Pub Date : 2026-01-26DOI: 10.1016/j.gr.2025.12.022
Xiang Ren , Yunpeng Dong , Inna Safonova , Shengsi Sun , Dengfeng He , Xiaoyan Zhao , Yuangang Yue , Bo Hui , Qiuming Pei , Baoping Gan
The southern East Kunlun Orogen (EKO) experienced protracted orogeny linked to the Proto-Tethys and Paleo-Tethys oceans. However, the evolution of the Proto-Tethys Ocean remains much less understood leaving the question of the timing of subduction initiation and magmatism. Here, we studied three early Paleozoic plutons exposed in the southern EKO: Kekesha (KKS) and Xialawen (XLW) gabbro-dioritic plutons and Longwakalu (LWKL) granitic pluton for geochronology, geochemistry and Sr-Nd-Hf isotopes. A KKS quartz diorite, XLW hornblende gabbro, and LWKL granite crystallized at 494, 470 and 477 Ma, respectively. The KKS gabbro-granodiorite series and XLW hornblende gabbros are enriched in light rare earth elements and large ion lithophile elements, but depleted in high strength field elements. The LWKL granites possess adakitic features: high Na2O content, Sr/Y and La/YbN ratios and differentiated heavy REEs. Isotopically, XLW hornblende gabbros and LWKL granites are less enriched isotopes (εNd(t) = −4.3 to −3.9; εHf(t) = −4.6 to +2.0) than KKS gabbro and granodiorite (εNd(t) = −7.0; εHf(t) = −7.2 to −4.8). Sr-Nd isotopic modeling suggests that KKS and XLW plutons were derived through partial melting of mantle wedge modified by different amounts of subducted terrigenous-dominated sediment derived melts. The LWKL adakitic granites were formed by high-pressure reworking of underplated arc-type intermediate rocks. The emplacement of early Paleozoic gabbro-granodiorite series and adakitic granites was related to subduction of the Proto-Tethys Ocean which started no later than ca. 500 Ma. Our new data along with available ages suggest that the supra-subduction magmatism of the Proto-Tethys Ocean in the southern EKO is episodic with peaks at ca. 495, 470, and 430 Ma. The first two episodes of magmatism mainly represent melting of enriched mantle wedge, and the third is the main pulse of magmatism formed by simultaneous melting of multiple sources of crustal rocks, subducted oceanic slab and mantle wedge.
{"title":"Episodic early Paleozoic arc magmatism of the Proto-Tethys Ocean: Evidence from geochronology, geochemistry and Sr-Nd-Hf isotopes of plutonic rocks in the southern East Kunlun Orogen","authors":"Xiang Ren , Yunpeng Dong , Inna Safonova , Shengsi Sun , Dengfeng He , Xiaoyan Zhao , Yuangang Yue , Bo Hui , Qiuming Pei , Baoping Gan","doi":"10.1016/j.gr.2025.12.022","DOIUrl":"10.1016/j.gr.2025.12.022","url":null,"abstract":"<div><div>The southern East Kunlun Orogen (EKO) experienced protracted orogeny linked to the Proto-Tethys and Paleo-Tethys oceans. However, the evolution of the Proto-Tethys Ocean remains much less understood leaving the question of the timing of subduction initiation and magmatism. Here, we studied three early Paleozoic plutons exposed in the southern EKO: Kekesha (KKS) and Xialawen (XLW) gabbro-dioritic plutons and Longwakalu (LWKL) granitic pluton for geochronology, geochemistry and Sr-Nd-Hf isotopes. A KKS quartz diorite, XLW hornblende gabbro, and LWKL granite crystallized at 494, 470 and 477 Ma, respectively. The KKS gabbro-granodiorite series and XLW hornblende gabbros are enriched in light rare earth elements and large ion lithophile elements, but depleted in high strength field elements. The LWKL granites possess adakitic features: high Na<sub>2</sub>O content, Sr/Y and La/Yb<sub>N</sub> ratios and differentiated heavy REEs. Isotopically, XLW hornblende gabbros and LWKL granites are less enriched isotopes (εNd<sub>(t)</sub> = −4.3 to −3.9; εHf<sub>(t)</sub> = −4.6 to +2.0) than KKS gabbro and granodiorite (εNd<sub>(t)</sub> = −7.0; εHf<sub>(t)</sub> = −7.2 to −4.8). Sr-Nd isotopic modeling suggests that KKS and XLW plutons were derived through partial melting of mantle wedge modified by different amounts of subducted terrigenous-dominated sediment derived melts. The LWKL adakitic granites were formed by high-pressure reworking of underplated arc-type intermediate rocks. The emplacement of early Paleozoic gabbro-granodiorite series and adakitic granites was related to subduction of the Proto-Tethys Ocean which started no later than ca. 500 Ma. Our new data along with available ages suggest that the supra-subduction magmatism of the Proto-Tethys Ocean in the southern EKO is episodic with peaks at ca. 495, 470, and 430 Ma. The first two episodes of magmatism mainly represent melting of enriched mantle wedge, and the third is the main pulse of magmatism formed by simultaneous melting of multiple sources of crustal rocks, subducted oceanic slab and mantle wedge.</div></div>","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"154 ","pages":"Pages 44-63"},"PeriodicalIF":7.2,"publicationDate":"2026-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146048047","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-25DOI: 10.1016/j.gr.2025.12.020
Mahdieh Shirmohammadi , Saied Pirasteh , Jonathan Li , Mohammad Sharifikia
Floods are among the most destructive natural hazards globally, with Iran, particularly Golestan Province, experiencing frequent and severe events. This study proposes an integrated Flood Susceptibility Mapping (FSM) framework that incorporates physical, environmental, and socioeconomic variables using advanced deep learning and machine learning techniques. A novel hybrid Convolutional Neural Network (CNN)- Long Short-Term Memory (LSTM)-Attention model was developed to capture spatiotemporal flood patterns using historical data from 2001 to 2019. Multi-source remote sensing data, including Sentinel-1 Synthetic Aperture Radar (SAR), Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS precipitation) and Landsat 7 Enhanced Thematic Mapper Plus (ETM+) Level-2 surface reflectance imagery, were processed via Google Earth Engine (GEE) and combined with topographic and environmental indices in ArcGIS. The model first generated a binary flood probability map, distinguishing between flooded and non-flooded areas. Next, validation was conducted using Global Positioning System (GPS)-based ground-truth points from the 2019 flood event. Then, over 200,000 high-confidence samples and key conditioning factors (elevation, slope, aspect, Standardized Precipitation Index (SPI), Topographic Wetness Index (TWI), lithology, river distance, river density, rainfall, and land use) were used to train a Random Forest (RF) model in Python and Geographical Information System (GIS) environments, producing a detailed FSM for 2024. Finally, to assess FSM future scenarios, we used precipitation projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6) EC-Earth3-Veg model under four Shared Socioeconomic Pathways (SSP) scenarios (SSP1-2.6 to SSP5-8.5) to model FSM from 2021 to 2100. Results indicate an increasing flood susceptibility in western and northern lowlands, with higher-risk zones expanding under high-emission scenarios. The RF model achieved an Area Under the Curve (of the ROC curve) (AUC) of 0.91, while the CNN-LSTM-Attention model showed high accuracy (99.5%) and strong spatial performance. This framework demonstrates potential for broader application in flood-prone regions globally, supporting climate-adaptive planning and mitigation.
