不可移动文物是我国文化遗产的重要组成部分，蕴含着极高的历史、艺术和文化价值。土遗址作为不可移动文物的一种，涵盖了从人类文明起源至今的多个历史时期，是我国重要的文化遗产。然而土遗址长期以来受到自然侵蚀和人类活动的双重影响，遭受了严重破坏。植被根系对土遗址的影响尤为显著，其根系生长过程中会对土遗址产生根劈作用，使得土体裂隙的扩大、土壤结构的破坏，从而加剧土遗址退化。本研究以陕西省榆林市榆阳区、神木市和佳县为研究区，该地区气候干旱，土遗址遗存较多，为了给该地区的土遗址提出更好的保护建议，本文应用深度学习模型对研究区植被类型进行提取分类，探讨了植被类型生态敏感性，构建土遗址综合生态敏感性模型，并划分小流域单元，提出了相应的保护模式。研究结论如下：

（1）根据研究区植被根系对土遗址的破坏程度，本文将植被类型对土遗址的影响从强到弱划分为乔木植被、沙地植被、灌木植被、农田植被和草地植被。为了获得较好的分类效果，本文利用哨兵2（Sentinel-2）影像对这五种类型植被进行提取，并采用四种机器学习模型和三种深度学习网络进行性分析。在机器学习模型中，随机森林表现出最高的精度，其准确率、召回率、F1值和Kappa系数分别为81.85%、82.02%、81.88%和0.78，优于其他模型。在深度学习部分，DeepLabV3+模型的精度最佳，优于U-Net和Transformer模型。为了进一步提高精度，本文在DeepLabV3+模型基础上进行了改进，采用了ResNet50主干网络、改进的DenseASPP模块及CoordAtt注意力机制进行特征融合，提出了DeepLabV3+RDF模型。与原DeepLabV3+模型相比，改进后的DeepLabV3+RDF模型在准确率、F1值和均交并比上分别提高了2.81%、2.84%和2.21%。在假彩色数据集下，DeepLabV3+RDF模型的准确率和F1值分别为89.03%和88.45%，显著优于随机森林模型。

（2）根据DeepLabV3+RDF模型提取的植被分类结果，结合层次分析法-CRITIC（AHP-CRITIC）组合定权法，构建了土遗址生态敏感性评价体系，从生物多样性、自然环境和人类活动等方面选择了13个指标，并利用地理信息和遥感技术进行单因子敏感性和综合生态敏感性分析。研究区内不敏感区和低敏感区面积约7855.39 km²，占46.99%，分布22处土遗址，适宜开发文化旅游设施；中敏感区面积5467.19 km²，占32.7%，分布14处土遗址；高敏感区面积3149.27 km²，占18.84%，分布15处土遗址，需加强保护；极高敏感区面积245.15 km²，占1.47%，分布4处土遗址，需严格限制开发。空间自相关分析结果表明，土遗址综合生态敏感性的全局莫兰指数为0.660，Z值为157.855，具有显著的空间正相关性和较强的空间聚集性。高-高聚集区主要分布在佳县以及榆阳区的西南部和东南部，低-低聚集区集中在神木市中西部和榆阳区中部。通过地理探测器分析，水域缓冲区对生态敏感性的解释力最高，q值为0.3081，植被类型的q值为0.1098，对土遗址综合生态敏感性的影响发挥着关键作用。交互探测结果表明，水域缓冲区与降雨侵蚀的交互作用对生态敏感性的解释力最强，q值达0.52，增强了对生态敏感性的解释能力。

（3）基于综合生态敏感性分析结果，本研究从生态保护与土遗址耦合的视角出发，采用K均值聚类法提出了土遗址保护模式，以小流域为划分单元，确定了92个核心土遗址小流域。根据流域属性分类指标，采用K均值聚类方法将流域划分为五种保护类型：土遗址+生态保护（16.30%）、土遗址+地形特征（15.22%）、土遗址+人口密集（13.04%）、矿产开采+土壤保护（34.78%）以及风力作用（20.65%）。对每种类型的综合生态敏感性进行计算与分级，并提出相应的保护与发展模式。其中，“土遗址+生态保护”为极高敏感性类型，应以生态保护为核心，采取水土保持措施并发展生态旅游；“土遗址+地形特征”为高敏感性类型，可以对北坡加强植被恢复，南坡则需人工干预，适度发展绿色能源；“土遗址+人口密集”为中敏感类型，应以文化旅游、社区参与和绿色产业为核心，避免过度开发；“矿产开采+土壤保护”为低敏感类型，需严格规划矿产开采，推动绿色能源和生态农业；“风力”敏感性最低，应建立防风林带，发展风力发电和绿色产业，推动经济增长。

