Application of Artificial Intelligence in Geological Surveys
Yu Jiang Chang, PE in Civil Engineering, Wei Yao Yu, PE in Applied Geology Taiwan
This study explores the application of artificial intelligence in geological surveys, including field investigations, drill core sample identification, report writing, and geological disaster risk assessment. The research utilized advanced language models and open-source AI platforms to automate and optimize the geological reporting and review process. The advantages and disadvantages of various AI applications were comparatively analyzed. Additionally, the revolutionary potential of smart glasses and drone image recognition technology in field data collection and real-time analysis was also explored.
Overview of AI Applications in Geological Surveying
The application of AI in geological surveying spans the entire process, from field investigation to report generation. AI technologies can significantly improve the efficiency and accuracy of data collection, while reducing human errors. In field investigations, AI-powered drones and smart glasses can quickly collect and analyze geological data. In sample identification, AI vision models can accurately identify rock types and geological structures. In the report generation stage, large language models can automate the writing process, ensuring consistency and completeness of the reports. Additionally, AI plays a crucial role in geological hazard risk assessment, as it can process large amounts of data and identify potential risks.
1
Field Investigation
AI-driven drones and smart glasses collect data
2
Sample Identification
AI vision models identify rock types and geological structures
3
Report Generation
Large language models automate the writing process
4
Risk Assessment
AI processes large data and identifies potential geological hazards
Traditional Geological Survey Methods
In Taiwan, traditional geological survey methods are subject to strict regulations under the Building Technical Regulations (Article 64). For buildings of five stories or more or for public use, regardless of height, a geological survey must be conducted. For smaller private buildings (four stories or less), a geological survey is usually not required, unless certain specific conditions are met. If the excavation depth exceeds 5 meters, or if there is no reliable data from adjacent plots, a comprehensive geological survey must be carried out. Additionally, if the building's footprint exceeds 600 square meters, a geological survey is also required. These traditional methods typically involve labor-intensive fieldwork, including manual drilling, recording results (such as standard penetration test N-values), and collecting rock samples.
Buildings of 5 stories or more
Comprehensive geological survey is required
Excavation depth exceeding 5 meters
Geological survey is required
Footprint exceeding 600 square meters
Geological survey is required
Geologically sensitive areas
Detailed geological safety assessment is needed
Challenges of Traditional Geological Survey Methods
Traditional geological survey methods face multiple challenges. First, these methods are highly time-consuming. Manual drilling, data recording, and sample collection, combined with laboratory analysis and final report writing, can lead to extended project timelines. Secondly, the reliance on human labor and interpretation introduces significant risk of human error. Misinterpretation of core samples, recording errors, or flawed analysis of historical geological data can impact the quality of geological reports, potentially introducing risks during construction. Furthermore, these traditional methods typically require a large workforce. Drilling teams are needed, and skilled engineers are required to analyze data and write reports. This dependence on human labor makes the process prone to delays and inconsistent results, especially in large-scale projects that require extensive fieldwork and data analysis.
1
Time-consuming
Manual drilling, data recording, and sample analysis take a long time
2
Risk of human error
Sample misinterpretation and data analysis errors can impact report quality
3
High labor demand
Requires a large workforce of workers and engineers, leading to delays and inconsistent results
4
Regulatory compliance complexity
Navigating a complex legal framework adds to project time and cost
Comparison of AI-driven and Traditional Methods
The emergence of artificial intelligence (AI) has provided potential solutions to many challenges faced by traditional geological survey methods. AI technologies can automate important parts of the survey process, reducing reliance on manual labor and lowering the risk of human error. For example, large language models (LLMs) can assist in writing geological reports by automatically generating text based on data inputs, greatly accelerating the report writing process. Additionally, AI tools with visual models can aid in sample analysis, improving the accuracy and consistency of geological interpretations. By automating the interpretation of core samples or geological structures, these tools reduce the potential for human error and provide more reliable results. The ability of AI to process large amounts of data also allows for more accurate historical geological analysis, reducing the time required for manual research and analysis of historical records.
Aspect
Traditional Method
AI-driven Method
Data Collection
Manual, labor-intensive field work
Automated through drones and visual recognition models
Report Generation
Manually written by geologists
Automatically generated using large language models
Time Efficiency
Time-consuming
Rapid data processing and analysis
Error Rate
High risk of human errors
Reduced errors through automated analysis
Regulatory Compliance
Requires expert review
AI ensures compliance with regulatory frameworks
Application of AI in Field Investigations
In the field of geological surveying, AI has become a valuable tool for automating and simplifying field work. Traditional field investigations require manual data collection, which is time-consuming and prone to human error. Artificial intelligence provides more efficient alternatives, utilizing tools such as drones and visual recognition models. These AI technologies can capture high-resolution images, identify geological features, and analyze data on-site in real-time, eliminating the need for multiple field visits. The integration of AI in field investigations allows for faster and more accurate data collection and immediate decision-making, significantly improving the overall workflow. For example, AI-powered drones can quickly scan large areas of terrain, identify potential geological anomalies, while smart glasses can provide real-time data analysis and augmented reality overlays for field geologists.
