谈玲,刘巧,夏景明. 2024. 融合DEM与FY-4A数据的ECMWF预报产品深度学习订正方法[J]. 气象学报, (0):-, doi:[doi]
融合DEM与FY-4A数据的ECMWF预报产品深度学习订正方法
Research on the revised deep learning method of ECMWF forecast product based on DEM and FY-4A data
投稿时间:2023-04-26  修订日期:2024-01-30
DOI:
中文关键词:  数值天气预报,误差订正,深度学习,多源数据融合,注意力机制
英文关键词:Numerical weather prediction, Error correction, Deep learning, Multi-source data fusion, Attention mechanism
基金项目:国家重点基础研究发展计划2021YFB2901990
作者单位邮编
谈玲 南京信息工程大学计算机学院 210044
刘巧 南京信息工程大学人工智能学院 210044
夏景明* 南京信息工程大学人工智能学院 210044
摘要点击次数: 4
全文下载次数: 4
中文摘要:
      精准的数值天气预报是精细化气象公共服务和商业服务的重要前提。欧洲中期天气预报中心(European Center for Medium Weather Forecasting, ECMWF)预报产品在全球被广泛采用,但始终存在系统性预报误差。针对数值气象预报中的误差问题和多源数据融合中的非线性映射等问题,本文设计了一个ECMWF数值预报产品的深度学习订正模型((Numerical Forecast Correction Network,NFC-Net)。NFC-Net引入FY-4A卫星数据、数字高程模型数据(Digital Elevation Model, DEM)和ERA5历史实况数据订正预报结果,利用多源数据空间分辨率对齐模块、时空特征提取模块解决多源异构数据特征的提取与融合问题,并通过UNet网络实现ECMWF预报产品的订正。为了评估所提算法性能,利用NFC-Net对ECMWF产品中的2 m温度和10 m风速两个天气要素开展订正试验,并将试验结果与ECMWF预报结果、ANO 方法订正结果、Convlstm方法订正结果、Fuse-CUnet方法订正结果和ERA5实况结果进行对比。试验结果显示NFC-Net模型订正的2 m温度和10 m风速的均方根误差(Root Mean Squared Error,RMSE)分别较ECMWF预报产品下降了49.71%和50.86%。表明在NFC-Net模型可以充分利用多源数据有效改善复杂地形条件下的订正结果。本文所提方法和模型可用于订正ECMWF预报结果,提升数值天气预报的精度。
英文摘要:
      Accurate numerical weather prediction is an important prerequisite for refined meteorological public and commercial services. ECMWF forecast products are widely used around the world, whereas systematic forecast errors always exist. As a correction of numerical prediction products, multi-source data fusion can effectively reduce prediction errors, which is also a typical high-dimensional nonlinear mapping problem. Due to the heterogeneity of geographic data, ground truth data, and satellite data, it is necessary to establish a mechanism to fully extract and utilize effective information from these data, while avoiding noise and redundancy of information. In recent years, deep learning methods have been extensively applied to data post-processing problems in meteorological field. Aiming at the errors in numerical weather prediction and the nonlinear mapping problem in multi-source data fusion, this thesis designs a correction deep learning model NFC-Net for numerical prediction products ECMWF, which mainly includes a multi-source data spatial resolution alignment module, a spatiotemporal feature extraction module, and a UNet correction module. NFC-Net optimizes and corrects the forecast results by integrating multi-source data such as FY-4A satellite data, DEM, and ERA5 historical truth data, and utilizes multi-source data spatial resolution alignment module and spatiotemporal feature extraction module to achieve the feature extraction and fusion for multi-source heterogeneous data. At the same time, this paper also proposes a spatial resolution alignment algorithm based on convolutional neural networks (UPS-MSR algorithm) and a dual self-attention mechanism (DSA). The UPS-MSR algorithm uses upsampling and multi-scale residual networks to achieve grid alignment of meteorological and geographic data with different resolutions, which can effectively avoid information loss. The DSAConvlstm network embedded with DSA module can balance the spatiotemporal correlation and element correlation when extracting features from high-dimensional meteorological information. To evaluate the performance of proposed method NFC-Net, correction experiments on the two weather elements of the 2 m temperature and 10 m wind speed in the ECMWF products were carried out, and the results are compared with the ECMWF forecast results, ANO, Convlstm, Fuse-CUnet and ERA5. The experiments showed that the root mean square error (RMSE) of the 2 m temperature and 10 m wind speed corrected by the NFC-Net model decreased by 49.71% and 50.86%, respectively, compared to the ECMWF forecast product. The experimental results indicate that the introduction of high-resolution DEM data in the NFC-Net model can obviously optimize the land surface process of the model, and the correction effect is more pronounced under complex terrain conditions. The use of FY-4A satellite data enable the model to obtain more three-dimensional information during correction. The application of DSA module can make the model pay more attention to variables that have strong correlation with correction elements, thereby significantly improving the quality of correction. The proposed method can prospectively be applied in the correction of ECMWF forecast results, promoting the accuracy of numerical weather prediction.
查看全文   查看/发表评论  下载PDF阅读器