Wednesday, August 5, 2009
Exhibit Hall NE & SE, Albuquerque Convention Center
Christian G. Andresen, Biological Sciences, University of Texas at El Paso, El Paso, TX and Craig E. Tweedie, Biology, University of Texas, El Paso, El Paso, TX
Background/Question/Methods The importance of developing accurate, low-cost, and time-efficient methods for arctic tundra monitoring is essential for spatial assessments and quantification of biogeochemical processes such as green house emissions. Coarse resolution satellite imagery (grid pixel > 10 m) have a limited capacity for mapping highly heterogeneous tundra ecosystems. Commercial high-spatial resolution imaging systems provide a new tool for mapping arctic vegetation at a finer scale. We incorporated Quickbird imagery (0.7 m pan, 2.8 m multispectral) and a compilation of ground-truth observations to develop a sub-meter tundra vegetation map for the Biocomplexity Experiment site (590 ha) near Barrow, Alaska.
Results/Conclusions
Plant species where grouped according to moisture levels at peak season and a total of seven plant communities where mapped. Using a pan-sharpened, orthorectified, atmospheric and radiometrically corrected Quickbird scene, we undertook a maximum-likelihood supervised classification approach based on ground data for the seven plant communities across the study region. To assess the accuracy we used a compilation of ground-truth observations from the Barrow-Arctic Information Database (BAID) resulting in an overall accuracy of 86.72%. The classification product with a grid pixel size of 0.7 m was successful in characterizing the extreme spatial heterogeneity of this tundra landscape. This study maps tundra at a high spatial resolution using satellite imagery, and improves the quantification and scalability of plot level measurements to the landscape scale.
