This page lists past seminars and presentations by STAR
scientists and visiting scientists. These seminars include the STAR
Science Forum and similar events. Presentation materials for
seminars will be provided when available.
NOCCG Seminar crosslisted with OneNOAA and STAR Seminars
Presenter(s): Dr. Emily Smail, CISESS-University of Maryland and SOCD
Sponsor(s): NOAA Ocean Color Coordinating Group (NOCCG). This seminar will not be recorded. Slides may be shared upon request (send email to the POC listed below).
Abstract: The Group on Earth Observations(GEO) is an international partnership working to improve the availability, access and use of Earth observations (EO) for the benefit of society. GEO's 109 Member Countries and 132 Participating Organizations work to actively improve and coordinate global Earth Observation systems and promote broad, open data sharing. GEO's priority engagement areas include the United Nations 2030 Agenda for Sustainable Development. GEO's initiative for oceans and coasts, GEO Blue Planet, is working to support Sustainable Development Goal (SDG) 14 "Conserve and sustainably use the oceans, seas and marine resources for sustainable development". This presentation will provide an overview of GEO Blue Planet's work in support of SDG 14 with a focus on activities to support the United Nations Environment Programme to develop the methodology for reporting on an SDG indicator for coastal eutrophication and applications for visualizing this data that are being developed in collaboration with Esri.
Bio(s): Emily Smail is the Executive Director of the GEO BluePlanet Initiative and a Senior Faculty Specialist at the NOAA-University of Maryland Cooperative Institute for Climate and Satellites.She also serves as the co-chair of the GEO AquaWatch Initiative's outreach and user Engagement working group and supports outreach and education efforts for the NOAA CoastWatch/OceanWatch program. Previously, Dr. Smail worked in informal science Education at the Waikiki Aquarium and policy in the United States Senate through the Knauss Marine Policy Fellowship Program. She received a B.S. in Biology from the Pennsylvania State University and a Ph.D. in Biology from the University of Southern California where her research focused on water quality and marine biogeochemistry.
Abstract: The NOAA global observing system (GOS) contains a large variety of observing platforms, including geostationary and polar-orbiting satellites, radiosonde,aircraft, surface stations, ships, buoys, etc. Despite the comprehensiveness of the observing system, many critical gaps exist in spatial/temporal coverage,spectral coverage, and resolution. To address these gaps, the NOAA/NESDIS Technology Maturation Program funded one of projects to explore use of emerging internet platforms (such as Loon high altitude balloons and SpaceX StarLink Satellites) for hosting remote sensing instruments. This talk summarizes feasibility assessment on potentials payload hosting opportunities that can benefit NOAA GOS system, which mainly focuses on Loon platforms and also extends to recent SpaceX StarLink constellations. First, the Loon platform characteristics and flight dynamics are comprehensively surveyed to explore the capability and limitation for Loon as a hosting platform. Second, by comparing GOES-16 Advanced Baseline Imager (ABI) with collocated Loon infrared thermometer measurements,we demonstrate that the Loon platform can served as a validation platform for future NOAA satellite sensors. Third, through simulation studies, observational geometry (e.g., footprint size, swath width, pointing accuracy) and weighting functions are studied for the scenarios that the Loon platform can host passive microwave instruments. More importantly, we demonstrate that balloon-based GPS radio occultation (RO) measurements can be complementary to current satellite based GPSRO systems. Efforts have been devoted to develop the capability of simulating the GPSRO slant path and bending angle from the balloon platform at~20 km, utilizing current constellation of Global Navigation Satellite Systems.Based on the calculations, the sampling characteristics and spatial and temporal coverage as well as the advantages and disadvantages are discussed.Based on this, the Observing System Simulation Experiment (OSSE) is designed to assess possible impacts on Global Forecasting System (GFS) forecasting capabilities by adding balloon-based GPSRO observations. The impacts are demonstrated and compared to those from space-based GPSRO observations. Finally,SpaceX StarLink constellation are simulated and potential hosting opportunities are discussed.
