Note that the data published in the 2002 State of the Nation’s Ecosystems Report as well as the 2003 and 2005 Web-Only Updates have been superseded by the 2008 Report and thus should be used with caution. For the most recent data, purchase the 2008 Report from Island Press.

The Indicator

As a means of illustrating the amount of forest providing different degrees of distance from non-forest cover, this indicator provides information on the percentage of forest surrounded by small, medium, and larger “neighborhoods” (defined below) containing at least 90% forest. The “percentage of forest” that meets a certain set of criteria is calculated by determining what fraction of “pixels” (squares of forest 30 meters, or about 100 feet, on a side) is in the center of a “window” that meets the criteria. Thus, the percentage of forest that has 90% or more forest cover within a radius of about 250 feet (the “immediate neighborhood,” about 5 acres) is determined by counting the number of pixels that are in the center of a 5-acre window that contains at least 90% forest.

The Data

Data Source/Collection Methodology: Data for this indicator were prepared by Kurt Riitters, USDA Forest Service (see http://www.srs.fs.fed.us/4803/landscapes/). The data are based on the National Land Cover Dataset, which is described in more detail in the technical note for the national extent indicator. This is a 30-meter resolution remote-sensing-based dataset that provides, among other things, forest/non-forest cover information for the lower 48 states. The unit of data is the pixel, which is a square approximately 30 meters on a side.

Data Manipulation: The data presented here are from a “moving window” analysis. In this approach, the algorithm describes many successive, overlapping “windows” of a certain size, making it possible to characterize the area surrounding each individual forest pixel, in addition to knowing its forest/non-forest status. As the window “moves” across the dataset, each pixel is used as the center of a window; thus, it is possible to determine how many forest pixels are surrounded by different amounts of forest.

Five window sizes were used for this analysis but only three are reported here. The three reported sizes are 2.25 hectares, referred to here as the “immediate neighborhood,” 5 acres, or “within a radius of about 250 feet”; 65.61 hectares, referred to here as the “local neighborhood,” 160 acres, or “with a radius of about one-quarter mile”; and 5314. 41 hectares, referred to here as the “larger neighborhood,” 13,000 acres, or “within a radius of about one and a half miles.” These sizes correspond to 25 pixels (a square of 5 x 5 pixels); 729 pixels (a square of 27 x 27 pixels) and 59,049 pixels (a square of 243 x 243 pixels). The other two window sizes were 7.29 hectares and 590.5 hectares. (Note: This analysis uses a square window, since each remote sensing pixel is square. Thus, the page text description of the “radius” of the “neighborhood” is an approximation to make the presentation clearer to a non-technical audience, and is written as if the window were round.)

The analysis on which the data presented here was based determines, for each pixel and window size, whether it is surrounded by at least 60% forest, at least 90% forest, or exactly 100% forest. For this report, the 90% criterion was chosen. The 90% criterion was selected based on considerations of data quality and previous experience with this analytical approach. The alternate interpretations, along with a detailed description of the methodology, are described in detail in K.H. Riitters et al. (submitted).

Table 3 presents the results of the full analysis, including all window sizes and all three degrees of forest cover. As in the original publication, the table uses the term “core” to refer to areas surrounded by 100% forest cover for the indicated window size, “interior” to refer to areas surrounded by at least 90% forest cover for the indicated window size, and “connected” for areas surrounded by at least 60% forest cover for the indicated window size. Data presented in the body of the report are indicated with an asterisk.

The satellite remote-sensing data presented here can, in theory, distinguish non-forest areas as small as 100 feet on a side (10,000 square feet) from adjacent forest pixels. In practice, the accuracy of doing this depends on the contrast between forest and non-forest land cover, which is, in general, quite good. In addition, geometry plays an important role in distinguishing non-forest land cover. For example, a clearing that fills several 100-foot by 100- foot pixels would probably be more easily detected than a winding road that may fill some pixels and only partially fill others. For further reading on habitat fragmentation, see other related indicators in this document and also Noss and Csuti (1997) and Wilcove et al. (1986).

References

Noss, R.F., and B. Csuti. 1997. Habitat fragmentation, pp. 269–304. In G.K. Meffe and R.C. Carroll (eds.), Principles of conservation biology. Second edition. Sunderland, MA: Sinauer Associates.

Riitters, K.H., et al. Fragmentation of continental United States forests. Submitted to Ecosystems.

Wilcove, D.S., C.H. McLellan, and A.P. Dobson. 1986. Habitat fragmentation in the temperate zone, pp. 237–256. In M.E. Soulé (ed.), Conservation biology: The science of scarcity and diversity. Sunderland, MA: Sinauer Associates.