Pub Date : 2025-07-23DOI: 10.1016/j.aiig.2025.100146
Vinicius L.S. Silva , Pablo Salinas , Claire E. Heaney , Matthew D. Jackson , Christopher C. Pain
We propose a novel type of nonlinear solver acceleration for systems of nonlinear partial differential equations (PDEs) that is based on online/adaptive learning. It is applied in the context of multiphase flow in porous media. The proposed method rely on four pillars: (i) dimensionless numbers as input parameters for the machine learning model, (ii) simplified numerical model (two-dimensional) for the offline training, (iii) dynamic control of a nonlinear solver tuning parameter (numerical relaxation), (iv) and online learning for real-time improvement of the machine learning model. This strategy decreases the number of nonlinear iterations by dynamically modifying a single global parameter, the relaxation factor, and by adaptively learning the attributes of each numerical model on-the-run. Furthermore, this work performs a sensitivity study in the dimensionless parameters (machine learning features), assess the efficacy of various machine learning models, demonstrate a decrease in nonlinear iterations using our method in more intricate, realistic three-dimensional models, and fully couple a machine learning model into an open-source multiphase flow simulator achieving up to 85% reduction in computational time.
{"title":"Online learning to accelerate nonlinear PDE solvers: Applied to multiphase porous media flow","authors":"Vinicius L.S. Silva , Pablo Salinas , Claire E. Heaney , Matthew D. Jackson , Christopher C. Pain","doi":"10.1016/j.aiig.2025.100146","DOIUrl":"10.1016/j.aiig.2025.100146","url":null,"abstract":"<div><div>We propose a novel type of nonlinear solver acceleration for systems of nonlinear partial differential equations (PDEs) that is based on online/adaptive learning. It is applied in the context of multiphase flow in porous media. The proposed method rely on four pillars: (i) dimensionless numbers as input parameters for the machine learning model, (ii) simplified numerical model (two-dimensional) for the offline training, (iii) dynamic control of a nonlinear solver tuning parameter (numerical relaxation), (iv) and online learning for real-time improvement of the machine learning model. This strategy decreases the number of nonlinear iterations by dynamically modifying a single global parameter, the relaxation factor, and by adaptively learning the attributes of each numerical model on-the-run. Furthermore, this work performs a sensitivity study in the dimensionless parameters (machine learning features), assess the efficacy of various machine learning models, demonstrate a decrease in nonlinear iterations using our method in more intricate, realistic three-dimensional models, and fully couple a machine learning model into an open-source multiphase flow simulator achieving up to 85% reduction in computational time.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100146"},"PeriodicalIF":0.0,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144702140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Maintaining high groundwater level (GWL) is important for preventing fires in peatlands. This study proposes GWL prediction using machine learning methods for forest plantations in Indonesian tropical peatlands. Deep neural networks (DNN) have been used for prediction; however, they have not been applied to groundwater prediction in Indonesian peatlands. Tropical peatland is characterized by high permeability and forest plantations are surrounded by several canals. By predicting daily differences in GWL, the GWL can be predicted with high accuracy. DNNs, random forests, support vector regression, and XGBoost were compared, all of which indicated similar errors. The SHAP value revealed that the precipitation falling on the hill rapidly seeps into the soil and flows into the canals, which agrees with the fact that the soil has high permeability. These findings can potentially be used to alleviate and manage future fires in peatlands.
{"title":"Prediction of groundwater level in Indonesian tropical peatland forest plantations using machine learning","authors":"Kazuo Yonekura , Sota Miyazaki , Masaatsu Aichi , Takafumi Nishizu , Masao Hasegawa , Katsuyuki Suzuki","doi":"10.1016/j.aiig.2025.100148","DOIUrl":"10.1016/j.aiig.2025.100148","url":null,"abstract":"<div><div>Maintaining high groundwater level (GWL) is important for preventing fires in peatlands. This study proposes GWL prediction using machine learning methods for forest plantations in Indonesian tropical peatlands. Deep neural networks (DNN) have been used for prediction; however, they have not been applied to groundwater prediction in Indonesian peatlands. Tropical peatland is characterized by high permeability and forest plantations are surrounded by several canals. By predicting daily differences in GWL, the GWL can be predicted with high accuracy. DNNs, random forests, support vector regression, and XGBoost were compared, all of which indicated similar errors. The SHAP value revealed that the precipitation falling on the hill rapidly seeps into the soil and flows into the canals, which agrees with the fact that the soil has high permeability. These findings can potentially be used to alleviate and manage future fires in peatlands.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100148"},"PeriodicalIF":0.0,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144714577","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-18DOI: 10.1016/j.aiig.2025.100147
Evet Naturinda , Fortunate Kemigyisha , Anthony Gidudu , Isa Kabenge , Emmanuel Omia , Jackline Aboth
Greenhouse gases (GHGs) from agriculture in Africa are among the world's fastest-growing emissions, with the livestock sector as the primary contributor. However, the methods for quantifying these emissions rely on manual and outdated data collection and processing approaches. Therefore, there is a need to develop more accurate and efficient methods of quantifying GHGs from livestock. This research developed a remote sensing and Artificial Intelligence (AI) based approach to quantify GHG emissions from cattle in the Kisombwa Ranching Scheme in Mubende District, central Uganda.
