Nafyad Serre Kawo, Jesse Korus, Yaser Kishawi, Erin Marie King Haacker, Aaron R. Mittelstet
{"title":"利用机器学习建立三维概率水成模型","authors":"Nafyad Serre Kawo, Jesse Korus, Yaser Kishawi, Erin Marie King Haacker, Aaron R. Mittelstet","doi":"10.1029/2023wr035910","DOIUrl":null,"url":null,"abstract":"Characterizing the 3D distribution of hydraulic properties in glacial sediments is challenging due to fine-scale heterogeneity and complexity. Borehole lithological data provide high vertical resolution but low horizontal resolution. Geophysical methods can fill gaps between boreholes, providing improved horizontal resolution but low vertical resolution. Machine learning can combine borehole and geophysical data to overcome these challenges. However, few studies have compared multiple machine learning methods for predicting hydrofacies in glacial aquifer systems. This study uses colocated airborne electromagnetic resistivity and borehole lithology data to train multiple machine learning models and predict the 3D distribution of hydrofacies in glacial deposits of eastern Nebraska, USA. Random Forest, Gradient Boosting Classifier, Extreme Gradient Boosting, Multilayer Perceptron, and Stacking Classifier were used to model 3D probabilistic distributions of hydrofacies (sand and clay) at a grid size of 200 m × 200 m × 3 m. Comparison of the predicted 3D hydrofacies models shows that the probability distributions and the contrasts between hydrofacies vary. The classification metrics show that the Stacking Classifier model performed better than other machine learning models in predicting hydrofacies. Multi-Layer Perceptron and Stacking Classifier models show sharp vertical transitions between the low and high sand probability while other machine learning models show gradual transitions. K-means clustering was used to translate the Stacking Classifier model into a 4-class hydraulic conductivity model. This study shows that machine learning methods advance our understanding of glacial hydrogeology by improving the vertical and horizontal resolution of hydrofacies distribution and resolving aquifer-aquifer and stream-aquifer connections.","PeriodicalId":23799,"journal":{"name":"Water Resources Research","volume":null,"pages":null},"PeriodicalIF":4.6000,"publicationDate":"2024-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Three-Dimensional Probabilistic Hydrofacies Modeling Using Machine Learning\",\"authors\":\"Nafyad Serre Kawo, Jesse Korus, Yaser Kishawi, Erin Marie King Haacker, Aaron R. Mittelstet\",\"doi\":\"10.1029/2023wr035910\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Characterizing the 3D distribution of hydraulic properties in glacial sediments is challenging due to fine-scale heterogeneity and complexity. Borehole lithological data provide high vertical resolution but low horizontal resolution. Geophysical methods can fill gaps between boreholes, providing improved horizontal resolution but low vertical resolution. Machine learning can combine borehole and geophysical data to overcome these challenges. However, few studies have compared multiple machine learning methods for predicting hydrofacies in glacial aquifer systems. This study uses colocated airborne electromagnetic resistivity and borehole lithology data to train multiple machine learning models and predict the 3D distribution of hydrofacies in glacial deposits of eastern Nebraska, USA. Random Forest, Gradient Boosting Classifier, Extreme Gradient Boosting, Multilayer Perceptron, and Stacking Classifier were used to model 3D probabilistic distributions of hydrofacies (sand and clay) at a grid size of 200 m × 200 m × 3 m. Comparison of the predicted 3D hydrofacies models shows that the probability distributions and the contrasts between hydrofacies vary. The classification metrics show that the Stacking Classifier model performed better than other machine learning models in predicting hydrofacies. Multi-Layer Perceptron and Stacking Classifier models show sharp vertical transitions between the low and high sand probability while other machine learning models show gradual transitions. K-means clustering was used to translate the Stacking Classifier model into a 4-class hydraulic conductivity model. This study shows that machine learning methods advance our understanding of glacial hydrogeology by improving the vertical and horizontal resolution of hydrofacies distribution and resolving aquifer-aquifer and stream-aquifer connections.