       本文从生态视角出发，系统分析了榆林北部地区土遗址生态敏感性的分布情况，考虑了植被根系对土遗址的影响。通过构建深度学习DeepLabV3+RDF模型提取植被类型并进行分级，进而依据植被分类结果构建土遗址生态敏感性模型。采用小流域划分法与聚类方法，提出了五种土遗址流域类型。针对不同类型的小流域及敏感性分区，提出了相应的土遗址保护模式与发展策略，以实现生态保护与文化遗址的可持续发展，为土遗址保护与生态安全提供了参考。

     Immovable cultural relics are an important part of China's cultural heritage and contain high historical, artistic and cultural values. As a kind of immovable cultural relics, earthen sites cover many historical periods from the origin of human civilization to the present, and are an important cultural heritage of China. However, earthen sites have long suffered serious damage from both natural erosion and human activities. The impact of vegetation roots on earth sites is particularly significant, and the growth of its root system can lead to root splitting of earth sites, resulting in the expansion of soil fissures and the destruction of soil structure, thus exacerbating the degradation of earth sites. In this study, Yulin City, Shaanxi Province, Yuyang District, Shenmu City and Jia County as the study area, due to the arid climate in the region and more remains of earth sites, in order to provide better protection suggestions for the earth sites in the region, this paper applies a deep learning model to extract and classify the vegetation types in the study area, discusses the ecological sensitivity of vegetation types, constructs a comprehensive ecological sensitivity model of the earth sites, and divides the sub-watershed unit, and proposes a corresponding. The model is divided into sub-watershed units, and the corresponding protection model is proposed. The conclusions of the study are as follows:

（1）Based on the degree of damage to the earth site by the vegetation root system in the study area, this paper classifies the impact of vegetation types on the earth site from strong to weak into trees, sandy vegetation, shrubs, farmland and grassland. To achieve optimal classification results, this study employs Sentinel-2 imagery to extract the five vegetation types and applies four machine learning models in conjunction with three deep learning networks for comprehensive analysis. Among the machine learning models, Random Forest showed the highest precision with an accuracy , recall, F1 value and Kappa coefficient of 81.85%, 82.02%, 81.88% and 0.78, respectively, which were better than the other models. In the deep learning part, the DeepLabV3+ model has the best accuracy, which is better than the U-Net and Transformer models. In order to further improve the accuracy, this paper improves on the DeepLabV3+ model by using ResNet50 backbone network, improved DenseASPP module and CoordAttention mechanism for feature fusion, and proposes the DeepLabV3+ RDF model. Compared with the original DeepLabV3+ model, the improved DeepLabV3+RDF model improves 2.81%, 2.84% and 2.21% in accuracy, F1 value and mean intersection and parallel ratio (MIOU), respectively. Under the false color dataset, the DeepLabV3+RDF model significantly outperforms the Random Forest model with 89.03% accuracy and 88.45% F1 value, respectively.

（2）Based on the vegetation classification results extracted from the DeepLabV3+RDF model, and combined with the hierarchical analysis method CRITIC (AHP-CRITIC) weighting approach, an ecological sensitivity evaluation system for earth sites was constructed. Thirteen indices were selected from aspects of biodiversity, the natural environment, and human activities. Geographic information and remote sensing technologies were used to perform both single-factor sensitivity and comprehensive ecological sensitivity analyses. In the study area, the insensitive and low-sensitive zones cover approximately 7,855.39 km² (46.99%) and contain 22 earth sites, which are suitable for cultural tourism development. The moderately sensitive zones span 5,467.19 km² (32.7%) and include 14 sites. The highly sensitive zones cover 3,149.27 km² (18.84%) and contain 15 sites, which require enhanced protection. The extremely highly sensitive zones cover 245.15 km² (1.47%) and include 4 sites, where development should be strictly restricted. Spatial autocorrelation analysis results show a global Moran’s I value of 0.660 and a Z value of 157.855, indicating significant positive spatial correlation and strong spatial aggregation. High-high aggregation areas are primarily located in the southwest and southeast of Yuyang District and Jiaxian County, while low-low aggregation areas are concentrated in the central and western parts of Shenmu City and central Yuyang District. Geodetector analysis revealed that the watershed buffer zone had the highest explanatory power for ecological sensitivity, with a q-value of 0.3081. Vegetation type, with a q-value of 0.1098, also played a significant role in influencing the integrated ecological sensitivity of the earth site. Interaction detection results showed that the interaction between watershed buffer and rainfall erosion had the strongest explanatory power for ecological sensitivity, with a q-value of 0.52, enhancing the overall explanatory power of ecological sensitivity.

