CLKA-LPO:A CNN-LSTM-KAN neural network for lightning potential and flash rate prediction based on atmospheric physical parametersOA
Lightning poses significant risks to infrastructure and safety.Existing forecasting models depend on LPI data but often ignore the impact of atmospheric parameters on lightning.These methods are simple and fail to predict complex lightning events effectively.This paper introduces a lightning forecasting model,CLKA-LPO,which uses physical parameters and a CNN-LSTM-KAN architecture to enhance prediction accuracy.The model adopts a CNN-LSTM-KAN architecture,integrating KANs(Kolmogorov–Arnold Networks)to replace traditional convolutional neural networks,and improve its ability to generalize,thus accurately forecasting lightning in given areas.The paper uses t-tests and Pearson correlation to assess the significance of correlation coefficients for 103 meteorological parameters,identifying 16 key predictors.Then,a machine learning–based thunderstorm forecasting model is developed to monitor lightning activity.To analyze the relationship between 103 parameters and the lightning flash rate,19 key physical parameters are selected.A deep learning model,CLKA-LPO,is then used to forecast ground and total lightning flashes.Evaluations show the CLKA-LPO model enhances the ground lightning prediction accuracy by 30%and total lightning prediction by 20%over traditional models like CNN,LSTM,Transformer,and CNN-LSTM.The CLKA-LPO model effectively correlates physical parameters with lightning events,offering precise forecasts for both ground and total lightning flashes,ensuring dependable lightning forecasts.
Hengyue Chen;Tongtong Xu;Yihan Du;Qiyuan Yin;Yihui Zhu;Zhiyong Luo;Yang Song;Min Xia;Shubin Zhao
School of Automation,Naning University of Information Science and Technology,Nanjing,China Collaborative Innovation Center on Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing,China School of Mathematical,Physical and Computational Sciences,University of Reading,Reading RG66UR,Berkshire,United KingdomSchool of Automation,Naning University of Information Science and Technology,Nanjing,China Collaborative Innovation Center on Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing,ChinaSchool of Computer Science School of Cyber Science and Engineering,Nanjing University of Information Science and Technology,Nanjing,ChinaKey Laboratory of Lightning,China Meteorological Administration,Beijing,China Guangdong Climate Center,Guangzhou,ChinaSchool of Automation,Naning University of Information Science and Technology,Nanjing,China Collaborative Innovation Center on Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing,ChinaKey Laboratory of Lightning,China Meteorological Administration,Beijing,ChinaSchool of Automation,Naning University of Information Science and Technology,Nanjing,China Collaborative Innovation Center on Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing,ChinaSchool of Automation,Naning University of Information Science and Technology,Nanjing,China Collaborative Innovation Center on Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing,ChinaNanjing LES Information Technology Co.,Ltd.Nanjing,China
天文与地球科学
Lightning potential predictionLightning flash rate predictionAtmospheric physical parametersKolmogorov-arnold(KAN)CLKA-LPO
《Atmospheric and Oceanic Science Letters》 2026 (3)
P.28-35,8
supported by the National Natural Science Foundation of China[grant number 12202210]the Open Grants of the Key Laboratory of Lightning[grant number 2023KELL-B006]the Startup Foundation for Introducing Talent of Nanjing University of Information Science and Technology[grant number 2022r095]the China Postdoctoral Science Foundation[grant number 2024M751481]。
评论