Dr. Likun Wang is now working in NOAA/NESDIS/STAR as a contract scientist employed by Riverside Technology, inc for Research Technology Maturation for the Exploitation of Emerging Technologies (RTMEE) Contract, including near space payload hosting platform assessment, Artificial Intelligence (AI) technology demonstration, and geostationary sounder proxy data simulations. With more than 15 years of progressive working experiences of NOAA's satellite sensors, Dr. Likun Wang has been responsible for the pre- and post-launch calibration testing data analysis, inter-calibration for post-launch instrument monitoring and assessment, ground processing software development, configuration and calibration parameter refining, and new algorithm design and integration. He currently serves the chair of World Meteorological Organization (WMO) sponsored Global Space-based Inter-Calibration System (GSICS) infrared sensor working group. Likun Wang received his B.S. degree in atmospheric sciences and the M.S. degree in meteorology from Peking University, Beijing, China, in 1996 and 1999, respectively, and the Ph.D. degree in atmospheric sciences from University of Alaska Fairbanks, in 2004.
Abstract: CubeSats are revolutionizing the way we make Earth Observations. These low-cost mini satellites are described as cube-shaped spacecrafts with units of 1U =10X10x10cm. SeaHawk, with a total weight of 5kg, is the first 3U CubeSat specifically designed to hold an ocean color instrument payload (HawkEye). SeaHawk was built as part of SOCON (Sustained Ocean Color Observations Using Nanosatellites, http://www.uncw.edu/socon), an ongoing “proof of concept” project at the University of North Carolina at Wilmington (UNCW). This seminar will provide an overview of SeaHawk's characteristics and mission status. HawkEye's specifications are similar to SeaWiFS (one of the most successful ocean color mission to date) in that it is an 8-band multi-spectral ocean color sensor,except band 7 was modified to improve atmospheric correction and the SNR is>50% that of SeaWiFS. However, HawkEye was designed to fit a 1U cube, it has ~130m spatial resolution, it does not saturate over land, and was built with low-cost, off-the-shelf materials. SeaHawk follows a sun-synchronous Low Earth Orbit at a nominal height of 575km, orbiting 15 times a day, with a swath of 216 x 480km and a repeat time of about 18 days. SeaHawk-1 was launched in December 2018 as part of SpaceX first ride-share mission. Once fully commissioned, data will be available at no cost through NASA's OBPG (https://oceancolor.gsfc.nasa.gov). The ocean color community will also be able to submit requests for image acquisition (e.g. for field support) through UNCW. This project was possible thanks to the Gordon and Betty Moore Foundation and a Space Act Agreement between NASA and UNCW.
Bio(s): Dr. Sara Rivero-Calle is a Post-Doctoral researcher at the University of North Carolina Wilmington. Her interest in bio-optics and satellite remote sensing started at the University of Puerto Rico, where she earned her M.Sc. degree in Biological Oceanography working on the ecology of sponges in mesophotic coral reefs using Autonomous Underwater Vehicles. She then earned a PhD degree in Oceanography from Johns Hopkins University, where she studied long term changes in the North Atlantic phytoplankton communities using the Continuous Plankton Recorder survey, the largest and longest ongoing phytoplankton sampling effort. Sara is interested in projects that involve large datasets, combining remote sensing, modeling and in situ data to answer large-scale ecological questions. She is currently working on the topic of finescale variability and subpixel variability, combining satellite products, ARGOfloats, HPLC pigments and numerical models. At UNCW, Sara is Science Lead and an active member of the Management team of SOCON: Sustained Ocean Color Observations Using Nanosatellites. SOCON recently built and launched SeaHawk-1,the first ocean color CubeSat mission.
Abstract: The field of Artificial Intelligence (AI), including applications in the environmental sciences, is evolving at an accelerating pace. Its progress has been made possible by developments in the computer sciences, the availability of larger and more comprehensive environmental data sets, and the ever-increasing availability of affordable computing power. The presentation will start with the early days of the field, including how the term was coined by John McCarthy. We will then cover the progression of the field, including its ups and downs, through a series of examples.
The American Meteorological Society AI workshops and conferences allow to track this progression. Expert systems were the method of choice in the eighties while Neural Networks took over in the nineties followed by a broadening of the methods including fuzzy logic, tree-based methods, genetic algorithms, support vector machines... At the 2019 and 2020 AMS AI conferences deep learning became by far the method of choice with 36% and over 50% of the presentations based on this new method. We will trace back this explosive growth to its roots including Imagenet, AlexNet and the importance of the datasets in a sense driving the development of these methods.