We trained a deep learning algorithm, You Only Look Once (YOLO) v4, to detect cattle from the Unmanned Aerial Vehicle (UAV) images of the study area and applied the Simple Online Real-time Tracker (SORT) algorithm for automated counting. Methane (CH4) and Nitrous Oxide (N2O) emissions from manure management and enteric fermentation were estimated using the number of cattle and the Tier 1 guidelines from the Intergovernmental Panel on Climate Change (IPCC). The total estimated emissions were 321,121.34 kg carbon dioxide equivalent (CO2eq) per year, with CH4 at 282,282.96 kg CO2eq per year (88 %) and N2O at 38,838.38 kg CO2eq per year (12 %). Enteric fermentation contributed the highest emissions, about 99 % of the total CH4 emissions and 87 % of the total GHGs.
The proposed remote sensing and AI-driven method achieved an average F1 score of 88.9 %, average precision of 97 %, and average recall of 82.9 % on the testing set of images. Therefore, these research findings demonstrate that remote sensing and AI are a more potent and efficient approach to upscale quantifying and reporting animal population and livestock GHG emissions for sustainable agriculture and climate change mitigation.
非洲农业排放的温室气体是世界上增长最快的排放源之一，其中畜牧业是主要排放源。然而，量化这些排放的方法依赖于人工和过时的数据收集和处理方法。因此，有必要开发更准确和有效的方法来量化牲畜的温室气体。这项研究开发了一种基于遥感和人工智能（AI）的方法，用于量化乌干达中部Mubende地区Kisombwa牧场计划中牛的温室气体排放。我们训练了一个深度学习算法You Only Look Once (YOLO) v4，从研究区域的无人机（UAV）图像中检测牛，并应用Simple Online Real-time Tracker （SORT）算法进行自动计数。利用牛的数量和政府间气候变化专门委员会（IPCC）的一级指南，估算了粪便管理和肠道发酵产生的甲烷（CH4）和氧化亚氮（N2O）排放。估计总排放量为每年321,121.34千克二氧化碳当量（CO2eq），其中CH4为每年282,282.96千克二氧化碳当量（88%），N2O为每年38,838.38千克二氧化碳当量（12%）。肠道发酵的排放最高，约占CH4排放总量的99%和温室气体排放总量的87%。本文提出的遥感和人工智能驱动方法在图像测试集上的平均F1分数为88.9%，平均精度为97%，平均召回率为82.9%。因此，这些研究结果表明，遥感和人工智能是对可持续农业和减缓气候变化的动物种群和牲畜温室气体排放进行高级量化和报告的更有力和有效的方法。
{"title":"Quantification of greenhouse gas emissions from livestock using remote sensing & artificial intelligence","authors":"Evet Naturinda , Fortunate Kemigyisha , Anthony Gidudu , Isa Kabenge , Emmanuel Omia , Jackline Aboth","doi":"10.1016/j.aiig.2025.100147","DOIUrl":"10.1016/j.aiig.2025.100147","url":null,"abstract":"<div><div>Greenhouse gases (GHGs) from agriculture in Africa are among the world's fastest-growing emissions, with the livestock sector as the primary contributor. However, the methods for quantifying these emissions rely on manual and outdated data collection and processing approaches. Therefore, there is a need to develop more accurate and efficient methods of quantifying GHGs from livestock. This research developed a remote sensing and Artificial Intelligence (AI) based approach to quantify GHG emissions from cattle in the Kisombwa Ranching Scheme in Mubende District, central Uganda.</div><div>We trained a deep learning algorithm, You Only Look Once (YOLO) v4, to detect cattle from the Unmanned Aerial Vehicle (UAV) images of the study area and applied the Simple Online Real-time Tracker (SORT) algorithm for automated counting. Methane (CH<sub>4)</sub> and Nitrous Oxide (N<sub>2</sub>O) emissions from manure management and enteric fermentation were estimated using the number of cattle and the Tier 1 guidelines from the Intergovernmental Panel on Climate Change (IPCC). The total estimated emissions were 321,121.34 kg carbon dioxide equivalent (CO<sub>2</sub>eq) per year, with CH<sub>4</sub> at 282,282.96 kg CO<sub>2</sub>eq per year (88 %) and N<sub>2</sub>O at 38,838.38 kg CO<sub>2</sub>eq per year (12 %). Enteric fermentation contributed the highest emissions, about 99 % of the total CH<sub>4</sub> emissions and 87 % of the total GHGs.</div><div>The proposed remote sensing and AI-driven method achieved an average F1 score of 88.9 %, average precision of 97 %, and average recall of 82.9 % on the testing set of images. Therefore, these research findings demonstrate that remote sensing and AI are a more potent and efficient approach to upscale quantifying and reporting animal population and livestock GHG emissions for sustainable agriculture and climate change mitigation.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100147"},"PeriodicalIF":0.0,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144702139","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-16DOI: 10.1016/j.aiig.2025.100145
Iver Martinsen , Steffen Aagaard Sørensen , Samuel Ortega , Fred Godtliebsen , Miguel Tejedor , Eirik Myrvoll-Nilsen
Foraminifera are shell-bearing microorganisms that are commonly found in marine deposits on the seabed. They are important indicators in many analyses, are used in climate change research, monitoring marine environments, evolutionary studies, and are also frequently used in the oil and gas industry. Although some research has focused on automating the classification of foraminifera images, few have addressed the uncertainty in these classifications. Although foraminifera classification is not a safety-critical task, estimating uncertainty is crucial to avoid misclassifications that could overlook rare and ecologically significant species that are informative indicators of the environment in which they lived. Uncertainty estimation in deep learning has gained significant attention and many methods have been developed. However, evaluating the performance of these methods in practical settings remains a challenge. To create a benchmark for uncertainty estimation in the classification of foraminifera, we administered a multiple choice questionnaire containing classification tasks to four senior geologists. By analyzing their responses, we generated human-derived uncertainty estimates