Objective Deeply excavated expansive soil canal slopes frequently exhibit significant deformation and poor stability due to the combined effects of excavation unloading, wet‒dry cycles, and groundwater level fluctuations. Under the influence of multiple environmental factors, the deformation mechanisms of these slopes are complex and demonstrate pronounced spatiotemporal heterogeneity. Deformation serves as a direct indicator of slope stability, making the analysis of deformation patterns and trends essential for stability assessment. In current engineering practice, deformation data are primarily obtained through discrete point-based monitoring, which captures information only at critical cross-sections or localized areas of the slope. This approach leaves large portions of the slope unmonitored, resulting in limited coverage and creating blind spots in deformation assessment. In addition, the dense deployment of monitoring points increases construction complexity and significantly raises project costs. Therefore, a key challenge in long-distance canal slope safety monitoring involves improving deformation monitoring coverage and cost-effectiveness while maintaining sufficient precision. InSAR technology provides high spatiotemporal resolution, broad coverage, non-contact monitoring capability, and strong adaptability to complex terrain, addressing many limitations associated with conventional point-based methods. This study integrates Small Baseline Subset InSAR (SBAS‒InSAR) with traditional deformation monitoring techniques to obtain multi-scale deformation data, enabling deformation analysis from regional scales to localized sections. In addition, the study develops a self-explaining neural network (SENN) model incorporating an attention mechanism to predict canal slope deformation. The model examines the deformation characteristics, dominant controlling factors, and evolutionary mechanisms of deeply excavated expansive soil slopes, facilitating multi-source data-driven analysis of deformation behavior and its influencing factors. This approach provides scientific support for the safety monitoring and stability evaluation of canal slopes. Methods First, SBAS‒InSAR was employed to process Sentinel‒1A satellite imagery to extract time-series deformation data. The derived displacements were projected onto the vertical direction and were cross-validated with ground-based vertical displacement measurements to evaluate the reliability of the SBAS‒InSAR results. Based on this validation, deformation rate thresholds were established to identify high-risk canal segments, which enabled an analysis of the overall slope deformation trends. Then, a SENN model incorporating an attention mechanism was developed to predict slope deformation. Key influencing factors, including groundwater level, canal water level, air temperature, precipitation, and time-dependent effects, were selected as input variables. The SENN model autonomously extracted critical features and dynamically assigned weights to each factor, predicting cumulative displacements along both the satellite Line-of-Sight (LOS) direction and the inclinometer A-direction. This approach enabled a comprehensive analysis of deformation patterns at the slope surface and at various subsurface depths while clarifying the dominant controlling factors governing expansive soil canal slope behavior. Results and Discussions The study focused on a deeply excavated expansive soil canal section (Stake 8+88612+921) of a major water diversion project and produced the following key findings: 1) The analysis of 128 Sentinel‒1A images from March 2017 to December 2021 revealed that the deformation data obtained through the SBAS‒InSAR method showed a discrepancy of less than 3 mm compared to surface vertical displacement measurements at monitoring points (fourth-stage berm at 9+363 and third-stage berm at 11+400), while the deformation trends remained consistent. This result confirmed the reliability of SBAS‒InSAR for monitoring canal slope deformation. 2) The study section exhibited LOS deformation rates ranging from ‒10 to 24 mm/a, which indicated overall uplift characteristics, with positive LOS deformation showing an annual increase. Using a threshold of 10 mm/a, critical deformation zones were identified, particularly within the 11+80612+921 section, where most areas experienced deformations exceeding 40 mm and reached a maximum value of 97 mm. 3) Inclinometer measurements indicated maximum displacements of 49.51 mm at a depth of 4.5 m below the orifice at the second-stage berm (11+715), 63.80 mm at 1 m below the first-stage berm (11+762), and 67.51 mm at a depth of 1.5 m at the third-stage berm (11+762), with deformation extending to 13.5 m below the orifice. 