基于LSTM_TCN模型的降雨型滑坡时间概率预测及气象预警建模

摘要:
如果滑坡发生时间信息不完备则会导致滑坡与降雨时序关系错误, 以至于降雨阈值模型精度偏低。以重庆市万州区1995-2015年所发生的降雨型滑坡为研究对象, 将区内严重缺失历史滑坡时间信息的恒合乡作为验证区, 提出了一种基于长短时记忆网络(LSTM)融合时域卷积网络(TCN)的模型方法。该方法通过模拟降雨型滑坡发生时间与降雨量间的非线性关系, 重建降雨型滑坡事件在某日发生的时间概率。将重建时间信息后的滑坡事件进行了验证与筛选, 应用于累积有效降雨量-降雨历时曲线的合理划分, 构建了滑坡气象预警模型。结果表明, 本方法所预测滑坡时间概率平均值达到90.33%, 高于人工神经网络(ANN)(71.17%)、LSTM(72.75%)和TCN(86.91%)的概率。利用预测概率高于90%的滑坡, 将验证区18个时间信息扩充至201个。基于扩充时间信息后的滑坡数据所构建的气象预警模型比仅利用历史滑坡事件具有更合理的预警分级, 在严重警告级别上有效预警率提升了42.86%。结果说明该方法可弥补野外调查中灾害数据时间信息不足的问题, 为降雨型滑坡气象预警工作提供数据支撑, 由此提高气象预警准确率。
- 降雨型滑坡 /
- 时间概率 /
- E-D有效降雨阈值模型 /
- TCN /
- LSTM /
- 滑坡气象预警
Abstract: Objective
Incomplete landslide timing information can result in inaccuracies in the temporal relationship between landslides and rainfall, consequently affecting the precision of a critical rainfall threshold model.
Methods
To address this issue, this study focuses on rainfall-induced landslides in the Wanzhou District of Chongqing from 1995 to 2015. The Henghe Township, lacking historical landslide data, serves as the verification area. We proposed a prediction model for the daily temporal probability of landslide occurrence on a certain day based on long short-term memory (LSTM) and a temporal convolutional network (TCN). This method was used to reconstruct the temporal information of rainfall-induced landslide events by simulating the nonlinear relationship between the duration of landslides and rainfall. After the reconstruction of temporal information, the landslide events were verified and selected and subsequently applied to a reasonable division of the E-D effective rainfall threshold curve to establish the landslide meteorological warning model.
Results
The results showed that the average temporal probability of rainfall-induced landslides predicted by the proposed method reached 90.33%, which was higher than that of the ANN (71.17%), LSTM (72.75%), and TCN (86.91%) models. Using temporal probabilities exceeding a 90% threshold, 18 data points, including 42 landslides in the verification area, are expanded to 201. Compared with using solely historical landslide events, the meteorological warning model based on expanded temporal information provides a more reasonable warning classification, and the effective warning rate at the severe warning level is increased by 42.86%.
Conclusion
This method can compensate for the shortage of landslide time information in field investigations and provide data support for early meteorological warning systems for rainfall-induced landslides, thus improving the accuracy of early meteorological warning systems.
- rainfall-induced landslide /
- temporal probability /
- E-D effective rainfall threshold model /
- TCN /
- LSTM /
- meteorological early warning for landslide
注释:

所有作者声明不存在利益冲突。

HTML全文

图 1 基于深度学习模型构建降雨型滑坡气象预警模型的流程图

Figure 1. Flow chart of the deep learning model for predicting the meteorological early warning modeling in landslide

下载: 全尺寸图片幻灯片

图 2 LSTM_TCN模型的网络结构

Figure 2. Structure of the LSTM_TCN model

下载: 全尺寸图片幻灯片

图 3 研究区概况及地质灾害分布图

Figure 3. Overview of the study area and geological hazards

下载: 全尺寸图片幻灯片

图 4 1995-2015年万州区的当日降雨量和灾害发生日

Figure 4. Daily rainfall and geological hazard occurrence dates in Wanzhou District, 1995 to 2015

