Developing a 1m res. Land Cover Classification for the State of Arizona
Previous efforts to create land use/land cover classifications have often been limited in either spatial resolution or extent and are frequently conducted using proprietary datasets. This effort aimed to create a repeatable, state-wide classification of basic land cover types (open water, roof, dense vegetation, etc.) utilizing open government data (NAIP and Landsat imagery) and open source software. Part of a larger effort by the Arizona Game and Fish Department to develop a comprehensive riparian classification layer for the entire state, this intermediary product is itself useful for a variety of disciplines and research fields. This presentation will address the methodology employed, the issues and challenges in analysis imposed by large datasets, and the benefits in repeatability and access with model development using open source utilities. As well, we'll touch on the technologies used including python packages rasterio (raster processing), scipy (machine learning), and geopandas (vector manipulation), and QGIS (preliminary data preparation/post-processing inspection).