\",\"PeriodicalId\":23799,\"journal\":{\"name\":\"Water Resources Research\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":4.6000,\"publicationDate\":\"2024-07-05\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Water Resources Research\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://doi.org/10.1029/2023wr035910\",\"RegionNum\":1,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"ENVIRONMENTAL SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Water Resources Research","FirstCategoryId":"89","ListUrlMain":"https://doi.org/10.1029/2023wr035910","RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENVIRONMENTAL SCIENCES","Score":null,"Total":0}
引用次数: 0
摘要
由于冰川沉积物的细尺度异质性和复杂性,确定其水力特性的三维分布具有挑战性。钻孔岩性数据可提供较高的垂直分辨率,但水平分辨率较低。地球物理方法可以填补钻孔之间的空白,提高水平分辨率,但垂直分辨率较低。机器学习可以结合钻孔和地球物理数据来克服这些挑战。然而,很少有研究比较多种机器学习方法来预测冰川含水层系统中的水成岩。本研究利用同位机载电磁电阻率和钻孔岩性数据来训练多个机器学习模型,并预测美国内布拉斯加州东部冰川沉积物中水成岩的三维分布。随机森林、梯度提升分类器、极端梯度提升、多层感知器和堆叠分类器被用于在 200 m × 200 m × 3 m 的网格大小上建立水成岩(砂和粘土)的三维概率分布模型。分类指标显示,堆叠分类器模型在预测水成层方面的表现优于其他机器学习模型。多层感知器和堆叠分类器模型在低砂和高砂概率之间显示出急剧的垂直过渡,而其他机器学习模型则显示出渐进的过渡。K-means 聚类用于将堆叠分类器模型转化为 4 级水力传导模型。这项研究表明,机器学习方法提高了水成岩分布的垂直和水平分辨率,并解决了含水层-含水层和溪流-含水层之间的联系问题,从而加深了我们对冰川水文地质学的理解。
Three-Dimensional Probabilistic Hydrofacies Modeling Using Machine Learning
Characterizing the 3D distribution of hydraulic properties in glacial sediments is challenging due to fine-scale heterogeneity and complexity. Borehole lithological data provide high vertical resolution but low horizontal resolution. Geophysical methods can fill gaps between boreholes, providing improved horizontal resolution but low vertical resolution. Machine learning can combine borehole and geophysical data to overcome these challenges. However, few studies have compared multiple machine learning methods for predicting hydrofacies in glacial aquifer systems. This study uses colocated airborne electromagnetic resistivity and borehole lithology data to train multiple machine learning models and predict the 3D distribution of hydrofacies in glacial deposits of eastern Nebraska, USA. Random Forest, Gradient Boosting Classifier, Extreme Gradient Boosting, Multilayer Perceptron, and Stacking Classifier were used to model 3D probabilistic distributions of hydrofacies (sand and clay) at a grid size of 200 m × 200 m × 3 m. Comparison of the predicted 3D hydrofacies models shows that the probability distributions and the contrasts between hydrofacies vary. The classification metrics show that the Stacking Classifier model performed better than other machine learning models in predicting hydrofacies. Multi-Layer Perceptron and Stacking Classifier models show sharp vertical transitions between the low and high sand probability while other machine learning models show gradual transitions. K-means clustering was used to translate the Stacking Classifier model into a 4-class hydraulic conductivity model. This study shows that machine learning methods advance our understanding of glacial hydrogeology by improving the vertical and horizontal resolution of hydrofacies distribution and resolving aquifer-aquifer and stream-aquifer connections.
期刊介绍:
Water Resources Research (WRR) is an interdisciplinary journal that focuses on hydrology and water resources. It publishes original research in the natural and social sciences of water. It emphasizes the role of water in the Earth system, including physical, chemical, biological, and ecological processes in water resources research and management, including social, policy, and public health implications. It encompasses observational, experimental, theoretical, analytical, numerical, and data-driven approaches that advance the science of water and its management. Submissions are evaluated for their novelty, accuracy, significance, and broader implications of the findings.