下载: 全尺寸图片幻灯片

图 5 LSTM_TCN模型在不同训练集上损失值和预测准确率的变化曲线

Figure 5. Variation curves of loss values and predicted accuracy of the LSTM_TCN model on different training datasets

下载: 全尺寸图片幻灯片

图 6 4种模型预测灾害事件的ROC曲线

Figure 6. ROC curves of four models predicting geological hazard events

下载: 全尺寸图片幻灯片

图 7 2012年1月-2020年8月恒合乡日降雨量和LSTM_TCN模型预测的滑坡在某日发生的时间概率

Figure 7. Daily rainfall and the geological hazard probability predicted by the LSTM_TCN model of Henghe Township from January 2012 to August 2020

下载: 全尺寸图片幻灯片

图 8 LSTM_TCN模型预测灾害事件的频率直方图(a)和筛选图(b)

Figure 8. LSTM_TCN model for predicting geological hazard events: Histogram (a) and filter diagram (b)

下载: 全尺寸图片幻灯片

图 9 降雨阈值曲线图

a.基于万州区历史滑坡事件; b.基于恒合乡历史滑坡事件; c.基于有效扩充后的滑坡事件

Figure 9. Rainfall threshold curve

下载: 全尺寸图片幻灯片

图 10 气象预警模型有效率混淆矩阵对比

a.基于历史滑坡事件划分的预警等级; b.基于扩充滑坡事件划分的预警等级。$ \frac{1}{14.3 \%}$.1代表滑坡灾害个数；14.3%代表有效预警率

Figure 10. Comparison of the efficient confusion matrix of meteorological early warning model

下载: 全尺寸图片幻灯片

表 1 验证区恒合乡降雨型灾害调查汇总

Table 1. Summary of rainfall-induced geological hazard in Henghe Township

编号	灾害名称	发生时间	编号	灾害名称	发生时间	编号	灾害名称	发生时间
1	向东村崩塌	2012/07/12	15	李子树崩塌	2019/06/29 2020/07/05	29	县道拐角大滑坡	2020/07/09
2	王爷庙桥崩塌	2013/07/19	16	恒一村崩塌	2020/07/07	30	县道拐角泥滑坡	2020/07/14
3	关家坪崩塌	2015/03/20	17	崩塌	2020/07/07	31	入城要道切坡滑坡	2020/07/13
4	黄显村2号崩塌	2015/04/19	18	崩塌	2020/07/09	32	乡道泥滑坡	2020/07/13
5	张氏沟崩塌	2015/05/29	19	马家坪滑坡	2020/7/7	33	乡道崩坡积滑坡	2020/06/22
6	黄显村1号崩塌	2015/06/17	20	李家坎滑坡	2020/7/7	34	要道宽滑坡	2020/07/09
7	汪家岭Ⅰ号崩塌	2015/06/17	21	下石坝滑坡	2020/7/17	35	三拐口滑坡	2020/07/13
8	大河坪崩塌	2015/06/17	22	黄草坪滑坡	2020/07/07	36	北岩质滑坡	2020/07/14
9	双河口-黄岭庄Ⅰ号	2015/07/15	23	三组滑坡1	2020/07/02	37	石桶寨滑坡	2020/07/13
10	先锋村	2020/07/04	24	三组滑坡2	2020/07/02	38	石桶寨岩质滑坡	2020/07/9
11	深沟子崩塌	2020/07/05	25	谯家坪滑坡	2020/07/05	39	先锋村滑坡	2020/06/22
12	崩塌	2020/07/05	26	玉都村滑坡	2020/07/02	40	先锋东滑坡	2020/07/05
13	危岩段	2020/07/05	27	杉木冲滑坡	2020/07/06	41	衫木冲滑坡	2020/07/05
14	崩塌	2020/07/05	28	县道滑坡	2020/07/05	42	道路滑坡	2020/07/05