洪水是全球最具破坏性的自然灾害之一，伊朗，特别是戈列斯坦省，经常发生严重的洪水。本研究提出了一个综合洪水易感性映射（FSM）框架，该框架使用先进的深度学习和机器学习技术，结合了物理、环境和社会经济变量。利用2001年至2019年的历史数据，开发了一种新的混合卷积神经网络(CNN)-长短期记忆(LSTM)-注意力模型来捕捉时空洪水模式。利用谷歌Earth Engine （GEE）对Sentinel-1合成孔径雷达（SAR）、Climate Hazards Group InfraRed Precipitation with Station data （CHIRPS Precipitation）和Landsat 7 Enhanced Thematic Mapper Plus (ETM+) Level-2地表反射率影像等多源遥感数据进行处理，并结合ArcGIS中的地形和环境指数。该模型首先生成了一个二元洪水概率图，区分了洪水区和非洪水区。接下来，利用2019年洪水事件中基于全球定位系统（GPS）的地面真实点进行验证。然后，使用超过20万个高置信度样本和关键条件因子（海拔、坡度、坡向、标准化降水指数（SPI）、地形湿度指数（TWI）、岩性、河流距离、河流密度、降雨量和土地利用）在Python和地理信息系统（GIS）环境中训练随机森林（RF）模型，生成2024年详细的FSM。最后，为了评估FSM的未来情景，我们使用了耦合模式比较项目第6阶段（CMIP6） EC-Earth3-Veg模型在四种共享社会经济路径（SSP）情景（SSP1-2.6至SSP5-8.5）下的降水预测来模拟FSM在2021年至2100年的情景。结果表明，在高排放情景下，西部和北部低地的洪水易感性增加，高风险区域扩大。RF模型的ROC曲线下面积（Area Under The Curve， AUC）为0.91，CNN-LSTM-Attention模型具有较高的准确率（99.5%）和较强的空间性能。该框架显示了在全球洪水易发地区更广泛应用的潜力，支持气候适应性规划和减灾。
{"title":"A novel CNN-LSTM-Attention model to forecast flood susceptibility under global climate scenarios","authors":"Mahdieh Shirmohammadi , Saied Pirasteh , Jonathan Li , Mohammad Sharifikia","doi":"10.1016/j.gr.2025.12.020","DOIUrl":"10.1016/j.gr.2025.12.020","url":null,"abstract":"<div><div>Floods are among the most destructive natural hazards globally, with Iran, particularly Golestan Province, experiencing frequent and severe events. This study proposes an integrated Flood Susceptibility Mapping (FSM) framework that incorporates physical, environmental, and socioeconomic variables using advanced deep learning and machine learning techniques. A novel hybrid Convolutional Neural Network (CNN)- Long Short-Term Memory (LSTM)-Attention model was developed to capture spatiotemporal flood patterns using historical data from 2001 to 2019. Multi-source remote sensing data, including Sentinel-1 Synthetic Aperture Radar (SAR), Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS precipitation) and Landsat 7 Enhanced Thematic Mapper Plus (ETM+) Level-2 surface reflectance imagery, were processed via Google Earth Engine (GEE) and combined with topographic and environmental indices in ArcGIS. The model first generated a binary flood probability map, distinguishing between flooded and non-flooded areas. Next, validation was conducted using Global Positioning System (GPS)-based ground-truth points from the 2019 flood event. Then, over 200,000 high-confidence samples and key conditioning factors (elevation, slope, aspect, Standardized Precipitation Index (SPI), Topographic Wetness Index (TWI), lithology, river distance, river density, rainfall, and land use) were used to train a Random Forest (RF) model in Python and Geographical Information System (GIS) environments, producing a detailed FSM for 2024. Finally, to assess FSM future scenarios, we used precipitation projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6) EC-Earth3-Veg model under four Shared Socioeconomic Pathways (SSP) scenarios (SSP1-2.6 to SSP5-8.5) to model FSM from 2021 to 2100. Results indicate an increasing flood susceptibility in western and northern lowlands, with higher-risk zones expanding under high-emission scenarios. The RF model achieved an Area Under the Curve (of the ROC curve) (AUC) of 0.91, while the CNN-LSTM-Attention model showed high accuracy (99.5%) and strong spatial performance. This framework demonstrates potential for broader application in flood-prone regions globally, supporting climate-adaptive planning and mitigation.</div></div>","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"154 ","pages":"Pages 82-102"},"PeriodicalIF":7.2,"publicationDate":"2026-01-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146048057","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-24DOI: 10.1016/j.gr.2026.01.003
Harsh K. Gupta, Rajesh Rekapalli
{"title":"Reply to the comment by Dr. K. Rajendran on “The 2008 Mw 7.9 Wenchuan, China earthquake: not a case of reservoir triggered seismicity” by Gupta and Rekapalli (2026), Gondwana Research, volume 151, pages 184–188","authors":"Harsh K. Gupta, Rajesh Rekapalli","doi":"10.1016/j.gr.2026.01.003","DOIUrl":"https://doi.org/10.1016/j.gr.2026.01.003","url":null,"abstract":"","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"13 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146048049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-24DOI: 10.1016/j.gr.2025.12.013