for a test set of 260 images of foraminifera and sediment grains. These uncertainty estimates served as a baseline for comparison when training neural networks in classification. We then trained multiple deep neural networks using a range of uncertainty quantification methods to classify and state the uncertainty about the classifications. The results of the deep learning uncertainty quantification methods were then analyzed and compared with the human benchmark, to see how the methods performed individually and how the methods aligned with humans. Our results show that human-level performance can be achieved with deep learning and that test-time data augmentation and ensembling can help improve both uncertainty estimation and classification performance. Our results also show that human uncertainty estimates are helpful indicators for detecting classification errors and that deep learning-based uncertainty estimates can improve calibration and classification accuracy.
{"title":"Quantifying uncertainty in foraminifera classification: How deep learning methods compare to human experts","authors":"Iver Martinsen , Steffen Aagaard Sørensen , Samuel Ortega , Fred Godtliebsen , Miguel Tejedor , Eirik Myrvoll-Nilsen","doi":"10.1016/j.aiig.2025.100145","DOIUrl":"10.1016/j.aiig.2025.100145","url":null,"abstract":"<div><div>Foraminifera are shell-bearing microorganisms that are commonly found in marine deposits on the seabed. They are important indicators in many analyses, are used in climate change research, monitoring marine environments, evolutionary studies, and are also frequently used in the oil and gas industry. Although some research has focused on automating the classification of foraminifera images, few have addressed the uncertainty in these classifications. Although foraminifera classification is not a safety-critical task, estimating uncertainty is crucial to avoid misclassifications that could overlook rare and ecologically significant species that are informative indicators of the environment in which they lived. Uncertainty estimation in deep learning has gained significant attention and many methods have been developed. However, evaluating the performance of these methods in practical settings remains a challenge. To create a benchmark for uncertainty estimation in the classification of foraminifera, we administered a multiple choice questionnaire containing classification tasks to four senior geologists. By analyzing their responses, we generated human-derived uncertainty estimates for a test set of 260 images of foraminifera and sediment grains. These uncertainty estimates served as a baseline for comparison when training neural networks in classification. We then trained multiple deep neural networks using a range of uncertainty quantification methods to classify and state the uncertainty about the classifications. The results of the deep learning uncertainty quantification methods were then analyzed and compared with the human benchmark, to see how the methods performed individually and how the methods aligned with humans. Our results show that human-level performance can be achieved with deep learning and that test-time data augmentation and ensembling can help improve both uncertainty estimation and classification performance. Our results also show that human uncertainty estimates are helpful indicators for detecting classification errors and that deep learning-based uncertainty estimates can improve calibration and classification accuracy.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100145"},"PeriodicalIF":0.0,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144694746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-15DOI: 10.1016/j.aiig.2025.100141
Michael Meadows, Karin Reinke, Simon Jones
Machine learning models are increasingly used to correct the vertical biases (mainly due to vegetation and buildings) in global Digital Elevation Models (DEMs), for downstream applications which need “bare earth” elevations. The predictive accuracy of these models has improved significantly as more flexible model architectures are developed and new explanatory datasets produced, leading to the recent release of three model-corrected DEMs (FABDEM, DiluviumDEM and FathomDEM). However, there has been relatively little focus so far on explaining or interrogating these models, especially important in this context given their downstream impact on many other applications (including natural hazard simulations). In this study we train five separate models (by land cover environment) to correct vertical biases in the Copernicus DEM and then explain them using SHapley Additive exPlanation (SHAP) values. Comparing the models, we find significant variation in terms of the specific input variables selected and their relative importance, suggesting that an ensemble of models (specialising by land cover) is likely preferable to a general model applied everywhere. Visualising the patterns learned by the models (using SHAP dependence plots) provides further insights, building confidence in some cases (where patterns are consistent with domain knowledge and past studies) and highlighting potentially problematic variables in others (such as proxy relationships which may not apply in new application sites). Our results have implications for future DEM error prediction studies, particularly in evaluating a very wide range of potential input variables (160 candidates) drawn from topographic, multispectral, Synthetic Aperture Radar, vegetation, climate and urbanisation datasets.
