A dataset of modeled nearshore water levels (WLs) was developed for three states (Virginia, Georgia, and Florida) along the U.S. Atlantic coast. Water levels, defined for this dataset as the linear sum of tides and non-tidal residuals (NTR), were produced by Muis and others (2016) using a global tide and surge model (GTSM) forced by global atmospheric fields. Water level outputs were extracted from the global grid at approximately 20 km resolution along the Atlantic coastline. These data were then statistically downscaled using a signal-specific set of corrections to improve skill in comparison to tide gauge observations (Parker and others, 2023). Projected water levels were forced by CMIP6 future period data. Four CMIP6 climate models were selected from the High-Resolution Model Intercomparison project (highresMIP; Haarsma and others, 2016) to sample variability in climate predictions. Similar modeled data for North Carolina and South Carolina are available from Barnard and others, 2023, at https://doi.org/10.5066/P9W91314)