基于RSIV-RF模型的凉山州泥石流易发性评价

摘要:
针对随机森林（RF）模型进行泥石流易发性评价过程中存在连续型因子依靠主观意识分级、随机选取的非泥石流样本准确度较低等问题，以位于四川西南部的凉山彝族自治州为研究区，提出基于统计学先验模型抽样的随机森林对研究区进行泥石流易发性评价分区。利用累计灾害频率等曲线的相对变化对连续型因子进行分级处理；采用粗糙集理论（RS）和信息量法（Ⅳ）计算加权信息量值，划定极低和低易发性区并从中选择负样本数据。通过袋外误差（OOB）变化曲线确定RF模型的最佳树棵数n_estimators和分裂特征数max_features，随后构建加权信息量-随机森林（RSIV-RF）模型预测凉山州泥石流易发性。进一步地，与从全区随机选择非泥石流样本的RF模型开展对比研究。结果表明，训练集和测试集下RSIV-RF模型的准确度分别为0.89，0.83，且对应的ROC曲线的AUC值分别为0.920，0.895，均高于单独的RF模型；RSIV-RF绘制的泥石流易发性评价图与历史灾害分布较为一致，较高和高易发性等级区域占研究区面积比为18.625%，包含了78.57%的泥石流点。性能评估和易发性统计结果均表明基于RSIV-RF能够解决单独模型存在的非泥石样本采样不准确的问题，其泥石流易发性预测精度更高，在凉山州地区泥石流易发性评价研究中具有较好的适应性。
- 随机森林(RF) /
- 不平衡数据集 /
- 加权信息量(RSIV) /
- 泥石流 /
- RSIV-RF模型 /
- 凉山州 /
- 易发性评价
Abstract: Objective
In employing the random forest (RF) model for debris flow susceptibility assessment, challenges arose, including subjectivity in classifying continuous factors and the low accuracy of randomly selected nondebris flow samples. Taking Liangshan Yi Autonomous Prefecture in southwestern Sichuan Province as the study area, a random forest based on statistical prior model sampling was proposed to evaluate the debris flow susceptibility in the study area.
Methods
Continuous factors are classified by the relative changes in cumulative disaster frequency and other curves. Rough set theory (RS) and the information value method (Ⅳ) were used to calculate the weighted information values, delimit the extremely low- and low-prone areas and selecting the negative sample data. The optimal number of trees n_estimators and the number of feature splits max_features for the RF model were determined from the out-of-bag error (OOB) change curves. Subsequently, a weighted information random forest (RSIV-RF) model was constructed to predict the vulnerability of debris flow in Liangshan Prefecture. Furthermore, a comparative analysis with the RF model randomly selecting non-debris flow samples revealed the superior performance of the RSIV-RF model.
Results
The results show that the accuracy of the RSIV-RF model in the training set and the test set is 0.89 and 0.83, respectively, and the AUC value of the corresponding ROC curve is 0.920 and 0.895, respectively, which are higher than that of the RF model alone. The assessment map of debris flow susceptibility drawn by RSIV-RF is consistent with the distribution of historical disasters. The areas with high and higher susceptibility levels account for 18.625% of the study area, including 78.57% of debris flow points.
Conclusion
The results of the performance evaluation and susceptibility statistics show that RSIV-RF can solve the problem of inaccurate sampling of nondebris samples in a single model, and its prediction accuracy of debris flow susceptibility is higher. It has good adaptability in the study of debris flow susceptibility evaluation in Liangshan Prefecture.
- random forest(RF) /
- unbalanced data set /
- weighted information quantity(RSIV) /
- debris flow /
- RSIV-RF model /
- Liangshan Prefecture /
- susceptibility evaluation
注释:

所有作者声明不存在利益冲突。

HTML全文

图 1 凉山州泥石流易发性评价流程

Figure 1. Evaluation process of debris flow susceptibility in Liangshan Prefecture