{"title":"Explaining machine learning models trained to predict Copernicus DEM errors in different land cover environments","authors":"Michael Meadows, Karin Reinke, Simon Jones","doi":"10.1016/j.aiig.2025.100141","DOIUrl":"10.1016/j.aiig.2025.100141","url":null,"abstract":"<div><div>Machine learning models are increasingly used to correct the vertical biases (mainly due to vegetation and buildings) in global Digital Elevation Models (DEMs), for downstream applications which need “bare earth” elevations. The predictive accuracy of these models has improved significantly as more flexible model architectures are developed and new explanatory datasets produced, leading to the recent release of three model-corrected DEMs (FABDEM, DiluviumDEM and FathomDEM). However, there has been relatively little focus so far on explaining or interrogating these models, especially important in this context given their downstream impact on many other applications (including natural hazard simulations). In this study we train five separate models (by land cover environment) to correct vertical biases in the Copernicus DEM and then explain them using SHapley Additive exPlanation (SHAP) values. Comparing the models, we find significant variation in terms of the specific input variables selected and their relative importance, suggesting that an ensemble of models (specialising by land cover) is likely preferable to a general model applied everywhere. Visualising the patterns learned by the models (using SHAP dependence plots) provides further insights, building confidence in some cases (where patterns are consistent with domain knowledge and past studies) and highlighting potentially problematic variables in others (such as proxy relationships which may not apply in new application sites). Our results have implications for future DEM error prediction studies, particularly in evaluating a very wide range of potential input variables (160 candidates) drawn from topographic, multispectral, Synthetic Aperture Radar, vegetation, climate and urbanisation datasets.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100141"},"PeriodicalIF":0.0,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144663596","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-13DOI: 10.1016/j.aiig.2025.100143
Christian Carhuancho , Edwin Villanueva , Christian Yarleque , Romel Erick Principe , Marcia Castromonte
In weather forecasting, generating atmospheric variables for regions with complex topography, such as the Andean regions with peaks reaching 6500 m above sea level, poses significant challenges. Traditional regional climate models often struggle to accurately represent the atmospheric behavior in such areas. Furthermore, the capability to produce high spatio-temporal resolution data (less than 27 km and hourly) is limited to a few institutions globally due to the substantial computational resources required. This study presents the results of atmospheric data generated using a new type of artificial intelligence (AI) models, aimed to reduce the computational cost of generating downscaled climate data using climate regional models like the Weather Research and Forecasting (WRF) model over the Andes. The WRF model was selected for this comparison due to its frequent use in simulating atmospheric variables in the Andes.
Our results demonstrate a higher downscaling performance for the four target weather variables studied (temperature, relative humidity, zonal and meridional wind) over coastal, mountain, and jungle regions. Moreover, this AI model offers several advantages, including lower computational costs compared to dynamic models like WRF and continuous improvement potential with additional training data.