下载: 导出CSV

表 2 不同降雨系数和累积有效降雨量与滑坡的相关性

Table 2. Correlation of different rainfall intensity factors and cumulative effective rainfall with landslides

a	R₀	E₁	E₂	E₃	E₄	E₅	E₆	E₇	E₈	E₉
a	R₀	相关系数
0.4	0.620	0.674	0.659	0.636	0.626	0.622	0.621	0.620	0.620	0.620
0.5	0.620	0.676	0.676	0.653	0.635	0.627	0.623	0.621	0.620	0.620
0.6	0.620	0.675	0.687	0.675	0.653	0.639	0.631	0.626	0.624	0.622
0.7	0.620	0.673	0.688	0.697	0.681	0.661	0.648	0.640	0.634	0.629
0.8	0.620	0.668	0.679	0.706	0.711	0.696	0.681	0.669	0.659	0.651
0.9	0.620	0.663	0.659	0.684	0.709	0.710	0.710	0.710	0.706	0.700
1.0	0.620	0.656	0.632	0.635	0.637	0.600	0.565	0.535	0.514	0.496
注：a.有效降雨系数；R₀.当日降雨量；E₁~E₉.前1~9 d累积有效降雨量

下载: 导出CSV

表 3 4种模型预测结果的对比

Table 3. Comparison of prediction results of four models

模型名称	ACC/%	MAE	RMSE
ANN模型	84.63	0.125	0.106
LSTM模型	90.29	0.065	0.056
TCN模型	90.85	0.047	0.042
LSTM_TCN模型	92.37	0.025	0.019
ACC.准确率；MAE.平均绝对误差; RMSE.均方根误差

下载: 导出CSV

表 4 4种模型预测滑坡在某日发生的时间概率对比

Table 4. Comparison of the probability of landslide occurrence on a certain day predicted by the four models

发生时间	灾害个数	ANN	LSTM	TCN	LSTM_TCN
2012/07/12	1	0.904 1	0.898 3	0.975 4	0.939 6
2015/06/17	3	0.878 1	0.886 5	0.900 4	0.931 6
2020/06/22	2	0.869 3	0.866 5	0.899 9	0.918 2
					
2020/07/04	1	0.936 9	0.856 6	0.925 8	0.949 8
2020/07/05	10	0.916 8	0.918 6	0.947 0	0.955 7
2020/07/06	1	0.923 6	0.801 8	0.941 2	0.968 9
2020/07/07	5	0.892 3	0.911 0	0.925 6	0.977 5
2020/07/09	3	0.907 6	0.890 7	0.925 6	0.988 7
平均值	—	0.711 7	0.727 5	0.869 1	0.903 3

下载: 导出CSV

表 5 基于不同数据划分的降雨阈值方程

Table 5. Rainfall threshold equation based on different data

阈值线	万州区滑坡	恒合乡滑坡	恒合乡扩充滑坡
20%阈值	E=17.13D^{0.136 8}	E=28.74D^{0.264 6}	E=32.35D^{0.181 5}
40%阈值	E=24.65D^{0.136 8}	E=34.61D^{0.264 6}	E=37.97D^{0.181 5}
60%阈值	E=34.43D^{0.136 8}	E=44.53D^{0.264 6}	E=44.56D^{0.181 5}
80%阈值	E=47.45D^{0.136 8}	E=52.62D^{0.264 6}	E=53.82D^{0.181 5}
注：D为降雨历时, d; E为累计有效降雨量, mm

下载: 导出CSV

表 6 降雨阈值模型的预警等级和恒合乡历史灾害实际风险等级对比

Table 6. Comparison between the warning levels of the rainfall threshold model and the actual risk levels of historical hazards in Henghe Township

历史灾害事件		实际风险等级	实际所处预警等级	基于恒合乡历史灾害阈值模型预警等级	基于扩充后恒合乡灾害阈值模型预警等级
发生时间	滑坡灾害个数	实际风险等级	实际所处预警等级	基于恒合乡历史灾害阈值模型预警等级	基于扩充后恒合乡灾害阈值模型预警等级
2012/07/12	1	极高	严重警告	严重警告	严重警告
2015/06/17	3	高	警告	特别注意	警告
2020/06/22	2	极高	严重警告	警告	严重警告
					
