Lidar-derived digital elevation models often contain a vertical bias due to vegetation. In areas with tidal influence the amount of bias can be ecologically significant, for example, by decreasing the expected inundation frequency. We generated a corrected digital elevation mode (DEM) for Suisun marsh using a modification of the Lidar Elevation Adjustment with NDVI (LEAN) technique (Buffington et al. 2016). GPS survey data (6912 points, collected across public and private land in 2018), Normalized Difference Vegetation Index (NDVI) derived from an airborne multispectral image (June 2018), a 1 m lidar DEM from September 2018, and a 1 m canopy surface model were used to generate models of predicted bias across the study domain. Due to the large differences in vegetation height and density between natural and diked wetlands, we calibrated a separate model for each cover type. The modeled predicted bias for each cover type was then subtracted from the original lidar DEM to generate a new DEM. Across all GPS points, mean initial lidar error was 22.5 cm (SD=17.5) and root-mean squared error (RMSE) was 28.5 cm. After correction with LEAN, mean error was 0 cm (SD=9.7) and RMSE was 9.7 cm, a 66 percent improvement in accuracy. Some ponds were partially flooded and had no lidar returns; to create a continuous coverage, we iteratively used the focal statistics tool with a 10 meter radius to expand the corrected elevation values into NoData areas until data gaps were covered. Large channels were masked out from the final DEM using the lidar returns and airborne imagery.