{"title":"Generating high-resolution climate data in the Andes using artificial intelligence: A lightweight alternative to the WRF model","authors":"Christian Carhuancho , Edwin Villanueva , Christian Yarleque , Romel Erick Principe , Marcia Castromonte","doi":"10.1016/j.aiig.2025.100143","DOIUrl":"10.1016/j.aiig.2025.100143","url":null,"abstract":"<div><div>In weather forecasting, generating atmospheric variables for regions with complex topography, such as the Andean regions with peaks reaching 6500 m above sea level, poses significant challenges. Traditional regional climate models often struggle to accurately represent the atmospheric behavior in such areas. Furthermore, the capability to produce high spatio-temporal resolution data (less than 27 km and hourly) is limited to a few institutions globally due to the substantial computational resources required. This study presents the results of atmospheric data generated using a new type of artificial intelligence (AI) models, aimed to reduce the computational cost of generating downscaled climate data using climate regional models like the Weather Research and Forecasting (WRF) model over the Andes. The WRF model was selected for this comparison due to its frequent use in simulating atmospheric variables in the Andes.</div><div>Our results demonstrate a higher downscaling performance for the four target weather variables studied (temperature, relative humidity, zonal and meridional wind) over coastal, mountain, and jungle regions. Moreover, this AI model offers several advantages, including lower computational costs compared to dynamic models like WRF and continuous improvement potential with additional training data.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100143"},"PeriodicalIF":0.0,"publicationDate":"2025-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144653503","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-05DOI: 10.1016/j.aiig.2025.100138
B.T. Gunel , Y.D. Pak , A.Ö. Herekeli , S. Gül , B. Kulga , E. Artun
Characterization and optimization of physical and chemical properties of drilling fluids are critical for the efficiency and success of drilling operations. In particular, maintaining the optimal levels of solids content is essential for achieving the most effective fluid performance. Proper management of solids content also reduces the risk of tool failures. Traditional solids content analysis methods, such as retort analysis, require substantial human intervention and time, which can lead to inaccuracies, time-management issues, and increased operational risks. In contrast to human-intensive methods, machine learning may offer a viable alternative for solids content estimation due to its pattern-recognition capability. In this study, a large set of laboratory reports of drilling-fluid analyses from 130 oil wells around the world were compiled to construct a comprehensive data set. The relationships among various rheological parameters were analyzed using statistical methods and machine learning algorithms. Several machine learning algorithms of diverse classes, namely linear (linear regression, ridge regression, and ElasticNet regression), kernel-based (support vector machine) and ensemble tree-based (gradient boosting, XGBoost, and random forests) algorithms, were trained and tuned to estimate solids content from other readily available drilling fluid properties. Input variables were kept consistent across all models for interpretation and comparison purposes. In the final stage, different evaluation metrics were employed to evaluate and compare the performance of different classes of machine learning models. Among all algorithms tested, random forests algorithm was found to be the best predictive model resulting in consistently high accuracy. Further optimization of the random forests model resulted in a mean absolute percentage error (MAPE) of 3.9% and 9.6% and R of 0.99 and 0.93 for the training and testing sets, respectively. Analysis of residuals, their histograms and Q-Q normality plots showed Gaussian distributions with residuals that are scattered around a mean of zero within error ranges of 1% and 4%, for training and testing, respectively. The selected model was further validated by applying the rheological measurements from mud samples taken from an offshore well from the Gulf of Mexico. The model was able to estimate total solids content in those four mud samples with an average absolute error of 1.08% of total solids content. The model was then used to develop a web-based graphical-user-interface (GUI) application, which can be practically used at the rig site by engineers to optimize drilling fluid programs. The proposed model can complement automation workflows that are designed to measure fundamental rheological properties in real time during drilling operations. While a st