4) LOS direction deformation was primarily influenced by groundwater level, canal water level, and time-dependent effects. Groundwater level exerted a stronger influence on right-bank deformation, whereas the canal water level more significantly affected left-bank deformation. At the third-stage berm (11+762), A-direction deformation was predominantly controlled by time-dependent effects, followed by temperature effects, while canal water level and rainfall exerted minimal influence. Groundwater level significantly affected surface soils but showed a reduced impact on deeper soil layers. Shallow deformations in expansive soil slopes were mainly controlled by groundwater level, temperature, and time-dependent effects, whereas displacements below a depth of 3.5 m showed a stronger correlation with temperature and time-dependent effects. 5) The 11+700 ~ 11+800 section exhibited continuous surface and internal deformation, with right-bank slopes showing significantly greater deformation than left-bank slopes due to higher groundwater levels. Deformation magnitudes were greater in the first to fourth stage slopes than in the fifth and sixth stage slopes. Deformation mechanisms varied with depth. Deeper soils, which were constrained by overburden pressure, were less affected by groundwater fluctuations, whereas shallow soils showed high sensitivity to both groundwater level variations and temperature changes. The study demonstrated that uplift deformation in this canal section resulted from the combined effects of excavation unloading and hydro-mechanical coupling in expansive soils. Conclusions The SBAS‒InSAR technique effectively monitors large-scale canal slope deformation trends by overcoming the spatial limitations inherent in conventional point-based monitoring methods, which often suffer from restricted coverage and sparse measurement points. The attention-mechanism-based SENN prediction model demonstrates high accuracy in forecasting slope deformation patterns while quantitatively evaluating the relative contributions of various influencing factors. This multi-source data integration approach provides comprehensive insights into the deformation mechanisms of expansive soil canal slopes and their controlling factors, providing a scientific basis for slope safety monitoring. The research outcomes serve as valuable references for the long-term monitoring and engineering management of expansive soil canal slopes and propose innovative methodologies for the safety monitoring of similar engineering projects. The demonstrated technical framework exhibits significant practical value for engineering applications, particularly in addressing deformation monitoring challenges in large-scale water conveyance infrastructure.
GongBiwei, ChengZhanlin, GuoXiling,et al.Research progress on the stability of expansive soil slope in the mi-ddle route of south-to-north water diversion project[J].Jou-rnal of Yangtze River Scientific Research Institute,2011,28(10):134‒140.
LiQingyun, ChengZhanlin, GongBiwei,et al.Failure mechanism and treatment technology of expansive soil slope of middle route project[J].Journal of Yangtze River Scienti-fic Research Institute,2009,26(11):1‒9.
LengTing, TangChaosheng, XuDan,et al.Advance on the engineering geological characteristics of expansive soil[J].Journal of Engineering Geology,2018,26(1):112‒128. doi:10.13544/j.cnki.jeg.2018.01.013
GuoJianhua, WangHanhui, LiShichang,et al.Evolution pa-ttern of cracks in expansive soil and the deformation of canal slopes in dry and wet environments[J].Rock and Soil Mechanics,2024,45(S1):433‒442.
HuangXiaohu, HuYulong, GuoFei,et al.Study on the mechanism of landslide promotion of expansive soil under the co-mbined action of rainfall and reservoir water[J].Advanced Engineering Science,2026,58(2):109‒122.
LiuLongwu, ZhengJianlong, MiaoWei.Researches on characteristics of swell and shrinking active zone of expansive soil and failure mechanism of landslides[J].Chinese Journal of Geotechnical Engineering,2008,30(1):28‒33.
HuJiang, LiXing.Analysis of spatiotemporal deformation characteristics of deep excavated expansive soil slopes[J].Rock and Soil Mechanics,2024,45(10):3071‒3080.
HuJiang, LiXing, MaFuheng.Deformation causes of deep excavated expansive soil canal slope during operation period[J].Journal of Yangtze River Scientific Research Institute,2023,40(11):160‒167. doi:10.11988/ckyyb.20220795
HuJiang, YangHongwei, LiXing,et al.Slope deformation ch-aracteristics and influencing factors of deeply excavated expansive soil canal slopes with high groundwater level during operation periods[J].Advances in Science and Technology of Water Resources,2022,42(5):94‒101.
XuQiang, ZhuXing, LiWeile,et al.Technical progress of space‒air‒ground collaborative monitoring of landslide[J].Acta Geodaetica et Cartographica Sinica,2022,51(7):1416‒1436.
KurkaM, GutjahrK H.Observation of expansive clay mo-vement with DInSAR[M]//Engineering Geology for Society and Territory—Volume 5.Cham:Springer International Publishing,2014:151‒154. doi:10.1007/978-3-319-09048-1_30
[26]
ZhangYongguang, TianFan, LiYingchun,et al.Application of time series InSAR to deformation monitoring in central line project of south-to-north water transfer[J].Journal of Yangtze River Scientific Research Institute,2021,38(8):72‒77. doi:10.11988/ckyyb.20191501
ZhangShuangcheng, SiJinzhao, XuYongfu,et al.Time-series InSAR for stability monitoring of Ankang Airport with expansive soil[J].Geomatics and Information Science of Wuhan Un-iversity,2021,46(10):1519‒1528.
