Pattern-Based Assessment of Land Cover Change on Continental Scale With Application to NLCD 2001-2006

We present a method for assessing land cover change on continental scale and with high spatial resolution. This is a post-classification method, but instead of tracking transitions of land cover classes on cell-by-cell basis the method measures the change at a tile level by quantifying a difference between local patterns of land cover at two different time steps. Pattern-based change assessment is well suited for the large scale survey as it addresses landscape dynamics rather than just simple land class transitions. A tile is defined as a local area consisting of large enough number of land cover cells to sample a distribution of landscape but small enough to detect change with high spatial resolution; 4.5 km× 4.5 km square tiles are used. The level of change is measured as the dissimilarity between motifs of tile patterns at two time steps and is calculated using information-theoretic metric called the Jensen-Shannon similarity. The method is able to discriminate between different types of change including change in geometric pattern, change in class composition, and numerous class transitions without significant changes in either pattern or composition. The methodology is applied to the National Land Cover Dataset (NLCD) to obtain a 2001-2006 change map of the conterminous U.S. The resultant map shows (in a high resolution of 3 km/cell) a spatial distribution of the degree to which the landscape has changed in this time period. Both, large regions (southeastern and Gulf regions, Pacific Northwest region, and the state of Maine) of heightened landscape dynamics, as well as small regions of sudden change due to fires, urban growth etc. are clearly identifiable from the map. A fully-featured online application for fast and convenient exploration of the change map together with original land cover maps in their full resolutions is available at http://sil.uc.edu/dataeye/.