Processed Data and Trained Models Supporting "Model-Sensitive Interannual Variability in Texas Coastal Subsurface Oxygen (2013–2024) Revealed by Machine Learning"

This Zenodo record is the article-specific processed-data and trained-model archive for the manuscript “Model-Sensitive Interannual Variability in Texas Coastal Subsurface Oxygen (2013–2024) Revealed by Machine Learning.” The deposited files were prepared specifically to support the analyses, model evaluations, interannual statistics, and scientific results reported in this article. This record is not intended as a general-purpose or standalone compilation of the original observational archives. The deposit contains the processed data products supporting Figures 1–5 and the three final trained machine-learning models: a multilayer perceptron, histogram gradient boosting, and Random Forest. The processed products include the one-minute Glider reconstruction table, common chronological validation predictions, mission-withholding and regional validation statistics, annual and pressure-binned oxygen products, hydrographic summaries, Mechanisms Controlling Hypoxia Acrobat track information, and Southeast Area Monitoring and Assessment Program station information. Oxygen values in the processed model tables are stored in µmol kg⁻¹. Values reported in the article in mL L⁻¹ were calculated sample by sample using in-situ seawater density. The deposit does not include publication figures, analysis source code, or complete copies of the original observational archives. Information on the original data sources and their public availability is provided in the accompanying README. This dataset and the trained models should be interpreted using the data screening, model configuration, validation design, and uncertainty limitations described in the associated article. Keywords dissolved oxygen; underwater gliders; machine learning; Texas continental shelf; Gulf of America; oxygen reconstruction; interannual variability; model comparison; ocean observations