This study presents the first national-scale compendium of annual lake trophic state (LTS) from 1984 through 2020, covering 55,662 lakes of at least 10 ha in the contiguous United States. The dataset, called LTS-US, is derived from Landsat surface reflectance data and in situ measurements of total phosphorus and true color. It is constructed using FAIR data principles, ensuring it is findable, accessible, interoperable, and reproducible. The dataset includes predictions for LTS for each lake, based on a combination of in situ measurements and remote sensing data. The LTS-US dataset provides a powerful tool for analyzing LTS at a national scale, allowing for the identification of macroscale patterns and trends in lake water quality. The dataset is structured in a tabular format, with each row representing a lake-year combination. The main dataset is contained in "ensemble_predictions.csv", which includes both categorical LTS predictions and probabilities for each LTS prediction. An additional dataset, "individual_predictions.csv", contains probabilities generated for each of the three modeling methodologies. The dataset is available at the Environmental Data Initiative (https://doi.org/10.6073/pasta/212a3172ac36e8dc6e1862f9c2522fa4). The study also includes technical validation, which shows that all three modeling approaches performed similarly in terms of overall and balanced accuracy, and AUC of ROC curves. The results suggest that the LTS-US dataset is a valuable resource for addressing basic and applied research questions about lake water quality at a suite of spatial and temporal scales.This study presents the first national-scale compendium of annual lake trophic state (LTS) from 1984 through 2020, covering 55,662 lakes of at least 10 ha in the contiguous United States. The dataset, called LTS-US, is derived from Landsat surface reflectance data and in situ measurements of total phosphorus and true color. It is constructed using FAIR data principles, ensuring it is findable, accessible, interoperable, and reproducible. The dataset includes predictions for LTS for each lake, based on a combination of in situ measurements and remote sensing data. The LTS-US dataset provides a powerful tool for analyzing LTS at a national scale, allowing for the identification of macroscale patterns and trends in lake water quality. The dataset is structured in a tabular format, with each row representing a lake-year combination. The main dataset is contained in "ensemble_predictions.csv", which includes both categorical LTS predictions and probabilities for each LTS prediction. An additional dataset, "individual_predictions.csv", contains probabilities generated for each of the three modeling methodologies. The dataset is available at the Environmental Data Initiative (https://doi.org/10.6073/pasta/212a3172ac36e8dc6e1862f9c2522fa4). The study also includes technical validation, which shows that all three modeling approaches performed similarly in terms of overall and balanced accuracy, and AUC of ROC curves. The results suggest that the LTS-US dataset is a valuable resource for addressing basic and applied research questions about lake water quality at a suite of spatial and temporal scales.