2nd NOAA AI Workshop a Resounding Success
9 April 2021 - NOAA hosted the 2nd Workshop on Leveraging AI in Environmental Sciences virtually from 30 July 2020 through 25 February 2021. The purpose of the workshop was to exchange ideas and experiences, discuss the future potential and limitations of AI as related to environmental sciences, and review AI-enabling technology, tools, and applications. The workshop featured technical talks, poster sessions, and panel discussions given by scientists from private, academic, and government institutions from the US and abroad.
Keynote addresses by leadership at NOAA discussed NOAA’s priorities and strategy for using AI and machine learning. Each technical session included a panel discussion where attendees could ask questions, provide feedback, and share additional useful AI techniques.
Six tutorials were presented by Kitware, Climate AI and LBL, NVIDIA, University of Oklahoma, and Microsoft on topics that included video and image analysis for marine environments, using machine learning algorithms to forecast El Nino, an introduction to deep learning techniques using PyTorch, machine learning applied to NWP model data, an introduction to the capabilities of the Azure platform, and a background of AI theory related to a fish detector model used for fisheries management.
Due to the virtual format, attendance was widespread across the US and in nearly 20 other countries. Lasting seven months, the workshop was dubbed a “Workshop in Slow Motion” and it allowed attendees to absorb the wealth of material presented. Allen Huang of the University of Wisconsin summed up what many people thought of the workshop: “Both the on-site and virtual workshops were very successful and effective. Every week I was always looking forward to tuning in to the talks to get fresh ideas and get educated.”
AMS Special Collection for the Workshop
The collection includes overview articles, as well as those covering a wide range of topics including image processing such as gap filling; estimation of radar reflectivity, soil moisture, and TC intensity; detection of tropical and extratropical cyclones, and intense mid-latitude convection; prediction of streamflow, TC rapid intensification, and tornados; development of neural networks to replace convective parameterization schemes; post-forecast correction of long-range forecasts of precipitation and surface temperature; postprocessing in the form of nonlinear averaging of precipitation and ocean wave ensemble forecasts; and explainable AI.
Many collaborations with the private sector, academia, and Government were made possible thanks to the workshop. Plans are underway at the new NOAA Center for Artificial Intelligence (NCAI) for NOAA’s 3rd Workshop on Leveraging AI next year.
Data, algorithms, and images presented on STAR websites are intended for experimental use only and are not supported on an operational basis. More information