基于U-Net改进的日平均2 m气温订正方法

doi:10.3969/j.issn.0255-8297.2025.01.004

应用科学学报 ›› 2025, Vol. 43 ›› Issue (1): 51-65.doi: 10.3969/j.issn.0255-8297.2025.01.004

基于U-Net改进的日平均2 m气温订正方法

王冰轮¹, 方巍^1,2,3

1. 南京信息工程大学计算机学院, 江苏南京 210044;
2. 南京气象科技创新研究院中国气象局交通气象重点开放实验室, 江苏南京 210041;
3. 苏州大学江苏省计算机信息处理技术重点实验室, 江苏苏州 215000

收稿日期:2024-07-17 出版日期:2025-01-30 发布日期:2025-01-24
通信作者: 方巍，教授，博导，研究方向为大数据分析、计算机视觉、数据挖掘、信息检索、AI+气象、软件工程。E-mail:hsfangwei@sina.com E-mail:hsfangwei@sina.com
基金资助:
国家自然科学基金（No.42075007，No.42475149）；中国气象局流域强降水重点开放实验室开放研究基金（No.2023BHR-Y14）；中国气象局交通气象重点开放实验室开放研究基金项目（北极阁基金项目）（No.BJG202306）资助

Improved Daily Average 2 m Temperature Correction Method Based on U-Net

WANG Binglun¹, FANG Wei^1,2,3

1. School of Computer, Nanjing University of Information Science & Technology, Nanjing 210044, Jiangsu, China;
2. Key Laboratory of Transportation Meteorology of China Meteorological Administration, Nanjing Joint Institute for Atmospheric Sciences, Nanjing 210041, Jiangsu, China;
3. Provincial Key Laboratory for Computer Information Processing Technology, Soochow University, Suzhou 215000, Jiangsu, China

Received:2024-07-17 Online:2025-01-30 Published:2025-01-24

摘要/Abstract

摘要： 针对数据订正常用的深度学习模型U-Net中不能充分学习空间特征以及图像细节信息丢失的问题，提出了S-CUnet 3+模型。S-CUnet 3+采取以下两个措施对U-Net进行改进：一是将原模型与能够学习图片全局特征的Swin Transformer有机结合起来；二是引入多尺度连接操作。模型还采用了预训练与微调的训练策略针对多个预报步长同时订正。7个预报步长的日平均2 m气温预报值订正的实验结果表明，S-CUnet 3+模型对所有预报步长的预报都有明显的订正效果，其中24 h预报步长的订正效果最好，平均绝对误差和均方根误差分别下降了50.64%和49.25%，且相比于基于历史资料的模式距平积分预报订正、分位数回归、岭回归、U-Net、CU-Net、Dense-CUnet和RA-UNet这7种订正方法，S-CUnet 3+取得了更好的订正效果。

关键词: 数据订正, 深度学习, Swin Transformer, 预训练, 微调

Abstract: In response to the limitations of the widely utilized deep learning model U-Net, which is unable to adequately learn spatial features and suffers from the loss of image detail information, the S-CUnet 3+ model has been proposed. S-CUnet 3+ enhances U-Net in two ways: firstly, it integrates the original model with the Swin Transformer, enabling it to learn global features of images, and secondly, it introduces multi-scale connection operations. The model also adopts pre-training and fine-tuning strategies to correct multiple forecast lead times simultaneously. Experimental results of correcting daily average 2 m temperature forecasts across seven lead times show that the S-CUnet 3+ model has a significant correction effect for all lead times, with the best correction effect at the 24-hour lead time. The mean absolute error and root mean square error are reduced by 50.64% and 49.25%, respectively. Moreover, S-CUnet 3+ outperforms seven existing correction methods: anomaly numerical-correction with observations, quantile regression, ridge regression, U-Net, CU-Net, Dense-CUnet, and RA-UNet.

Key words: data correction, deep learning, Swin Transformer, pre-training, finetuning

中图分类号:

TP301.6

王冰轮, 方巍. 基于U-Net改进的日平均2 m气温订正方法[J]. 应用科学学报, 2025, 43(1): 51-65.

WANG Binglun, FANG Wei. Improved Daily Average 2 m Temperature Correction Method Based on U-Net[J]. Journal of Applied Sciences, 2025, 43(1): 51-65.

