Catch the Replay: Meet a Data Scientist with Ben Kirtman

Dr. Ben Kirtman, Deputy Director, University of Miami Institute for Data Science & Computing

Catch the Replay: Meet a Data Scientist with Ben…


The second Meet a Data Scientist talk, given on November 18, 2020, featured Dr. Ben Kirtman, a Professor of Atmospheric Sciences at the University of Miami Rosenstiel School of Marine and Atmospheric Science. Dr. Kirtman uses atmosphere-ocean general circulation models to study the predictability and variability of the Earth’s climate system. Dr. Kirtman also teaches graduate courses on the general circulation of the atmosphere and El Niño/Southern Oscillation, and climate prediction and predictability. Originally from Santa Barbara, California, Dr. Kirtman shared a funny story about his early start with computers and his (harmless) hacking efforts that got also his [also brilliant] Dad in a little ‘hot water’.  Relating subsequently to having to monitor flooding in his basement (related to El Niño) as a punishment, he’s come a long way! He’s understandably proud of the fact that the University of Miami has been the lead in the multi-agency, multi-institutional effort to improve NOAA’s seasonal and sub-seasonal operative forecasts since 2015.  His NMME (North-American Multi-Model Ensemble, on the chart below in light pink NCAR_CCSM4) forecasting system consists of coupled models from US modeling centers including NOAA/NCEP, NOAA/GFDL, IRI, NCAR, NASA, and from Canada’s modeling center. It has proven extremely effective at quantifying prediction uncertainty due to uncertainty in model formulation, and has proven to produce better prediction quality (on average) than any single model ensemble.

Download “Predicting Environmental Hazards: Where Data Science and Computing Meet” Powerpoint .pdf

Ben Kirtman slide NMME

The talk went from touching on the disparate data sources that go into forecasting El Niño and La Niña, to using machine learning to capture the Gulfstream’s interactions with the atmosphere (parameterization/parametrization), and ended with questions from the audience on the future of machine learning in climate prediction.

Ben Kirtman slides

Dr, Kirtman concluded by emphasizing that “What question are you trying to solve?” should drive what data/methods you use. He conveyed that new climate model(s) will be a blend with machine learning, which may lead to breakthroughs on how models work.

Ben Kirtman slides

Dr. Kirtman also teaches dynamic meteorology and atmospheric thermodynamics to undergraduates, and he mentors graduate students in the Meteorology and Physical Oceanography graduate program, as well as post-doctoral researchers.

His research is a wide-ranging program designed to understand and quantify the limits of climate predictability from days to decades. His research also involves understanding how the climate will change in response to changes in anthropogenic (e.g., greenhouse gases) and natural (e.g., volcanoes) forcing. This research involves hypothesis testing numerical experiments, using sophisticated state-of-the-art climate models and experimental real-time prediction. His group uses and has access to a suite of climate models, climate data, and high-performance computational platforms.