Davis Kaimalayil Ephsy, Selvaraju Raja
{"title":"Seasonal variation and distribution of microplastics in surface water and sediments of Coimbatore Lakes, India","authors":"Davis Kaimalayil Ephsy, Selvaraju Raja","doi":"10.1016/j.gr.2025.12.013","DOIUrl":"https://doi.org/10.1016/j.gr.2025.12.013","url":null,"abstract":"","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"2 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033602","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-24DOI: 10.1016/j.gr.2025.12.015
Misha A.J.B. Whittingham, Danielle Fraser, Hillary C. Maddin
{"title":"Regional-scale trends in floral community change through the Pennsylvanian of the Maritimes Basin, Atlantic Canada","authors":"Misha A.J.B. Whittingham, Danielle Fraser, Hillary C. Maddin","doi":"10.1016/j.gr.2025.12.015","DOIUrl":"https://doi.org/10.1016/j.gr.2025.12.015","url":null,"abstract":"","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"254 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033601","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.gr.2026.01.002
Kusala Rajendran
{"title":"Comment on “The 2008 Mw 7.9 Wenchuan, China earthquake: not a case of reservoir triggered seismicity” by Gupta and Rekhapalli (2026), Gondwana Research, Volume 151, pages 184–188","authors":"Kusala Rajendran","doi":"10.1016/j.gr.2026.01.002","DOIUrl":"https://doi.org/10.1016/j.gr.2026.01.002","url":null,"abstract":"","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"117 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.gr.2025.12.019
Priyansha Gupta, Mahua Saha, Chayanika Rathore, V. Suneel, Jacob de Boer, Anita Garg
{"title":"Microplastics and PAH contamination in the Eastern Arabian Sea: A synergistic environmental hazard","authors":"Priyansha Gupta, Mahua Saha, Chayanika Rathore, V. Suneel, Jacob de Boer, Anita Garg","doi":"10.1016/j.gr.2025.12.019","DOIUrl":"https://doi.org/10.1016/j.gr.2025.12.019","url":null,"abstract":"","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"66 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.gr.2025.12.016
Xiao-Fei Qiu , Da Wang , Xi-Run Tong , Shi-Wen Xie , Nian-Wen Wu , Fei Liu , Yu-Sheng Wan
The oldest rocks provide direct constraints on the nature of the first crust on Earth and the earliest magmatic process, which is significant for understanding the physical and chemical properties of our planet’s early stage. On the modern Earth, Hadean to Eoarchean crustal rocks have been identified in less than ten areas worldwide. Due to relatively poor preservation of early Archean rocks, major controversies exist on the tectonic mechanisms responsible for the formation of the continent during the early Earth. Therefore, identification of new Eoarchean or even Hadean crustal exposures would provide key information for understanding the formation and evolution of early continental crust and its geodynamic driver in the early Earth. Here, we report the new identification of Eoarchean trondhjemitic gneisses in the Muzidian Gneiss Complex (MGC) in the northern margin of Yangtze Craton. Zircon SHRIMP U-Pb ages of 3855 ± 7 Ma and 3851 ± 6 Ma suggest this trondhjemite unit in the MGC is the oldest known igneous rock in Eurasia. Zircon Hf isotopic compositions indicate that the MGC gneisses were formed from reworking of pre-existing Hadean crust older than 4.1 Ga. These newly recognized rocks in the MGC mark an important, Hadean crust derived, ancient gneiss complex, which is isotopically comparable to the Acasta Gneiss Complex in the currently established global Eoarchean geological record. Our findings indicate that at least some of the earliest crustal rocks might have originated from an early-differentiated, incompatible element-enriched protocrust in the Hadean.