While the AI methods have changed considerably over the years, the topics not so much. The first AMS AI conference in 1998 included talks on precipitation predictions, satellite retrieval and pattern recognition, climate classification and prediction, image processing, decision aids and natural language systems. We will introduce selected environmental applications and methods developed at the Conrad Blucher Institute (CBI) to provide local operational predictions including for water levels, coastal flooding, coastal fog, and a model designed and implemented to predict the cold stunning of sea turtles. These methods take advantage of the flexibility of AI to combine real-time environmental measurements and numerical weather predictions, typically from NOAA, as the predictors to different types of AI models.
We are expecting the fast growth of AI/ML to continue and as the method is becoming one of the main approaches to better predict and gather a deeper understanding of a wide variety of complex and nonlinear processes in the earth sciences. The presentation will conclude with the introduction of some of the present AI related research questions such as the quantification of uncertainties, interpretability, incorporating domain-knowledge in model design and the further potential for AI applications in the environmental sciences.
Bio(s): Philippe Tissot is the Interim Director of the Conrad Blucher Institute and an Associate Research Professor at Texas A&M University-Corpus Christi. For the past 20 years, his research has focused on the development of artificial intelligence methods and other models for the analysis and predictions of environmental systems and coastal physical processes. Projects have included the development and implementation of predictive models supporting navigation and coastal management. Other studies have included the modeling and impact of relative sea level rise and storm surge, the spatial variability of subsidence at the regional scale, tidal studies and local hydrodynamic models. His team's models have been used for over a decade for the prediction of cold stunning of sea turtles allowing to interrupt navigation ahead of these events and other preparation by local stakeholders. Other work has included ML predictions of thunderstorms and the development of ML algorithms to take advantage of 3D point clouds of marsh environments and urban runoff water quality modeling. Dr. Tissot has authored or co-authored over 40 peer reviewed articles, 200 proceedings, abstracts and technical presentations, a Physical Science textbook for future K-12 teachers, and 2 US Patents. Professor Tissot is a member and former chair of the American Meteorological Society Committee on Artificial Intelligence Applications to Environmental Science.
Abstract: This introductory talk provides basic information about mostly used machine learning (ML) techniques and some ML applications developed to enhance different components of Numerical Weather Prediction (NWP) systems. Basic groups of ML applications that have been already developed for NWP systems are overviewed.Major challenges that NWP currently faces are discussed. It is shown that many of these problems can be resolved or alleviated using ML techniques. ML applications developed for NWP model initialization/data assimilation, model improvements, and model output post processing are discussed. Several examples of such application (ML satellite retrieval algorithm, ML fast parameterizations of subgrid processes, and ML nonlinear ensembles) are introduced to illustrate the capabilities of ML techniques. Advantages and limitations of ML techniques are discussed.
Bio(s): Dr. Vladimir M. Krasnopolsky got his M.S. in Theoretical and Computational Physics and Ph. D. in Theoretical Nuclear Physics from the Moscow State University (Russia). He worked in the field of theoretical nuclear physics at the Institute of Nuclear Physics (Moscow State University) before coming to the US in 1989. Since 1990 he has been working in the field of numerical weather and climate prediction and AI applications. Vladimir works on applications of remote sensing and satellite data in meteorology, oceanography, and numerical weather and climate prediction. Dr. Krasnopolsky also works with various machine learning techniques. He developed multiple neural network applications for numerical weather and climate prediction. Dr. Krasnopolsky published two books, two book chapters, over 70 papers in refereed scientific journals. He is a member (formerly Chair) of the Committee on “Computational and Artificial Intelligence Applications in Environmental Science” of American Meteorological Society, a member of the IEEE/Computational Intelligence Society Task Force “Computational intelligence in earth and environmental sciences”, and a member of the International Neural Network Society Working Group “Computational intelligence in earth and environmental sciences”. In 2018 Vladimir was awarded AMS Distinguished Scientific Committee award for “Contributions to advancing the application of artificial neural networks to earth science problems and in particular emulations of complex multidimensional mappings.”
