Date: 20th January 2026, Tuesday (11:00am – 12:00pm)
Presenter: Jianyu Liang (CCRS)
Topic: A Machine-Learning Observation Operator for Satellite Observations Without Radiative Transfer Models

Abstract:
In data assimilation, observation operators convert model variables into what instruments observe. The development of observation operators usually relies on writing computer code to represent physical relationships, which can be time-consuming and slow down the use of new observations. In this study, I will present how to use machine learning to build an observation operator for satellite observations without prior physical knowledge of the radiative transfer process. Using this ML-based observation operator to assimilate additional satellite microwave observations can reduce the atmospheric temperature error in the model.

Speaker Profile:
Dr Jianyu Liang is a Research Scientist at CCRS. At CCRS, his main research includes radar quality control, Quantitative Precipitation Estimation, and nowcasting. He obtained his MSc in Atmospheric Sciences at the Institute of Atmospheric Physics, China, in 2008 and his PhD from York University, Canada, in 2019. Before joining CCRS, he was a postdoctoral researcher at RIKEN R-CCS, Japan. His postdoctoral work mainly focuses on data assimilation, including assimilating satellite observations, developing machine-learning-based observation operators, investigating non-Gaussian data assimilation methods, and applying data assimilation techniques to study the Venus atmosphere.

Date: 27th January 2026, Tuesday (11:00am – 12:00pm)
Presenter: Yi Ming (Boston College)
Topic: Relative Contributions of Hadley vs. Walker Circulation to Tropical Circulation Weakening

Abstract:
The tropical time-mean circulation is widely expected to slow down in a warmer climate. An often cited explanation for this is the thermodynamic scaling argument proposed by Held and Soden (2006), which was later invoked by Vecchi and Soden (2007) to explain the projected weakening of the Walker circulation (the zonal component of the tropical circulation). Drawing upon three papers currently being developed within my group, I will discuss the physical meanings and utilities of three commonly used metrics of circulation strength. I will then present a more complete picture of how the Hadley (the meridional component) and Walker circulations respond to warming—an issue central to understanding the future large-scale conditions that govern extreme weather events in Singapore and Southeast Asia.

Speaker Profile:
Prof Yi Ming is the Institute Professor of Climate Science and Society at Boston College. He uses climate models, observations and theories to elucidate the physical mechanisms governing Earth’s climate system and applies the fundamental understanding to practical issues of societal and policy importance.

Date: 3rd February 2026 2026, Tuesday (11:00am – 12:00pm)
Presenter: Chanh Kieu (CCRS)
Topic: Quantifying the atmospheric predictability with machine learning: global versus regional models

Abstract:
The atmosphere is known to possess limited predictability, with time scales ranging from hours to months depending on the weather system. Traditional physics-based models suggest that large-scale dynamics have a predictability limit of approximately two weeks. However, determining the predictability range for different regional weather and climate extremes remains an open question, as it depends strongly on system-specific characteristics. In this study, I present a machine learning (ML)–based approach that can help quantify atmospheric predictability at both regional and global scales. Using a deep learning architecture based on a ConvLSTM model, we show that different time-stepping strategies can substantially influence model performance and predictability. For global prediction, the incremental iteration approach in which the future state is obtained by recursively applying the AI model over short time steps shows strong sensitivity to factors such as input variable selection, the number of input frames, and the forecast lead time. In contrast, the direct iteration approach for which the current state is mapped directly to the desired lead time provides markedly better forecast skill and extends the predictability horizon. At the regional scale, the predictability limit is however strongly controlled by the lateral boundary conditions. In fact, the update frequency of lateral boundary conditions during incremental autoregressive rolling can bypass the advantages gained from the direct iteration strategy. These findings not only provide guidance for implementing ML models for weather and climate forecasting but also offer a framework for exploring the practical limits of predictability in chaotic systems beyond what traditional approaches allow.

Speaker Profile:
Dr Chanh Kieu is a new scientist joining CCRS in January 2026 to work on AI/ML model development for regional weather and climate applications. He received his B.S. degree in 2002 from Vietnam National University and Ph.D. in Atmospheric Science in 2008 from the University of Maryland, College Park, US. Dr Kieu’s main research interests include atmospheric predictability, tropical cyclone/meteorology modeling and theory, AI/ML applications, nonlinear dynamical systems, and ensemble data assimilation. Previously, he was an Associate Professor at Indiana University Bloomington (2015–2025) and a member of the NOAA/NCEP Hurricane Modeling Development team (2011–2014).

Date: 26th February 2026, Tuesday (10:00am – 11:00am)
Presenter: Greg McFarquhar (University of Oklahoma)
Topic: The Cooperative Institute for Severe and High Impact Weather Research and Operations (CIWRO): Impetus, Accomplishments, Impacts, and Future Plans

Abstract:
The Cooperative Institute for Severe and High Impact Weather Research and Operations (CIWRO) at the University of Oklahoma (OU) is a global leader in meteorological research, bridging partnerships between OU and the National Oceanic and Atmospheric Administration (NOAA) to transform scientific understanding into products that provide innovative, life-saving forecasts that reduce the impacts of extreme weather events on communities and ecosystems. By transitioning research findings into operational products, CIWRO’s work improves weather forecasts and warnings, ultimately saving lives, protecting property, and reducing the economic impacts of storms. The research at CIWRO focuses on advancing fundamental knowledge of weather radar, multi-scale processes, and subseasonal to seasonal (S2S) predictions by integrating improved observations, modeling, data assimilation and the study of social science under the guise of five key research themes: 1) weather radar and observations research and development (R&D); 2) mesoscale and storm-scale modeling R&D; 3) forecast applications improvements R&D; 4) S2S prediction for extreme weather events; and 5) social and socioeconomic impacts of high-impact weather systems. In this presentation, examples of research performed under all five themes will be presented. Research to operations (R2O) and operations to research (O2R) play a key role in driving CIWRO’s mission. Examples of some of the 55 research products developed by CIWRO that have been transitioned to operations will also be provided. Education, outreach and training are also vital components of CIWRO, building capacity to support NOAA’s future research and operational activities. Examples of these activities that will be discussed include the following: funded graduate student, undergraduate student and postdoctoral research projects; classroom visits to K-12 schools, summer camps and after school programs; creation and loaning of travelling trunks to schools (pre-assembled comprehensive lesson plans with supplies); curriculum development for middle schools; teacher workshops; and consumption of a minimum of 81,943 hours of CIWRO-produced training by National Weather Service (NWS) forecasters. Future research and educational opportunities within CIWRO will also be discussed.

Speaker Profile:
Prof Greg McFarquhar is Director of the Cooperative Institute for Severe and High Impact Weather Research and Operations (CIWRO) and Professor at the University of Oklahoma. He completed his B.Sc./M.Sc./Ph.D. at the University of Toronto. He previously worked at the University of Illinois, National Center for Atmospheric Research, and Scripps Institution of Oceanography. He is a Fellow of the American Meteorological Society and American Geophysical Union and President of the International Commission on Clouds and Precipitation. His researches cloud properties and processes. He has conducted 36 airborne field campaigns measuring clouds. He is Editor for the American Meteorological Society Monograph Collection, and Atmospheric Chemistry and Physics. He has 263 refereed journal publications and has received over $10 Million of research funding.