参考文献

[1] 程胡华, 王益柏, 赵亮, 等. 相似偏差订正法在短期温度预报中的应用研究[J]. 气象研究与应用, 2020, 41(3): 21-26. Cheng H H, Wang Y B, Zhao L, et al. Application of similar deviation correction method in short-term temperature forecast [J]. Journal of Meteorological Research and Application, 2020, 41(3): 21-26. (in Chinese)
[2] 张延彪, 陈明轩, 韩雷, 等. 数值天气预报多要素深度学习融合订正方法[J]. 气象学报, 2022, 80(1): 153-167. Zhang Y B, Chen M X, Han L, et al. Multi-element deep learning fusion correction method for numerical weather prediction [J]. Acta Meteorologica Sinica, 2022, 80(1): 153-167. (in Chinese)
[3] Li X Y, Li Z. Evaluation of bias correction techniques for generating high-resolution daily temperature projections from CMIP6 models [J]. Climate Dynamics, 2023, 61(7): 3893-3910.
[4] Tong Y, Gao X J, Han Z Y, et al. Bias correction of temperature and precipitation over China for RCM simulations using the QM and QDM methods [J]. Climate Dynamics, 2021, 57(5): 1425-1443.
[5] Enayati M, Bozorg-Haddad O, Bazrafshan J, et al. Bias correction capabilities of quantile mapping methods for rainfall and temperature variables [J]. Journal of Water and Climate Change, 2021, 12(2): 401-419.
[6] 范江琳, 陈朝平, 曹萍萍, 等. 四川地区ECWMF模式晴雨预报订正试验研究[J]. 暴雨灾害, 2022, 41(1): 58-65. Fan J L, Chen C P, Cao P P, et al. Experimental study on revision of ECWMF clear-rainy forecast in Sichuan region [J]. Torrential Rain and Disasters, 2022, 41(1): 58-65. (in Chinese)
[7] Han M, Wu Q S, Liu H J, et al. Correction method by introducing cloud cover forecast factor in model temperature forecast [J]. Frontiers in Earth Science, 2023, 11: 1099344.
[8] 邱学兴, 王东勇, 陈宝峰. T639模式预报系统误差统计和订正方法研究[J]. 气象, 2012, 38(5): 526-532. Qiu X X, Wang D Y, Chen B F. The statistics and correction of T639 model forecast system errors [J]. Meteorological Monthly, 2012, 38(5): 526-532. (in Chinese)
[9] 林健玲, 金龙, 彭海燕. 区域降水数值预报产品人工神经网络释用预报研究[J]. 气象科技, 2006, 34(1): 12-17. Lin J L, Jin L, Peng H Y. Application of numerical forecast products to regional rainfall forecasting by artificial neural network [J]. Meteorological Science and Technology, 2006, 34(1): 12-17. (in Chinese)
[10] Cho D J, Yoo C H, Im J H, et al. Comparative assessment of various machine learningbased bias correction methods for numerical weather prediction model forecasts of extreme air temperatures in urban areas [J]. Earth and Space Science, 2020, 7(4): 740-757.
[11] Watt-Meyer O, Brenowitz N D, Clark S K, et al. Correcting weather and climate models by machine learning nudged historical simulations [J]. Geophysical Research Letters, 2021, 48(15): e2021GL092555.
[12] 李德伦, 肖志祥, 谢宁新, 等. 机器学习中混合特征选择对模式预报广西春夏气温的订正研究[J]. 成都信息工程大学学报, 2023, 38(5): 602-609. Li D L, Xiao Z X, Xie N X, et al. A study on the adjusting spring and summer surface air temperature of ECMWF model by a hybrid feature selection method in machine learning of Guangxi [J]. Journal of Chengdu University of Information Technology, 2023, 38(5): 602-609. (in Chinese)
[13] 方巍, 庞林, 王楠, 等. 人工智能在短临降水预报中应用研究综述[J]. 南京信息工程大学学报(自然科学版), 2020, 12(4): 406-420. Fang W, Pang L, Wang N, et al. Survey on the application of artificial intelligence in precipitation nowcasting [J]. Journal of Nanjing University of Information Science & Technology (Natural Science Edition), 2020, 12(4): 406-420. (in Chinese)
[14] Hu Y F, Yin F K, Zhang W M. Deep learning-based precipitation bias correction approach for Yin-He global spectral model [J]. Meteorological Applications, 2021, 28(5): 2032-2045.
[15] You X X, Liang Z M, Wang Y Q, et al. A study on loss function against data imbalance in deep learning correction of precipitation forecasts [J]. Atmospheric Research, 2023, 281: 106500- 106510.
[16] Zhu Y H, Zhi X F, Lyu Y, et al. Forecast calibrations of surface air temperature over Xinjiang based on U-net neural network [J]. Frontiers in Environmental Science, 2022, 10: 1011321.
[17] Hess P, Boers N. Deep learning for improving numerical weather prediction of heavy rainfall [J]. Journal of Advances in Modeling Earth Systems, 2022, 14(3): 2765-2775.
[18] Han L, Chen M X, Chen K K, et al. A deep learning method for bias correction of ECMWF 24-240 h forecasts [J]. Advances in Atmospheric sciences, 2021, 38(9): 1444-1459.
[19] Fei T H, Huang B H, Wang X, et al. A hybrid deep learning model for the bias correction of SST numerical forecast products using satellite data [J]. Remote Sensing, 2022, 14(6): 1339-1357.
[20] Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation [C]//Medical Image Computing and Computer-Assisted Intervention-MICCAI, 2015: 234-241.
[21] Liu Z, Lin Y T, Cao Y, et al. Swin Transformer: hierarchical vision Transformer using shifted windows [C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV), 2021: 9992-10002.
[22] Huang H M, Lin L F, Tong R F, et al. UNet 3: a full-scale connected UNet for medical image segmentation [C]//IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020: 1055-1059.
[23] 胡莹莹, 庞林, 王启光. 基于深度学习的7-15 d温度格点预报偏差订正[J]. 应用气象学报, 2023, 34(4): 426-437. Hu Y Y, Pang L, Wang Q G. Application of deep learning bias correction method to temperature grid forecast of 7-15 days [J]. Journal of Applied Meteorological Science, 2023, 34(4): 426-437. (in Chinese)
[24] Han K, Wang Y H, Chen H T, et al. A survey on vision transformer [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(1): 87-110.
[25] Muñoz-Sabater J, Dutra E, Agustí-Panareda A, et al. ERA5-Land: a state-of-the-art global reanalysis dataset for land applications [J]. Earth System Science Data, 2021, 13(9): 4349-4383.
[26] He D X, Zhou Z M, Kang Z P, et al. Numerical studies on forecast error correction of GRAPES model with variational approach [J]. Advances in Meteorology, 2019(1): 2856289.
[27] Hersbach H, Bell B, Berrisford P, et al. The ERA5 global reanalysis [J]. Quarterly Journal of the Royal Meteorological Society, 2020, 146(730): 1999-2049.

