Using object-based image analysis to conduct high-resolution conifer extraction at regional spatial scales

Abstract

The distribution and abundance of pinyon ( Pinus monophylla ) and juniper ( Juniperus osteosperma, J. occidentalis ) trees (hereinafter, "pinyon-juniper") in sagebrush ( Artemisia spp.) ecosystems of the Great Basin in the Western United States has increased substantially since the late 1800s. Distributional expansion and infill of pinyon-juniper into sagebrush ecosystems threatens the ecological function and economic viability of these ecosystems within the Great Basin, and is now a major contemporary challenge facing land and wildlife managers. Particularly, pinyon-juniper encroachment into intact sagebrush ecosystems has been identified as a primary threat facing populations of greater sage-grouse ( Centrocercus urophasianus ; hereinafter, "sage-grouse"), which is a sagebrush obligate species. Even seemingly innocuous scatterings of isolated pinyon-juniper in an otherwise intact sagebrush landscape can negatively affect survival and reproduction of sage-grouse. Therefore, accurate and high-resolution maps of pinyon-juniper distribution and abundance (indexed by canopy cover) across broad geographic extents would help guide land management decisions that better target areas for pinyon-juniper removal projects (for example, fuel reduction, habitat improvement for sage-grouse, and other sagebrush species) and facilitate science that further quantifies ecological effects of pinyon-juniper encroachment on sage-grouse populations and sagebrush ecosystem processes . Hence, we mapped pinyon-juniper (referred to as conifers for actual mapping) at a 1 × 1-meter (m) high resolution across the entire range of previously mapped sage-grouse habitat in Nevada and northeastern California. We used digital orthophoto quad tiles from National Agriculture Imagery Program (2010, 2013) as base imagery, and then classified conifers using automated feature extraction methodology with the program Feature Analyst ™ . This method relies on machine learning algorithms that extract features from imagery based on their spectral and spatial signatures. We classified conifers in 6,230 tiles and then tested for errors of omission and commission using confusion matrices. Accuracy ranged from 79.1 to 96.8, with an overall accuracy of 84.3 percent across all mapped areas. An estimated accuracy coefficient (kappa) indicated substantial to nearly perfect agreement, which varied across mapped areas. For this mapping process across the entire mapping extent, four sets of products are available at https://doi.org/10.5066/F7348HVC , including (1) a shapefile representing accuracy results linked to mapping subunits; (2) binary rasters representing conifer presence or absence at a 1 × 1 m resolution; (3) a 30 × 30 m resolution raster representing percentages of conifer canopy cover within each cell from 0 to 100; and (4) 1 × 1 m resolution canopy cover classification rasters derived from a 50-m-radius moving window analysis. The latter two products can be reclassified in a geographic information system (GIS) into user-specified bins to meet different objectives, which include approximations for phases of encroachment. These products complement, and in some cases improve upon, existing conifer maps in the Western United States, and will help facilitate sage-grouse habitat management and sagebrush ecosystem restoration.