Reproducibility package — Early-season surface urban heat contrast in two Tier-2 Gangetic cities (Purnea and Katihar, Bihar, India): a reproducible Landsat 8/9 and Dynamic World workflow with cross-product uncertainty audit, 2016–2026

Reproducibility package for the manuscript Early-season surface urban heat contrast in two Tier-2 Gangetic cities (Purnea and Katihar, Bihar, India): a reproducible Landsat 8/9 and Dynamic World workflow with cross-product uncertainty audit, 2016-2026, submitted to the SPIE Journal of Applied Remote Sensing. This deposit contains everything a reviewer or independent researcher needs to (i) audit every numerical claim in the manuscript without re-running the Earth Engine pipeline, and (ii) reproduce every Earth Engine output from open data in a single execution. Contents (zipped): manuscript preprint PDF; five supplementary CSV tables (S1 per-scene Landsat inventory, S2 reproducibility checklist, S3 annual primary- and full-window metrics, S4 Landsat-MODIS matched 1 km values, S5 SUHI reference-ring 5 km values); six 600 dpi figures; three reproducibility-critical scripts (Earth Engine extraction, statistical reduction, matplotlib figure generation in JARS house style); pinned Conda environment.yml and pip requirements.txt; and a documentation layer including reproducibility checklist, data dictionary, processing pipeline narrative with Mermaid diagram, FAIR-compliance audit, naming conventions, provenance record, validation and uncertainty notes, reviewer-facing transparency statement, release notes, and Zenodo description text. Methodology summary: Landsat 8/9 Collection-2 Level-2 ST_B10 with QA_PIXEL bits 1-5 masking and ST_QA retained as retrieval-uncertainty diagnostic; primary 9 February-5 March and full 1 February-31 March windows; Dynamic World v1 annual modal labels; SUHI defined as Dynamic World built (label 6) inside the municipality minus Dynamic World non-urban classes (labels 1-5) in a 5 km external ring; MODIS MOD11A2 Collection 6.1 Terra cross-product check at 1 km grid; descriptive 95% Student t-intervals and 2,000-iteration bootstrap RMSE intervals; Pearson r with Fisher z CI and Spearman rho reported as exploratory because n = 10-11 per city. NOT included: raw Bhuvan municipal/ward GeoJSON layers (provenance documented separately); intermediate Landsat GeoTIFFs (fully reproducible from open sources). License: Data and documentation under CC BY 4.0; code under MIT. Access during peer review: restricted with a shared reviewer link; the deposit will be opened on manuscript acceptance.