Data and code for: Land-cover product choice alters assessments of urban riverfront change: Evidence from four Chinese cities

This dataset contains the data and analysis code supporting a comparison of CLCD v1.0.5 and GLC_FCS30D v2 along urban riverfronts in Fuzhou, Guangzhou, Hangzhou and Nanning, China. The study uses the 2010 and 2020 classifications for one predefined 15 km main-river reach in each city. The files include land-cover windows, frozen study geometries and class mappings, pixel- and edge-level records, derived tables, GIS layers, and source code. The analyses cover product-specific land–water interfaces, strictly paired interface states, and land-cover transitions in two landward distance bands. Supplementary workflows examine common source classes, proximity between interfaces, reference water geometry, leave-one-segment-out sensitivity, and half-cell shifts of the analysis grid. Figure source data and the submitted supplementary tables are also included. The historical-imagery records contain the acquired Landsat and Sentinel windows, source metadata, archived AI observations, and eligibility decisions. No two-date trajectory met the predefined main-reference criteria; these records therefore do not provide an independent estimate of either product’s accuracy. The README identifies the calculation entry points and software requirements. File checksums and execution records document the archived inputs and the checks performed when preparing this deposit. Third-party data retain their original licences and source attribution.