Description: This land cover classification product contains 5 classes (values of ‘0’ are NoData):Tree Canopy (elevated vegetation (>= 15 ft) not described in the forb and shrub layer)Low Vegetation / Grass (shrub and forb layer)Bare Earth & Impervious Surfaces (any and all manmade and natural)Open WaterOtherMethods: The classification method for the primary, contiguous portions of the unincorporated land was based upon a neural network classificaiton using the NAIP imagery and LIDAR-derived feature height data. A convolution neural network (CNN) incorporating the 4-band NAIP imagery and the feature height data was utilized to classify the categories of interest to the County of San Diego. The process included extracting training data from the NAIP imagery and feature height data sets, feature training, image classification, and two accuracy checks. One of the accuracy checks was based on visual review of the high-spatial resolution NAIP imagery and the other was based on cross-tabulation with existing image classification data set (described below). The accuracy check ensured that the product met accuracy expectations (at least 85%-90% accuracy for tree canopy estimate). Tree canopy was classified based upon spectral-radiometric information in the NAIP imagery, as well as the feature height being 15 ft or greater. The tree canopy classification was the primary focus of the project. The final deliverable product has the same spatial resolution as the NAIP imagery (i.e., 0.6 m). For the island polygons in the western portion of the study area, the land cover classification was derived using data already available from the 2014 Urban Tree Canopy Assessment (UTCA), and the neural network classification approach was not used. The existing UTCA classification product was recoded using the below crosswalk, and also merged/aggregated into the same raster grid as the SD County image classification product provided by Imagis.UTCA "Tree Canopy" = CNN classification "Tree Canopy" (elevated vegetation (>= 15 ft) not described in the forb and shrub layer)UTCA "Grass/Shrub" = CNN classification "Grass/Vegetation" (shrub and forb layer)UTCA "Bare Earth" = CNN classification "Bare Ground" (any and all manmade and natural)UTCA "Water" = CNN classification "Open Water"UTCA "Building" = CNN classification "Other"UTCA "Roads" = CNN classification "Bare Ground" (any and all manmade and natural)UTCA "Other Paved Surfaces" = CNN classification "Bare Ground" (any and all manmade and natural)The neural network model (DeepLab 3+ in TensorFlow) was trained on 7 NAIP/nDSM combination TIFF tiles, with the raster-based labels for those areas derived from the 2014 UTCA classification product.The 2016 NAIP 4-band imagery and LiDAR acquisitions of San Diego County were used for this classification product, as mentioned above. DSMs were extracted from the LiDAR data by Imagis, and normalized DSMs (nDSMs) were generated by subtracting the DEMs (provided by the County) from the DSMs. The nDSMs data tiles with feature height information were combined with the NAIP imagery and utilized with the trained TensorFlow model and perform the image classification on the primary, conginuous portions of the unincorporated land.