Meeting summaries and links in reverse chronological order.
Recording: On Microsoft Stream for Dept of Interior employees.
Abstract: The North American Bat Monitoring Program (NABat) is a multi-agency coordinated monitoring program that uses standardized protocols to gather data on the 47 bat species in the U.S. and Canada. Bats are notoriously difficult to monitor, however some monitoring methods, such as passive acoustic sensors, have seen success in recent years. As of 2021, NABat has collected over 8.7 million acoustic samples of bats from a wide range of habitats. Using this data, we have begun developing an open source reference library and deep learning environment to better serve the conservation needs of the bat community.
Bio: Ben Gotthold is a Computer Scientist with the North American Bat Monitoring Program in Fort Collins, Colorado. Ben has worked to design and build the cloud based technology stack that serves the diverse needs of this large scale monitoring effort.
Title: Turning nature into numbers: Advancing Ecoscience through the lens of remote sensing and artificial intelligence/machine learning
Recording: On Microsoft Stream for Dept of Interior employees
Abstract: The EcoAI project supports USGS science initiatives within dryland and high-latitude ecosystems through the integration of field observations, remote sensing, and artificial intelligence/machine learning (AI/ML). Dr. Neal Pastick will discuss how EcoAI makes use of cloud and high-performance computing resources (e.g., USGS HPCs, Amazon Web Services, Google Cloud Platform) and AI/ML to solve complex environmental problems. Specifically, this talk will focus on the USGS Rangeland Exotic Plant Monitoring System that was developed to fingerprint the abundance, spread, and impacts of invasive plants in the western United States. This system currently leverages Harmonized Landsat and Sentinel-2 data and machine and deep learning techniques to assess the past, present, and future abundance of exotic annual grasses and the resistance of sagebrush ecosystems to invasion. If time allows, Dr. Pastick will also present research aimed at better understanding high-latitude ecosystems and the development of a global lake monitoring system for assessing changes in surface area, color, and temperature using Alaska as a pilot study area.
Bios: Neal Pastick is a Research Physical Scientist with the U.S. Geological Survey (USGS) – Earth Resources Observation and Science (EROS) Center and a Ph.D. graduate from the Natural Resources Science and Management Program at the University of Minnesota – Twin Cities. He has conducted remote sensing research for environmental science applications over the past 11 years, with an emphasis on characterizing changing aquatic and terrestrial landscapes through the assimilation of field studies, remote sensing, and AI/ML.
Recording: On Microsoft Stream for Dept of Interior employees
Title: Integrating Deep Learning Methods into a National High-Resolution Land Cover Mapping
Nate Herold is a Physical Scientist with NOAA's Office for Coastal Management (OCM) where he manages the Coastal Change Analysis Program (C-CAP), NOAA's national land cover mapping and monitoring effort. He has been with OCM for 18 years, is intimately involved with Data and Tool offerings within our Digital Coast site, and is currently the chair of the Multi-Resolution Land Characteristics (MRLC) Consortium, helping to coordinate the land cover mapping work from all major federal agencies.
Chris Robinson is a Senior Remote Sensing Analyst with LynkerTechnologies on contract at the NOAA Office for Coastal Management (OCM) where he serves as the lead technical analyst for the Coastal Change Analysis Program (C-CAP). Chris has over 20 years of experience working with high-resolution imagery for the development of coastal land cover and habitat data.
Title: Assessing the Artificial Intelligence (AI) Readiness of USGS Data
This presentation provides an overview of a pilot program at the USGS for understanding the “state of the data” or data maturity as a result of institutional, system, and data level policies and practices. The pilot program includes a component for evaluating specific data characteristics for supporting AI applications. The presentation will share the initial AI readiness evaluation approach and the preliminary results.
Sophie Hou, Contractor to USGS, CSS Science Data Management Branch
Title: Semantics & machine reasoning: the (other) AI road to EarthMAP?
