School of Earth Sciences,Yunnan University
Yunnan University Engineering Research Center of High-resolution Satellite Remote Sensing, Yunnan University
Institute of International Rivers and Ecol-Security, Yunnan University

目的目的怒江州是典型的高山峡谷地区,地质灾害(滑坡)频发,严重制约着当地的发展。为解决这一问题,方法方法综合考虑怒江州实际情况,从气象水文、地形地貌、地层岩性、植被生态和人类活动5个方面选取坡向,高程等14个影响因子,判断滑坡与各影响因子间相关性,构建评价指标体系,对最大熵(maximumentropy,MaxEnt)模型的特征类(featurecombination,FC)和正则化乘数(regularizationmultiplier,RM)参数进行优化,对比优化前后小样本赤池信息量准则(akaikeinformationcriterion,AIC)、遗漏率(omissionrate,OR)和AUC(areaundercurve),并基于优化的MaxEnt模型预测滑坡灾害的发生,实现怒江州滑坡易发性评价。结果结果结果表明:优化后的MaxEnt模型在研究区滑坡易发性预测中适用性优秀(AUC=0.913);运用刀切法(Jackknife)计算各影响因子对易发性的影响程度,高程(S3,23.2%)、坡度(S9,22.4%)、居民点密度(S5,14.2%)、距河流距离(S13,13.7%)、距道路距离(S4,9.6%)和岩性(S7,8.7%)是前六位影响因子,累计贡献度达91.8%;极高、高、中、低滑坡易发性等级的空间占比分别为4.88%,8.96%,18.40%,67.76%,县域中极高和高易发区占比最大的是泸水市,整体上看,极高、高易发区主要沿河流和道路分布于峡谷中,低易发区主要分布于人类活动少、河谷不发育的区域。结论结论优化后的MaxEnt模型更适合怒江州滑坡易发性预测,研究结果可为怒江州防灾减灾与土地利用规划提供参考。

Objectives Nujiang Prefecture is a typical alpine canyon area, frequent landscape geological di‑sasters seriously restricts local development. Methods To solve this problem, taking into account the actual situation in Nujiang Prefecture,14 influencing factors such as slope direction and elevation were selected from five aspects of meteorology and hydrology, topography and geo ⁃ morphology, stratigraphic lithology, vegetation ecology and human activities to judge the correlation between landslides and each influencing factor, and build an evaluation index system. The MaxEnt model feature combination (FC) and regulariza‑tion multi ⁃ plier (RM) parameters were optimized, and the sample of the akaike information criterion(AICc), Omission Rate (OR) and AUC (Area Under Curve) value before optimization were compared with that after opitimazation,and the occurrence of landslide hazards based on the optimized MaxEnt model was predictedto realize the landslide susceptibility evaluation in Nujiang Prefecture. Results The optimized MaxEnt model has excellent applicability in predicting landslide susceptibility in the study area (AUC=0.913)). Jackknife method was used to calculate the influence degree of each influencing factor on suscepti‑bility. Elevation (S3,23.2%), slope (S9,22.4%), settlement density (S5,14.2%), distance from river(S13,13.7%), distance from road (S4,9.6%) and lithology (S7,8.7%) tare the top six factors, with a cu‑mulative contribution of 91.8%. The spatial proportions of extremely high, high, medium and low landslide susceptibility levels were 4.88%, 8.96%, 18.40% and 67.76%, respectively. The highest proportion of ex‑tremely high and high susceptibility areas was found in Lushui City. On the whole, extremely high and high susceptibility areas were mainly distributed in valleys along rivers and roads, while low susceptibility areas were mainly distributed in areas with little human activities and undeveloped river valleys. Conclusions The optimized MaxEnt model is more suitable for landslide sensitivity prediction in Nujiang Prefecture, and the research results can provide reference for disaster prevention, and land use planning in Nujiang Prefecture.

国家自然科学基金资助项目(41161070);云南省科技厅-云南大学联合基金重点资助项目(2019FY003017);云南大学大湄公河次区域气候变化研究省创新团队项目(2019HC027);中国地质调查局项目(DD20221824)

李益敏,向倩英,邓选伦,等.基于优化MaxEnt模型的怒江州滑坡易发性评价[J].河南理工大学学报(自然科学版),2025,44(1):57‑67.

LI Y M, XIANG Q Y, DENG X L, et al. Evaluation of landslide susceptibility in Nujiang Prefecture based on opti‑mized MaxEnt model[J]. Journal of Henan Polytechnic University (Natural Science),2025,44(1):57-67.