PARENT SESSION
Poster Session: Remote Sensing

A notable limitation on mapping large-area forest structure using TM imagery and ground plot data. ^*XU, YAGUANG ¹, ^*PRATHER, JOHN ¹, ^*HAMPTON, HAYDEE ¹, DICKSON, BRETT ², AUMACK, ETHAM ¹ and SISK, THOMAS ¹, ¹ Center for Environmental Science and Education, Northern Arizona University, Flagstaff, AZ, USA² Department of Fishery and Wildlife Biology, Colorado State University, Fort Collins, CO, USA

ABSTRACT- Maps of forest structural attributes such as tree density and basal area form the spatial data foundation required for fire behavior and wildlife habitat analysis (or modeling) at landscape and regional scales. TM imagery coupled with training data from ground plot measurements is a major source of spatial data used to build such structural attribute maps. Over two years we mapped forest structural attributes across two million acres of the ponderosa pine dominated forest in northern Arizona using regression tree analysis, with and without committee models. We encountered a considerable limitation in using TM imagery for mapping forest structure, although the technical approach using the two types of data is usually considered an efficient way to map structure over large forested areas. Our results demonstrate that where forest was very dense (> 800 trees / ha) or extremely sparse (< 100 trees / ha), accuracy of the structural maps decreased significantly. We were unable to erase this problem by changing predictive layers, ground plot training datasets, or analytical methods. Thus, we conclude that this problem decrease in accuracy was caused by insensitivity of TM imagery to the areas with very low tree cover, and weak penetrating capacity of TM imagery to very high tree cover areas. A canopy cover map derived from USGS Digital Orthophoto Quads (DOQs) was used to illustrate this conclusion. The consequences of the genetic drawback of TM imagery may result in profound effects on some fire behavior modeling and wildlife habitat analysis in which the structural maps are used as primary input layers.

KEY WORDS: forest , mapping, structure, TM, limitation