下载: 全尺寸图片幻灯片

图 2 研究区凉山州地理位置及灾害分布

a.研究区位置; b.研究区行政区划; c.基本地理情况

Figure 2. Geological location and disaster distribution in Liangshan Prefecture

下载: 全尺寸图片幻灯片

图 3 连续性型因子分级曲线

a.高程; b.坡度; c.坡向; d.归一化植被指数; e.道路距离; f.水系距离; g.降雨量; h.土地利用

Figure 3. Evaluation factor status grading curve

下载: 全尺寸图片幻灯片

图 4 RSIV负样本分布

Figure 4. RSIV negative sample distribution

下载: 全尺寸图片幻灯片

图 5 随机森林的袋外误差曲线

Figure 5. Out-of-pocket error curve of random forest

下载: 全尺寸图片幻灯片

图 6 RSIV-RF研究区泥石流易发性评价图

Figure 6. Assessment map of debris flow susceptibility in the RSIV-RF study area

下载: 全尺寸图片幻灯片

图 7 2种模型训练集和测试集下的AUC曲线

Figure 7. AUC curves of the two model training sets and test sets

下载: 全尺寸图片幻灯片

表 1 指标相关性系数计算

Table 1. Calculation of the index correlation coefficient

指标因子	降雨量	坡度	距道路距离	距河流距离	坡向	高程	土地利用	NDVI
降雨量	1.000
坡度	0.101	1.000
距道路距离	-0.110	-0.020	1.000
距河流距离	0.003	-0.011	0.340	1.000
坡向	-0.123	-0.113	0.098	-0.125	1.000
高程	-0.486	-0.102	0.447	0.301	0.113	1.000
土地利用	-0.057	0.037	0.140	-0.062	0.115	0.119	1.000
NDVI	0.124	0.304	0.058	0.132	-0.009	-0.044	-0.034	1.000

下载: 导出CSV

表 2 指标因子RS权重系数计算

Table 2. Calculation of the index factor RS weight coefficient

指标因子	属性重要度	权重系数	权重大小排序
降雨量	0.065 5	0.104 8	6
坡度	0.093 3	0.149 2	3
距道路距离	0.055 6	0.088 9	7
距河流距离	0.049 6	0.079 4	8
坡向	0.117 1	0.187 3	1
高程	0.103 2	0.165 1	2
土地利用	0.071 4	0.114 3	4
NDVI	0.069 4	0.111 1	5

下载: 导出CSV

表 3 指标因子加权信息量计算结果

Table 3. Calculation results of weighted information content of index factors

指标因子	二级状态	信息量值	加权信息量值
降雨量/mm	900	-1.492 5	-0.156 4
	950	-2.644 4	-0.277 0
	1050	-0.136 6	-0.014 3
	1150	0.179 1	0.018 8
	>1150	1.549 3	0.162 3
坡度/(°)	[0, 15)	-0.502 4	-0.075 0
	[15, 20)	0.196 1	0.029 3
	[20, 25)	0.431 7	0.064 4
	[25, 45)	-0.272 6	-0.040 7
	[45, 60]	-0.779 1	-0.116 2
道路距离/m	[0, 300)	2.126 7	0.189 0
	[300, 700)	0.868 6	0.077 2
	[700, 1 000)	-0.114 5	-0.010 2
	[1 000, 1500)	0.728 4	0.064 7
	[1 500, 2 000]	-0.344 8	-0.030 6
	>2 000	-0.461 9	-0.041 1
河流距离/m	[0, 200)	1.631 7	0.129 5
	[200, 300)	2.012 7	0.159 7
	[300, 600)	1.458 9	0.115 8
	[600, 1 000)	1.360 6	0.108 0
	[1 000, 1 500)	0.641 1	0.050 9
	[1 500, 2 000]	0.081 5	0.006 5
	>2 000	-0.539 5	-0.042 8
坡向	-1	0.000 0	0.000 0
	N-NE 0°~67.5°	-0.379 8	-0.071 1
	E-S 67.5°~202.5°	0.099 3	0.018 6
	SW-W 202.5°~292.5°	0.161 0	0.030 1
	WE-N 292.5°~360°	-0.154 7	-0.029 0
归一植被指数(NDVI)	0.1	1.146 0	0.127 3
	0.2	1.024 0	0.113 8
	0.4	0.564 8	0.062 8
	0.6	0.302 3	0.033 6
	>0.6	-1.137 3	-0.126 4
土地利用	耕地	0.903 0	0.103 2
	林地	-0.696 0	-0.079 5
	草地	0.038 5	0.004 4
	水域	0.552 9	0.063 2
	建筑用地	1.390 8	0.159 0
	未利用土地	0.000 0	0.000 0
高程/m	1220	2.090 9	0.345 2
	1630	0.993 0	0.163 9
	2000	0.798 0	0.131 7
	2600	-0.117 7	-0.019 4
	3000	-1.321 2	-0.218 1
	>3000	-2.608 9	-0.430 7

