About This Project

NOAA's Gulf of Alaska heatwave index tracks excess ocean heat as a warning that nutritious lipid-rich copepods (zooplankton), a cornerstone of the food web, may be scarce. Temperature tells only part of the story. We hypothesize that the copepods' gene activity gives a clearer and more detailed warning. We will build an open-source LLM workflow to sift through complex sequencing data from 40 dormant Neocalanus flemingeri females (2015–2021) for early markers of marine food shortages.

Ask the Scientists

What is the context of this research?

NOAA's Gulf of Alaska Ecosystem Status Report uses marine heatwave cumulative intensity (MHCI), the total excess heat over a season's heatwave days, as a headline warning: hot seasons favor smaller, less fatty copepods, leaving fish, seabirds and salmon short of food (Suryan et al. 2021). The index assumes every copepod responds to heat the same way.

We study one species, Neocalanus flemingeri, the dominant large copepod on the northern Gulf shelf. It feeds in spring, building the fat that carries it through six months of dormancy (diapause) at depth. The Lenz lab has RNA sequencing, which measures how active each gene is, for 40 dormant females from eight collections (2015–2021), each followed by spring monitoring counts the next year. Our years span two heatwaves, a recovery and a cool year. Their gene activity differs by year, and the heatwave years 2015, 2016 and 2019 stand out.

What is the significance of this project?

NOAA's Gulf of Alaska Ecosystem Status Report informs fishery harvest limits and tracks ocean warming using the MHCI. The index treats a hot season as a sign of poor recruitment of lipid-rich copepods, but that link is inferred and ignores physiological acclimatization. We will test gene expression profiles of dormant N. flemingeri females from two heatwaves and the recovery between. If gene activity improves the prediction, it can support sustainable fishery management here and in other ecosystems.

Turning more than 30 GB of raw sequence data into scientific insight is intractable without sophisticated tooling, and large language models offer valuable leverage. The workflow we're building integrates these powerful AI tools with research data while keeping a human-in-the-loop paradigm that ensures oversight and is flexible enough to be applied to many fields of research.

What are the goals of the project?

Hypothesis: Monitoring copepod gene activity in addition to the marine heatwave index (MHCI) can provide better indicators of marine food shortages than the index alone.

We will compare the predictive power of the heatwave index and the gene activity separately and together. Success will be measured by how well each predicts the number of N. flemingeri counted the following spring. Those copepods are the next generation, and the food that fish, seabirds and salmon depend on.

We will also build an open-source AI workflow that gives researchers a structured, portable way to bring in large datasets. An LLM writes queries and code that produce charts and tables for researchers to review. The workflow also includes tools to check results and produce figures and tables ready for publication. Over 2027, with no new fieldwork, we will build the workflow and use it to run the comparison.