Landslide Susceptibility Assessment Model Based on Positive Sample Buffering with Transformer-CNN Integration：A Case Study of Baoji City

QU Wei; BIAN Zice; LI Jiuyuan; TANG Xingyou; CHEN Peinan

doi:10.13203/j.whugis20250295

您当前的位置：

首页 >

文章列表页 >

Landslide Susceptibility Assessment Model Based on Positive Sample Buffering with Transformer-CNN Integration：A Case Study of Baoji City

Photogrammetry and Remote Sensing | 更新时间：2026-11-05

- Landslide Susceptibility Assessment Model Based on Positive Sample Buffering with Transformer-CNN Integration：A Case Study of Baoji City
- Geomatics and Information Science of Wuhan University Vol. 51, Issue 4, Pages: 726-740(2026)
- 作者机构：
  
  长安大学地质工程与测绘学院，陕西西安，710054
- 作者简介：
- 基金信息：
- DOI：10.13203/j.whugis20250295
  CLC： P237
- Received：14 October 2025，
  
  Published：05 April 2026
- 稿件说明：
移动端阅览
瞿伟,边子策,李久元，等.融合Transformer与CNN的正样本缓冲滑坡易发性评价模型：以宝鸡市为例[J].武汉大学学报（信息科学版）,2026,51(4):726-740.

QU Wei,BIAN Zice,LI Jiuyuan,et al.Landslide Susceptibility Assessment Model Based on Positive Sample Buffering with Transformer-CNN Integration:A Case Study of Baoji City[J].Geomatics and Information Science of Wuhan University,2026,51(4):726-740.
瞿伟,边子策,李久元，等.融合Transformer与CNN的正样本缓冲滑坡易发性评价模型：以宝鸡市为例[J].武汉大学学报（信息科学版）,2026,51(4):726-740. DOI：10.13203/j.whugis20250295

QU Wei,BIAN Zice,LI Jiuyuan,et al.Landslide Susceptibility Assessment Model Based on Positive Sample Buffering with Transformer-CNN Integration:A Case Study of Baoji City[J].Geomatics and Information Science of Wuhan University,2026,51(4):726-740. DOI：10.13203/j.whugis20250295

摘要

宝鸡市滑坡灾害频发，但当前该区域滑坡编录数据样本较少，同时若仅采用滑坡点样本训练模型会导致模型的空间代表性不足，从而制约对该区域滑坡易发性评价的精度。针对此，结合卷积神经网络模型（convolutional neural network

CNN ）的局部特征提取优势和注意力机制Transformer的全局建模能力，提出一种融合Transformer 与CNN 的正样本缓冲滑坡易发性评价模型。依据滑坡规模设置90~130 m动态缓冲区扩展正样本，并综合考虑地形地貌、地质条件、水文气象和人类工程等选取13类滑坡影响因子，通过多重共线性分析后构建了滑坡评价体系。研究结果显示，采用缓冲区将随机森林、CNN、Transformer、Transformer-CNN模型四者的ROC曲线下的面积(area under the curve，AUC)从0.834、0.852、0.847、0.875分别提升至0.883、0.913、0.926、0.959。此外，Transformer-CNN相较CNN、Transformer，未缓冲时AUC分别从0.852、0.847提升至0.875，进行缓冲时分别从0.913、0.926提升至0.959；基于夏普利加性解释算法可解释性分析进一步揭示出岩性、年降雨、坡向三类因子对滑坡易发性预测贡献度最大，贡献度分别达0.55、0.47、0.43，且三者交互效应显著，为锁定区域滑坡高易发区提供了可量化的依据。

Abstract

Objectives

Landslide susceptibility mapping (LSM) serves as a fundamental tool for geohazard prevention and regional planning. However

the accuracy of LSM in complex geological environments

such as Baoji City

is often constrained by two primary challenges: The scarcity of historical landslide samples leading to poor spatial representation

and the limitation of traditional convolutional neural networks (CNN) in capturing long-range spatial dependencies. To address these issues

we propose a novel framework that integrates a positive sample buffering strategy with a hybrid Transformer-CNN model. The primary objective is to enhance the spatial representation of landslide samples and effectively fuse local and global features to improve the predictive performance and interpretability of LSM.