下载: 导出CSV

表 4 RSIV-RF模型易发性分区结果统计

Table 4. Statistics of susceptibility zoning results of the RSIV-RF model

易发性分区	栅格数	占栅格比例/%	泥石流数	泥石流比例/%	灾害密度
极低易发区	36 061 783	54.316	6	2.381	0.044
低易发区	8 153 595	12.281	16	6.349	0.517
中易发区	10 288 057	15.496	32	12.698	0.819
较高易发区	8 104 859	12.208	87	34.524	2.828
高易发区	4 021 161	6.057	111	44.048	7.273

下载: 导出CSV

表 5 RF模型和RSIV-RF模型的评价性能统计

Table 5. Evaluation performance statistics of the RF model and RSIV-RF model

数据集	准确率		均方误差		Kappa系数
数据集	RF	RSIV-RF	RF	RSIV-RF	RF	RSIV-RF
训练集	0.87	0.89	0.12	0.10	0.62	0.70
测试集	0.76	0.83	0.23	0.16	0.55	0.63

下载: 导出CSV

参考文献(29)

[1]	LI C, MA Y, HE Y. Sensitivity analysis of debris flow to environmental factors: A case of Longxi River basin in Dujiangyan, Sichuan Province[J]. The Chinese Journal of Geological Hazard and Control, 2020, 31(5): 32-39.
[2]	PAOLA R, MAURO R, BRUCE D M, et al. A review of statistically-based landslide susceptibility models[J]. Earth Science Reviews, 2018, 180: 60-91. doi: 10.1016/j.earscirev.2018.03.001
[3]	ULRICH K, BENJAMIN G, GHAZANFAR K, et al. GIS-based landslide susceptibility mapping for the 2005 Kashmir-Earthquake region[J]. Geomorphology, 2008, 101(4): 631-642. doi: 10.1016/j.geomorph.2008.03.003
[4]	AYKUT A, SERHAT D, FILKRI B. Landslide susceptibility mapping for a landslide-prone area(Findikli, NE of Turkey) by likelihood-frequency ratio and weighted linear combination models[J]. Environmental Geology, 2008, 54(6): 1127-1143. doi: 10.1007/s00254-007-0882-8
[5]	KAYASTHA P, DHITAL M R, DE S F. Application of the analytical hierarchy process(AHP) for landslide susceptibility mapping: A case study from the Tinau watershed, westNepal[J]. Computers & Geosciences, 2013, 52: 398-408.
[6]	SHARAD K G, DERICKS P S, MANOJ T. Selection of weightages for causative factors used in preparation of landslide susceptibility zonation(LSZ)[J]. Geomatics Natural Hazards & Risk, 2018, 9(1): 471-487.
[7]	张俊, 殷坤龙, 王佳佳, 等. 三峡库区万州区滑坡灾害易发性评价研究[J]. 岩石力学与工程学报, 2016, 35(2): 284-296. ZHANG J, YIN K L, WANG J J, et al. Evaluation of landslide susceptibility in Wanzhou District of the Three Gorges Reservoir Area[J]. Chinese Journal of Rock Mechanics and Engineering, 2016, 35(2): 284-296. (in Chinese with English abstract)
[8]	王念秦, 薛瑶琼, 李少兵, 等. 基于粗糙集理论的泥石流易发性综合评判模型[J]. 水土保持研究, 2014, 21(3): 246-250. WANG N Q, XUE Y Q, LI S B, et al. A comprehensive evaluation model of debris flow susceptibility based on rough set theory[J]. Soil and Water Conservation Research, 2014, 21(3): 246-250. (in Chinese with English abstract)
[9]	刘彦花, 叶国华. 基于粗糙集与GIS的滑坡地质灾害风险评估: 以广西梧州为例[J]. 灾害学, 2015, 30(2): 108-114. LIU Y H, YE G H. Landslide geological hazard risk assessment based on rough set and GIS: A case study of Wuzhou, Guangxi[J]. Catastrophology, 2015, 30(2): 108-114. (in Chinese with English abstract)