（3）Based on the results of the comprehensive ecological sensitivity analysis, this study proposes a model of earth site protection from the perspective of coupling ecological protection and earth sites using the K-mean clustering method, and identifies 92 core earth site sub-watersheds using sub-watersheds as the dividing unit. Based on the classification index of watershed attributes, the K-mean clustering method was used to classify the watersheds into five protection types: earth site + ecological protection (16.30%), earth site + topographic features (15.22%), earth site + dense population (13.04%), mineral extraction + soil protection (34.78%), and wind action (20.65%).The comprehensive ecological sensitivity of each type was calculated and graded, and the corresponding protection and development model was proposed. Among them, "earth site + ecological protection" is a very high sensitivity type, which should take ecological protection as the core, adopt soil and water conservation measures and develop eco-tourism; "earth site + topographic features" is a high sensitivity type, which can strengthen vegetation restoration on the north slope, while the south slope needs artificial intervention and moderate development of green tourism. Artificial intervention and moderate development of green energy are needed on the southern slope; "earth site + dense population" is a medium-sensitive type, which should focus on cultural tourism, community participation and green industry to avoid over-development; "mineral mining + soil protection" is a low-sensitive type, which needs strict planning of mineral "Mineral Mining + Soil Protection" is a low-sensitivity type that requires strict planning of mineral mining and the promotion of green energy and eco-agriculture; "Wind Power" is the least sensitive type, and windbreaks should be established to develop wind power generation and green industries to promote economic growth.

     This paper systematically analyzes the distribution of ecological sensitivity of earth sites in the northern region of Yulin from an ecological perspective, and considers the influence of vegetation on earth sites. The vegetation types are extracted and graded by constructing a deep learning DeepLabV3+RDF model, and then the ecological sensitivity model of earth sites is constructed based on the vegetation results. Five types of earth site watersheds were proposed by using the sub-watershed division method and clustering method. According to the different types of sub-watersheds and sensitivity areas, corresponding protection models and development strategies are proposed to realize the ecological protection and sustainable development of cultural heritage sites, which provides a reference for the protection and ecological safety of earth heritage sites.

[1] 陈同滨, 王力军. 不可移动文物保护规划十年[J]. 中国文化遗产, 2004(03):108–111.

[2] 黄克忠. 岩土文物建筑的保护[M]. 岩土文物建筑的保护, 1998.

[3] 刘炜. 我国北方7省室外土遗址病害分布特征研究[D]. 西北大学, 2012.

[4] 樊东, 黄光琦, 杨海龙. 预防性保护理念下的土遗址保护——以统万城遗址为例[J]. 中国文化遗产, 2020(03):54–57.

[5] 汪万福, 武发思, 陈拓, 等. 遗产地植物与遗产保护间关系研究进展[J]. 敦煌研究, 2011(06):101-108+132.

[6] 杜昱民. 青海明长城防御体系及典型遗址易损性评价[D]. 兰州大学, 2019.

[7] 杨小菊, 武发思, 贾荣亮, 等. 植物对岩土文物的作用及保护技术研究进展[J]. 文物保护与考古科学, 2023, 35(05):150–164.

[8] 武发思, 朱非清, 汪万福, 等. 日本高松冢古坟微生物病害及其防治研究概述[J]. 文物保护与考古科学, 2019, 31(03):26–35.