2020/07/04	1	极高	严重警告	警告	严重警告
2020/07/05	10	极高	严重警告	严重警告	严重警告
2020/07/06	1	极高	严重警告	警告	严重警告
2020/07/07	5	极高	严重警告	严重警告	严重警告
2020/07/09	3	极高	严重警告	特别注意	警告

下载: 导出CSV

表 7 滑坡气象预警各等级有效预警率对比

Table 7. Comparison of effective warning rates for different levels of landslide meteorological warning systems

滑坡预警等级	注意	特别注意	警告	严重警告
阈值线	20%阈值	40%阈值	60%阈值	80%阈值
滑坡风险等级	低	中	高	极高
历史有效预警率/%	33.33	50	0	42.86
扩充后有效预警率/%	50	100	100	85.71

下载: 导出CSV

参考文献(46)

[1]	殷坤龙, 张桂荣, 龚日祥, 等. 浙江省突发性地质灾害预警预报[M]. 武汉: 中国地质大学出版社, 2005: 42-47. YIN K L, ZHANG G R, GONG R X, et al. Early warning and prediction of sudden geological hazards in Zhejiang Province[M]. Wuhan: China University of Geosciences Press, 2005: 42-47. (in Chinese)
[2]	陈丽霞, 殷坤龙, 刘长春. 降雨重现期及其用于滑坡概率分析的探讨[J]. 工程地质学报, 2012, 20(5): 745-750. https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ201205014.htm CHEN L X, YIN K L, LIU C C. Return period statistics of extreme rainfall and application to landslide probability analysis[J]. Journal of Engineering Geology, 2012, 20(5): 745-750. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ201205014.htm
[3]	LOPEZ S J, CORONA C, STOFFEL M, et al. High-resolution fingerprints of past landsliding and spatially explicit, probabilistic assessment of future reactivations: Aiguettes landslide, southeastern French Alps[J]. Tectonophysics, 2013, 602: 335-369.
[4]	ALTHUWAYNEE O F, PRADHAN B, AHMAD N. Estimation of rainfall threshold and its use in landslide hazard mapping of Kuala Lumpur metropolitan and surrounding areas[J]. Landslides, 2014, 12(5): 861-875.
[5]	GUMBEL E J. Multivariate extreme distributions[J]. Bulletin of the International Statistical Institute, 1960, 39(2): 471-475.
[6]	HALLINAN A J. A review of the Weibull distribution[J]. Journal of Quality Technology, 1993, 25(2): 85-93. doi: 10.1080/00224065.1993.11979431
[7]	徐勇, 连志鹏, 李德营, 等. 鄂西南地区五峰县凉风洞滑坡灾害风险分析[J]. 长江科学院院报, 2016, 33(1): 51-56. https://www.cnki.com.cn/Article/CJFDTOTAL-CJKB201610010.htm XU Y, LIAN Z P, LI D Y, et al. Risk analysis of Liangfengdong landslide in Wufeng County, Southwest Hubei Province[J]. Journal of Yangtze River Scientific Research Institute, 2016, 33(1): 51-56. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-CJKB201610010.htm
[8]	王芳, 殷坤龙, 桂蕾, 等. 不同日降雨工况下万州区滑坡灾害危险性分析[J]. 地质科技通报, 2018, 37(1): 190-195. https://dzkjqb.cug.edu.cn/article/id/9528 WANG F, YIN K L, GUI L, et al. Landslide hazard analysis under different daily rainfall conditions in Wanzhou District[J]. Geological Science and Technology Information, 2018, 37(1): 190-195. (in Chinese with English abstract) https://dzkjqb.cug.edu.cn/article/id/9528
[9]	赵海燕, 殷坤龙, 陈丽霞, 等. 基于有效降雨阈值的澧源镇滑坡灾害危险性分析[J]. 地质科技通报, 2020, 39(4): 85-93. doi: 10.19509/j.cnki.dzkq.2020.0431 ZHAO H Y, YIN K L, CHEN L X, et al. Landslide hazard analysis of Liyuan Town based on effective rainfall threshold[J]. Bulletin of Geological Science and Technology, 2020, 39(4): 85-93. (in Chinese with English abstract) doi: 10.19509/j.cnki.dzkq.2020.0431
[10]	SEGONI S, PICIULLO L, GARIANO S L. A review of the recent literature on rainfall thresholds for landslide occurrence[J]. Landslides, 2018, 15(8): 1483-1501. doi: 10.1007/s10346-018-0966-4