{"title":"The oldest rock in the Eurasian continent was reworked from Hadean protocrust","authors":"Xiao-Fei Qiu , Da Wang , Xi-Run Tong , Shi-Wen Xie , Nian-Wen Wu , Fei Liu , Yu-Sheng Wan","doi":"10.1016/j.gr.2025.12.016","DOIUrl":"10.1016/j.gr.2025.12.016","url":null,"abstract":"<div><div>The oldest rocks provide direct constraints on the nature of the first crust on Earth and the earliest magmatic process, which is significant for understanding the physical and chemical properties of our planet’s early stage. On the modern Earth, Hadean to Eoarchean crustal rocks have been identified in less than ten areas worldwide. Due to relatively poor preservation of early Archean rocks, major controversies exist on the tectonic mechanisms responsible for the formation of the continent during the early Earth. Therefore, identification of new Eoarchean or even Hadean crustal exposures would provide key information for understanding the formation and evolution of early continental crust and its geodynamic driver in the early Earth. Here, we report the new identification of Eoarchean trondhjemitic gneisses in the Muzidian Gneiss Complex (MGC) in the northern margin of Yangtze Craton. Zircon SHRIMP U-Pb ages of 3855 ± 7 Ma and 3851 ± 6 Ma suggest this trondhjemite unit in the MGC is the oldest known igneous rock in Eurasia. Zircon Hf isotopic compositions indicate that the MGC gneisses were formed from reworking of pre-existing Hadean crust older than 4.1 Ga. These newly recognized rocks in the MGC mark an important, Hadean crust derived, ancient gneiss complex, which is isotopically comparable to the Acasta Gneiss Complex in the currently established global Eoarchean geological record. Our findings indicate that at least some of the earliest crustal rocks might have originated from an early-differentiated, incompatible element-enriched protocrust in the Hadean.</div></div>","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"154 ","pages":"Pages 35-43"},"PeriodicalIF":7.2,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033933","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.gr.2025.12.012
Tehseen Zafar, Shuguang Song, Hafiz Ur Rehman, Hamed Gamaleldien, Abiola Oyebamiji, Zaheen Ullah, Umar Farooq Jadoon, Muhammad Farhan, Mohamed Zaki Khedr, Irfan Maqbool Bhat, Fatemeh Sepidbar, Fatemeh Nouri, Amjad Hussain, Zahid Hussain, Mabrouk Sami
{"title":"Reply to the comment on Zafar et al., 2025: “Retrieving petrogenetic source, compositional diversity and tectono-magmatic scenario of Tethyan sediment-derived magmatic flare-up: A tale from petrochemical and multi-isotopic (Sr–Nd–B–Hf) systematics” by Bhat, 2025","authors":"Tehseen Zafar, Shuguang Song, Hafiz Ur Rehman, Hamed Gamaleldien, Abiola Oyebamiji, Zaheen Ullah, Umar Farooq Jadoon, Muhammad Farhan, Mohamed Zaki Khedr, Irfan Maqbool Bhat, Fatemeh Sepidbar, Fatemeh Nouri, Amjad Hussain, Zahid Hussain, Mabrouk Sami","doi":"10.1016/j.gr.2025.12.012","DOIUrl":"https://doi.org/10.1016/j.gr.2025.12.012","url":null,"abstract":"","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"271 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033610","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.gr.2025.12.014
Jian Ji, Junhan Deng, Hongzhi Cui, Bin Tong, Xintao Tang, Te Pei
Landslides are among the most destructive geological hazards, highlighting the urgent need for accurate landslide susceptibility mapping (LSM) to support risk reduction and mitigation strategies. Here, we systematically assess the performance of individual machine learning (ML) models and their logistic regression (LR)-coupled counterparts, with a particular focus on the influence of raster resolution on model accuracy. A total of 10 landslide conditioning factors were selected to construct both individual and coupling models, while correlation analysis and SHAP-based feature attribution were applied to ensure input independence and enhance interpretability. Hyperparameters were