结合数据扩增与残差收缩网络的地震短临预测

doi:10.12196/j.issn.1000-3274.2022.02.006

地震 ›› 2022, Vol. 42 ›› Issue (2): 74-88.doi: 10.12196/j.issn.1000-3274.2022.02.006

结合数据扩增与残差收缩网络的地震短临预测

张翔¹, 孙宪坤¹, 胡峻², 尹京苑², 熊玉洁¹

1.上海工程技术大学电子电气工程学院, 上海 201620;
2.上海市地震局, 上海 200062

收稿日期:2021-10-25 修回日期:2021-12-01 出版日期:2022-04-30 发布日期:2023-02-17
通讯作者: 孙宪坤, 副教授。 E-mail: xksun@sues.edu.cn
作者简介:张翔(1996-), 男, 江苏扬州人, 硕士研究生, 主要从事模式识别与智能系统研究。
基金资助:
国家重点研发计划项目(2019YFC1509202); 国家自然科学基金项目(62006150); 上海青年科技英才扬帆计划项目(19YF1418400)

Short-term and Imminent Earthquake Prediction Combined with Data Augmentation and Residual Shrinkage Network

ZHANG Xiang¹, SUN Xian-kun¹, HU Jun², YIN Jing-yuan², XIONG Yu-jie¹

1. School of Electronic and Electrical Engineering, Shanghai University of Engineering and Science, Shanghai 201620, China;
2. Earthquake Administration of Shanghai Municipality, Shanghai 200062, China

Received:2021-10-25 Revised:2021-12-01 Online:2022-04-30 Published:2023-02-17

摘要/Abstract

摘要： 中强地震发生前存在地表异常增温现象, 热红外信息可以成为地震短临预测的途径之一。然而, 地震预测研究常存在可供分析的震例样本不足的问题。本文基于MODIS地表温度数据, 先通过深度卷积生成对抗网络(DCGAN)对地震前的地温数据做扩增处理, 再将扩增后的地温数据输入深度残差收缩网络进行特征提取并预测未来短期内是否存在发生5级及以上地震事件的可能性。实验针对中国地震较多的中西部地区, 将地表温度数据依照地震实际发生情况标记为发生地震数据和未发生地震数据, 样本比为3∶1, 分不同的预测时间段进行比较。结果显示, 5日预测的准确率最高为73.86%, 正确预测发生占实际发生的比例为68.09%。多次实验准确率曲线趋向稳定, 证明该预测方法有很好的实用性。基于MODIS数据结合数据扩增与残差收缩网络的预测方法为短临中强地震预测研究提供了一种新思路。

关键词: 地震预测, MODIS, 数据扩增, 深度卷积生成对抗网络, 深度残差收缩网络

Abstract: Anomalies of temperature increase will occur before moderate and strong earthquakes, and thermal infrared information can be one of the ways to predict the short-term and imminent earthquakes. However, earthquake prediction research often has the problem of insufficient earthquake samples for analysis. Based on the MODIS surface temperature data, the deep convolutional generative adversarial network (DCGAN) is used to amplify the pre-earthquake ground temperature data, and then the amplified ground temperature data is input into the deep residual shrinkage network for feature extraction, and predict whether there is the possibility of M≥5 earthquake in the short-term future. The experiment is aimed at the central and western regions of China where earthquakes often occur, and the surface temperature data is marked according to the actual occurrence of the earthquake as the data of the occurrence of earthquakes and the data of no earthquakes. The sample ratio is 3∶1, and the comparison is divided into different prediction time periods. The results show that the highest accuracy rate of 5-day prediction is 73.86%, and the proportion of correct predictions to actual occurrences is 68.09%. The accuracy rate curve of multiple experiments tends to be stable, which proves that the prediction method has good practicability. Based on MODIS data, the prediction method combining data augmentation and residual shrinkage network provides a new idea for short-term, moderate-strong earthquake prediction research.

Key words: Earthquake prediction, MODIS, Data augmentation, Deep convolution generation adversarial network, Deep residual shrinkage network

中图分类号:

P315.75

张翔, 孙宪坤, 胡峻, 尹京苑, 熊玉洁. 结合数据扩增与残差收缩网络的地震短临预测[J]. 地震, 2022, 42(2): 74-88.

ZHANG Xiang, SUN Xian-kun, HU Jun, YIN Jing-yuan, XIONG Yu-jie. Short-term and Imminent Earthquake Prediction Combined with Data Augmentation and Residual Shrinkage Network[J]. EARTHQUAKE, 2022, 42(2): 74-88.