[9] TROTTA G, SAVO V, CICINELLI E, et al. Colonization and damages of Ailanthus altissima (Mill.) Swingle on archaeological structures: Evidence from the Aurelian Walls in Rome (Italy)[J]. International Biodeterioration & Biodegradation, 2020, 153:105054.

[10] 赵晓文. 植物对土遗址裂隙的影响及其作用机理研究[D]. 西北农林科技大学, 2014.

[11] 武发思, 杜维波, 杨小菊, 等. 瓜州锁阳城遗址植物群落特征及与土遗址保护间关系[J]. 西北大学学报(自然科学版), 2022, 52(04):667–677.

[12] H.-Y. WANG, D.-L. PAN, D.-S. XIA. Fast algorithm for two-dimensional Otsu adaptive threshold algorithm[J]. ACTA AUTOMATICA SINICA, 2007, 33(9):968–971.

[13] 李丹, 陈水森, 陈修治. 高光谱遥感数据植被信息提取方法[J]. 农业工程学报, 2010, 26(07):181-185+386.

[14] GILBERTSON J K, KEMP J, VAN NIEKERK A. Effect of pan-sharpening multi-temporal Landsat 8 imagery for crop type differentiation using different classification techniques[J]. Computers and Electronics in Agriculture, 2017, 134:151–159.

[15] CHAN J C-W, PAELINCKX D. Evaluation of Random Forest and Adaboost tree-based ensemble classification and spectral band selection for ecotope mapping using airborne hyperspectral imagery[J]. Remote Sensing of Environment, 2008, 112(6):2999–3011.

[16] HINTON G E, SALAKHUTDINOV R R. Reducing the Dimensionality of Data with Neural Networks[J]. Science, 2006, 313(5786): 504-507.

[17] 戴鹏钦, 丁丽霞, 刘丽娟, 等. 基于FCN的无人机可见光影像树种分类[J]. 激光与光电子学进展, 2020, 57(10):36–45.

[18] YANG M-D, TSENG H-H, HSU Y-C, et al. Semantic Segmentation Using Deep Learning with Vegetation Indices for Rice Lodging Identification in Multi-date UAV Visible Images[J]. Remote Sensing, 2020, 12(4):633.

[19] MORALES G, KEMPER G, SEVILLANO G, et al. Automatic Segmentation of Mauritia flexuosa in Unmanned Aerial Vehicle (UAV) Imagery Using Deep Learning[J]. Forests, 2018, 9(12):736.

[20] 刘春娟, 辛钰强, 吴小所, 等. 双注意力引导的U-Net++遥感图像语义分割模型[J/OL]. 北京航空航天大学学报, 1-13[2025-04-11]

[21] 孙常建, 尚永福, 王石岩, et al. 基于语义分割网络的冬小麦遥感分类及变化分析[J]. 测绘通报, 2024(10):151–156.

[22] 欧阳志云,王效科,苗鸿. 中国生态环境敏感性及其区域差异规律研究[J]. 生态学报, 2000(01):10–13.

[23] 麦克哈格, 芮经纬. 设计结合自然 / (美)伊恩·伦诺克斯·麦克哈格(Ian Lennox McHarg)著 ; 芮经纬译[M]. Tian jin: 天津大学出版社, 2006.

[24] 郑文发. 基于GIS的城镇居住用地生态适宜性评价研究[D]. 华东师范大学, 2010.

[25] 康娅琳. 武汉市城市湖泊湿地生态敏感性研究[D]. 华中农业大学, 2010.

[26] HORNE R, HICKEY J. Ecological sensitivity of Australian rainforests to selective logging[J]. Australian Journal of Ecology, 1991, 16(1):119–129.

[27] BUTT M A, NISAR K, MAHMOOD S A, et al. Toward GIS-Based Approach for Identification of Ecological Sensitivity Areas: Multi-Criteria Evaluation Technique for Promotion of Tourism in Soon Valley, Pakistan[J]. Journal of the Indian Society of Remote Sensing, 2019, 47(9):1527–1536.

[28] POLETTO F, FARINA B, CARCIONE J M. Sensitivity of seismic properties to temperature variations in a geothermal reservoir[J]. Geothermics, 2018, 76:149–163.

