GAGE/SAGE Plenary Webinar: New Approaches to Processing Big Geophysical and Geospatial Datasets
Thursday, October 22, 2020, 2:00pm ET
Please join us for a virtual GAGE/SAGE Plenary Webinar on October 22, 2020 at 2 PM Eastern. The plenary session is New Approaches to Processing Big Geophysical and Geospatial Datasets. The in-person GAGE/SAGE Science Workshop was postponed from August 2020 to August 2021. In the meantime, several of the planned speakers are presenting in this webinar series so we learn more about these subjects before next year.
Presenters: Dr. Lindsey Heagy, UC-Berkeley and Dr. Michael Olsen, Oregon State University. Abstracts are below the webinar link and call-in info.
- Dr. Heagy will present "Community Driven Development of Open Source Tools for Simulations and Inversions of Geophysical Data"
- Dr. Olsen will present "From Big to Small, You Can Do It All! Efficient Approaches to Analyze Rockfall Activity from Point Clouds
Please click the link below to register for the webinar:
After registering, you will receive a confirmation email containing information about joining the webinar.
Plenary Abstract: New Approaches to Processing Big Geophysical and Geospatial Datasets
‘Big data’ is challenging our computational capabilities in terms of volume, velocity and variety (3Vs). Increases in the spatial and temporal resolution of our geophysical observations have produced dramatic increases in the amount of data for geoscientists to analyze. Each stage of the life cycle of geospatial big data can present issues for our community: (1) data acquisition, (2) compilation and management, and (3) data analysis, visualization and distribution. This session seeks to showcase strategies that are overcoming these obstacles and providing new insights into Earth behavior. For instance, machine learning presents a host of new opportunities for our community, with increasing emphasis on technical rigor, common benchmarks, and repeatability. In addition, enhanced understanding can also be achieved via data amalgamation, high-performance computing, data mining, and dataset-integration strategies. Vast datasets also create opportunities for improved statistical characterization and improved quantification of uncertainties. We are particularly interested in highlighting techniques likely to be scalable and transferable to other problems.
Dr. Heagy's Abstract: Community Driven Development of Open Source Tools for Simulations and Inversions of Geophysical Data
Open communities in astrophysics, scientific computing, machine learning, and many other domains demonstrate the power of collaborative efforts to develop open-source software that facilitates research in each of their respective domains (e.g. Astropy, SciPy, Scikit-learn, etc.). Not only do open tools facilitate the reproducibility of scientific work, but they also streamline the exchange of ideas between researchers, even across domains. In 2013, we started SimPEG as an effort to build an open-source framework and community around numerical simulations and gradient-based inversions in geophysics. The SimPEG software supports forward simulations and inversions across a range of geophysical methods including magnetics, gravity, direct current resistivity, induced polarization, electromagnetics (time domain, frequency domain and natural source methods) as well as fluid flow. We leverage tools in the Pangeo ecosystem, including Jupyter and Dask (for parallelization), to enable scalable, interactive computation (locally, on the cloud, and HPC centers). The community has grown and brings together researchers working on advancing inversion methods (e.g. joint inversions, compact regularizations, large-scale inversions), as well as those focussed on using geophysical data in applications that span groundwater, mineral exploration, tectonic studies and near-surface applications such as agriculture. In this talk, I will share examples from active research projects led by members of the community and discuss how we have built a suite of resources that combine SimPEG with tools in the Jupyter ecosystem to enable interactive exploration of geophysical simulations and inversions.
Dr. Olsen's Abstract: From Big to Small, You Can Do It All! Efficient Approaches to Analyze Rockfall Activity from Point Clouds
Important infrastructure such as highways in the Pacific Northwest traverse particularly unstable terrain throughout much of the state, resulting in maintenance, system unreliability due to frequent closures and restrictions, and safety hazards due to landslides and rockfalls. Seismic activity significantly amplifies these negative economic and community impacts. This presentation will discuss efficient data processing strategies to utilize point cloud data to analyze rockfall activity at rock outcrops from repeat lidar or Uncrewed Aircraft Systems (UAS) Structure from Motion/MultiView Stereo (SfM/MVS) photogrammetric surveys that enable one to simultaneously look closely at detailed features on the slope as well as from afar to evaluate an entire corridor. These technologies enable personnel to safely capture data for active rock slopes that may be otherwise inaccessible. Nevertheless, the “Big Data” generated through these technologies combined with other data sources provide immense challenges in terms of acquisition and data processing. This presentation will discuss a suite of tools enabling efficient analyses of detailed remote sensing data to quantify and visually communicate rockfall hazards. These tools include efficient surface modeling algorithms, rockfall cluster detection to produce magnitude-frequency relationships, the Rockfall Activity Index - a morphological-based assessment technique, and seismically-induced rockfall debris estimates. Such approaches yield many important safety benefits, enable mitigation of geohazards before catastrophic events occur, and improved responses after major seismic events.