Methods

First

to overcome the limitations of point-based sampling which fails to reflect the actual spatial extent of landslides

a dynamic buffering strategy was implemented. Based on the scale of historical landslides

variable buffer radii ranging from 90 m to 130 m were applied to expand positive samples from isolated points into spatial polygons. Second

a total of 13 conditioning factors were selected

including topography (elevation

slope

aspect

plan curvature

profile curvature)

geology (lithology

distance to faults)

hydrology (distance to rivers

topographic wetness index

rainfall)

ecological environment (normalized difference vegetation index) and human activities (land use

distance to roads). Multicollinearity analysis

utilizing both Pearson correlation coefficients and variance inflation factor (VIF)

was conducted to ensure the statistical independence of these factors. Third

a Transformer-CNN coupled model was constructed. The architecture utilizes a CNN backbone with a 15×15 input window to extract high-dimensional local features

which are then serialized and fed into a Transformer encoder. The Transformer employs a multi-head self-attention mechanism to dynamically weigh the importance of different spatial positions

thereby capturing global non-linear interactions among factors. To rigorously evaluate the proposed framework

a total of eight experimental scenarios were designed

comparing the performance of random forest

standalone CNN

standalone Transformer

and the coupled Transformer-CNN model under both buffered and unbuffered sample conditions. Furthermore

the Shapley additive explanations (SHAP) algorithm was integrated to interpret the "“black-box”decision process of the optimal Transformer-CNN model

quantifying the contribution of each factor to landslide susceptibility.

Results

Experimental validation demonstrate significant improvements in both model robustness and prediction accuracy. (1) Effectiveness of buffering strategy: The application of the buffering strategy universally enhanced the performance of all tested models. Specifically

the area under the curve values for RF

CNN

Transformer

and Transformer-CNN increased from 0.834

0.852

0.847

and 0.875 (unbuffered) to 0.883

0.913

0.926

and 0.959 (buffered)

respectively. (2) Superiority of the hybrid model: The proposed Transformer-CNN model consistently outperformed single-structure models in both experimental settings. In the unbuffered scenario

it improved the AUC from 0.852 (CNN) and 0.847 (Transformer) to 0.875; in the buffered scenario

it further elevated the AUC from 0.913 (CNN) and 0.926 (Transformer) to 0.959. (3) Interpretability analysis: Based on SHAP interpretability analysis

it was further revealed that lithology

annual precipitation

and slope aspect contributed most significantly to predicting landslide susceptibility in Baoji City

with contributions of 0.55

0.47

and 0.43

respectively. The interaction effects among these three factors were significant.

Conclusions

The positive sample buffering strategy effectively mitigates the issue of insufficient spatial representation in historical inventories. The Transformer-CNN coupled architecture successfully overcomes the intrinsic limitations of single-structure models

offering a state-of-the-art solution for complex non-linear geological modeling. This study establishes a robust framework for landslide susceptibility assessment in data-scarce regions.These findings provide actionable scientific insights for local disaster mitigation and land-use planning.

关键词

Keywords

references

MERGHADI A ， YUNUS A P ， DOU J ， et al . Machine Learning Methods for Landslide Susceptibility Studies： A Comparative Overview of Algorithm Performance ［J］. Earth-Science Reviews ， 2020 ， 207 ： 103225 .

ADO M ， AMITAB K ， MAJI A K ， et al . Landslide Susceptibility Mapping Using Machine Learning： A Literature Survey ［J］. Remote Sensing ， 2022 ， 14 （ 13 ）： 3029 .

JIANG Z Y ， WANG M ， LIU K . Comparisons of Convolutional Neural Network and Other Machine Learning Methods in Landslide Susceptibility Assessment： A Case Study in Pingwu ［J］. Remote Sensing ， 2023 ， 15 （ 3 ）： 798 .

郭飞，赖鹏，黄发明，等 . 基于知识图谱的滑坡易发性评价文献综述及研究进展［J］. 地球科学， 2024 ， 49 （ 5 ）： 1584 - 1606 .

GUO Fei ， LAI Peng ， HUANG Faming ， et al . Literature Review and Research Progress of Landslide Susceptibility Mapping Based on Knowledge Graph ［J］. Earth Science ， 2024 ， 49 （ 5 ）： 1584 - 1606 .

赵占骜，王继周，毛曦，等 . 多维CNN耦合的滑坡易发性评价方法［J］. 武汉大学学报（信息科学版）， 2024 ， 49 （ 8 ）： 1466 - 1481 .

ZHAO Zhan'ao ， WANG Jizhou ， MAO Xi ， et al . A Multi-dimensional CNN Coupled Landslide Susceptibility Assessment Method ［J］. Geomatics and Information Science of Wuhan University ， 2024 ， 49 （ 8 ）： 1466 - 1481 .

