Comparison of Land Cover Classifications between Ground Observations and Remote Sensing Data in San Diego, California

Abstract

Land cover and its changes are intertwined with human and natural activities. Accurate land cover data is invaluable for land use decisions as well as understanding local and global land cover change. This study explored the local land cover trends of an area of interest in San Diego, California, with the goal of evaluating consistency between ground observations, manual land cover classifications, and land cover maps from remote sensing sources. Ground observations were taken using the Global Learning and Observations to Benefit the Environment (GLOBE) Observer application. Remote sensing sources were accessed through Earth Map, and included the World Cover, Dynamic World, ESRI, and Meta/WRI Global Canopy Height maps of land cover and tree canopy based on satellite imagery. While there was general agreement between ground observations and remote sensing sources over the area as a whole, small inconsistencies were found at nearly all sampling points, and blatant differences were found at three of thirty-seven sampling points between the chosen remote sensing sources as well as between ground observations and remote sensing maps. These inconsistencies were most obvious with the Dynamic World and ESRI datasets, which tended to overgeneralize urban land cover compared to the World Cover dataset. The discrepancies between sources describing the land cover of the same area show the need for caution when examining remote sensing sources and the challenges of remote sensing data, which are increased with differing classification measures.