thunderstorm gale, hail and precipitation nowcasting. Compared to the operational products of the Severe Weather Automatic Nowcast (SWAN) system, AI products provided obvious improvement in the prediction of echo evolution. The intelligent prediction ability of radar echo decreased with time at a lower speed than that of the traditional extrapolation method, and the prediction for longer lead times was improved. Verification showed that AI precipitation forecast had a higher skill than SWAN to a certain extent. During 2019 trial operational application period, the intelligent forecast model indicated thunderstorm gale(hail) 77min(37min) ahead in average. The research and application of AI techniques in the early warning of severe convective weather in Chongqing have achieved initial results.
文章编号:201911220411     中图分类号:    文献标志码:
基金项目:重庆市基础研究与前沿探索项目(cstc2018jcyjAX0434)、重庆市气象局创新团队项目(ZHCXTD-201803)
引用文本:
顾建峰,周国兵,刘伯骏,聂磊,张亚萍,张勇,吴胜刚,0.[en_title][J].Meteor Mon,():-.
gu jian feng,zhou guo bing,liu bo jun,nie lei,zhang ya ping,zhang yong,wu sheng gang,0.Study on Artificial Intelligence Technology and its Application to Chongqing Operational Nowcasting[J].Meteor Mon,():-.