ZHANG Q ， WANG T . Deep Learning for Exploring Landslides with Remote Sensing and Geo-Environmental Data： Frameworks， Progress， Challenges， and Opportunities ［J］. Remote Sensing ， 2024 ， 16 （ 8 ）： 1344 .

王守华，王睿菘，孙希延，等 . 基于CF-CNN-LSTM模型的滑坡易发性评价［J］. 自然灾害学报， 2024 ， 33 （ 5 ）： 84 - 95 .

WANG Shouhua ， WANG Ruisong ， SUN Xiyan ， et al . Landslide Susceptibility Evaluation Based on CF-CNN-LSTM Model ［J］. Journal of Natural Disasters ， 2024 ， 33 （ 5 ）： 84 - 95 .

伊廷婧汶，黄才生，覃勇，等 . 基于北斗监测数据的高陡边坡变形Transformer-CNN预测模型［J］. 重庆大学学报， 2025 ， 48 （ 10 ）： 81 - 94 .

YI Tingjingwen ， HUANG Caisheng ， QIN Yong ， et al . Transformer-CNN Prediction Model of High and Steep Slope Deformation Based on Beidou Detection Data ［J］. Journal of Chongqing University （Natural Science Edition）， 2025 ， 48 （ 10 ）： 81 - 94 .

秦佳松，李为乐，周胜森，等 . 考虑InSAR地表形变的滑坡易发性评价负样本选取与结果优化［J］. 武汉大学学报（信息科学版）， 2025 ， DOI： 10.13203/j.whugis20240494 http://dx.doi.org/10.13203/j.whugis20240494 .

QIN Jiasong ， LI Weile ， ZHOU Shengsen ， et al . Optimizing Negative Sample Selection and Results for Landslide Susceptibility Assessment Using InSAR Surface Deformation ［J］. Geomatics and Information Science of Wuhan University ， 2025 ， DOI： 10.13203/j.whugis20240494 http://dx.doi.org/10.13203/j.whugis20240494 .

杨仲康，张世殊，邓建辉，等 . 斜坡单元尺寸和滑坡取样方式对区域滑坡易发性评价的影响［J］. 岩石力学与工程学报， 2025 ， 44 （ 3 ）： 602 - 617 .

YANG Zhongkang ， ZHANG Shishu ， DENG Jianhui ， et al . The Impact of Slope Unit Scale and Landslide Sampling Methods on Regional Landslide Susceptibility Assessment ［J］. Chinese Journal of Rock Mechanics and Engineering ， 2025 ， 44 （ 3 ）： 602 - 617 .

刘纪平，梁恩婕，徐胜华，等 . 顾及样本优化选择的多核支持向量机滑坡灾害易发性分析评价［J］. 测绘学报， 2022 ， 51 （ 10 ）： 2034 - 2045 .

LIU Jiping ， LIANG Enjie ， XU Shenghua ， et al . Multi-Kernel Support Vector Machine Considering Sample Optimization Selection for Analysis and Evaluation of Landslide Disaster Susceptibility ［J］. Acta Geodaetica et Cartographica Sinica ， 2022 ， 51 （ 10 ）： 2034 - 2045 .

武雪玲，杨经宇，牛瑞卿 . 一种结合SMOTE和卷积神经网络的滑坡易发性评价方法［J］. 武汉大学学报（信息科学版）， 2020 ， 45 （ 8 ）： 1223 - 1232 .

WU Xueling ， YANG Jingyu ， NIU Ruiqing . A Landslide Susceptibility Assessment Method Using SMOTE and Convolutional Neural Network ［J］. Geomatics and Information Science of Wuhan University ， 2020 ， 45 （ 8 ）： 1223 - 1232 .

孔嘉旭，庄建琦，彭建兵，等 . 基于信息量和卷积神经网络的黄土高原滑坡易发性评价［J］. 地球科学， 2023 ， 48 （ 5 ）： 1711 - 1729 .

KONG Jiaxu ， ZHUANG Jianqi ， PENG Jianbing ， et al . Evaluation of Landslide Susceptibility in Chinese Loess Plateau Based on IV-RF and IV-CNN Coupling Models ［J］. Earth Science ， 2023 ， 48 （ 5 ）： 1711 - 1729 .

朱阿兴，闾国年，周成虎，等 . 地理相似性：地理学的第三定律？［J］. 地球信息科学学报， 2020 ， 22 （ 4 ）： 673 - 679 .

