Print PageWednesday, August 10, 2011
Exhibit Hall 3, Austin Convention Center
Background/Question/Methods Land cover change has had a global impact over the last half century as a result of both natural and human effects. In the Chihuahuan desert, the conversion of desert grassland to shrubland appears to have altered the provision of ecosystem goods and services with important social, economic, and ecological implications. To improve large-scale change detection capacity and land management practices in the Chihuahuan desert, improved high-spatial resolution land cover maps are required – especially in Chihuahuan Desert mountain landscapes, which are important to regional biodiversity, sensitive to change, and poorly studied relative to lowland landscapes. The use of satellite imagery in conjunction with remote sensing techniques has become an important and efficient tool for classifying and mapping land cover over large areas. This study combined field data collection, IKONOS satellite imagery and other geospatial data, remote sensing, and geographic information system analysis to produce a land cover classification for the Indio Mountains Research Station (IMRS) in West Texas. IMRS is the primary field research station owned and managed by the University of Texas at El Paso and is approximately 15,992 hectares in size. Field surveys that derived plant community and other environmental data were conducted during the summers of 2008 and 2009. A total of 802 sites were sampled. Cluster analysis and non-metric multidimensional scaling ordinations of plant species cover were computed using PCOrd statistical software to determine the range of vegetated land cover classes present at IMRS. A supervised classification of a 2007 IKONOS image was conducted using ENVI 4.4 and a subset of the field sites from each vegetation class. An accuracy assessment was calculated using a confusion matrix of the remaining sites not used during the classification. Land cover was correlated with a range of environmental data including geology, NDVI, slope, aspect, and elevation using ArcGIS 9.2.
Results/Conclusions A total of 6 vegetation classes were derived from the cluster analysis and ordination. These 6 classes were combined with non-vegetated land cover classes to derive a10-class land cover map of IMRS. The overall accuracy of the land cover map is 78.7% with a Kappa statistic of 0.716. The error of commission and omission are 86% and 80%, respectively. Correlation of environmental variables shows that geology was a factor in the distribution of one of the vegetation classes. This land cover map is being used as a baseline to study decade-time scale land cover change within discrete land cover classes.