[29] GHOSH S, DAS CHATTERJEE N, DINDA S. Urban ecological security assessment and forecasting using integrated DEMATEL-ANP and CA-Markov models: A case study on Kolkata Metropolitan Area, India[J]. Sustainable Cities and Society, 2021, 68:102773.

[30] ADAMCZYK J, DYLEWSKI R. Analysis of the sensitivity of the ecological effects for the investment based on the thermal insulation of the building: A Polish case study[J]. Journal of Cleaner Production, 2017, 162:856–864.

[31] WIIK E, BENNION H, SAYER C D, et al. Ecological sensitivity of marl lakes to nutrient enrichment: evidence from Hawes Water, UK[J]. Freshwater Biology, 2015, 60(11):2226–2247.

[32] ROSE K C, BIERWAGEN B, BRIDGHAM S D, et al. Indicators of the effects of climate change on freshwater ecosystems[J]. Climatic Change, 2023, 173(23):1–20.

[33] KWIZERA V, GENÇER N S, GÜNDOĞDU K S, et al. Impacts of Rural Land Use and Insect Ecological Sensitivity in Burundi[J]. Journal of Agriculture and Ecology Research International, 2022:60–71.

[34] LIN Y, RONG Y, LI L, et al. Spatiotemporal impacts of climate change and human activities on water resources and ecological sensitivity in the Mekong subregion in Cambodia[J]. Environmental Science and Pollution Research, 2023, 30(2):4023–4043.

[35] YU G, XIE M, LIANG J, et al. A GIS-based 3D slope stability analysis method based on the assumed normal stress on the slip surface[J]. Scientific Reports, 2020, 10(1):4442.

[36] 李振亚, 魏伟, 周亮, 等.中国陆地生态敏感性时空演变特征[J]. 地理学报, 2022, 77(01):150–163.

[37] 李明慧, 周启刚, 孟浩斌, 等. 基于最小累积阻力模型的三峡库区重庆段生态安全格局构建[J]. 长江流域资源与环境, 2021, 30(08):1916–1926.

[38] 王曦, 程三友, 李英杰, 等. 基于Landsat TM和OLI数据的30年间土地利用变化及生态敏感性研究——以西安市秦岭段鄠邑区为例[J]. 第四纪研究, 2022, 42(06):1655–1672.

[39] 王浩, 马星, 杜勇. 基于生态系统服务重要性和生态敏感性的广东省生态安全格局构建[J]. 生态学报, 2021, 41(05):1705–1715.

[40] 胡西武, 贾天朝. 基于生态敏感性与景观连通性的三江源国家公园生态安全格局构建与优化[J]. 长江流域资源与环境, 2023, 32(08):1724–1735.

[41] 方臣, 匡华, 贾琦琪, 等. 基于生态系统服务重要性和生态敏感性的武汉市生态安全格局评价[J]. 环境工程技术学报, 2022, 12(05):1446–1454.

[42] 何苏玲, 邹凤琼, 王金亮. 基于AHP和MSE赋权法的龙南县生态敏感性评价[J]. 生态学杂志, 2021, 40(09):2927–2935.

[43] 李芮芝, 胡希军, 杜心宇, 等. 基于SRP模型的南雄丹霞梧桐自然保护区生态脆弱性评价[J]. 西北林学院学报, 2021, 36(05):152–160.

[44] 郑群明, 陈子奇, 陈奕昊, 等. 生态敏感性约束下的国家公园游憩适宜性评价研究——以南山国家公园为例[J]. 生态经济, 2024, 40(11):119–127.

[45] 霍海鹰, 任书樣, 彭可平, 等. 山西省煤矿区生态现状评价[J]. 煤炭工程, 2024, 56(01):32–40.

[46] 刘海龙, 王炜桥, 王跃飞, 等. 汾河流域生态敏感性综合评价及时空演变特征[J]. 生态学报, 2021, 41(10):3952–3964.

[47] 魏婵娟, 蒙吉军. 中国土地资源生态敏感性评价与空间格局分析[J]. 北京大学学报(自然科学版), 2022, 58(01):157–168.

[48] 王永祥, 魏佳轩, 周娟, 等. 基于GIS和AHP方法的藏东南生态敏感性研究——以林芝地区为例（英文）[J]. Journal of Resources and Ecology, 2023, 14(01):158–166.

