Marybeth Arcodia, Rosenstiel: Harnessing Data Science

Marybeth Arcodia, Rosenstiel: Harnessing Data Science

by Marybeth Arcodia

Applied Sciences & Engineering Science — General & History GPResearch and information: generalGPHData science and analysis
🎙 Marybeth Arcodia 👥 336 📅 October 9, 2025 ⏱ 64 min 👁 94 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

forecast of opportunitysea surface salinityneural networkexplainable AIsubseasonal

Summary

Dr. Marybeth Arcodia presents her research on harnessing data science to improve climate prediction, focusing on forecasts of opportunity. She explains that the climate system is chaotic, but certain states offer windows of predictability. Using neural networks trained on large climate model ensembles, she identifies sea surface salinity anomalies in the Caribbean and Gulf of Mexico as predictors for Midwest precipitation at subseasonal timescales. Explainable AI techniques reveal the physical basis for these predictions, building trust. She also discusses a real-time forecasting tool for the West Coast, using tropical precipitation patterns and climate modes like the MJO and ENSO. Additionally, she applies machine learning to predict coral heat stress in the Florida Keys, using random forests to forecast the onset of bleaching conditions. The talk concludes with future directions, including AI-based weather models and interdisciplinary collaboration at the Rosenstiel School and IDSC.

142 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides valuable insights into the application of machine learning to climate prediction, highlighting the concept of forecasts of opportunity. The argumentation is solid, supported by peer-reviewed research and ongoing projects. The speaker clearly explains the methodology, including the use of neural networks, explainable AI, and validation against observations. She also addresses limitations, such as the use of climate model data and the need for real-world validation. The presentation is well-structured, with clear examples and visual aids.

87 words

Title / Content Match

The title accurately reflects the content, focusing on the application of data science to climate prediction.

Quality & Reliability

8/10

The talk presents peer-reviewed research and ongoing projects, with clear methodology and validation against observations. The speaker is an expert in the field, and the content is consistent with current scientific understanding. Some limitations are acknowledged, but the overall reliability is high.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • Arcodia et al. (2025) - Sea surface salinity as a subseasonal predictor — Mentioned as a paper published earlier this year
  • NOAA Coral Reef Watch — Mentioned as the current prediction system for coral heat stress
  • WHAM (Water Cycle and Heat Flux Model) — Mentioned as a moisture tracking algorithm used for validation

Concurring Sources

Contribution & Novelties

The talk presents novel applications of machine learning to climate prediction, particularly the use of sea surface salinity as a predictor and the development of real-time forecasting tools. The emphasis on explainable AI to build trust and uncover physical mechanisms is a valuable contribution.

Pour aller plus loin :

89 words

Radar Profile

The radar profile shows high scores in information quantity, quality, and reliability, with a slightly lower technical level. This indicates a well-balanced presentation that is both informative and credible, though it may require some background knowledge to fully grasp the technical details.

Reliability 8/10

💬 No comments were provided for analysis.