参考文献

[1] 陈棋福, 王克林. 2008年汶川地震与中国的地震预报[J]. 世界地震译丛, 2010, 3(5): 34-54.
CHEN Qi-fu, WANG Ke-lin. The 2008 Wenchuan earthquake and earthquake prediction in China[J]. World Earthquake Translation Collection, 2010, 3(5): 34-54 (in Chinese).
[2] 徐程, 邹金慧. BP神经网络在地震预测中的应用研究[J]. 自动化仪表, 2012, 33(6): 12-14.
XU Cheng, ZOU Jin-hui. Research and application of BP neural network in earthquake prediction[J]. Automation Instrumentation, 2012, 33(6): 12-14 (in Chinese).
[3] 强祖基, 赁常恭, 李玲芝, 等. 卫星热红外图像亮温异常短临震兆[J]. 中国科学(D辑: 地球科学), 1998, 28(6): 446-455.
QIANG Zu-ji, LIN Chang-gong, LI Ling-zhi, et al. Satellitic thermal infrared brightness temperature anomaly image: short-term and impending earthquake precursors[J]. Science in China (Series D: Earth Science), 1998, 28(6): 446-455 (in Chinese).
[4] 闫丽莉, 屈春燕, 温少妍, 等. 卫星热红外亮温、气温及地温观测的年变变化对比研究[J]. 地震学报, 2012, 34(2): 257-266.
YAN Li-li, QU Chun-yan, WEN Shao-yan, et al. A comparison study on annual variation of thermal infrared brightness temperature and land temperature from meteorological stations[J]. Acta Seismologica Sinica, 2012, 34(2): 257-266 (in Chinese).
[5] 李勇, 何洋波, 马丽, 等. 用分类回归树方法研究年亮温异常变[J]. 北京师范大学学报(自然科学版), 2003, 39(2): 143-146.
LI Yong, HE Yang-bo, MA Li, et al. A study on abnormality brightness by classification and regression tree methods[J]. Journal of Beijing Normal University (Natural Science), 2003, 39(2): 143-146 (in Chinese).
[6] 熊攀. 小波方法在地震遥感信息提取中的应用[D]. 北京: 中国地震局地震预测研究所, 2009.
XIONG Pan. Wavelet-based methods for detecting earthquake information in remote sensing data[D]. Beijing: Institute of Earthquake Prediction, China Earthquake Administration, 2009 (in Chinese).
[7] 陈顺云, 刘培洵, 刘力强, 等. 遥感与实测地表温度的对比分析及在地震研究中的意义[J]. 地球物理学报, 2011, 54(3): 747-755.
CHEN Shun-yun, LIU Pei-xun, LIU Li-qiang, et al. Comparative analysis between land surface temperatures obtained by field measurement and satellite remote sensing and its implication in earthquake research[J]. Chinese Journal of Geophysics, 2011, 54(3): 747-755 (in Chinese).
[8] 解涛, 康春丽, 卢军, 等. 2012年彝良M_S5.7和M_S5.6级地震热辐射亮温异常分析[J]. 地球物理学进展, 2013, 28(5): 2322-2327.
XIE Tao, KANG Chun-li, LU Jun, et al. Study of brightness temperature anomalies of Yiliang (China) M_S5.7 and M_S5.6 earthquakes on 7 Sep., 2012[J]. Progress in Geophysics, 2013, 28(5): 2322-2327 (in Chinese).
[9] Moustra M, Avraamides M, Christodoulou C. Artificial neural networks for earthquake prediction using time series magnitude data or Seismic Electric Signals[J]. Expert Systems with Applications, 2011, 38(12): 15032-15039.
[10] Asencio-Cortés G, Martínez-álvarez F, Troncoso A, et al. Medium-large earthquake magnitude prediction in Tokyo with artificial neural networks[J]. Neural Computing and Applications, 2017, 28(3): 1043-1055.
[11] Li R, Lu X, Li S, et al. DEP: A deep learning model for earthquake prediction[C]. In Proceedings of the 2020 IEEE International Joint Conference on Neural Networks, 2020,IJCNN’20: 1-8.
[12] Huang J P, Wang X A, Zhao Y, et al. Large earthquake magnitude prediction in Taiwan based on deep learning neural network[J]. Neural Network World, 2018, 28(2): 149-160.
[13] Nicolis O, Plaza F, Salas R. Prediction of intensity and location of seismic events using deep learning[J]. Spatial Statistics, 2021, 42: 100442.
[14] 宋冬梅, 臧琳, 单新建, 等. 基于LST年趋势背景场的地震热异常提取算法[J]. 地震地质, 2016, 38(3): 680-695.
SONG Dong-mei, ZANG Lin, SHAN Xin-jian, et al. A study on the algorithm for extracting earthquake thermal infrared anomalies based on the yearly trend of LST[J]. Seismology and Geology, 2016, 38(3): 680-695 (in Chinese).
[15] 李冠东, 张春菊, 高飞, 等. 双卷积池化结构的3D-CNN高光谱遥感影像分类方法[J]. 中国图象图形学报, 2019, 24(4): 639-654.
LI Guan-dong, ZHANG Chun-ju, GAO Fei, et al. Doubleconvpool-structured 3D-CNN for hyperspectral remote sensing image classification[J]. Journal of Image and Graphics, 2019, 24(4): 639-654 (in Chinese).
[16] Chen Y S, Jiang H L, Li C Y. Deep feature extraction and classification of hyperspectral images based on convolutional neural networks[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(10): 6232-6251.
[17] Tramutoli V, Aliano C, Corrado R, et al. On the possible origin of thermal infrared radiation (TIR) anomalies in earthquake-prone areas observed using robust satellite techniques (RST)[J]. Chemical Geology, 2013, 339: 157-168.
[18] 李雪松, 李劲华, 吕智涵. 基于改进深度残差收缩网络的轴承故障诊断[J]. 青岛大学学报(自然科学版), 2022, 35(2): 38-43+50.
LI Xue-song, LI Jin-hua, LÜ Zhi-han. Bearing fault diagnosis based on improved deep residual shrinkage network[J]. Journal of Qingdao University(Natural Science Edition), 2022, 35(2): 38-43+50 (in Chinese).
[19] Peng Y H. Super-resolution reconstruction using multiconnection deep residual network combined an improved loss function for single-frame image[J]. Multimedia Tools and Applications, 2020, 79(3): 9351-9362.
[20] Ni J, Shi P, Zhao Y, et al. Fixed-time output consensus tracking for high-order multi-agent systems with directed network topology and packet dropout[J]. IEEE/CAA Journal of Automatica Sinica, 2021, 8(4): 817-836.
[21] 杜圣杰, 贾晓芬, 黄友锐, 等. 面向CNN模型图像分类任务的高效激活函数设计[J]. 红外与激光工程, 2022, 51(3): 493-501.
DU Sheng-jie, JIA Xiao-fen, HUANG You-rui,et al. High efficient activation function design for CNN model image classification task[J]. Infrared and Laser Engineering, 2022, 51(3): 493-501 (in Chinese).
[22] 刘成勇, 乔文杰, 陈蜀喆, 等. 基于LSTM与注意力机制的船舶航迹预测模型研究[J]. 中国航海, 2021, 44(4): 94-100.
LIU Cheng-yong, QIAO Wen-jie, CHEN Shu-zhe, et al. On ship track prediction with LSTM and attention mechanism[J]. Navigation of China, 2021, 44(4): 94-100 (in Chinese).