{"title":"Machine learning assisted estimation of total solids content of drilling fluids","authors":"B.T. Gunel , Y.D. Pak , A.Ö. Herekeli , S. Gül , B. Kulga , E. Artun","doi":"10.1016/j.aiig.2025.100138","DOIUrl":"10.1016/j.aiig.2025.100138","url":null,"abstract":"<div><div>Characterization and optimization of physical and chemical properties of drilling fluids are critical for the efficiency and success of drilling operations. In particular, maintaining the optimal levels of solids content is essential for achieving the most effective fluid performance. Proper management of solids content also reduces the risk of tool failures. Traditional solids content analysis methods, such as retort analysis, require substantial human intervention and time, which can lead to inaccuracies, time-management issues, and increased operational risks. In contrast to human-intensive methods, machine learning may offer a viable alternative for solids content estimation due to its pattern-recognition capability. In this study, a large set of laboratory reports of drilling-fluid analyses from 130 oil wells around the world were compiled to construct a comprehensive data set. The relationships among various rheological parameters were analyzed using statistical methods and machine learning algorithms. Several machine learning algorithms of diverse classes, namely linear (linear regression, ridge regression, and ElasticNet regression), kernel-based (support vector machine) and ensemble tree-based (gradient boosting, XGBoost, and random forests) algorithms, were trained and tuned to estimate solids content from other readily available drilling fluid properties. Input variables were kept consistent across all models for interpretation and comparison purposes. In the final stage, different evaluation metrics were employed to evaluate and compare the performance of different classes of machine learning models. Among all algorithms tested, random forests algorithm was found to be the best predictive model resulting in consistently high accuracy. Further optimization of the random forests model resulted in a mean absolute percentage error (MAPE) of 3.9% and 9.6% and R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span> of 0.99 and 0.93 for the training and testing sets, respectively. Analysis of residuals, their histograms and Q-Q normality plots showed Gaussian distributions with residuals that are scattered around a mean of zero within error ranges of <span><math><mo>±</mo></math></span>1% and <span><math><mo>±</mo></math></span>4%, for training and testing, respectively. The selected model was further validated by applying the rheological measurements from mud samples taken from an offshore well from the Gulf of Mexico. The model was able to estimate total solids content in those four mud samples with an average absolute error of 1.08% of total solids content. The model was then used to develop a web-based graphical-user-interface (GUI) application, which can be practically used at the rig site by engineers to optimize drilling fluid programs. The proposed model can complement automation workflows that are designed to measure fundamental rheological properties in real time during drilling operations. While a st","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100138"},"PeriodicalIF":0.0,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144581266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-05DOI: 10.1016/j.aiig.2025.100144
Oriyomi Raheem , Misael M. Morales , Wen Pan , Carlos Torres-Verdín
Capillary pressure plays a crucial role in determining the spatial distribution of oil and gas, particularly in medium-to-low permeability reservoirs, where it is closely linked to the rock's pore structure and wettability. In these environments, pore structure is the primary factor influencing capillary pressure, with different pore types affecting fluid transport through varying degrees of hydrocarbon saturation. One of the main challenges in characterizing pore structure is how to use data from core plugs to establish a relationship with microscopic pore and throat properties, enabling more accurate predictions of capillary pressure. While special core analysis laboratory experiments are effective, they are time-consuming and expensive. In contrast, nuclear magnetic resonance (NMR) measurements, which provide information on pore body size distribution, are faster and can be leveraged to estimate capillary pressure using machine learning algorithms. Recently, artificial intelligence methods have also been applied to capillary pressure prediction (Qi et al., 2024).
Currently, no readily applicable predictive model exists for estimating an entire capillary pressure curve directly from standard petrophysical logs and core data. Although porescale imaging and network modeling techniques can compute capillary pressure from micro-CT rock images (Øren and Bakke, 2003; Valvatne and Blunt, 2004), these approaches are time-consuming, limited to small sample volumes, and not yet practical for routine reservoir evaluation. In this study, we introduce rock classification techniques and implement a data-driven machine learning (ML) method to estimate saturation-dependent capillary pressure from core petrophysical properties. The new model integrates cumulative NMR data and densely resampled core measurements as training data, with prediction errors quantified throughout the process. To approach the common condition of sparsely sampled training data, we transformed the prediction problem into an overdetermined one by applying composite fitting to both capillary pressure and pore throat size distribution, and Gaussian cumulative distribution fitting to the NMR measurements, generating evenly sampled data points. Using these preprocessed input features, we performed classification based on the natural logarithm of the permeability-to-porosity ratio to cluster distinct rock types. For each rock class, we applied regression techniques—such as random forest (RF), k-nearest neighbors (k-NN), extreme gradient boosting (XGB), and artificial neural networks (ANN)—to estimate the logarithm of capillary pressure. The methods were tested on blind core samples, and performance comparisons among different estimation methods