Abstract: Despite widespread growth in open data and machine learning, substantial challenges remain to the reusability and interoperability of scientific data and models. Since 2007, the Artificial Intelligence for Ecosystem Services (ARIES) project has been developing infrastructure for integrated, multidisciplinary scientific modeling using two AI tools – semantics and machine reasoning. These automate the assembly of multidisciplinary scientific data and models appropriate to the user’s context (i.e., location and spatiotemporal scale) of interest. Semantics apply consistent terminology to data and model components, enabling a computer system to recognize compatible data/model elements. Interdisciplinary semantics are particularly challenging to develop and apply, but ARIES has demonstrated that robust, modular, interdisciplinary semantics are possible. Machine reasoning enables a computer system to make choices when presented with alternative options – i.e., to use a particular model or dataset in a given application. A semantic web system like ARIES provides an environment for scientists to add new data and models to a global ecosystem for coupling, testing, adjusting, and reusing models – in particular, specifying appropriate conditions for model reuse. At the same time, a simple web interface provides access to data and models for a location and time period of interest, enabling non-technical users (like DOI resource managers) to run models, explore results and management tradeoffs, and view full model provenance. ARIES has been used to address diverse scientific and natural resource management questions globally. Although substantial work remains to achieve large-scale application, ARIES’ underlying technology may provide inspiration to what an integrated, AI-enabled system like EarthMAP could achieve.
Bio: Ken Bagstad is a Research Economist in the Geosciences & Environmental Change Science Center in Denver. His research interests span the modeling and valuation of ecosystem services, bridging the worlds of economic and natural capital accounting, and ecoinformatics. Since 2007 he has been actively involved in the Artificial Intelligence for Ecosystem Services (ARIES) project, an international collaboration to build a semantic web application supporting networked, automated multidisciplinary modeling for decision making.
Title: Injecting process knowledge into neural networks for more accurate predictions
Abstract: We have applied Process-Guided Deep Learning (PGDL) to water temperature prediction in several recent studies, supporting fisheries assessment in hundreds of lakes in the Upper Midwest and informing timed releases of cold water from reservoirs into streams of the Delaware River Basin. Our PGDL models, which integrate physics knowledge into neural networks, outperform baseline deep-learning and process-based models with respect to prediction accuracy and reliable detection of threshold exceedances. A rapidly growing community is applying similar methods to modeling tasks in other fields, from climate to translational biology, and the approach holds promise for numerous USGS-relevant applications. In this talk I will dive into the details of the neural network structures, physical constraints, and training methods that are responsible for the success of PGDL models to date.
Bio: Alison Appling has been a water data scientist with the US Geological Survey since 2015. She has a bachelor’s degree in Symbolic Systems from Stanford University and a PhD in Ecology from Duke University. Her research addresses the movement of energy, carbon, and nutrients through rivers, lakes, and floodplains, with an emphasis on using data science and machine learning to improve the estimation and prediction of water quality variables.
Recording: 200908-AIML-recording.mp4 (People with access to the Microsoft Team can also stream the recording from the Recordings tab, or go to the GS-CDI Channel)
Abstract: The Cloud Hosting Solutions (CHS) program is now offering and actively supporting the utilization of various artificial intelligence and machine learning (AI-ML) services. Matt Kuckuk will describe the kinds of support that are or will be provided to CHS customers. Matt will describe his recommendations for how investigators can identify use cases that are most likely to benefit from application of AI-ML techniques, and how they can begin to determine what standard algorithms to evaluate. He’ll discuss, for example, how “scientific” use cases and “operational” use cases differ in terms of their requirements. Finally, he will describe how to engage with the CHS AI-ML team to get support for new proposed use cases and applications.
Matt Kuckuk recently joined the CHS team after decades leading AI, ML and data analytics practice teams of up to 200 data scientists and developers in large consulting companies. He has implemented a wide variety of AI-ML applications and research projects for public sector as well as commercial organizations. He is now focused on creating and sustaining the AI-ML capability within CHS to advance the USGS mission.
Gage-Cam is a low cost, custom built wireless web camera paired with a custom deep learning algorithm that allows for a computer vision method to measure water surface elevation (stage). This project is a joint venture between Web Informatics and Mapping and the New York Water Science Center. Today's topic will be a short presentation on Gage-cam's design, capabilities, and prototyping. This will be followed by an open forum discussion on the technology and engineering behind the sensor, emerging methods in AI and single board "Lite-Tech" based devices researched by NYWSC-AI.
