Derived data and code for "Baseline bias shapes apparent skill changes under rainfall-input scaling in an urban flood surrogate"

This archive contains public-safe derived data, analysis code, and verification records supporting the manuscript "Baseline bias shapes apparent skill changes under rainfall-input scaling in an urban flood surrogate." The package supports integrity checks and selected derived-analysis reruns for: the fixed-reference baseline–response boundary check; storage- and detection-threshold sensitivities; the separate 60-min auxiliary analysis; lead-time × rainfall-factor × tier sensitivity; internal-pathway aggregate diagnostics; rank-preserving area- and depth-index-matched spatial controls; and seed-level CSI direction counts (Supplementary Table S7g). Every file is covered by a SHA-256 inventory, and a self-contained verifier (verify_public_package.py) recomputes the recorded checks. The archive does not contain raw hydraulic inputs, the hydraulic solver or run controls, the source regional-frequency workbook, model checkpoints, restricted high-resolution response banks, or licensed basemap and satellite imagery. It therefore cannot reproduce model training, full inference, hydraulic simulations, or the map-based figures. The restricted source workbook and held-out forcing archive are represented by SHA-256 records and public-safe aggregate tables. Derived data and documentation are licensed under CC BY 4.0; Python code is licensed under MIT (see LICENSE). This work was supported by an Electronics and Telecommunications Research Institute (ETRI) grant funded by the Korean government [26ZR1300, Development of Technology for the Urban Extreme Rainfall Response Platform].