Reproducibility package for a 30-year TELEMAC-2D Lake Ontario water-level hindcast (1992–2021)

This record supports the manuscript Separating boundary-inherited skill from hydrodynamic skill in a 30-year Lake Ontario water-level hindcast: seasonal delay, high-water bias and mesh sensitivity. The record contains the model set-up, extracted gauge series, boundary-only benchmark code, derived results, and provenance materials supporting a 30-year (1992–2021) depth-averaged TELEMAC-2D water-level hindcast of Lake Ontario. Model performance is evaluated at five water-level gauges (Kingston, Cobourg, Toronto, Burlington, and Port Weller) against observations and a cross-validated boundary-only benchmark based on prescribed Niagara River discharge and St. Lawrence River water level at Brockville. The package includes the base mesh and spatial Manning field; boundary-condition and liquid-boundary files; documented steering files for M1 (ERA5, uniform Manning n = 0.030), M2 (ERA5, spatial friction, reference), and M3 (CaSR v3.1, spatial friction, 1992–2001); observed and simulated daily gauge series; observation quality-control information; benchmark and analysis code; numerical-sensitivity outputs; gauge-extraction-node sensitivity; and derived metrics underlying the article tables and figures. Production steering files are documented from the surviving working template and production configuration. Later TELEMAC-2D v9.1 re-runs and the 2019 controlled numerical-sensitivity runs are included separately for provenance and sensitivity analysis and are not presented as the analysed production M1–M3 runs. Atmospheric forcing fields interpolated to the model mesh, restart files, and full-field TELEMAC result files are not included because of their volume. See README.md for experiment definitions, provenance, source-data terms, licensing, conventions, and reproducibility instructions, and documentation/FILE_INVENTORY.csv for the description and SHA-256 checksum of every deposited file.