optimized via Bayesian search. Results indicate that slope, lithology, and elevation exert the strongest controls on landslide occurrence, and that deep learning (DL) architectures outperform other individual models. Crucially, all LR-coupled models yielded significant gains over their standalone equivalents, with AUC improvements of 4.4% (DNN_LR), 6.1% (BP_NN_LR), 6.0% (XGBoost_LR), 3.9% (RF_LR), and 5.1% (SVM_LR). DL-based hybrids achieved the highest predictive accuracy, although LR tended to overpredict low-risk zones. Across multiple raster resolutions, coupled models, particularly DNN_LR and BP_NN_LR, exhibited strong robustness and spatial generalizability. Overall, we propose a novel LR-ML coupling framework that integrates the transparency and efficiency of LR, a lightweight model with superior linear modeling capacity, with the representational power of non-linear meta-learners (RF, SVM, XGBoost, BP_NN, and DNN). LR provides efficient preliminary predictions and refines label quality via targeted non-landslide sampling, yielding high-quality training inputs for subsequent learning. This integration effectively mitigates overfitting, enhances interpretability, and reduces computational demand, while maintaining stability across scales. Collectively, our findings establish LR-ML as a robust and scalable framework for large-scale LSM.
{"title":"Comparison of different machine learning models coupling with logistic regression for landslide susceptibility mapping","authors":"Jian Ji, Junhan Deng, Hongzhi Cui, Bin Tong, Xintao Tang, Te Pei","doi":"10.1016/j.gr.2025.12.014","DOIUrl":"https://doi.org/10.1016/j.gr.2025.12.014","url":null,"abstract":"Landslides are among the most destructive geological hazards, highlighting the urgent need for accurate landslide susceptibility mapping (LSM) to support risk reduction and mitigation strategies. Here, we systematically assess the performance of individual machine learning (ML) models and their logistic regression (LR)-coupled counterparts, with a particular focus on the influence of raster resolution on model accuracy. A total of 10 landslide conditioning factors were selected to construct both individual and coupling models, while correlation analysis and SHAP-based feature attribution were applied to ensure input independence and enhance interpretability. Hyperparameters were optimized via Bayesian search. Results indicate that slope, lithology, and elevation exert the strongest controls on landslide occurrence, and that deep learning (DL) architectures outperform other individual models. Crucially, all LR-coupled models yielded significant gains over their standalone equivalents, with AUC improvements of 4.4% (DNN_LR), 6.1% (BP_NN_LR), 6.0% (XGBoost_LR), 3.9% (RF_LR), and 5.1% (SVM_LR). DL-based hybrids achieved the highest predictive accuracy, although LR tended to overpredict low-risk zones. Across multiple raster resolutions, coupled models, particularly DNN_LR and BP_NN_LR, exhibited strong robustness and spatial generalizability. Overall, we propose a novel LR-ML coupling framework that integrates the transparency and efficiency of LR, a lightweight model with superior linear modeling capacity, with the representational power of non-linear <ce:italic>meta</ce:italic>-learners (RF, SVM, XGBoost, BP_NN, and DNN). LR provides efficient preliminary predictions and refines label quality via targeted non-landslide sampling, yielding high-quality training inputs for subsequent learning. This integration effectively mitigates overfitting, enhances interpretability, and reduces computational demand, while maintaining stability across scales. Collectively, our findings establish LR-ML as a robust and scalable framework for large-scale LSM.","PeriodicalId":12761,"journal":{"name":"Gondwana Research","volume":"79 1","pages":""},"PeriodicalIF":6.1,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146048089","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号