[11]	CAINE N. The rainfall intensity: Duration control of shallow landslides and debris flows[J]. Geografiska Annaler(Series A, Physical Geography), 1980, 62(1/2): 23-27. doi: 10.2307/520449
[12]	GUZZETⅡ F, CROSTA G, DETTI R, et al. Stone: A computer program for the three-dimensional simulation for rock-falls[J]. Computers and Geosciences, 2002, 28(9): 1079-1093. doi: 10.1016/S0098-3004(02)00025-0
[13]	吴益平, 张秋霞, 唐辉明, 等. 基于有效降雨强度的滑坡灾害危险性预警[J]. 地球科学, 2014, 39(7): 889-895. https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX201407011.htm WU Y P, ZHANG Q X, TANG H M, et al. Landslide hazard warning based on effective rainfall intensity[J]. Earth Science, 2014, 39(7): 889-895. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX201407011.htm
[14]	GUZZETTI F, PERUCCACCI S, ROSSI M, et al. Rainfall thresholds for the initiation of landslides in central and southern Europe[J]. Meteorology and Atmospheric Physics, 2007, 98(3/4): 239-267.
[15]	HONG Y, HIURA H, SHINO K, et al. The influence of intense rainfall on the activity of large-scale crystalline schist landslides in Shikoku Island, Japan[J]. Landslides, 2005, 3(2): 97-105.
[16]	MATHEW J, BABU D G, KUNDU S, et al. Integrating intensity-duration-based rainfall threshold and antecedent rainfall-based probability estimate towards generating early warning for rainfall-induced landslides in parts of the Garhwal Himalaya, India[J]. Landslides, 2013, 11(4): 575-588.
[17]	刘谢攀, 殷坤龙, 肖常贵, 等. 基于I-D-R阈值模型的滑坡气象预警[J/OL]. 地球科学, 2022, 1-15. https://doi.org/10.3799/dqkx.2022.233 LIU X P, YIN K L, XIAO C G, et al. Meteorological early warning of landslide based on I-D-R threshold model[J/OL]. Earth Science, 2022, 1-15. https://doi.org/10.3799/dqkx.2022.233. (in Chinese with English abstract)
[18]	CROZIER M J. Landslides: Causes, consequences & environment[M]. London: Croom Helm, 1986.
[19]	黄发明, 陈佳武, 范宣梅, 等. 降雨型滑坡时间概率的逻辑回归拟合及连续概率滑坡危险性建模[J/OL]. 地球科学, 2021, 1-25. https://doi.org/10.3799/dqkx.2021.164 HUANG F M, CHEN J W, FAN X M, et al. Logistic regression fitting of rainfal-induced landslide occurrence probability and continuous landslide hazard prediction modelling[J/OL]. Earth Science, 2021, 1-25. https://doi.org/10.3799/dqkx.2021.164. (in Chinese with English abstract)
[20]	黄发明, 曹中山, 姚池, 等. 基于决策树和有效降雨强度的区域降雨型滑坡危险性预警建模[J]. 浙江大学学报(工学版), 2021, 55(3): 472-482. HUANG F M, CAO Z S, YAO C, et al. Landslides hazard warning based on decision tree and effective rainfall intensity[J]. Journal of Zhejiang University(Engineering Science), 2021, 55(3): 472-482. (in Chinese with English abstract)
[21]	COROMINAS J, MOYA J. A review of assessing landslide frequency for hazard zoning purposes[J]. Engineering Geology, 2008, 102(3/4): 193-213.
[22]	黄发明, 陈佳武, 唐志鹏, 等. 不同空间分辨率和不同训练测试集比例的滑坡易发性预测不确定性[J]. 岩石力学与工程学报, 2021, 40(6): 1155-1169. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202106008.htm HUANG F M, CHEN J W, TANG Z P, et al. Uncertainties of landslide susceptibility prediction due to different spatial resolutions and different proportions of training and testing datasets[J]. Chinese Journal of Rock Mechanics and Engineering, 2021, 40(6): 1155-1169. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202106008.htm