ZHU Axing ， LÜ Guonian ， ZHOU Chenghu ， et al . Geographic Similarity： Third Law of Geography？［J］. Journal of Geo-Information Science ， 2020 ， 22 （ 4 ）： 673 - 679 .

HE Y ， ZHAO Z A ， YANG W ， et al . A Unified Network of Information Considering Superimposed Landslide Factors Sequence and Pixel Spatial Neighbourhood for Landslide Susceptibility Mapping ［J］. International Journal of Applied Earth Observation and Geoinformation ， 2021 ， 104 ： 102508 .

刘坚，李树林，陈涛 . 基于优化随机森林模型的滑坡易发性评价［J］. 武汉大学学报（信息科学版）， 2018 ， 43 （ 7 ）： 1085 - 1091 .

LIU Jian ， LI Shulin ， CHEN Tao . Landslide Susceptibility Assesment Based on Optimized Random Forest Model ［J］. Geomatics and Information Science of Wuhan University ， 2018 ， 43 （ 7 ）： 1085 - 1091 .

瞿伟，李昕，李久元，等 . 一种改进自适应步长人工蜂群优化随机森林的滑坡风险性评价方法［J］. 大地测量与地球动力学， 2024 ， 44 （ 12 ）： 1211 - 1219 .

QU Wei ， LI Xin ， LI Jiuyuan ， et al . A Landslide Risk Assessment Method with Improved Adaptive Step Size Artificial Bee Colony Optimization Random Forest ［J］. Journal of Geodesy and Geodynamics ， 2024 ， 44 （ 12 ）： 1211 - 1219 .

郑德凤，高敏，闫成林，等 . 基于卷积神经网络的滑坡易发性评价：以辽南仙人洞国家级自然保护区为例［J］. 地球科学， 2024 ， 49 （ 5 ）： 1654 - 1664 .

ZHENG Defeng ， GAO Min ， YAN Chenglin ， et al . Susceptibility Assessment of Landslides Based on Convolutional Neural Network Model： A Case Study from Xianrendong National Nature Reserve in Southern Liaoning Province ［J］. Earth Science ， 2024 ， 49 （ 5 ）： 1654 - 1664 .

ZHAO Z ， CHEN J H . A Robust Discretization Method of Factor Screening for Landslide Susceptibility Mapping Using Convolution Neural Network， Random Forest， and Logistic Regression Models ［J］. International Journal of Digital Earth ， 2023 ， 16 （ 1 ）： 408 - 429 .

LI Y ， ZENG J B ， SHAN S G ， et al . Occlusion Aware Facial Expression Recognition Using CNN with Attention Mechanism ［J］. IEEE Transactions on Image Processing ， 2019 ， 28 （ 5 ）： 2439 - 2450 .

丁明涛，江辉，李振洪，等 . 融合卷积模糊神经网络与SHAP特征优化的地震诱发滑坡易发性评价［J］. 武汉大学学报（信息科学版）， 2025 ， DOI： 10.13203/J.whugis20250017 http://dx.doi.org/10.13203/J.whugis20250017 .

DING Mingtao ， JIANG Hui ， LI Zhenhong ， et al . Earthquake-Induced Landslide Susceptibility Assessment by Fusing Convolutional Fuzzy Neural Network and SHAP Feature Optimization Strategy ［J］. Geomatics and Information Science of Wuhan University ， 2025 ， DOI： 10.13203/J.whugis20250017 http://dx.doi.org/10.13203/J.whugis20250017 .

李文权，邓安鑫，郑炎，等 . 基于机器学习的中型城市居民出行方式选择行为研究［J］. 交通运输系统工程与信息， 2024 ， 24 （ 2 ）： 13 - 23 .

LI Wenquan ， DENG Anxin ， ZHENG Yan ， et al . Analysis of Residents' Travel Mode Choice in Medium-Sized City Based on Machine Learning ［J］. Journal of Transportation Systems Engineering and Information Technology ， 2024 ， 24 （ 2 ）： 13 - 23 .

李磊 . 宝鸡市滑坡发育特征与易发性评价研究［D］. 北京：中国地质大学（北京）， 2023 .

LI Lei . Study on Development Characteristics and Vulnerability Evaluation of Landslide in Baoji City ［D］. Beijing ： China University of Geosciences（Beijing）， 2023 .

