China
Africa
Explore chapters and articles related to this topic
Applications of Physics-Informed Scientific Machine Learning in Subsurface Science: A Survey
Published in Anuj Karpatne, Ramakrishnan Kannan, Vipin Kumar, Knowledge-Guided Machine Learning, 2023
Alexander Y. Sun, Hongkyu Yoon, Chung-Yan Shih, Zhi Zhong
Fast advances in machine learning (ML) technologies have revolutionized predictive and prescriptive analytics in recent years. Significant interests exist in harnessing this new generation of ML tools for Earth system studies (Reichstein et al. 2019; Sun and Scanlon, 2019; Bergen et al., 2019). Unlike many other sectors, however, subsurface formations are often poorly characterized and scarcely monitored, thus relying extensively upon geological and geofluid modeling to generate spatially and temporally continuous “images” of the subsurface. A conventional workflow may consist of (a) geologic modeling, which seeks to provide a 3D representation of the geosystem under study by fusing qualitative interpretation of the geological structure, stratigraphy, and sedimentological facies, as well as quantitative data on geologic properties; and (b) fluid and geomechanical modeling, which describes the fluid flow, mass transport, and formation deformations through physics-based governing equations and the accompanying initial/boundary/forcing conditions. Once established, the workflow is used to generate 3D “images” of the subsurface processes for inference and/or prediction (Figure 5.1).
How virtual reality can help visualise and assess geohazards
Published in International Journal of Digital Earth, 2019
Hans-Balder Havenith, Philippe Cerfontaine, Anne-Sophie Mreyen
As GIS data generally only contain spatial (possibly with limited discrete temporal) components, they can cover much wider areas than the modelling in- and outputs. The highly changing spatial scales of maps and models to be represented is one of the reasons why most handling software are unable to visualise and process the ‘other’ type of data and outputs with the adapted spatial/temporal resolution/extent (e.g. GIS cannot efficiently show data over varying depths, whereas modelling software generally does not include mapping tools). A compromise is proposed by a third type of modelling techniques that can be grouped together under the general term of 3D visualisation tools or, more specifically, of 3D geomodellers. This software can represent in the same time large maps and much smaller cross-sections or 3D numerical models representing simulation outputs. The geological modelling (or simply geomodelling) software is generally not used to create the data, but it helps representing in- and outputs in the 3D space. In addition, this software allows for some pre-processing of information needed for the numerical models and for the development of 3D volumes on the basis of points, lines or surface data distributed within a 3D space. As volumes are the core part of 3D geomodels, geomodelling tools must be able to visualise efficiently the 3D space. Therefore, their 3D visualisation capabilities generally exceed by far those of GIS or original numerical modelling tools. Geomodels also allow for 3D spatial and temporal analyses (if the required data are included in the model). Some workflows related to local geohazard studies involving also geomodels are presented in Section 4 (case studies).