[23]	YAO W, ZENG Z, LIAN C, et al. Training enhanced reservoir computing predictor for landslide displacement[J]. Engineering Geology, 2015, 188: 101-109. doi: 10.1016/j.enggeo.2014.11.008
[24]	黄发明, 李金凤, 王俊宇, 等. 考虑线状环境因子适宜性和不同机器学习模型的滑坡易发性预测建模规律[J]. 地质科技通报, 2022, 41(2): 44-59. doi: 10.19509/j.cnki.dzkq.2022.0010 HUANG F M, LI J F, WANG J Y, et al. Modelling rules of landslides usceptibility prediction considering the suitability of linear environmental factors and different machine learning models[J]. Bulletin of Geological Science and Technology, 2022, 41(2): 44-59. (in Chinese with English abstract) doi: 10.19509/j.cnki.dzkq.2022.0010
[25]	FALASCHI F, GIACOMELLI F, FEDERICI P R, et al. Logistic regression versus artificial neural networks: Landslide susceptibility evaluation in a sample area of the Serchio River valley, Italy[J]. Natural Hazards, 2009, 50(3): 551-569. doi: 10.1007/s11069-009-9356-5
[26]	PARK S, CHOI C, KIM B, et al. Landslide susceptibility mapping using frequency ratio, analytic hierarchy process, logistic regression, and artificial neural network methods at the Inje area, Korea[J]. Environmental Earth Sciences, 2013, 68(5): 1443-1464. doi: 10.1007/s12665-012-1842-5
[27]	田乃满, 兰恒星, 伍宇明, 等. 人工神经网络和决策树模型在滑坡易发性分析中的性能对比[J]. 地球信息科学学报, 2020, 22(12): 2304-2316. https://www.cnki.com.cn/Article/CJFDTOTAL-DQXX202012004.htm TIAN N M, LAN H X, WU Y M, et al. Performance comparison of BP artificial neural network and CART decision tree model in landslide susceptibility prediction[J]. Journal of Geo-Information Science, 2020, 22(12): 2304-2316. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-DQXX202012004.htm
[28]	YAO J, QIN S, QIAO S, et al. Application of a two-step sampling strategy based on deep neural network for landslide susceptibility mapping[J]. Bulletin of Engineering Geology and the Environment, 2022, 81(4): 1-20.
[29]	吴琪, 周创兵, 黄发明, 等. 基于双重注意力机制的滑坡识别方法优化[J]. 地质科技通报, 2022, 41(2): 246-253. doi: 10.19509/j.cnki.dzkq.2022.0053 WU Q, ZHOU C B, HUANG F M, et al. Optimization of the landslide identification method based on adual attention mechanism[J]. Bulletin of Geological Science and Technology, 2022, 41(2): 246-253. (in Chinese with English abstract) doi: 10.19509/j.cnki.dzkq.2022.0053
[30]	HUANG F, ZHANG J, ZHOU C, et al. A deep learning algorithm using a fully connected sparse autoencoder neural network for landslide susceptibility prediction[J]. Landslides, 2020, 17(1): 217-229. doi: 10.1007/s10346-019-01274-9
[31]	ZHU L, HUANG L, FAN L, et al. Landslide susceptibility prediction modeling based on remote sensing and a novel deep learning algorithm of a cascade-parallel recurrent neural network[J]. Sensors, 2020, 20(6): 1576-1587. doi: 10.3390/s20061576
[32]	黄发明, 陈彬, 毛达雄, 等. 基于自筛选深度学习的滑坡易发性预测建模及其可解释性[J]. 地球科学, 2022, 48(5): 1696-1710. https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX202305003.htm HUANG F M, CHEN B, MAO D X, et al. Landslide susceptibility prediction modeling and interpretability based on self-screening deep learning model[J]. Earth Science, 2022, 48(5): 1696-1710. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX202305003.htm
