Using Classification and Regression Tree and Valley Bottom Modeling Techniques to Identify Riparian Vegetation in Pinal County, Arizona

Abstract

The ecological value and functionality of riparian systems along ephemeral, intermittent, and perennial streams in the Southwest is well established. In Pinal County, Arizona the existing datasets available to environmental managers and governing bodies drastically underestimate the extent and presence of riparian zones. This study addresses the issue through the use of remote sensing land cover classification techniques. Landsat 8 data, topographic data, and high-resolution color infrared (CIR) imagery, and several derived vegetation indices are used to construct a classification and regression tree (CART) model. Using training data, the CART model is used for the identification and delineation of basic land cover classes across the County. Woody annual and perennial species are identified and associated to riparian zones using a valley bottom model (VBM) developed by the United States Department of Agriculture. The CART model (kappa value of 0.76) found that 929 square-miles of annual vegetation and 651 square-miles of perennial vegetation are present across Pinal County. Annual and perennial vegetation classifications are assessed for density using a 0.33 acre moving window. The density values for both classes are then used in conjunction to differentiate upland, xeroriparian, mesoriparian, and hydropriarian vegetation zones. Vegetation zones are clipped to regions where the VBM identifies valley bottom probability to be 62 percent or greater. The results generated provide a sufficiently comprehensive dataset that gives County managers and environmental professionals improved insight into the presence and distribution of important riparian habitats.