A New Source for Land Cover Change Validation: Wal-Mart from Space

We introduce an event data set of the location and opening dates for 3,043 Wal-Mart stores as a means for validating land cover change-related products at medium (30 m) to coarse (1 km) resolutions throughout the conterminous United States (US). As validation data, these Wal-Mart stores and distribution centers share several favorable attributes, including construction atop a diverse array of vegetated environments, wide dispersion across the entire country, building and parking lot footprints that measure between 100 m and 500 m on a side, and construction dates that span much of the remote sensing record (1964-2005). To generate the data set, we geo-coded the full Wal-Mart store address list, combined these locations with a listing of Wal-Mart store opening dates, and geo-located the building footprints of 30 Wal-Mart stores using cost-free high-resolution (4 m) imagery available from internet search engines. Twenty-five stores constructed in North Carolina and Virginia between 1987 and 2002 served to validate a single scene (WRS2 p16 r035, 180 km per side) of the new Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) product - a 28.5 m resolution forest disturbance map which is in production for the conterminous US. Disturbance events were clearly discernable in the LEDAPS beta product at all 25 of the validation sites. In addition, we selected five Wal-Mart sites constructed between 2000-2005 in Maine, North Carolina, Oklahoma, and California to validate the University of Maryland?s 250 m Moderate Resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index 16-day time series (MOD44C). These five construction events are evident in the time series. At a Wal-Mart distribution center in Gordonsville, Virginia, a similar construction signature is present at 1 km resolution for the MOD13A2 enhanced vegetation index 16-day time series. These results demonstrate a new approach for validating land cover change related products by combining an unusual disturbance event data set with free high-resolution internet-based images.