[10]	陈骏骏, 秦胜伍, 李广杰, 等. 基于RS-IVM的吉林省泥石流灾害易发性评价研究[J]. 应用基础与工程科学学报, 2021, 29(6): 1359-1371. CHEN J J, QIN S W, LI G J, et al. Evaluation of debris flow susceptibility in Jilin Province based on RS-IVM[J]. Journal of Applied Basic and Engineering Sciences, 2021, 29(6): 1359-1371. (in Chinese with English abstract)
[11]	DOU Q, QIN S W, ZHANG Y C, et al. A method for improving controlling factors based on information fusion for debris flow susceptibility mapping: A case study in Jilin Province, China[J]. Entropy, 2019, 21(7): 695. doi: 10.3390/e21070695
[12]	王世宝, 庄建琦, 樊宏宇, 等. 基于频率比与集成学习的滑坡易发性评价: 以金沙江上游巴塘-德格河段为例[J]. 工程地质学报, 2021, 30(3): 817-828. WANG S B, ZHUANG J Q, FAN H Y, et al. Landslide susceptibility evaluation based on frequency ratio and ensemble learning: Taking the Batang-Dege reach in the upper reaches of the Jinsha River as an example[J]. Chinese Journal of Engineering Geology, 2021, 30(3): 817-828(in Chinese with English abstract)
[13]	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. Prediction and modeling law of landslide susceptibility 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
[14]	张钟远, 邓明国, 徐世光, 等. 镇康县滑坡易发性评价模型对比研究[J]. 岩石力学与工程学报, 2022, 41(1): 157-171. ZHANG Z Y, DENG M G, XU S G, et al. Comparative study on evaluation models of landslide susceptibility in Zhenkang County[J]. Chinese Journal of Rock Mechanics and Engineering, 2022, 41(1): 157-171. (in Chinese with English abstract)
[15]	黄发明, 胡松雁, 闫学涯, 等. 基于机器学习的滑坡易发性预测建模及其主控因子识别[J]. 地质科技通报, 2022, 41(2): 79-90. doi: 10.19509/j.cnki.dzkq.2021.0087 HUANG F M, HU S Y, YAN X Y, et al. Machine learning-based landslide susceptibility prediction modeling and identification of main controlling factors[J]. Bulletin of Geological Science and Technology, 2022, 41(2): 79-90. (in Chinese with English abstract) doi: 10.19509/j.cnki.dzkq.2021.0087
[16]	尹留志. 关于非平衡数据特征问题的研究[D]. 合肥: 中国科学技术大学, 2014. YIN L Z. Research on the problem of unbalanced data characteristics[D]. Hefei: University of Science and Technology of China, 2014. (in Chinese with English abstract)
[17]	戴福初, 姚鑫, 谭国焕. 滑坡灾害空间预测支持向量机模型及其应用[J]. 地学前缘, 2007, 14(6): 153-159. DAI F C, YAO X, TAN G H. Support vector machine model for landslide disaster spatial prediction and its application[J]. Earth Science Frontiers, 2007, 14(6): 153-159. (in Chinese with English abstract)
[18]	徐伟, 郑玄, 欧文, 等. 四川凉山州地质灾害灾情特征与主要致灾类型[J/OL]. 中国地质灾害与防治学报, 2023: 1-12. XU W, ZHENG X, OU W, et al. Characteristics of losses of geological disasters and major disaster types in Liangshan Prefecture, Sichuan Province[J/OL]. The Chinese Journal of Geological Hazard and Control, 2023: 1-12. (in Chinese with English abstract)
[19]	SAHA A K, GUPTA R P, SARKAR I, et al. An approach for GIS-based statistical landslide susceptibility zonation with a case study in the Himalayas[J]. Landslides, 2005, 2(1): 61-69. doi: 10.1007/s10346-004-0039-8
[20]	PAWLAK Z. Rough set approach to knowledge-based decision support[J]. European Journal of Operational Research, 1997, 99(1): 58-57. doi: 10.1016/S0377-2217(96)00383-9