结合数据扩增与残差收缩网络的地震短临预测

Short-term and Imminent Earthquake Prediction Combined with Data Augmentation and Residual Shrinkage Network

PDF (PC)

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]	袁爱璟, 王伟君, 彭菲, 闫坤, 寇华东. 机器学习在地震预测中的应用进展[J]. 地震, 2021, 41(1): 51-66.
[2]	马士振, 刘宏志, 牟磊育. 在地脉动数据上应用分类算法的地震预测实验[J]. 地震, 2020, 40(1): 159-171.
[3]	许忠淮. 应重视大地震预测物理基础的研究[J]. 地震, 2019, 39(2): 11-18.
[4]	张琰, 吴忠良, 李佳威. 地震预测预报研究议程的演变: 文献计量分析的启示[J]. 地震, 2019, 39(2): 159-173.
[5]	李艳娥, 陈学忠. 2008年汶川M_S8.0地震在2013年芦山M_S7.0地震和2014年康定M_S6.3地震破裂区引起的库仑破裂应力[J]. 地震, 2019, 39(1): 136-145.
[6]	李丽, 王霞, 宋美琴, 吴昊昱, 梁艳. 大同窗异常判定及预测效能初步分析[J]. 地震, 2018, 38(4): 142-150.
[7]	尹祥础, 刘月, 张浪平, 李文军, 袁帅, 张小涛. “泛西南”地区加卸载响应比异常及大震活动预测与追踪[J]. 地震, 2017, 37(4): 37-49.
[8]	马禾青, 杨明芝. 地震活动多参数综合变量的异常分析方法以青海两次7级地震为例[J]. 地震, 2017, 37(4): 50-57.
[9]	李玉江,邵志刚,季灵运,陈连旺. 加勒比板块多学科研究综述及其对地震预测的启示[J]. 地震, 2017, 37(2): 17-31.
[10]	陈长云, 郑智江, 李腊月, 畅柳. 南北地震带中南段典型强震震前跨断层形变特征分析[J]. 地震, 2016, 36(4): 47-60.
[11]	王俊, 邵志刚, 孙小龙, 王熠熙, 向阳, 王喜龙. 川滇地区强震前流体异常特征与预测指标体系初探[J]. 地震, 2016, 36(4): 109-119.
[12]	张琳琳, 聂晓红, 刘建明, 魏芸芸, 刘萍. 基于AHP的新疆南天山西段中强地震预测指标体系及效能评估[J]. 地震, 2016, 36(4): 205-214.
[13]	齐玉妍, 孙丽娜, 邱玉荣, 李姜, 金学申. K_cv值在河北及邻区地震预测中的应用研究[J]. 地震, 2015, 35(2): 101-110.
[14]	周筑宝, 唐松花. 最小耗能原理与地震预测问题[J]. 地震, 2014, 34(4): 50-60.
[15]	王鹏, 郑建常. 地震力标度及其在华北地区的应用[J]. 地震, 2013, 33(1): 111-118.