This dataset provides a global annual 1 km fractional land-cover product from 1982 to 2024 based on the FROM-GLC Plus framework. The framework integrates multi-source satellite imagery, spatio-temporal sample migration, time-series reconstruction, knowledge-guided inference, and multimodal observations to support scalable mapping from 30 m to meter/sub-meter resolution and from annual to daily temporal frequency. The dataset was generated by aggregating high-resolution annual land-cover maps to 1 km resolution. Each 1 km pixel stores the proportional cover of major land-cover classes, allowing mixed land-cover composition to be represented explicitly. It is suitable for long-term land-cover change analysis, ecosystem monitoring, biodiversity assessment, and Earth system applications.