AI-Powered Drones
Capture high-resolution geological imagery
Smart Glasses
Provide real-time data analysis and AR overlays
Real-Time Analysis
AI systems analyze geological features in real-time
Application of AI in Drill Sample Identification
The application of AI in drill sample identification has revolutionized the way geologists classify and analyze rock cores and other subsurface materials. Traditionally, this process required manual inspection and expert interpretation, but AI can automate much of the work. Visual models, such as Google Lens, can now be used to differentiate rock types by analyzing sample images. Google Lens uses advanced pattern recognition algorithms to identify specific geological features, allowing for more precise and efficient rock sample classification. Additionally, wearable technologies like Meta's smart glasses can customize AI-driven models to assist geologists in real-time sample identification. These smart glasses are equipped with visual recognition tools that can automatically identify various rock types on-site. ChatGPT's visual models go even further, integrating text and image data to fully automate the workflow, from petrological analysis to report writing.
Image Capture
Capture high-resolution images of rock samples using cameras or smart glasses
AI Analysis
AI visual models analyze the images, identifying rock types and features
Result Generation
AI system generates detailed rock classification reports
Expert Review
Geologists review the AI results and make adjustments as needed
Application of AI in Geological Disaster Risk Assessment
AI plays a crucial role in geological disaster risk assessment. A case in Taiwan in 2024 highlighted the limitations of traditional monitoring systems. An eastbound train derailed after colliding with a large boulder that had fallen onto the tracks. The existing monitoring system relied on human operators to monitor camera footage to detect obstacles, but human oversight failed to detect the danger in time. This incident underscored the inefficiency and potential risks of manual monitoring, especially for tasks requiring continuous attention. AI systems, such as OpenAI's vision models, can replace these human operators for the monotonous but critical monitoring tasks. By continuously scanning and analyzing camera footage to detect signs of geological disasters (such as landslides, rock falls, or debris on the tracks), AI systems can immediately detect anomalies and provide real-time alerts to authorities. This automation not only reduces the risk of human error, but also allows for more efficient and effective monitoring of critical infrastructure.
Landslide Detection
AI systems analyze terrain changes to predict potential landslides
Earthquake Risk Assessment
AI processes earthquake data to evaluate regional seismic risk
Flood Forecasting
AI analyzes weather and terrain data to predict flood risk
Real-time Alert System
AI-driven system provides immediate geological disaster alerts
Case Study: AI Application in Rock Classification
In our recent field experiments, we utilized various AI tools to improve the accuracy and efficiency of rock classification. The main AI platforms used in this study include Google Lens, OpenAI's GPT-4, and other large language models (LLMs) such as Claude (developed by Anthropic), Gemini (developed by Google), and Wenxin Yiyan (developed by Baidu). The initial experiments used Google Lens as a low-level AI tool for basic rock classification. While the tool was able to perform some basic pattern recognition tasks, the results were often vague and lacked the precision required for geological analysis. Subsequently, we used OpenAI's GPT-4, which was able to accurately classify rock samples as basalt. GPT-4 successfully distinguished rock types with a high degree of certainty, demonstrating its advanced analytical capabilities compared to simpler AI models. We further tested Claude, Gemini, and Wenxin Yiyan, and found that these models were generally able to identify the lithology of the rocks.
Google Lens
Basic pattern recognition, with vague results
GPT-4
Highly accurate rock classification, with detailed analysis
Other LLMs
Able to identify lithology, but with varying speed and accuracy
Case Study: AI Application in Core Box Analysis
In addition to rock classification, we have expanded our analysis to the core boxes on the drilling site. We provided the AI models with images of core samples and asked them to try to identify the rock types from the core boxes. The initial results were not very satisfactory, as the AI had difficulty making confident classifications from a single photo. However, when we presented the model with multiple images (from distant views of the core box to close-up views of individual rock samples), the accurate rock type classification success rate rose to over 70%. This suggests that LLMs have the potential to improve when provided with more detailed, multi-angle data. Our case study also explored how LLMs can identify geological structures, such as bedding planes, anticlines, and interbedded sandstone and conglomerate formations. When fed images of these structures, the model was able to distinguish these structures with significant accuracy. This capability is crucial for the field work of geologists who require precise structural data.