[49] 王硕. 辽宁省生态敏感性评价研究[J]. 国土与自然资源研究, 2021(01):35–38.

[50] 黄晓芬. 文物保护的思想[J]. 考古与文物, 1995(02):86–90.

[51] 世界各国文物保护的历史发展概况[J]. 瞭望新闻周刊, 1994(03):17–18.

[52] 王石斌. 北方土遗址的病害成因与环境区划研究[D]. 兰州大学, 2009.

[53] 周环. 潮湿环境土遗址的加固保护研究[D]. 浙江大学, 2008.

[54] 田林. 大遗址遗迹保护问题研究[D]. 天津大学, 2004.

[55] PARK H y. HERITAGE TOURISM[J]. Annals of Tourism Research, 2010, 37(1):116–135.

[56] GAN Z, LONG L, ZHANG D. STUDY ON CONSERVATION AND RESTORATION OF LIUGOU ANCIENT TOWN EARTHEN SITES IN BEIJING (CHINA)[J]. HERITAGE2020 (3DPast | RISK-Terra) International Conference, 2020, XLIV-M-1-2020:1097–1101.

[57] PORTER D, MEHROTRA A, DEJONG M, et al. Material and Seismic Assessment of the Great House at Casa Grande Ruins National Monument, Arizona[J]. Journal of Architectural Engineering, 2020, 26(1): 05019007.

[58] RICHARDS J, VILES H, GUO Q. The importance of wind as a driver of earthen heritage deterioration in dryland environments[J]. Geomorphology, 2020, 369:107363.

[59] MASOUD NAKHAEI, MARIANA CORREIA. Assessment of vulnerability and site adaptive capacity to the risk of climate change: the case of Tchogha Zanbil World Heritage earthen site in Iran[J]. Journal of Cultural Heritage Management and Sustainable Development, 2022, 12(2):107–125.

[60] 鲍小会. 中国现代文物保护意识的形成[J]. 文博, 2000(03):75–80.

[61] 鲜乔蓥. 民国初期的文物保护政策与措施[J]. 西华大学学报(哲学社会科学版), 2008(02):46–49.

[62] 李最雄 王旭东. 古代土建筑遗址保护加固研究的新进展[J]. 敦煌研究, 1997(04):169–174.

[63] 郭青林, 裴强强, 王彦武, 等. 我国土遗址保护研究新进展[J]. 石窟与土遗址保护研究, 2024, 3(03):18–31.

[64] CHEN W W, ZHANG Q Y, LIU H W, et al. Feasibility of protecting earthen sites by infiltration of modified polyvinyl alcohol[J]. Construction and Building Materials, 2019, 204:410–418.

[65] 张博. 不同气候环境下土遗址防风化技术适应性研究[D]. 兰州大学, 2021.

[66] 李黎, 陈锐, 邵明申, 等. 经PS加固土遗址水饱和强度及加固效果的环境影响研究[J]. 岩石力学与工程学报, 2009, 28(05):1074–1080.

[67] 周双林, 原思训. 有机硅改性丙烯酸树脂非水分散体的制备及在土遗址保护中的试用[J]. 文物保护与考古科学, 2004(04):50–52.

[68] ZHANG J, CHEN W, LI Z, et al. Study on workability and durability of calcined ginger nuts-based grouts used in anchoring conservation of earthen sites[J]. Journal of Cultural Heritage, 2015, 16(6):831–837.

[69] WANG N, ZHANG J, WANG Y, et al. Experimental Study on Mechanical Properties of Grout–Soil Interface in Anchor System of Rammed Earthen Sites[J]. International Journal of Geomechanics, 2020, 20(6):4020064.

[70] 孙满利, 李最雄, 王旭东, 等. 交河故城垛泥墙体裂隙注浆工艺研究[J]. 文物保护与考古科学, 2013, 25(01):1–5.

[71] 张景科, 王南, 樊孟, 等. 烧料礓石改性遗址土裂隙注浆材料龄期性能试验研究[J]. 岩石力学与工程学报, 2018, 37(01):220–229.

[72] PEI Q, ZHANG B, SHANG D, et al. Characteristics of Temperature Field of Rammed Earth Wall in Arid Environment[J]. Coatings, 2022, 12(6):735.

