Publishing and Intuitive Visual Analysis of Rock Samples in Large Geologic Repositories.
Abstract
A key practical need of state and federal geological repositories, university collections, and natural history museums, is to expose their resources to researchers seeking geological information. A flexible and intuitive visual data exploration system, integrated with tools for cartographic and statistical analysis, is needed to help researchers identify samples of interest and communicate their findings. We present our experience using Survey Analysis via Visual Exploration (SuAVE, http://suave.sdsc.edu) for several mineral sample and rock core collections from the US Geological Survey, the British Geologic Survey, and academic repositories.
SuAVE is a free open source visual analysis platform that lets users publish and share their image collections (surveys) online, explore them visually and statistically, display them on a map, compare and annotate individual samples or distribution patterns, and communicate analytical results as a series of annotated data views. Its central feature is the ability to simultaneously analyze photographs and metadata about specimens or other objects. Users can explore visual patterns in the entire collection (as opposed to a more typical way of browsing a collection page by page or image by image), and zoom in to individual photos or subsets. This allows geologists to zero in on samples of interests using any combination of metadata elements (geologic age, depth, formation, sample type, and more), while viewing all images in a subset and accessing associated resources, such as analytical reports. For additional analysis, SuAVE integrates with Jupyter notebooks. Users can launch Jupyter notebooks residing on several Jupyterhubs directly from SuAVE, to perform image processing, generate additional variables and compute statistical models based on sample metadata or data extracted from photographs, or run image classification algorithms. Results, such as derived variables, keywords, class labels, or statistical estimates, are then automatically added to SuAVE for additional visual analysis. This loosely-coupled integration approach to exploration and in-depth data science analysis of rock and similar collections has the potential to improve accessibility and analytical re-use of such collections in a manner that is easy to communicate and reproduce.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMIN21B..18Z
- Keywords:
-
- 0850 Geoscience education research;
- EDUCATION;
- 9820 Techniques applicable in three or more fields;
- GENERAL OR MISCELLANEOUS;
- 1992 Virtual globes;
- INFORMATICS;
- 1994 Visualization and portrayal;
- INFORMATICS