毛管压力对油气的空间分布起着至关重要的作用，特别是在中低渗透储层中，毛管压力与岩石的孔隙结构和润湿性密切相关。在这些环境中，孔隙结构是影响毛管压力的主要因素，不同孔隙类型通过不同程度的烃饱和度影响流体的输运。表征孔隙结构的主要挑战之一是如何利用岩心桥塞的数据建立微观孔隙和喉道特性的关系，从而更准确地预测毛管压力。虽然特殊的岩心分析实验室实验是有效的，但它们耗时且昂贵。相比之下，核磁共振（NMR）测量可以提供孔体大小分布的信息，速度更快，并且可以利用机器学习算法来估计毛细管压力。最近，人工智能方法也被应用于毛细管压力预测（Qi et al., 2024）。目前，还没有现成的预测模型可以直接从标准岩石物理测井和岩心数据中估计整个毛管压力曲线。虽然孔隙尺度成像和网络建模技术可以从微ct岩石图像中计算毛细压力(Øren和Bakke, 2003；Valvatne和Blunt， 2004)，这些方法耗时长，仅限于小样本量，还不能用于常规储层评价。在这项研究中，我们引入了岩石分类技术，并实现了一种数据驱动的机器学习（ML）方法，通过岩心岩石物理性质来估计与饱和度相关的毛管压力。新模型将累积核磁共振数据和密集重采样的岩心测量数据作为训练数据，并在整个过程中量化预测误差。为了接近稀疏采样训练数据的常见情况，我们通过对毛细管压力和孔喉大小分布进行复合拟合，并对NMR T2测量值进行高斯累积分布拟合，将预测问题转化为过确定问题，生成均匀采样的数据点。利用这些预处理的输入特征，我们根据渗透率-孔隙度比（ln（k/ϕ））的自然对数进行分类，以聚类不同的岩石类型。对于每个岩石类别，我们应用回归技术——如随机森林（RF）、k近邻（k-NN）、极端梯度增强（XGB）和人工神经网络（ANN）——来估计毛细管压力的对数。对盲岩心样本进行了测试，并基于预测的相对标准误差对不同估计方法进行了性能比较。结果表明，核磁共振数据对岩石孔隙结构较为敏感，对毛细管压力和孔喉大小分布的预测有显著改善。对于毛细管压力和孔喉大小分布，极端梯度增强和随机森林模型的平均估计误差分别为5%和10%，表现最好。相比之下，当NMR T2数据被排除作为输入特征时，预测误差增加到25%。使用传统的高斯模型拟合和更高分辨率的重采样确保了训练数据覆盖了广泛的变异性。将核磁共振T2数据作为输入特征增强了模型捕捉非常规岩石中多峰的能力，使预测问题过度确定。通过向量输入特征预测向量函数，有效降低了预测误差。该解释工作流程可用于构建具有代表性的分类模型，并在广泛的饱和度范围内估计毛细管压力。
{"title":"Improved estimation of two-phase capillary pressure with nuclear magnetic resonance measurements via machine learning","authors":"Oriyomi Raheem , Misael M. Morales , Wen Pan , Carlos Torres-Verdín","doi":"10.1016/j.aiig.2025.100144","DOIUrl":"10.1016/j.aiig.2025.100144","url":null,"abstract":"<div><div>Capillary pressure plays a crucial role in determining the spatial distribution of oil and gas, particularly in medium-to-low permeability reservoirs, where it is closely linked to the rock's pore structure and wettability. In these environments, pore structure is the primary factor influencing capillary pressure, with different pore types affecting fluid transport through varying degrees of hydrocarbon saturation. One of the main challenges in characterizing pore structure is how to use data from core plugs to establish a relationship with microscopic pore and throat properties, enabling more accurate predictions of capillary pressure. While special core analysis laboratory experiments are effective, they are time-consuming and expensive. In contrast, nuclear magnetic resonance (NMR) measurements, which provide information on pore body size distribution, are faster and can be leveraged to estimate capillary pressure using machine learning algorithms. Recently, artificial intelligence methods have also been applied to capillary pressure prediction (Qi et al., 2024).</div><div>Currently, no readily applicable predictive model exists for estimating an entire capillary pressure curve directly from standard petrophysical logs and core data. Although porescale imaging and network modeling techniques can compute capillary pressure from micro-CT rock images (Øren and Bakke, 2003; Valvatne and Blunt, 2004), these approaches are time-consuming, limited to small sample volumes, and not yet practical for routine reservoir evaluation. In this study, we introduce rock classification techniques and implement a data-driven machine learning (ML) method to estimate saturation-dependent capillary pressure from core petrophysical properties. The new model integrates cumulative NMR data and densely resampled core measurements as training data, with prediction errors quantified throughout the process. To approach the common condition of sparsely sampled training data, we transformed the prediction problem into an overdetermined one by applying composite fitting to both capillary pressure and pore throat size distribution, and Gaussian cumulative distribution fitting to the NMR <span><math><mrow><msub><mi>T</mi><mn>2</mn></msub></mrow></math></span> measurements, generating evenly sampled data points. Using these preprocessed input features, we performed classification based on the natural logarithm of the permeability-to-porosity ratio <span><math><mrow><mo>(</mo><mrow><mi>ln</mi><mrow><mo>(</mo><mrow><mi>k</mi><mo>/</mo><mi>ϕ</mi></mrow><mo>)</mo></mrow></mrow><mo>)</mo></mrow></math></span> to cluster distinct rock types. For each rock class, we applied regression techniques—such as random forest (RF), k-nearest neighbors (k-NN), extreme gradient boosting (XGB), and artificial neural networks (ANN)—to estimate the logarithm of capillary pressure. The methods were tested on blind core samples, and performance comparisons among different estimation methods ","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100144"},"PeriodicalIF":0.0,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633212","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-04DOI: 10.1016/j.aiig.2025.100140
Vinh V. Le , HongGiang Nguyen , Nguyen Huu Ngu
This paper presents a deep learning architecture combined with exploratory data analysis to estimate maximum wall deflection in deep excavations. Six major geotechnical parameters were studied. Statistical methods, such as pair plots and Pearson correlation, highlighted excavation depth (correlation coefficient = 0.82) as the most significant factor. For method prediction, five deep learning models (CNN, LSTM, BiLSTM, CNN-LSTM, and CNN-BiLSTM) were built. The CNN-BiLSTM model excelled in training performance (R2 = 0.98, RMSE = 0.02), while BiLSTM reached superior testing results (R2 = 0.85, RMSE = 0.06), suggesting greater generalization ability. Based on the feature importance analysis from model weights, excavation depth, stiffness ratio, and bracing spacing were ranked as the highest contributors. This point verified a lack of prediction bias on residual plots and high model agreement with measured values on Taylor diagrams (correlation coefficient 0.92). The effectiveness of integrated techniques was reliably assured for predicting wall deformation. This approach facilitates more accurate and efficient geotechnical design and provides engineers with improved tools for risk evaluation and decision-making in deep excavation projects.