Daniel Beckman holds a bachelors degree from the University of Colorado in Ecology and Evolutionary Biology and minors in Chemistry and Computer Science. He currently attends Graduate school at the University of Colorado School of Engineering and Applied Science where he studies Machine Learning and Artificial Intelligence. He has worked in data for almost two decades in a variety of fields including, counterintelligence, research & development, forensic chemistry, and genomics. Daniel joined the USGS in 2017 and WIM in 2018. Currently, he works in cloud integration.
Recording: 200714-AIML-recording.mp4 (People with access to the Microsoft Team can also stream the recording from the Recordings tab, or go to the GS-CDI Channel)
Natalya Rapstine will give an overview of new USGS Tallgrass supercomputer designed to support machine learning and deep learning workflows at scale and deep learning software and tools for data science workflows.
Natalya Rapstine is a computational scientist in the Science Analytics and Synthesis (SAS), Advanced Research Computing (ARC) group of the Core Science Systems. She has a bachelor degree in Earth Science from Rice University and a MS in Statistics from Colorado School of Mines, and she has been with the USGS since 2016. Her expertise is in high performance computing, machine and deep learning applications for advancement of science at U.S. Geological Survey.
Inseok Heo is a data scientist in Envision Engineering AWS.
Inseok received a PhD from the department of Electrical Engineering in the University of Wisconsin Madison in 2015. He specializes in speech and audio signal processing and machine learning. In his career, he developed and worked on single/multi channel noise reduction, beamforming, and Alexa wakeword recognition/detection for Amazon Echo device.
Amogh Gaikwad is a Solutions Architect, specializing in AI/ML, for AWS Federal customers and is part of the specialist team for Analytics. Prior to his role at AWS, Amogh has worked as a software developer, developing enterprise applications. Through his role at AWS his has created ML solutions to help federal customers migrate their AI/ML workloads to AWS.
Amogh has received his Master’s Degree in Computer Science specializing in Big Data Analytics and Machine Learning from George Mason University
Phillip Dawson is a geophysicist with the U.S. Geological Survey’s Volcano Science Center, focusing on theoretical and experimental investigations of active volcanism and volcanic processes. He currently works on the Seismology of Magmatic Injection project at the California Volcano Observatory, Menlo Park, California. This project is dedicated to understanding the underlying physics driving volcanic seismicity and processes through the use of detailed field experiments and the application, modification, and extension of existing seismic methods and theories.
Michael Furlong, NASA-Ames Intelligent Robotics Group
Jack Eggleston, USGS WMA Hydrologic Remote Sensing Branch
John Stock, USGS Innovation Center
Abstract: The availability of high-resolution satellite imagery, combined with machine learning analysis to rapidly process the satellite imagery, provides the USGS with a new capability to map natural resources at the national scale. The new capability is made possible by technology progress in these areas:
1 - Daily national imagery at <1 to 5 m pixel size from commercial providers
2 - High-performance computing (USGS high-performance computing or Cloud)
3 - Artificial intelligence and machine learning (AI-ML) tools to automatically process the imagery
USGS is working to build enterprise capability in each of these 3 areas and has a growing focus on development of AI-ML tools. In this presentation, two USGS projects that rely on collaborations with external partners to develop AI/ML tools to map water extent will be discussed. In one of these projects USGS is collaborating with the NASA-Ames Intelligent Robotics Group to use its Deep Earth Learning Training, and Analysis (DELTA) software. The DELTA software will be presented including description of its early implementation on the USGS TallGrass supercomputing system.
Abstract: This presentation provides an overview of how we use a recurrent autoencoder neural network to encode sequential Californiagolden California golden eagle telemetry data. The encoding is followed by an unsupervisedclustering unsupervised clustering technique, Deep Embedded Clustering (DEC), to iteratively clusterthe cluster the data into a chosen number of behavior classes. We apply the method tosimulated to simulated movement data sets and telemetry data for a Golden Eagle. The DECachieves DEC achieves better unsupervised clustering accuracy scores for the simulated datasets as compared to the baseline K-means clustering result.