Abstract: The success of a machine learning system depends on not only the choice of machine learning algorithm but also on how the the whole machine learning pipeline is constructed. In this presentation, the key components of the machine learning pipeline, including problem definition, preprocessing, choosing appropriate algorithms, training, evaluation, and interpretation will be described. Common approaches in the atmospheric sciences for each component will be explained and linked with examples from machine learning applications in the atmospheric sciences. Finally, challenges of transitioning machine learning systems to operational use will be discussed.
Bio(s): David John Gagne is a Machine Learning Scientist in the Computational Information Systems Laboratory (CISL) and the Research Applications Laboratory (RAL) at the National Center for Atmospheric Research (NCAR). His research focuses on developing machine learning systems to improve the prediction and understanding of high impact weather, and to enhance weather and climate models. During his time at NCAR, he has collaborated with interdisciplinary teams to produce machine learning systems to study hail, tornadoes, hurricanes, and renewable energy. He has also developed short courses and hackathons to provide atmospheric scientists hands-on experience with machine learning. Gagne received his Ph.D. in meteorology from the University of Oklahoma in 2016 and completed an Advanced Study Program postdoctoral fellowship at NCAR in 2018. In addition to his duties at NCAR, he also serves as chair of the American Meteorological Society Artificial Intelligence Committee.
Abstract: Data-driven algorithms, in particular neural networks, can emulate the effect of sub-grid scale processes in coarse-resolution climate models if trained on high-resolution climate simulations. However, they may violate key physical constraints, lack interpretability, and make large errors when evaluated outside of their training set. First, we show that nonlinear physical constraints can be enforced in neural networks, either approximately by adapting the loss function or to within machine precision by adapting the architecture. As these physical constraints are insufficient to guarantee generalizability, we additionally propose a framework to incorporate physical rescalings within the neural network: By aligning the distributions of both input and output variables across climates, we transform extrapolation into interpolation and significantly improve the ability of neural networks to generalize to unseen climates. Third, we present recent tools designed to interpret machine-learning parametrizations of convection, which we leverage to show that two sets of neural networks trained on different datasets behave consistently with observations. Our interpretability tools can further diagnose the stability of machine-learning parametrizations when coupled to atmospheric fluid dynamics, which helps the ultimate goal of improving the performance and stability of coupled online climate simulations.
Bio(s): Tom Beucler is a project scientist affiliated with UC Irvine and Columbia University. He is interested in atmospheric physics, machine learning and climate risk analysis.
Abstract: I describe the application of convolutional neural networks (CNN), a type of deep-learning method, to high-impact weather. CNNs are specially designed to learn directly from spatial grids, which improves both skill and interpretability. Specifically, I develop and test CNNs for two tasks. The first is tornado prediction, where two CNNs predict next-hour tornado occurrence for a given storm, using datasets similar to those used by forecasters in real-time operations. The tornado models achieve an area under the receiver-operating-characteristic curve (AUC) of 0.94 and critical success index (CSI) of ~0.3. This is competitive with a machine-learning model currently used in operations, which suggests that the CNNs would also be suitable for operations. Specialized machine-learning-interpretation methods highlight the importance of a deep reflectivity core and strong mesocyclone, as well as low-level instability and wind shear in the surrounding environment. Also, interpretation methods suggest that a rear-flank downdraft with too much precipitation and negative buoyancy can lead to tornadogenesis failure, which corroborates some previous literature. The second task is front detection, where a CNN draws warm and cold fronts in reanalysis data. I use the CNN-detected fronts to create a 40-year climatology over North America. On a large scale, fronts are most common in the mid-latitude cyclone track, which migrates poleward from winter to summer, equatorward during El Niño, and poleward during La Niña. Also, the cyclone track appears to be migrating poleward as a consequence of global warming. These results are broadly consistent with the few pre-existing climatologies, although there are some discrepancies that should be investigated in the future. Overall, I demonstrate that deep learning can be used to advance both the prediction and understanding of high-impact weather.
Bio(s): Dr. Ryan Lagerquist recently graduated with a Ph.D. in Meteorology from the University of Oklahoma. He has been researching machine-learning applications in atmospheric science for 8 years with organizations including Environment Canada, the University of Alberta, Google, NCAR, and CIMMS. Ryan begins a postdoc with the Cooperative Institute for Research in the Atmosphere (CIRA) in June. Ryan is also program co-chair of the 2021 Artificial Intelligence conference at the AMS annual meeting.
