基于卷积神经网络的雷达回波外推方法
|
摘要
雷达回波外推技术目前被广泛应用于临近预报中。针对传统雷达回波外推方法存在外推时效较短,对雷达资料数据利用率不高的问题,采取深度学习的方法,提出了一种基于输入的动态卷积神经网络(DCNN-I)模型。根据相邻时刻的雷达回波图像之间相关性强的特点,该网络模型中增加了动态子网络和概率预测层,建立了卷积核与输入图像的映射关系,使卷积核在网络测试阶段仍然能够根据输入雷达回波图像的不同而变化,增强了预测图像与输入图像之间的关联。以南京、杭州、厦门三地的雷达数据为样本进行实验,实验结果表明,与传统的雷达回波外推方法相比,所提方法能够获得更高的预测图像准确率,并且有效延长外推时效。
Extrapolation technique of weather radar echo possesses a widely application prospects in short-term nowcast.The traditional methods of radar echo extrapolation are difficult to obtain long limitation period and have low utilization rate of radar data. This problem is researched from deep learning perspective in this paper, and a new model named Dynamic Convolutional Neural Network based on Input(DCNN-I) was proposed. According to the strong correlation between weather radar echo images at adjacent times, dynamic sub-network and probability prediction layer were added, and a function was created that maped the convolution kernels to the input, through which the convolution kernels could be updated based on the input weather radar echo images during the testing. In the experiments of radar data from Nanjing, Hangzhuo and Xiamen,this method achieved higher accuracy of prediction images compared with traditional methods, and extended the limitation period of exploration effectively.
Extrapolation technique of weather radar echo possesses a widely application prospects in short-term nowcast.The traditional methods of radar echo extrapolation are difficult to obtain long limitation period and have low utilization rate of radar data. This problem is researched from deep learning perspective in this paper, and a new model named Dynamic Convolutional Neural Network based on Input(DCNN-I) was proposed. According to the strong correlation between weather radar echo images at adjacent times, dynamic sub-network and probability prediction layer were added, and a function was created that maped the convolution kernels to the input, through which the convolution kernels could be updated based on the input weather radar echo images during the testing. In the experiments of radar data from Nanjing, Hangzhuo and Xiamen,this method achieved higher accuracy of prediction images compared with traditional methods, and extended the limitation period of exploration effectively.
引文
[1]DOVIAK R J,ZRNIC D S.Doppler Radar&Weather Observations[M].Waltham,MA:Academic Press,2014:1-9.
[2]张沛源,杨洪平,胡绍萍.新一代天气雷达在临近预报和灾害性天气警报中的应用[J].气象,2008,34(1):3-11.(ZHANG PY,YANG H P,HU S P.Applications of new generation weather radar to nowcasting and warning of severe weather[J].Meteorological Monthly,2008,34(1):3-11.)
[3]OTSUKA S,TUERHONG G,KIKUCHI R,et al.Precipitation nowcasting with three-dimensional space-time extrapolation of dense and frequent phased-array weather radar observations[J].Weather and Forecasting,2016,31(1):329-340.
[4]RINEHART R E,GARVEY E T.Three-dimensional storm motion detection by conventional weather radar[J].Nature,1978,273(5660):287-289.
[5]李英俊,韩雷.基于三维雷达图像数据的风暴体追踪算法研究[J].计算机应用,2008,28(4):1078-1080.(LI Y J,HAN L.Storm tracking algorithm developmentbased on the three-dimensional radar image data[J].Journal of Computer Applications,2008,28(4):1078-1080.)
[6]LIANG Q Q,FENG Y R,DENG W J,et al.A composite approach of radar echo extrapolation based on TREC vectors in combination with model-predicted winds[J].Advances in Atmospheric Sciences,2010,27(5):1119-1130.
[7]FLETCHER T D,ANDRIEU H,HAMEL P.Understanding,management and modelling of urban hydrology and its consequences for receiving waters:a state of the art[J].Advances in Water Resources,2013,51(1):261-279.
[8]张亚萍,程明虎,夏文梅,等.天气雷达回波运动场估测及在降水临近预报中的应用[J].气象学报,2006,64(5):631-646.(ZHANG Y P,CHENG M H,XIA W M,et al.Estimation of weather radar echo motion field and its application to precipitation nowcasting[J].Acta Meteor Sinica,2006,64(5):631-646.)
[9]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Imagenet classification with deep convolutional neural networks[C]//Proceedings of the25th International Conference on Neural Information Processing Systems.[S.l.]:Curran Associates Inc.,2012,1:1097-1105.
[10]常亮,邓小明,周明全,等.图像理解中的卷积神经网络[J].自动化学报,2016,42(9):1300-1312.(CHANG L,DENG X M,ZHOU MQ,et al.Convolutional neural networks in image understanding[J].Acta Automatic Sinica,2016,42(9):1300-1312.)
[11]李倩玉,蒋建国,齐美彬.基于改进深层网络的人脸识别算法[J].电子学报,2017,45(3):619-625.(LI Q Y,JIANG J G,QIM B.Face recognition algorithm based on improved deep networks[J].Acta Electronica Sinica,2017,45(3):619-625.)
[12]Le CUN Y,HUANG F J.Large-scale learning with SVM and convolutional for generic object categorization[C]//CVPR'06:Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEEComputer Society,2006,1:284-291.
[13]张志强,刘黎平,王红艳.三维可视化技术在雷达三维组网产品显示中的运用[J].气象科技,2010,38(5):605-608.(ZHANGZ Q,LIU L P,WANG H Y.Application of 3D visualization technology to display of Doppler radar networking products[J].Metrological Science and Technology,2010,38(5):605-608.)