{"title":"Deep learning approaches for estimating maximum wall deflection in excavations with inconsistent clay stratigraphy","authors":"Vinh V. Le , HongGiang Nguyen , Nguyen Huu Ngu","doi":"10.1016/j.aiig.2025.100140","DOIUrl":"10.1016/j.aiig.2025.100140","url":null,"abstract":"<div><div>This paper presents a deep learning architecture combined with exploratory data analysis to estimate maximum wall deflection in deep excavations. Six major geotechnical parameters were studied. Statistical methods, such as pair plots and Pearson correlation, highlighted excavation depth (correlation coefficient = 0.82) as the most significant factor. For method prediction, five deep learning models (CNN, LSTM, BiLSTM, CNN-LSTM, and CNN-BiLSTM) were built. The CNN-BiLSTM model excelled in training performance (R<sup>2</sup> = 0.98, RMSE = 0.02), while BiLSTM reached superior testing results (R<sup>2</sup> = 0.85, RMSE = 0.06), suggesting greater generalization ability. Based on the feature importance analysis from model weights, excavation depth, stiffness ratio, and bracing spacing were ranked as the highest contributors. This point verified a lack of prediction bias on residual plots and high model agreement with measured values on Taylor diagrams (correlation coefficient 0.92). The effectiveness of integrated techniques was reliably assured for predicting wall deformation. This approach facilitates more accurate and efficient geotechnical design and provides engineers with improved tools for risk evaluation and decision-making in deep excavation projects.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100140"},"PeriodicalIF":0.0,"publicationDate":"2025-07-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144581267","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-30DOI: 10.1016/j.aiig.2025.100142
Baoling Gui, Anshuman Bhardwaj, Lydia Sam
Rapid urbanization and land-use changes are placing immense pressure on resources, infrastructure, and environmental sustainability. To address these, accurate urban simulation models are essential for sustainable development and governance. Among them, Cellular Automata (CA) models have become key tools for predicting urban expansion, optimizing land-use planning, and supporting data-driven decision-making. This review provides a comprehensive examination of the development of urban cellular automata (UCA) models, presenting a new framework to enhance individual UCA sub-modules within the context of emerging technologies, sustainable environments, and public governance. By addressing gaps in prior UCA modelling reviews—particularly in the integration and optimization of UCA sub-module technologies—this framework is designed to simplify UCA model understanding and development. We systematically review pioneering case studies, deconstruct current UCA operational processes, and explore modern technologies, such as big data and artificial intelligence, to optimize these sub-modules further. We discuss current limitations within UCA models and propose future pathways, emphasizing the necessity of comprehensive analyses for effective UCA simulations. Proposed solutions include strengthening our understanding of urban growth mechanisms, examining spatial positioning and temporal evolution dynamics, and enhancing urban geographic simulations with deep learning techniques to support sustainable transitions in public governance. These improvements offer data-driven decision support for environmental management, advancing policies that foster sustainable urban development.
{"title":"Cellular automata models for simulation and prediction of urban land use change: Development and prospects","authors":"Baoling Gui, Anshuman Bhardwaj, Lydia Sam","doi":"10.1016/j.aiig.2025.100142","DOIUrl":"10.1016/j.aiig.2025.100142","url":null,"abstract":"<div><div>Rapid urbanization and land-use changes are placing immense pressure on resources, infrastructure, and environmental sustainability. To address these, accurate urban simulation models are essential for sustainable development and governance. Among them, Cellular Automata (CA) models have become key tools for predicting urban expansion, optimizing land-use planning, and supporting data-driven decision-making. This review provides a comprehensive examination of the development of urban cellular automata (UCA) models, presenting a new framework to enhance individual UCA sub-modules within the context of emerging technologies, sustainable environments, and public governance. By addressing gaps in prior UCA modelling reviews—particularly in the integration and optimization of UCA sub-module technologies—this framework is designed to simplify UCA model understanding and development. We systematically review pioneering case studies, deconstruct current UCA operational processes, and explore modern technologies, such as big data and artificial intelligence, to optimize these sub-modules further. We discuss current limitations within UCA models and propose future pathways, emphasizing the necessity of comprehensive analyses for effective UCA simulations. Proposed solutions include strengthening our understanding of urban growth mechanisms, examining spatial positioning and temporal evolution dynamics, and enhancing urban geographic simulations with deep learning techniques to support sustainable transitions in public governance. These improvements offer data-driven decision support for environmental management, advancing policies that foster sustainable urban development.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 2","pages":"Article 100142"},"PeriodicalIF":0.0,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144518220","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","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