[73] WANG X, PEI Q, GUO Q, et al. Stress mechanism for the rammed layer interfaces of earthen heritage sites with different treatments[J]. Journal of Cultural Heritage, 2019, 39:110–119.

[74] WANG X, ZHANG B, PEI Q, et al. Experimental studies on sacrificial layer in conservation of earthen sites[J]. Journal of Cultural Heritage, 2020, 41:74–83.

[75] 王旭东. 基于风险管理理论的莫高窟监测预警体系构建与预防性保护探索[J]. 敦煌研究, 2015(01):104–110.

[76] 刘海. 西安城墙预防性保护研究[D]. 西北大学, 2020.

[77] 彭守璋. 中国1km逐月潜在蒸散发数据集（1901-2023）. 国家青藏高原科学数据中心, 2024.

[78] 车涛,戴礼云,李新. 中国雪深长时间序列数据集（1979-2024）. 国家青藏高原科学数据中心, 2015.

[79] 彭守璋. 中国1km分辨率逐月降水量数据集（1901-2023）. 国家青藏高原科学数据中心, 2024.

[80] YANG J, HUANG X. The 30 m annual land cover dataset and its dynamics in China from 1990 to 2019[J]. Earth System Science Data, 2021, 13(8):3907–3925.

[81] VAPNIK V N. Statistical learning theory[M]. New York: Wiley, 1998.

[82] OOMMEN T, MISRA D, TWARAKAVI N K C, et al. An Objective Analysis of Support Vector Machine Based Classification for Remote Sensing[J]. Mathematical Geosciences, 2008, 40(4):409–424.

[83] ATKINSON P M, MASSARI R. GENERALISED LINEAR MODELLING OF SUSCEPTIBILITY TO LANDSLIDING IN THE CENTRAL APENNINES, ITALY[J]. Computers & Geosciences, 1998, 24(4):373–385.

[84] CHEN T, GUESTRIN C. XGBoost : A Scalable Tree Boosting System [C]//Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 2016: 785-794.

[85] FRIEDMAN J H. Greedy function approximation: A gradient boosting machine[J]. The Annals of Statistics, 2001, 29(5):1189–1232.

[86] FRIEDMAN J H. Stochastic gradient boosting[J]. Computational Statistics & Data Analysis, 2002, 38(4):367–378.

[87] FAN Z, XU Y, ZHANG D. Local linear discriminant analysis framework using sample neighbors[J]. IEEE transactions on neural networks, 2011, 22(7):1119–1132.

[88] BREIMAN L. Random Forests[J]. Machine Learning, 2001, 45(1):5–32.

[89] KARAKAS G, CAN R, KOCAMAN S, et al. LANDSLIDE SUSCEPTIBILITY MAPPING WITH RANDOM FOREST MODEL FOR ORDU, TURKEY[J]. HERITAGE2020 (3DPast | RISK-Terra) International Conference, 2020, XLIII-B3-2020:1229–1236.

[90] SUN X, CHEN J, LI Y, et al. Landslide Susceptibility Mapping along a Rapidly Uplifting River Valley of the Upper Jinsha River, Southeastern Tibetan Plateau, China[J]. Remote Sensing, 2022, 14(7):1730.

[91] RONNEBERGER O, FISCHER P, BROX T. U-Net: Convolutional Networks for Biomedical Image Segmentation [C]//Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18. Springer international publishing, 2015: 234-241.

[92] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale [J]. arXiv preprint arXiv:2010.11929, 2020.

[93] CHEN L-C, PAPANDREOU G, KOKKINOS I, et al. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs[J]. IEEE transactions on pattern analysis and machine intelligence, 2018, 40(4):834–848.

[94] 陈彦光. 基于Moran统计量的空间自相关理论发展和方法改进[J]. 地理研究, 2009, 28(06):1449–1463.

[95] 张广纳, 邵景安, 王金亮, 等. 三峡库区重庆段农村面源污染时空格局演变特征[J]. 自然资源学报, 2015, 30(07):1197–1209.

[96] 环办生态〔2017〕48号 关于印发《生态保护红线划定指南》的通知 [Z].2017 2-2[M].

[97] 王劲峰, 徐成东. 地理探测器:原理与展望[J]. 地理学报, 2017, 72(01):116–134.

