Annual National Land Cover Database Collection 1.0: 39-year record of land cover and land cover change products across the conterminous U.S.

To meet the growing need for timely, frequent, and consistent long-term land surface information, the Annual NLCD Collection 1.0 integrates methodologies from the National Land Cover Database (NLCD) and the Land Change Monitoring, Assessment, and Projection (LCMAP) initiative with modern geospatial deep-learning approaches. The resulting product suite provides six annual 30-m products for the conterminous United States (CONUS) spanning 1985–2023: Land Cover, Land Cover Change, Land Cover Confidence, Fractional Impervious Surface, Impervious Descriptor, and Spectral Change Day of Year. The overall accuracy of the NLCD 16-class thematic land cover product was 64.7 ± 0.5% across the study period and increased to 76.9 ± 0.4% when agreement included a match with either the primary or alternate reference label. Overall accuracy remained stable over time ranging from 64.5% to 65.0%. Even with increased temporal resolution, Annual NLCD maintained comparable accuracy to that of previous NLCD products mapped at 2–3-year intervals. Change detection accuracy for Annual NLCD was low, consistent with the difficulty of detecting and dating annual land cover transitions. User's and producer's accuracies varied considerably across change themes. Water gain and loss, forest loss, and shrubland gain were the better-detected change types, while most other change themes had accuracy <40% when allowing a ± 2-year buffer to define agreement. Despite its finer thematic classification scheme, the Annual NLCD product substantially improved change accuracy relative to LCMAP. Based on the Annual NLCD maps, approximately 19.4% of CONUS experienced at least one land cover change during 1985–2023. Annual NLCD products provide a consistent annual record of land cover and land use information to support diverse downstream applications.