Search for: All records

One of the areas where Large Language Models (LLMs) show promise is for automated qualitative coding, typically framed as a text classification task in natural language processing (NLP). Their demonstrated ability to leverage in-context learning to operate well even in data-scarce settings poses the question of whether collecting and annotating large-scale data for training qualitative coding models is still beneficial. In this paper, we empirically investigate the performance of LLMs designed for use in prompting-based in-context learning settings, and draw a comparison to models that have been trained using the traditional pretraining--finetuning paradigm with task-specific annotated data, specifically for tasks involving qualitative coding of classroom dialog. Compared to other domains where NLP studies are typically situated, classroom dialog is much more natural and therefore messier. Moreover, tasks in this domain are nuanced and theoretically grounded and require a deep understanding of the conversational context. We provide a comprehensive evaluation across five datasets, including tasks such as talkmove prediction and collaborative problem solving skill identification. Our findings show that task-specific finetuning strongly outperforms in-context learning, showing the continuing need for high-quality annotated training datasets. 

Subtropical seagrass meadows play a major role in the coastal carbon cycle, but the nature of air–water CO₂exchanges over these ecosystems is still poorly understood. The complex physical forcing of air–water exchange in coastal waters challenges our ability to quantify bulk exchanges of CO₂and water (evaporation), emphasizing the need for direct measurements. We describe the first direct measurements of evaporation and CO₂flux over a calcifying seagrass meadow near Bob Allen Keys, Florida. Over the 78‐d study, CO₂emissions were 36% greater during the day than at night, and the site was a net CO₂source to the atmosphere of 0.27 ± 0.17 μmol m⁻²s⁻¹(x̅ ± standard deviation). A quarter (23%) of the diurnal variability in CO₂flux was caused by the effect of changing water temperature on gas solubility. Furthermore, evaporation rates were ~ 10 times greater than precipitation, causing a 14% increase in salinity, a potential precursor of seagrass die‐offs. Evaporation rates were not correlated with solar radiation, but instead with air–water temperature gradient and wind shear. We also confirm the role of convective forcing on night‐time enhancement and day‐time suppression of gas transfer. At this site, temperature trends are regulated by solar heating, combined with shallow water depth and relatively consistent air temperature. Our findings indicate that evaporation and air–water CO₂exchange over shallow, tropical, and subtropical seagrass ecosystems may be fundamentally different than in submerged vegetated environments elsewhere, in part due to the complex physical forcing of coastal air–sea gas transfer.

 

Coastal vegetated habitats like seagrass meadows can mitigate anthropogenic carbon emissions by sequestering CO₂as “blue carbon” (BC). Already, some coastal ecosystems are actively managed to enhance BC storage, with associated BC stocks included in national greenhouse gas inventories. However, the extent to which BC burial fluxes are enhanced or counteracted by other carbon fluxes, especially air‐water CO₂flux (FCO₂) remains poorly understood. In this study, we synthesized all available direct FCO₂measurements over seagrass meadows made using atmospheric Eddy Covariance, across a globally representative range of ecotypes. Of the four sites with seasonal data coverage, two were net CO₂sources, with average FCO₂equivalent to 44%–115% of the global average BC burial rate. At the remaining sites, net CO₂uptake was 101%–888% of average BC burial. A wavelet coherence analysis demonstrated that FCO₂was most strongly related to physical factors like temperature, wind, and tides. In particular, tidal forcing was a key driver of global‐scale patterns in FCO₂, likely due to a combination of lateral carbon exchange, bottom‐driven turbulence, and pore‐water pumping. Lastly, sea‐surface drag coefficients were always greater than the prediction for the open ocean, supporting a universal enhancement of gas‐transfer in shallow coastal waters. Our study points to the need for a more comprehensive approach to BC assessments, considering not only organic carbon storage, but also air‐water CO₂exchange, and its complex biogeochemical and physical drivers.