ORNL News Releases <newsreleases@ornl.gov>

Feb 4 at 9:38 AM

Modeling—Mapping Arctic vegetation

A team of scientists led by Oak Ridge National Laboratory used machine learning methods to generate a high-resolution map of vegetation growing in the remote reaches of the Alaskan tundra. They used artificial intelligence to combine satellite imagery, synthetic aperture radar, topography and other data to produce a more accurate representation of changing Arctic plant communities and compared it with traditional ground-based measurements. “We used unsupervised classification techniques with the ORNL-developed Mapcurves scheme to label the types of vegetation,” said ORNL’s Forrest Hoffman. “Then, we trained a deep neural network to produce a new map of vegetation communities with 95 percent accuracy compared to field observations.” The new data will inform field sampling expeditions as well as ecosystem modeling and analysis as part of the Department of Energy’s Next-Generation Ecosystem Experiments-Arctic project. These machine learning methods, published in Remote Sensing, can be applied to other regions. [Contact: Kim Askey, (865) 576-2841; askeyka@ornl.gov]

Image: https://www.ornl.gov/sites/default/files/rs2019_highlight_plot_3d.png

Caption: Researchers used machine learning methods on the ORNL Compute and Data Environment for Science, or CADES, to map vegetation communities in the Kougarok Watershed on the Seward Peninsula of Alaska. The colors denote different types of vegetation, such as willow, sedge and dryas lichen. Credit: Jitendra Kumar/Oak Ridge National Laboratory, U.S. Dept. of Energy

Commuting—Oh, the places you’ll go

Oak Ridge National Laboratory geospatial scientists who study the movement of people are using advanced machine learning methods to better predict home-to-work commuting patterns. Commuting models, which are based on publicly available information such as population and labor force data, can influence plans to reduce traffic and pollution and drive infrastructure choices. “Traditional commuting models are not able to easily accommodate new and different kinds of input data such as income and family structure,” said ORNL’s April Morton, co-author of a study presented at GIScience. “Machine learning models help overcome this barrier because the algorithms can easily accept new data, quickly re-train themselves and update predictions about commuting patterns.” Their study focused on a small region in East Tennessee, but the method could be applied anywhere in the United States. [Contact: Sara Shoemaker, (865) 576-9219; shoemakerms@ornl.gov]

Image: https://www.ornl.gov/sites/default/files/study_area_one_dest_2.jpg


Caption: ORNL scientists used commuting behavior data from East Tennessee to demonstrate how machine learning models can easily accept new data, quickly re-train themselves and update predictions about commuting patterns. Credit: April Morton/Oak Ridge National Laboratory, U.S. Dept. of Energy


Neutrons—Fueling better power


A University of South Carolina research team is investigating the oxygen reduction performance of energy conversion materials called perovskites by using neutron diffraction at Oak Ridge National Laboratory’s Spallation Neutron Source. Perovskites are core components of solid oxide fuel cells, which can be utilized for distributed power generation in remote areas or for backup power at data centers. Neutrons’ sensitivity to light elements like oxygen allow them to accurately probe the perovskites’ structures and reveal how they influence the fuel cell’s performance. Using a furnace in the VULCAN beamline, the team mimicked a fuel cell’s typical environmental conditions.VULCAN’s unique high-temperature capability allowed us to see the perovskites’ structures in their operating conditions,” said USC’s Kevin Huang, the corresponding author. “Better understanding this structure-property relationship could allow us to enable better power generation performance by optimizing materials.” The team’s research was published in Applied Materials & Interfaces. — Josh Witt [Contact: Kelley Smith, (865) 576-5668; smithks@ornl.gov]


Image: https://www.ornl.gov/sites/default/files/Neutron-Fueling_better_power_image1.jpg


Caption: Researchers analyzed the oxygen structure (highlighted in red) found in a perovskite’s crystal structure at room temperature, 500°C and 900°C using neutron scattering at ORNL’s Spallation Neutron Source. Analyzing how these structures impact solid oxide fuel cells could lead to the development of new, improved materials. Credit: Kevin Huang/University of South Carolina


Materials—Quelling corrosion


Oak Ridge National Laboratory scientists analyzed more than 50 years of data showing puzzlingly inconsistent trends about corrosion of structural alloys in molten salts and found one factor mattered most—salt purity. Typically, alloys used in high-temperature environments are high in chromium, which forms barrier layers that slow corrosion. However, these barriers are unstable in nuclear reactors and concentrating solar power plants that use molten salts for heat transfer and storage. Absent the protective barriers, alloy chromium leeches into the salt, aided by oxidizing impurities. “We’ve got to throw out a lot of what we think about traditional corrosion science when we’re talking molten salts,” said ORNL materials scientist Stephen Raiman, who performed the analysis with ORNL data scientist Matt Sangkeun Lee. “Salt purification makes a big difference. Pure salts have much lower corrosion rate than impure salts.” Their published understanding will guide future experiments and models aimed at advancing robust, resilient materials. [Contact: Dawn Levy, (865) 576-6448; levyd@ornl.gov]


Image: https://www.ornl.gov/sites/default/files/story%20tip%20image%20BW%20only.jpg


Caption: At the salt–metal interface, thermodynamic forces drive chromium from the bulk of a nickel alloy, leaving a porous, weakened layer. Impurities in the salt drive further corrosion of the structural material. Credit: Stephen Raiman/Oak Ridge National Laboratory, U.S. Dept. of Energy