[33]	QIN S, JIAO J, WANG S. A nonlinear dynamical model of landslide evolution[J]. Geomorphology, 2002, 43(1): 77-85.
[34]	YANG B B, YIN K L, LACASSE S, et al. Time series analysis and long short-term memory neural network to predict landslide displacement[J]. Landslides, 2019, 16(4): 677-694. doi: 10.1007/s10346-018-01127-x
[35]	BAI S, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[J/OL]. 2018. http:doi.org/10.48550/arXiv.1803.01271
[36]	BENGIO Y, SIMARD P, FRASCONI P. Learning long-term dependencies with gradient descent is difficult[J]. IEEE Transactions on Neural Networks, 1994, 5(2): 157-166. doi: 10.1109/72.279181
[37]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. doi: 10.1162/neco.1997.9.8.1735
[38]	KOLEN J K, KREMER S C. Gradient flow in recurrent nets: The difficulty of learning long term dependencies[M]. Hyderaba, India: Wiley-IEEE Press, 2001.
[39]	Anon. Landslides: Causes, consequences and environment[M]. London: Routledge Kegan and Paul, 1986: 185-189.
[40]	MERGHADI A, ABDERRAHMANE B, BUI D T. Landslide susceptibility assessment at Mila Basin(Algeria): A comparative assessment of prediction capability of advanced machine learning methods[J]. ISPRS International Journal of Geo-information, 2018, 7(7): 268-298. doi: 10.3390/ijgi7070268
[41]	罗路广, 裴向军, 黄润秋, 等. GIS支持下CF与Logistic回归模型耦合的九寨沟景区滑坡易发性评价[J]. 工程地质学报, 2021, 29(2): 526-535. https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202102022.htm LUO L G, PEI X J, HUANG R Q, et al. Landslide susceptibility assessment in Jiuzhaigou scenic area with GIS based on certainty factor and logistic regression model[J]. Journal of Engineering Geology, 2021, 29(2): 526-535. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202102022.htm
[42]	SHAO H, JIANG H, ZHAO H, et al. A novel deep autoencoder feature learning method for rotating machinery fault diagnosis[J]. Mechanical Systems and Signal Processing, 2017, 95: 187-204. doi: 10.1016/j.ymssp.2017.03.034
[43]	YANG B, LEI Y, JIA F, et al. An intelligent fault diagnosis approach based on transfer learning from laboratory bearings to locomotive bearings[J]. Mechanical Systems and Signal Processing, 2019, 122: 692-706. doi: 10.1016/j.ymssp.2018.12.051
[44]	林巍, 李远耀, 徐勇, 等. 湖南慈利县滑坡灾害的临界降雨量阈值研究[J]. 长江科学院院报, 2020, 37(2): 48-54. https://www.cnki.com.cn/Article/CJFDTOTAL-CJKB202002011.htm LIN W, LI Y Y, XU Y, et al. Rainfall thresholds of rainfall-triggered landslides in Cili County, Hunan Province[J]. Journal of Yangtze River Scientific Research Institute, 2020, 37(2): 48-54. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-CJKB202002011.htm
[45]	PERUCCACCI S, BRUNETTI M T, LUCIANI S, et al. Lithological and seasonal control on rainfall thresholds for the possible initiation of landslides in central Italy[J]. Geomorphology, 2012, 139/140: 79-90.
[46]	苏晨旭, 田钦, 刘本朝, 等. 江西省龙南县滑坡易发性评价[J]. 科学技术与工程, 2019, 19(17): 91-99. https://www.cnki.com.cn/Article/CJFDTOTAL-KXJS201917012.htm SUN C X, TIAN Q, LIU B C, et al. Regional landslide susceptibility assessment for Longnan County in Jiangxi Province[J]. Science Technology and Engineering, 2019, 19(17): 91-99. (in Chinese with English abstract) https://www.cnki.com.cn/Article/CJFDTOTAL-KXJS201917012.htm