1
Single Image Analysis
AI had difficulty making accurate classifications from a single photo
2
Multi-Angle Image Input
Provided distant and close-up views
3
Improved AI Analysis
Accurate classification rate rose to over 70%
4
Geological Structure Identification
AI successfully identified complex geological structures
Geospatial Data Management and Drone Applications
The potential applications of AI extend beyond static image analysis. We have also experimented with images captured by drones, integrating this data with Google's terrain maps to assess geological features. When provided with additional regional geological maps, LLMs have demonstrated powerful capabilities in generating detailed geological analyses. This approach holds great promise for pre-disaster potential analysis, providing key insights on geological hazards (such as landslide or flood risk) to local governments and disaster management agencies, especially during typhoon seasons. Furthermore, innovations like Meta's smart glasses offer future potential for real-time geological analysis. While current image resolutions and AI model integration are still in early stages, the combination of high-resolution real-time imagery and LLM APIs portends that real-time rock classification may soon become a reality. This development will be particularly valuable in situations requiring rapid geological assessments, such as fieldwork in disaster-prone areas or during emergency response operations.
Drone Imagery Integration
Combining drone-captured imagery with Google terrain maps to provide a comprehensive geological view
Pre-Disaster Analysis
AI analysis of geological data to predict potential disaster risks, such as landslides and floods
Smart Glasses Applications
Future potential for real-time rock classification and geological analysis
Emergency Response Support
AI-assisted rapid geological assessment to support disaster response decision-making
Application of AI in the Third-Party Review Process
In Taiwan, the review process for geological reports, especially those related to geologically sensitive areas, is governed by the Geology Act and relevant administrative orders, such as the Operating Guidelines for Geological Safety Assessment in Geologically Sensitive Areas. These regulations specify the format, content, and procedural requirements for the geological safety assessment of geologically sensitive areas (such as fault lines, landslide-prone areas, groundwater recharge areas, and geological heritage sites). The third-party review process is typically carried out by professional associations, such as the Civil Engineering Association or the Applied Geology Technicians Association. These organizations are responsible for ensuring that the geological investigations meet the required standards, including technical content and legal compliance. However, this review process is often time-consuming, and the quality of the review may vary depending on the individual reviewer's expertise and personal style. Differences in interpretation and requirements among different reviewers can lead to inconsistencies in the assessment process.
Report Submission
Geological reports are submitted to professional associations for review
AI Preliminary Review
The AI system conducts a preliminary analysis, checking for format and content completeness
Expert Manual Review
Experts review the AI analysis results and conduct in-depth technical evaluation
Final Review Report
The final review report is generated by combining the AI analysis and expert opinions
Application of Large Language Models in Automated Review
To address these challenges, large language models (LLMs) can be used to automate the preliminary stages of the review process. By integrating AI into these workflows, the manual labor required for reviewing geological reports can be significantly reduced, and the impact of individual reviewer biases can be minimized. For example, the author of this paper has extensive experience in reviewing structural assessments, soil and water conservation reports, seismic assessments, and geological reports. Over the years, tens of thousands of review comments have been generated, which can now be used to train LLMs to perform reviews efficiently. LLMs can be particularly useful for automating the initial review by analyzing the text content of geological reports and comparing them to established guidelines and review patterns. These models can identify discrepancies, flag incomplete sections, and provide structured feedback based on past reviews. In doing so, they create a standardized approach to the review process, minimize human errors, and ensure that each report complies with the legal requirements specified in the "Operating Guidelines for Geological Investigation and Geological Safety Assessment in Geologically Sensitive Areas in Taiwan".
1
Text Analysis
LLMs analyze report content, identify key information, and detect potential issues
2
Compliance Checks
AI systems automatically check if reports meet regulatory requirements
3
Standardized Feedback
Generate structured feedback based on past review experience
4
Efficiency Improvement
Significantly reduce the time required for the initial review, improving overall review efficiency
Limitations of AI Review Systems
However, while large language models (LLMs) are effective in text-based review, they currently face limitations in analyzing technical drawings and charts, which are often critical components of geological reports. For example, human reviewers' expertise is still indispensable when reviewing the structural integrity or geological features depicted in blueprints or cross-section diagrams. Therefore, AI can greatly simplify the review process, but cannot completely replace human oversight, especially when dealing with complex visual data. AI systems may also struggle with unstructured data, such as hand-drawn sketches or non-standard format charts. Additionally, AI may have difficulty understanding the subtle context or implicit information in geological reports, which typically requires years of field experience to interpret correctly. Therefore, while AI can significantly improve the efficiency of the review process, the judgment and experience of human experts remains key to ensuring a comprehensive and accurate review.
AI Advantages
AI Limitations
Role of Human Experts
AI Ethics Considerations
Ethical considerations are paramount in the application of AI in any field. This is especially true in the realm of geoscience, where the impact of decisions based on AI analysis can have significant consequences.
Thank you for your time
Thank you for your time, we hope this presentation has helped you understand the application of artificial intelligence in geological surveys.
We are excited to work with you to explore how we can leverage AI to optimize your geological survey workflows.