Submitted to: Environmental Management
Publication Type: Peer reviewed journal
Publication Acceptance Date: 1/27/2011
Publication Date: 5/1/2011
Publication URL: hdl.handle.net/10113/49306
Citation: Madsen, M.D., Zvirzden, D.L., Davis, B.D., Petersen, S.L., Roundy, B.A. 2011. Feature extraction techniques for estimating pinon and juniper tree cover and density, and comparison with field based management surveys. Environmental Management. 47:766-776. Interpretive Summary: Since European settlement the western United States has seen a dramatic expansion of piñon (Pinus spp.) and juniper (Juniperus spp.) (P-J) woodlands into surrounding ecosystem types. Expansion of these woodlands can result in ecological and socioeconomic degradation. While degradation is site dependent, P-J canopy cover and tree density are correlated with the degree woodland encroachment is controlling ecosystem function. Consequently, land managers are actively involved in monitoring these parameters in order to develop appropriate restoration treatments. Field-based P-J monitoring approaches are limited, due to the heterogeneity of rangeland systems which makes it is difficult to extrapolate these data beyond the area where the measurements were made, and the lack of resources required to monitor woodland encroachment over extensive and rugged terrain. The purpose of this study was to 1) develop an effective and efficient method for accurately quantifying P-J tree canopy cover and density directly from high resolution photographs; and 2) compare feature extracted data to a typical field-based dataset collected by Utah’s Division of Wildlife Resources Range Trend Project (DWR-RTP). Tree cover was estimated from aerial-photography using Feature Analyst®. Tree density was estimated from the resultant cover map (which had trees represented as polygons) as the sum of the total number of individual polygons after isolation using a negative buffer post-processing technique. Results indicated that the proposed feature extraction techniques accurately estimated tree canopy cover and were comparable to the DWR-RTP. Feature extraction estimates of tree density tended to underestimate the number of trees at a site, primarily due to the techniques inability to detect immature trees. Feature extraction techniques proposed in this study could be used for a host of monitoring applications, such as woodland encroachment of all but immature trees, fuel loads, timber value, wildlife habitat, and grazing suitability.
Technical Abstract: In western North America, expansion and stand infilling by piñon (Pinus spp.) and juniper (Juniperus spp.) (P-J) trees constitutes one of the greatest afforestations of our time. Feature extracted data acquired from remotely sensed imagery can help managers rapidly and accurately assess this expansion at broad landscape-scales. The objectives of this study were to: 1) develop an effective and efficient method for accurately quantifying P-J tree canopy cover and density directly from high resolution photographs; and 2) compare feature extracted data to typical in-situ datasets used by land managers. Tree cover was extracted from aerial-photography using Feature Analyst®. Tree density was calculated as the sum of the total number of individual polygons (trees) within the tree cover output file after isolation using a negative buffer post-processing technique. Feature extracted data were compared to ground reference measurements and estimates from Utah’s Division of Wildlife Resources Range Trend Project (DWR-RTP). Results indicated that the proposed feature extraction techniques for cover and density were highly correlated to ground reference and DWR-RTP datasets. Feature extracted estimates of cover generally showed a near 1:1 relationship to these data, while tree density was underestimated; however, after calibration for juvenile trees a near 1:1 relationship was realized. Feature extraction techniques used in this study provide an efficient method for assessing important rangeland indicators, including: density, cover, and extent of P-J tree encroachment. Correlations found between field and feature extracted data provide evidence to support extrapolation between the two approaches when assessing woodland encroachment.