Land Surface Models are now increasingly being run at the kilometer scale (k-scale) due to advances in computational power and the availability of global k-scale datasets for configuring and validating simulations at such high spatial resolutions. In this work, we performed a 40-year E3SM Land Model version 2 (ELMv2) simulation over the Contiguous United States at 1 km and evaluated the simulation against multiple benchmark datasets. The simulation was driven by the NLDAS atmospheric forcing from 2000-2019, which was recycled, and the first 20 years of the simulation were excluded from the analysis. Key enhancements to the default ELMv2 configuration include (1) the implementation of recently developed k-scale surface parameters from Li et al. (2024), (2) the solar radiation parameterization scheme from Hao et al. (2021) to account for sub-grid topography, and (3) the updated infiltration scheme from Xu et al. (2024) to represent pluvial inundation better. By calibrating the model based on benchmark datasets of water table depth and surface water fraction, we significantly improved the simulation accuracy of these variables. Additional benchmark datasets were used to evaluate the simulation of soil moisture, evapotranspiration, runoff, snow processes, and land surface temperature. Furthermore, we quantified the information loss when the 1 km ELMv2 simulation is coarsened to a 12 km resolution.