基于CLDAS和ConvLSTM的海南岛区域网格温度短临预报  

Nowcasting of Grid Temperature in Hainan Island Based on CLDAS and ConvLSTM

作　　者：罗小川 唐荣年[1] 宁雨珂 李创[1] Luo Xiaochuan;Tang Rongnian;Ning Yuke;Li Chuang(School of Mechanical and Electrical Engineering,Hainan University,Haikou 570228,China)

机构地区：[1]海南大学机电工程学院,海南海口570228

出　　处：《海南大学学报（自然科学版）》2021年第3期297-305,共9页Natural Science Journal of Hainan University

基　　金：海南省重点研发计划(ZDYF2020041)。

摘　　要：利用中国气象局陆面数据同化系统实时产品数据集(CLDAS),结合ConvLSTM网络提出了新的网格温度短临预报方法,通过实验实现了海南岛区域网格温度的1 h分辨率短临预报.在模型测试当中,12 h连续预报的网格误差超过2℃的比率约11.5%,验证了本次研究所提出的模型应用于网格温度短临预报的可能性.In the report,combined with the ConvLSTM network,the China Meteorological Administration’s land surface data assimilation system real-time product dataset(CLDAS)was used to propose a new grid temperature short-term forecast method for the Hainan Island area.With the simulation experiment,a one-hour resolution short-term forecast of the grid temperature in Hainan Island was achieved,in which the grid error of the 12-hour continuous forecast exceeds 2℃at a rate of about 11.5%.The results verified the possibility of applying the model proposed in this study to short-term grid temperature prediction.

关 键 词：ConvLSTM CLDAS 2 m温度 短临预报 

分 类 号：S716.1[农业科学—林学]