ZhuWu, DouHao, YinNazheng,et al.Deformation characteristics analysis of the expansive soil slope by integrating of InSAR and SSA techniques:A case study of the South-to-North Water Diversion Project[J].Acta Geodaetica et Cartographica Sinica,2022,51(10):2083‒2092. doi:10.11947/j.AGCS.2022.20220357
LiZiyuan, HuJun, ZhangXing,et al.Characterization of elastoplastic behavior and retrieval of active zone depth for expansive soil slopes in the middle-route channel head of the South-to-North Water Diversion Project,China,using InSAR time series[J].Remote Sensing of Environment,2023,295:113666. doi:10.1016/j.rse.2023.113666
[33]
LiuHui, LiShihuan, MiaoChangwei,et al.Multi-platform InSAR three-dimensional deformation monitoring of He-nan Changge section of south-to-north water diversion pr-oject[J].Journal of North China University of Water Res-ources and Electric Power(Natural Science Edition),2025,46(5):1‒13.
BerardinoP, FornaroG, LanariR,et al.A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms[J].IEEE Transactions on Geoscience and Remote Sensing,2002,40(11):2375‒2383. doi:10.1109/tgrs.2002.803792
[36]
ZhuJianjun, LiZhiwei, HuJun.Research progress and me-thods of InSAR for deformation monitoring[J].Acta Geodaetica et Cartographica Sinica,2017,46(10):1717‒1733.
Alvarez‒MelisD, JaakkolaT S.Towards robust interpretability with self-explaining neural networks[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems.Montréal:Curran Associates,Inc.,2018:7775‒7784.
[39]
DhurandharA, ChenP, LussR,et al.Explanations based on the missing:Towards contrastive explanations with pertin-ent negatives[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems.Montréal:Curran Associates,Inc.,2018:892‒902.
[40]
ZhangZaixi, LiuQi, WangHao,et al.ProtGNN:Towards self-explaining graph neural networks[J].Proceedings of the AAAI Conference on Artificial Intelligence,2022,36(8):9127‒9135. doi:10.1609/aaai.v36i8.20898
[41]
ZhuZhangli, RaoYuan, WuYuan,et al.Research progress of attention mechanism in deep learning[J].Journal of Chinese Information Processing,2019,33(6):1‒11. doi:10.3969/j.issn.1003-0077.2019.06.001
RenQiubing, LiMingchao, ShenYang,et al.Dynamic monitoring model for dam deformation with spatiotemporal co-upling correlation characteristics[J].Journal of Hydroelect-ric Engineering,2021,40(10):160‒172.
ZhangWenfeng.Study on slope deformation of expansive soil canal section in the middle route of south-to-north water diversion project[J].Yellow River,2019,41(7):131‒135. doi:10.3969/j.issn.1000-1379.2019.07.028
HuJiang, ZhangJikang, YuMengxue,et al.Characteristics analysis and prediction of slope deformation of deeply-excavated expansive soil canal[J].Hydro‒Science and Engineering,2021(4):1‒9. doi:10.12170/20201201001
WangXiaobo, CaiYaojun, LiLiang,et al.Failure characteristics and mechanism of excavated expansive soil slope of MRP of South-to-North Water Diversion[J].Yangtze River,2015,46(1):26‒29. doi:10.16232/j.cnki.1001-4179.2015.01.007
MaXiaowen, AiYingbo.Review of soil behaviors under excavation unloading of foundation pits[J].Chinese Journal of Geotechnical Engineering,2011,33(S1):189‒193.
LiuJingde, LiQingyun, GongBiwei.Swelling properties of expansive rock in Middle Route Project of South-to-North Water Diversion[J].Chinese Journal of Geotechnical Engineering,2011,33(5):826‒830.
HuJiang, MaFuheng, LiXing.Analysis of deformation formation causes and reinforcement disposal effects in deep excavated expansive soil canal slope:Case study[J].International Journal of Geomechanics,2023,23(5):05023003. doi:10.1061/ijgnai.gmeng-8265
[57]
HuJiang, LiXing.Deformation mechanism and treatment effect of deeply excavated expansive soil slopes with high groundwater level:Case study of MR‒SNWTP,China[J].Transportation Geotechnics,2024,46:101253. doi:10.1016/j.trgeo.2024.101253