[21]	刘彦花, 叶国华. 基于粗糙集与GIS的滑坡地质灾害风险评估: 以广西梧州为例[J]. 灾害学, 2015, 30(2): 108-114. doi: 10.3969/j.issn.1000-811X.2015.02.021 LIU Y H, YE G H. Risk assessment of landslide geological hazards based on rough sets and GIS: Taking Wuzhou, Guangxi as an example[J]. Disaster Science, 2015, 30(2): 108-114. (in Chinese with English abstract) doi: 10.3969/j.issn.1000-811X.2015.02.021
[22]	郎秋玲, 张以晨, 张继权, 等. 基于组合赋权理论的泥石流孕灾因子分析[J]. 灾害学, 2019, 34(1): 68-72. LANG Q L, ZHANG Y C, ZHANG J Q, et al. Analysis of debris flow hazard producing factors based on combination weighting theory[J]. Disaster Science, 2019, 34(1): 68-72. (in Chinese with English abstract)
[23]	LEO B. Random forests[J]. Machine Learning, 2001, 45(1): 5-32. doi: 10.1023/A:1010933404324
[24]	闫举生, 谭建民. 基于不同因子分级法的滑坡易发性评价: 以湖北远安县为例[J]. 中国地质灾害与防治学报, 2019, 30(1): 52-60. YAN J S, TAN J M. Landslide susceptibility evaluation based on different factor classification methods: Taking Yuanan County, Hubei as an example[J]. Chinese Journal of Geological Hazards and Prevention, 2019, 30(1): 52-60. (in Chinese with English abstract)
[25]	李郎平, 兰恒星, 郭长宝, 等. 基于改进频率比法的川藏铁路沿线及邻区地质灾害易发性分区评价[J]. 现代地质, 2017, 31(5): 911-929. LI L P, LAN H X, GUO C B, et al. Zoning evaluation of geological disaster susceptibility along Sichuan-Tibet railway and adjacent areas based on improved frequency ratio method[J]. Geoscience, 2017, 31(5): 911-929. (in Chinese with English abstract)
[26]	郭子正, 殷坤龙, 黄发明, 等. 基于滑坡分类和加权频率比模型的滑坡易发性评价[J]. 岩石力学与工程学报, 2019, 38(2): 287-300. GUO Z Z, YIN K L, HUANG F M, et al. Landslide susceptibility evaluation based on landslide classification and weighted frequency ratio model[J]. Journal of Rock Mechanics and Engineering, 2019, 38(2): 287-300. (in Chinese with English abstract)
[27]	吴润泽, 胡旭东, 梅红波, 等. 基于随机森林的滑坡空间易发性评价: 以三峡库区湖北段为例[J]. 地球科学, 2021, 46(1): 321-330. WU R Z, HU X D, MEI H B, et al. Evaluation of landslide spatial susceptibility based on random forest: A case study of Hubei section of the Three Gorges Reservoir Area[J]. Earth Science, 2021, 46(1): 321-330. (in Chinese with English abstract)
[28]	刘月, 王宁涛, 周超, 等. 基于ROC曲线与确定性系数法集成模型的三峡库区奉节县滑坡易发性评价[J]. 安全与环境工程, 2020, 27(4): 61-70. LIU Y, WANG N T, ZHOU C, el at. Evaluation of landslide susceptibility in Fengjie County, Three Gorges Reservoir Area based on integrated model of ROC curve and deterministic coefficient method[J]. Safety and Environmental Engineering, 2020, 27(4): 61-70. (in Chinese with English abstract)
[29]	连志鹏, 徐勇, 付圣, 等. 采用多模型融合方法评价滑坡灾害易发性: 以湖北省五峰县为例[J]. 地质科技通报, 2020, 39(3): 178-186. doi: 10.19509/j.cnki.dzkq.2020.0319 LIAN Z P, XU Y, FU S, et al. Evaluation of landslide hazard susceptibility using multi model fusion method: Taking Wufeng County of Hubei Province as an example[J]. Bulletin of Geological Science and Technology, 2020, 39(3): 178-186. (in Chinese with English abstract) doi: 10.19509/j.cnki.dzkq.2020.0319