石玲，王涛，辛鹏 . 陕西省宝鸡市地质灾害发育特征［J］. 地质力学学报， 2013 ， 19 （ 4 ）： 351 - 363 .

SHI Ling ， WANG Tao ， XIN Peng . Development Characteristics of the Geo-Harzards in Baoji City， Shaanxi Province ［J］. Journal of Geomechanics ， 2013 ， 19 （ 4 ）： 351 - 363 .

LI L ， XU C ， YANG Z Q ， et al . An Inventory of Large-Scale Landslides in Baoji City， Shaanxi Province， China ［J］. Data ， 2022 ， 7 （ 8 ）： 114 .

ZHUANG J Q ， PENG J B ， WANG G H ， et al . Distribution and Characteristics of Landslide in Loess Plateau： A Case Study in Shaanxi Province ［J］. Engineering Geology ， 2018 ， 236 ： 89 - 96 .

唐晓娜 . 基于卷积神经网络和综合指数模型的吕梁市滑坡灾害易发性评价［D］. 太原：太原理工大学， 2019 .

TANG Xiaona . Evaluation of Landslide Susceptibility in Lüliang City Based on Convolutional Neural Network and Comprehensive Index Model ［D］. Taiyuan ： Taiyuan University of Technology ， 2019 .

YANG S ， LI D Y ， SUN Y Q ， et al . Effect of Landslide Spatial Representation and Raster Resolution on the Landslide Susceptibility Assessment ［J］. Environmental Earth Sciences ， 2024 ， 83 （ 4 ）： 132 .

王世宝，庄建琦，郑佳，等 . 基于深度学习的CZ铁路康定-理塘段滑坡易发性评价［J］. 工程地质学报， 2022 ， 30 （ 3 ）： 908 - 919 .

WANG Shibao ， ZHUANG Jianqi ， ZHENG Jia ， et al . Landslide Susceptibility Evaluation Based on Deep Learning Along Kangding-Litang Section of CZ Railway ［J］. Journal of Engineering Geology ， 2022 ， 30 （ 3 ）： 908 - 919 .

DORMANN C F ， ELITH J ， BACHER S ， et al . Collinearity： A Review of Methods to Deal with It and a Simulation Study Evaluating Their Performance ［J］. Ecography ， 2013 ， 36 （ 1 ）： 27 - 46 .

陈飞，蔡超，李小双，等 . 基于信息量与神经网络模型的滑坡易发性评价［J］. 岩石力学与工程学报， 2020 ， 39 （ S1 ）： 2859 - 2870 .

CHEN Fei ， CAI Chao ， LI Xiaoshuang ， et al . Evaluation of Landslide Susceptibility Based on Information Quantity and Neural Network Model ［J］. Chinese Journal of Rock Mechanics and Engineering ， 2020 ， 39 （ S1 ）： 2859 - 2870 .

LU J G ， HE Y ， ZHANG L F ， et al . A Synergistic CNN-DF Method for Landslide Susceptibility Assessment ［J］. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing ， 2025 ， 18 ： 6584 - 6599 .

LIU S L ， WANG L Q ， ZHANG W G ， et al . A Comprehensive Review of Machine Learning-Based Methods in Landslide Susceptibility Mapping ［J］. Geological Journal ， 2023 ， 58 （ 6 ）： 2283 - 2301 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Deep Learning-Based Joint Local and Non-local Phase Filtering for InSAR Images

Spatiotemporal Evolution and Explainable Analysis of Grassland Vulnerability in the Hohhot-Baotou-Ordos-Yulin Urban Agglomeration

InSAR Detection and Spatiotemporal Characteristics of Active Landslides in the Maqin Section of the Upper Yellow River

A Crime Prediction Model Incorporating Regional Spatial Similarity Characteristics and Spatiotemporal Characteristics of Events

A Parameter Genetic Algorithm for Evolutionary Calibration in Coupled Dynamics Models of Geological Hazard Chain

Related Author

WANG Jie

LIU Jiahang

LING Xinpeng

DUAN Zexian

DU Peijun

ZHAO Wenpeng

GUO Shanchuan

LI Chunqiang

Related Institution

Key Laboratory of Space Photoelectric Detection and Sensing of Ministry of Industry and Information Technology

College of Astronautics, Nanjing University of Aeronautics and Astronautics

Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural Resources

Jiangsu Provincial Key Laboratory for Advanced Remote Sensing and Geographic Information Technology

School of Geography and Ocean Science, Nanjing University

AI问答

⁰