基于U-Net改进的日平均2 m气温订正方法

Improved Daily Average 2 m Temperature Correction Method Based on U-Net

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	胡汪鑫, 成英超, 何玉林, 黄哲学, 蔡占川. 脉冲神经网络基准测试及类脑训练框架性能评估[J]. 应用科学学报, 2025, 43(1): 169-182.
[2]	王莹笑, 杨彦红, 谭云峰. 面向视频的人脸特征计算方法[J]. 应用科学学报, 2025, 43(1): 137-153.
[3]	刘珈宁, 张思佳, 张正龙, 王祎涵, 安宗诗. 基于多跳机制的扩散图谱推荐模型[J]. 应用科学学报, 2025, 43(1): 123-136.
[4]	贾志洋, 许兆, 冷艳梅, 闻新, 龚浩宇. 基于并行优化CBAM的轻量级故障诊断模型[J]. 应用科学学报, 2025, 43(1): 94-109.
[5]	段新涛, 白鹿伟, 徐凯欧, 张萌, 保梦茹, 武银行, 秦川. 基于卷积神经网络的轻量高效图像隐写[J]. 应用科学学报, 2025, 43(1): 80-93.
[6]	魏伟, 金成功, 杨龙, 周默, 孟祥主, 冯慧. 基于预训练大语言模型的实体关系抽取框架及其应用[J]. 应用科学学报, 2025, 43(1): 20-34.
[7]	唐意恒, 王永雄, 王哲, 张晓理. 基于子领域适应和时空学习的脑电情感识别[J]. 应用科学学报, 2024, 42(6): 1016-1026.
[8]	贺加贝, 周菊香, 甘健侯, 吴迪, 温晓宇. 基于多任务学习的课堂表情分类模型[J]. 应用科学学报, 2024, 42(6): 947-961.
[9]	栗莎, 王永雄, 王哲, 陈旭, 何嘉欣. 融合局部和全局特征的铸件缺陷检测[J]. 应用科学学报, 2024, 42(5): 757-768.
[10]	王子驰, 李斌, 冯国瑞, 张新鹏. 数字图像隐写分析综述[J]. 应用科学学报, 2024, 42(5): 723-732.
[11]	王娜, 朱会娟, 宋香梅, 冯霞. 一种基于领域自适应的智能合约安全分析框架[J]. 应用科学学报, 2024, 42(4): 585-597.
[12]	李子杰, 张姝, 欧阳昭相, 王俊, 吴迪. 基于场因子分解的xDeepFM推荐模型[J]. 应用科学学报, 2024, 42(3): 513-524.
[13]	沈坤烨, 周晓飞, 费晓波, 陈雨中, 张继勇, 颜成钢. 基于边缘感知深度残差网络的带钢表面缺陷检测[J]. 应用科学学报, 2023, 41(6): 978-988.
[14]	张春森, 朱江乐, 张学芬, 刘旭东, 史书. 融合自注意力机制与生成对抗网络的DEM空洞填充[J]. 应用科学学报, 2023, 41(5): 789-800.
[15]	黄显培, 孟庆祥. 基于多特征卷积神经网络的哨兵二号影像地物分类[J]. 应用科学学报, 2023, 41(5): 766-776.