Arcodia Research Group
Dr. Marybeth Arcodia earned a Bachelor of Arts in Mathematics from Georgetown University in 2014 before completing a Ph.D. in Atmospheric Science at the University of Miami Rosenstiel School of Marine, Atmospheric and Earth Science in 2021. She is now an Assistant Professor joint between the Rosenstiel School of Marine, Atmospheric, and Earth Sciences Department of Atmospheric Sciences and the Frost Institute for Data Science and Computing at the University of Miami.
Her research bridges Earth system predictability and prediction, integrating atmospheric science and AI-based techniques to explore variability and change across weather-to-climate scales. Her work focuses on localized impacts in future climates to further our understanding of the stressed climate system and aid in advancing preparedness for climate risk. She is a member of the US CLIVAR Predictability, Predictions, and Applications Interface Panel and the Working Group on Climate Data and Predictions for Coastal Solutions.
Dr. Marybeth Arcodia
Assistant Professor
marcodia@earth.miami.edu
Forecasting rainfall three to four weeks out is one of the hardest problems in weather and climate — too far ahead for day-to-day weather, too soon for seasonal signals to take over. Chi-Jui uses AI/machine learning to predict it as a full probability distribution, and to spot "forecasts of opportunity": the moments when the climate system is especially predictable. The aim is practical guidance — telling forecasters not just what to expect, but when and where to trust it.
Chi-Jui Chen
Postdoctoral Research Fellow
cxc2918@earth.miami.edu
Juliette's project examines whether ocean surface conditions in the Gulf of Mexico and Caribbean Sea can improve subseasonal (weeks 1–4) summer precipitation predictions for the United States Midwest. Using explainable neural networks trained on varied ocean surface layers (0–1 m to 0–10 m). She is evaluating forecast skill using sea surface temperature and salinity anomalies as predictors, while exploring the physical mechanisms linking ocean-atmosphere-land interactions to Midwest precipitation variability.
Juliette Rocha
2nd Year Ph.D. Student
julietterocha@miami.edu
Nimay's initial research focused on using a Convolutional Neural Network (CNN) to predict subseasonal (Week 3-4) precipitation in Miami. More importantly, this work pairs traditional machine learning with explainable AI (XAI) to not only improve prediction at this timescale, but to also pinpoint the specific geographic regions that serve as sources of predictability for these rainfall forecasts. Building on preliminary results, his is extending this research to test why certain sources of predictability are present, checking if the regions flagged by XAI correspond to actual physical climate mechanisms. To do this, he is developing a hybrid CNN-Vision-Transformer architecture to identify predictability sources the original CNN may have missed and to compare their respective attribution patterns.
Nimay Mahajan
1st Year PhD Student
nrm5729@miami.edu