NOAA CoastWatch/OceanWatch provides easy access for everyone to global and regional satellite data products for use in understanding, managing and protecting ocean and coastal resources and for assessing impacts of environmental change in ecosystems, weather, and climate. A CoastWatch objective is to educate and train users about using satellite data and CoastWatch products. The CoastWatch Data Portal is a collection of services that facilitate the discovery and utilization of satellite ocean remote sensing products. In this seminar, Michael Soracco will provide an overview of these services and demonstrate the map viewer for the CoastWatch Data Portal. You will learn how to display, explore, and access sea surface temperature, salinity, color, wind, synthetic aperture radar and sea level anomaly data products available through NOAA CoastWatch. https://coastwatch.noaa.gov/cw/index.html
Bio(s): Michael Soracco is the HelpDesk Coordinator for NOAA CoastWatch/OceanWatch/PolarWatch. He specializes in user access by developing and maintaining data products, the data portal (and website) and runs the helpdesk. Michael studied aerospace engineering for his B.AE at Georgia Institute of Technology and earned his Master's degree in Business at Central Michigan University. He served as a commissioned officer in the NOAA Corps servicing equatorial moored buoys, conducting coastal hydrographic surveys, and as an ocean remote sensing operations officer. He has developed and taught the NOAA GIS course and is a fundamental source of the CoastWatch program's “corporate memory.“
STAR Science Seminars Note: This seminar will be presented online only.
Presenter(s): Tapio Schneider, Caltech and NASA/JPL
Sponsor(s): STAR Science Seminar Series Host: Imme Ebert-Uphoff, CIRA
While climate change is certain, precisely how climate will change is less clear. But breakthroughs in the accuracy of climate projections and in the quantification of their uncertainties are now within reach, thanks to advances in the computational and data sciences and in the availability of Earth observations from space and from the ground. I will survey the design of a new Earth system model (ESM), under development by the Climate Modeling Alliance (CliMA) of Caltech, MIT, Jet Propulsion Laboratory, and the Naval Postgraduate School. The talk will cover key new concepts in the ESM, including turbulence, convection, and cloud parameterizations and fast and efficient algorithms for assimilating data and quantifying uncertainties through a three-step process involving calibration, emulation, and sampling.
Bio(s): Tapio Schneider is the Theodore Y. Wu Professor of Environmental Science and Engineering at Caltech and a Senior Research Scientist at NASA's Jet Propulsion Laboratory. His research has elucidated how rainfall extremes change with climate, how changes in cloud cover can destabilize the climate system, and how winds and weather on planetary bodies such as Jupiter and Titan come about. He is currently leading the Climate Modeling Alliance (clima.caltech.edu), which is building a new Earth system model that automatically learns from diverse data sources.
Abstract: A nonlinear ensemble averaging technique is demonstrated using neural networks applied to one year (2017) of Global ocean Wave Ensemble forecast System (GWES) data provided by NCEP. Post-processing algorithms are developed based on multilayer perceptron neural networks (NN) trained with altimeter data to improve the global forecast skill, from nowcast to forecast ranges up to 10 days, including significant wave height (Hs) and wind speed (U10). NNs are applied as an alternative to the typical use of the arithmetic ensemble mean (EM). The novel method shows that one single NN model with 140 neurons is able to improve the error metrics for the whole globe while covering all forecast ranges analyzed. The bias of the widely used EM of GWES that varies from -10% to 10% for Hs compared to altimeters can be reduced to values within 5%. The RMSE of day-10 forecasts from the NN simulations indicated a gain of two days in predictability when compared to the EM, using a reasonably simple post-processing model with low computational cost.
Bio(s): Dr. Ricardo Campos has been studying and working in the fields of Ocean Engineering, Physical Oceanography, and Meteorology for the last 17 years, more specifically with ocean waves. He has recently concluded postdoctoral research at the University of Maryland in collaboration with EMC/NCEP/NOAA, working on neural networks to improve wave forecasts. Ricardo is currently a principal investigator and visiting professor at the Centre for Marine Technology and Ocean Engineering (CENTEC/IST) in Portugal.