[98] 巩国丽, 刘纪远, 邵全琴. 草地覆盖度变化对生态系统防风固沙服务的影响分析——以内蒙古典型草原区为例[J]. 地球信息科学学报, 2014, 16(03):426–434.

[99] 申陆, 田美荣, 高吉喜, 等. 浑善达克沙漠化防治生态功能区防风固沙功能的时空变化及驱动力[J]. 应用生态学报, 2016, 27(01):73–82.

[100]迟文峰, 白文科, 刘正佳, 等. 基于RWEQ模型的内蒙古高原土壤风蚀研究[J]. 生态环境学报, 2018, 27(06):1024–1033.

[101]YANG M, YU K, ZHANG C, et al. DenseASPP for Semantic Segmentation in Street Scenes[C] //. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2018 - 2018: 3684–3692.

[102]XU Y, LIU R, XUE C, et al. Ecological Sensitivity Evaluation and Explanatory Power Analysis of the Giant Panda National Park in China[J]. Ecological Indicators, 2023, 146:109792.

[103]鲁敏, 穆回港, 谭蕾, 等. 基于GIS的济西国家湿地公园生态敏感性评价[J]. 中国海洋大学学报(自然科学版), 2022, 52(12):95–103.

[104]SHI Y, FAN X, DING X, et al. An assessment of ecological sensitivity and landscape pattern in abandoned mining land[J]. Sustainability, 2024, 16(3):1105.

[105]汪佳灿, 张红艳, 谢聪颖, 等. 基于GIS的大运河文化带非遗廊道构建适宜性评价研究[J]. 生态学报, 2025(11):1–18.

[106]柴莎莎, 刘玉, 任艳敏, 等. 生态脆弱区拟建光伏基地的生态敏感性评价及分区研究——以祁连县为例[J]. 生态学报, 2025(09):1–13.

[107]高梦雯, 胡业翠, 李向, 等. 基于生态系统服务重要性和环境敏感性的喀斯特山区生态安全格局构建——以广西河池为例[J]. 生态学报, 2021, 41(07):2596–2608.

[108]邓雪, 李家铭, 曾浩健, 等. 层次分析法权重计算方法分析及其应用研究[J]. 数学的实践与认识, 2012, 42(07):93–100.

[109]孙苑苑, 王琳, 王晋. 黄河三角洲自然保护区生态敏感性评价[J]. 中国海洋大学学报(自然科学版), 2017, 47(11):96–102.

[110] DIAKOULAKI D, MAVROTAS G, PAPAYANNAKIS L. Determining objective weights in multiple criteria problems: The critic method[J]. Computers & Operations Research, 1995, 22(7):763–770.

[111] 高丽霞, 闫治攀, 李喜香, 等. 基于指纹图谱结合AHP-CRITIC综合评分法优选紫连生肌凝胶剂提取工艺[J]. 中国中医药信息杂志, 2023, 30(08):136–141.

[112] 朱峰, 张宏伟. 基于“AHP+熵权法”的CW-TOPSIS冲击地压评判模型[J]. 中国安全科学学报, 2017, 27(01):128–133.

[113] 李益敏, 管成文, 朱军. 基于GIS的星云湖流域生态敏感性评价[J]. 水土保持研究, 2017, 24(05):266-271+278.

[114] DUAN SHU-HUAI. Unit for Beijing Information System of Soil and Water Conservation-Small Watershed Division in Beijing Mountain Area[J]. Res Soil Water Conserv, 2004, 2: 45-46.

[115] SPAKE R, LASSEUR R, CROUZAT E, et al. Unpacking ecosystem service bundles: Towards predictive mapping of synergies and trade-offs between ecosystem services[J]. Global Environmental Change, 2017, 47:37–50.

[116] CAPÓ M, PÉREZ A, LOZANO J A. An efficient approximation to the K-means clustering for massive data[J]. Knowledge-Based Systems, 2017, 117:56–69.

[117] MU Jiaqi, ZHUANG Jianqi, WANG Shibao, KONG Jiaxu, DU Chenhui, MU Jiaqi, ZHUANG Jianqi, et al. Evaluation of landslide susceptibility in Ya’an City based on depth neural network model[J]. The Chinese Journal of Geological Hazard and Control, 2023, 34(3):157–168.