[14]LU G Y,WONG D W.An adaptive inverse-distance weighting spatial interpolation technique[J].Computers&Geosciences,2008,34(9):1044-1055.
[15]GLOROT X,BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[EB/OL].[2017-03-01].http://www.weblio.jp/redirect?url=http%3A%2F%2Fjmlr.org%2Fproceedings%2Fpapers%2Fv9%2Fglorot10a%2Fglorot10a.pdf&etd=3e03a9174d723be0.
[16]王改利,赵翠光,刘黎平,等.雷达回波外推预报的误差分析[J].高原气象,2013,32(3):874-883.(WANG G L,ZHAOC G,LIU L P,et al.Error analysis of radar echo extrapolation[J].Plateau Meteorology,2013,32(3):874-883.)
[17]UIJLENHOET R,POMEROY J H.Raindrop size distribution and radar reflectivity-rain rate relationships for radar hydrology[J].Hydrology&Earth System Sciences&Discussions,2001,5(4):3012-3018.
[18]SHI X,CHEN Z,WANG H,et al.Convolutional LSTM network:a machine learning approach for precipitation nowcasting[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems.Cambridge,MA:MIT Press,2015,1:802-810.
[2]张沛源,杨洪平,胡绍萍.新一代天气雷达在临近预报和灾害性天气警报中的应用[J].气象,2008,34(1):3-11.(ZHANG PY,YANG H P,HU S P.Applications of new generation weather radar to nowcasting and warning of severe weather[J].Meteorological Monthly,2008,34(1):3-11.)
[3]OTSUKA S,TUERHONG G,KIKUCHI R,et al.Precipitation nowcasting with three-dimensional space-time extrapolation of dense and frequent phased-array weather radar observations[J].Weather and Forecasting,2016,31(1):329-340.
[4]RINEHART R E,GARVEY E T.Three-dimensional storm motion detection by conventional weather radar[J].Nature,1978,273(5660):287-289.
[5]李英俊,韩雷.基于三维雷达图像数据的风暴体追踪算法研究[J].计算机应用,2008,28(4):1078-1080.(LI Y J,HAN L.Storm tracking algorithm developmentbased on the three-dimensional radar image data[J].Journal of Computer Applications,2008,28(4):1078-1080.)
[6]LIANG Q Q,FENG Y R,DENG W J,et al.A composite approach of radar echo extrapolation based on TREC vectors in combination with model-predicted winds[J].Advances in Atmospheric Sciences,2010,27(5):1119-1130.
[7]FLETCHER T D,ANDRIEU H,HAMEL P.Understanding,management and modelling of urban hydrology and its consequences for receiving waters:a state of the art[J].Advances in Water Resources,2013,51(1):261-279.
[8]张亚萍,程明虎,夏文梅,等.天气雷达回波运动场估测及在降水临近预报中的应用[J].气象学报,2006,64(5):631-646.(ZHANG Y P,CHENG M H,XIA W M,et al.Estimation of weather radar echo motion field and its application to precipitation nowcasting[J].Acta Meteor Sinica,2006,64(5):631-646.)
[9]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Imagenet classification with deep convolutional neural networks[C]//Proceedings of the25th International Conference on Neural Information Processing Systems.[S.l.]:Curran Associates Inc.,2012,1:1097-1105.
[10]常亮,邓小明,周明全,等.图像理解中的卷积神经网络[J].自动化学报,2016,42(9):1300-1312.(CHANG L,DENG X M,ZHOU MQ,et al.Convolutional neural networks in image understanding[J].Acta Automatic Sinica,2016,42(9):1300-1312.)
[11]李倩玉,蒋建国,齐美彬.基于改进深层网络的人脸识别算法[J].电子学报,2017,45(3):619-625.(LI Q Y,JIANG J G,QIM B.Face recognition algorithm based on improved deep networks[J].Acta Electronica Sinica,2017,45(3):619-625.)
[12]Le CUN Y,HUANG F J.Large-scale learning with SVM and convolutional for generic object categorization[C]//CVPR'06:Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEEComputer Society,2006,1:284-291.
[13]张志强,刘黎平,王红艳.三维可视化技术在雷达三维组网产品显示中的运用[J].气象科技,2010,38(5):605-608.(ZHANGZ Q,LIU L P,WANG H Y.Application of 3D visualization technology to display of Doppler radar networking products[J].Metrological Science and Technology,2010,38(5):605-608.)
[14]LU G Y,WONG D W.An adaptive inverse-distance weighting spatial interpolation technique[J].Computers&Geosciences,2008,34(9):1044-1055.
[15]GLOROT X,BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[EB/OL].[2017-03-01].http://www.weblio.jp/redirect?url=http%3A%2F%2Fjmlr.org%2Fproceedings%2Fpapers%2Fv9%2Fglorot10a%2Fglorot10a.pdf&etd=3e03a9174d723be0.
[16]王改利,赵翠光,刘黎平,等.雷达回波外推预报的误差分析[J].高原气象,2013,32(3):874-883.(WANG G L,ZHAOC G,LIU L P,et al.Error analysis of radar echo extrapolation[J].Plateau Meteorology,2013,32(3):874-883.)
[17]UIJLENHOET R,POMEROY J H.Raindrop size distribution and radar reflectivity-rain rate relationships for radar hydrology[J].Hydrology&Earth System Sciences&Discussions,2001,5(4):3012-3018.
[18]SHI X,CHEN Z,WANG H,et al.Convolutional LSTM network:a machine learning approach for precipitation nowcasting[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems.Cambridge,MA:MIT Press,2015,1:802-810.