Methods

Researchers with the United States Geological Survey's (USGS) Land Cover Trends Project created a dataset for the conterminous U.S. designed to characterize the historical state of the nation's land surface between 1973 and 2000. The dataset was needed to analyze patterns, rates, and trends in land-use/land-cover (LULC) change and to assess the causes and potential consequences of land change across the country (Loveland and others, 2002). The dataset shows the geographic variability and characteristics of 1973-2000 U.S. landscape change and provides a scientific foundation for assessing the environmental consequences of land change. Over 60 scientific papers have been published and the results have been the basis for collaborative studies of the environmental consequences of change with scientists from many organizations including the U.S. Geological Survey, U.S. Environmental Protection Agency, and National Aeronautics and Space Administration, National Oceanic and Atmospheric Association, National Science Foundation, U.S. Fish and Wildlife Service, and academia.

The U.S. has a number of programs that use statistical surveys and remote sensing approaches to generate land cover information. The Forest Inventory and Analysis Program of the U.S. Forest Service provides the information needed to assess status and trends in America's forests (Gillespie, 1999). The U.S. Department of Agriculture's Census of Agriculture is the leading source of facts and figures about American agriculture (USDA, 2009). The U.S. Department of Agriculture's National Resources Inventory is a statistical survey of land use and natural resource conditions and trends on non-Federal lands (USDA, 2009). And only recently, the USGS's National Land Cover Database became a source of 30-meter resolution, Landsat-based, spatial land cover data for the Nation, containing layers for thematic land cover, percent developed imperviousness, and percent tree canopy density (Fry and others, 2011). Each effort contributes to our understanding of the land use and land cover, but none offered a complete, nationally comprehensive assessment of 1973 to 2000 land change based on methods that are spatially and temporally consistent across the U.S.

The dataset was created using a statistical sampling approach because it was a much more cost efficient method for characterizing LULC change over as large an area as the U.S. (Stehman and others, 2003). Ecological regions were used as a geographic framework for selecting sample blocks across the conterminous U.S. with a total of 2,688 sample blocks randomly selected from 84 Level III ecoregions (Omernik, 1987). Researchers manually interpreted Landsat MSS, TM and ETM+ imagery for five dates (1973, 1980, 1986, 1992, and 2000) and then used the land cover data to derive change statistics for each interval (1973-1980, 1980-1986, 1986-1992, and 1992-2000) and for the entire study period (1973-2000), ultimately serving as the basis for ecoregion-based change estimates used as the primary land-change metrics in reporting.

The change estimates are used to determine: (1) the predominant types of LULC conversions occurring within each ecoregion, (2) the estimated rates of change for these conversions, and (3) whether the types and rates of change are constant or variable across time. The analysis of change, which was conducted in several ecoregion publications (Sleeter and others, 2012), also involved looking for spatial correlations between conversion types and selected environmental factors, such as terrain characteristics, proximity to urban development, economic conditions, etc., in order to improve our understanding of potential drivers of change (Raumann and Soulard, 2007; Napton and others, 2010).

The Land Cover Trends project was first proposed by Loveland and others (1999), and the project methodology was later published by Loveland and others (2002). Additional details on sampling design were provided by Stehman and others (2003). Sleeter and others (2012) also provided a detailed description of each facet of the Land Cover Trends methodology. Here, we make a similar attempt to summarize the ecoregion framework, sampling strategy, source imagery, classification system, and manual interpretation process.

Map showing the locations of the 2,688 sample blocks that comprise the Land Cover Trends dataset. The red squares indicate locations of the 10 km x 10 km and/or 20 km x 20 km sample blocks. The base map is the 1992 National Land Cover Dataset (Vogelmann and others, 2001).

An example of the 5 dates of land-use and land-cover data available for each sample block. The data shown here is for sample 74_0134 in the Mississippi Valley Loess Plains ecoregion, located on the northeastern edge of Memphis, Tennessee. Note-the Landsat imagery are not available for download as part of this paper.