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.