White, R.A.; Piraino, K.; Shortridge, A., and Arbogast, A.F., 2019. Measurement of vegetation change in critical dune sites along the eastern shores of Lake Michigan from 1938 to 2014 with object-based image analysis. Journal of Coastal Research, 35(4), 842–851. Coconut Creek (Florida), ISSN 0749-0208.
Coastal sand dunes are common on coastlines around the world and are often heavily managed to control erosion or to enable their dynamic nature. The largest body of freshwater coastal dunes in the world occurs on the eastern shore of Lake Michigan in North America. This dune system is one of the most heavily utilized landscapes in the Great Lakes region and has tremendous economic and cultural significance. As a result, they are managed in association with Michigan's Sand Dune Protection and Management Act, which seeks, in part, to preserve their diversity, quality, and function. To achieve these goals, it is essential to understand the dunes' geomorphic evolution and behaviors. Prior research has thoroughly investigated their geomorphic development in the past 5000 years; however, this record is based on 14C and optical dates with inherent chronological uncertainty. In contrast, little is known about the patterns or rates of vegetation expansion and contraction in this period. High-resolution remote sensing data may provide new insights into the spatial conditions of the landscape not available through field data collection. Land cover change mapping provides an estimate of the extent of vegetation change in these environments. The use of black and white historical photographs are difficult for these studies because of their lack of spectral information. Geographic object-based image analysis (GEOBIA) incorporates spatial context into the classification process and can improve classification resolution and accuracy from such images. A GEOBIA technique was applied to estimate the extent of vegetation change from 1938 to 2014 in coastal dune systems along much of the eastern coast of Lake Michigan. Results show that 14 of 16 study sites experienced significant vegetation increase; however, classification accuracies depend on the successful co-registration of the images and land cover heterogeneity. These results provide essential baseline information for maintaining the diversity, quality, and function of this dynamic landscape.
