Sediment core analysis using artificial intelligence.

Area of Science:

• Geology
• Machine Learning
• Stratigraphy

Background:

• Subsurface stratigraphic modeling is vital for environmental and economic applications.
• Sediment core analysis requires specialized sedimentological expertise, posing a bottleneck.
• Automated methods using Machine Learning (ML) and Deep Learning (DL) can streamline this process.

Purpose of the Study:

• To develop a novel deep-learning-based approach for automatic semantic segmentation of sediment core images.
• To enable rapid characterization of sediment cores for improved subsurface reconstructions.
• To overcome limitations of traditional sedimentological analysis by leveraging AI.

Main Methods:

• Utilized a dataset of high-resolution digital images from Holocene-age continuous sediment cores.
• Applied a deep learning model, specifically convolutional neural networks (CNNs), for semantic segmentation.
• Defined six sedimentary facies associations as target classes to enhance interpretative accuracy.

Main Results:

• Successfully developed an automated model for semantic segmentation of sediment core images.
• Demonstrated the model's capability to rapidly characterize sediment cores across diverse depositional environments.
• Showcased the potential for immediate guidance in stratigraphic correlation and subsurface modeling.

Conclusions:

• The proposed deep learning approach offers an efficient and automated solution for sediment core characterization.
• This method can significantly reduce the time and expertise required for subsurface stratigraphic modeling.
• Facilitates more accessible and rapid subsurface reconstructions for various applications.