Friday, August 7, 2009 - 10:50 AM

COS 127-9: A remote sensing landscape analysis of Sudden Aspen Decline (SAD) in southern Utah’s Cedar Mountain

Chad Oukrop1, Ronald J. Ryel1, Dale Bartos2, Doug Ramsey1, David Evans1, and A. Joshua Leffler1. (1) Utah State University, (2) Rocky Mountain Research Station

Background/Question/Methods                                                                                                                                                                                                Quaking aspen (Populus tremuloides Michx.) is the most widespread deciduous tree species in North America (Baker, 1925).  In the Intermountain West, aspen communities are highly valued multiple use ecosystems, noted for forage production, understory diversity, wildlife habitat, timber, water resources, and aesthetics (Bartos, 2001).  However, aspen communities in the Intermountain Region of western United States are in evident decline, with certain areas experiencing sudden aspen decline (SAD) over the past decade, a phenomena defined by rapid overstory mortality with little or no regeneration.  Recently, landowners on Cedar Mountain in southern Utah have expressed concern over the occurrence of SAD across portions of the Mountain and surrounding area.  Land managers lack critical information on the extent and magnitude of SAD, and could utilize detailed spatial information to plan, implement, and monitor aspen restorations projects.  The main objective of this study was to produce a map of the present aspen health status across the Cedar Mountain landscape using multi-spectral imagery, digital elevation models, and ancillary biophysical datasets. 

Results/Conclusions
During the 2008 summer, a total of 94 points from a 900 m systematic grid were sampled within an aspen cover layer extract from the 2005 SWReGAP analysis, verified for aspen cover using 2006 NAIP imagery, and assigned to one of four aspen health classes: 1) healthy, 2) dying, 3) dead, and 4) seral.  A supervised classification for Cedar Mountain was conducted using classification and regression tree (CART) analysis using See5 data mining software.  The predictor layers used in the model consisted of Landsat reflectance data, tassel cap transformations (BGW), vegetation indices (NDVI), 30-m Land Form data, and topographic ancillary data (slope, aspect, elevation) extracted from 30-m DEMs.  The classification algorithm produced an aspen health map for Cedar Mountain consisting of 4 cover classes with an overall accuracy of 73.7%.  Healthy aspen cover was the most abundant cover type, followed by dying, seral, and dead aspen stands (49%, 28%, 16%, and 7%, respectively).   Damaged aspen stands (dying and dead) illustrating SAD characteristics accounted for 35% (4,210 ha) of the aspen cover on Cedar Mountain.  This spatial product will be available to land managers of Cedar Mountain to assist in locating and implementing restoration projects on declining aspen stands.