This dataset is a grid map for an area bounding the Delaware River Basin at 30 meter resolution. Each pixel value represents an index of forest fragmentation for the surrounding 590.49 ha.(81 x 81 pixel) analysis window. The dataset was created by combining published maps of forest area density, and forest area connectivity, from the National Land-Cover Pattern Database (NLCPD), and a published map of land-cover from the Multi-Resolution Land Characteristics (MRLC) Consortium. One of twelve categories of fragmentation was identified by using a model similar to Riitters et al. (2000). The data set describes one aspect of forest fragmentation at one scale. The forest fragmentation index is designed to distinguish among types of fragmentation (e.g., edges on the interior versus the exterior of a forest patch) and it also reflects differences in the absolute amount of forest present. No distinction was drawn between "natural" and "human-caused" fragmentation.
These data were created as part of a national analysis of land cover pattern. A series of derived maps (different indices, different window sizes) is available. They are proposed for consideration by landscape analysts, as independent variables (e.g., covariates, stratification rules), and as "context" information for finer-scale investigations, for example field plots. Caveats include: Maps in this series are naturally correlated since all came from one map. No warranty or claim is made of the utility of this map for any particular purpose, this is considered to be a research dataset. The forest fragmentation index map portrays considers only forest versus non-forest cover types.
Dates of satellite imagery.
None. Acknowlegement of the U.S. Forest Service would be appreciated in products derived from these data. To support further development of the database, please send a postcard or reprint.
11 Campus Boulevard, Suite 200
The basic data from which this map was ultimately derived were provided by the Multi-Resolution Land Characteristics Consortium, hosted by the EROS Data Center.
There is no information about the accuracy of of the forest fragmentation index. The statistical accuracy of the map is related to the statistical accuracy of the maps from which it is derived (see: Source_Information sections). This index is one of several which could have been used instead, and standard textbooks discuss relative strengths and weaknesses. Like other indices, this index is best interpreted as a relative value, in comparison to other pixels on the same map. Comparisons with other maps prepared from different data sources or at different scale are tenuous.
All cell values are in the range [0,14]. A code (9) was assigned to water, ice, or bare rock types (on the MRLC land cover map). A missing value code (0) was assigned to locations which were outside the extent of the MRLC land cover map. A missing value code (10) was assigned to locations for which the classification model failed to assign a fragmentation index value.
Every non-null cell in the original land cover map has a value representing the index or a missing data value.
The positional accuracy of the derived map is the same as the positional accuracy of the MRLC land- cover map.
Forest connectivity index map derived from the MRLC land cover map.
Forest area density index map derived from the MRLC land cover map.
Derivation of forest fragmentation index from forest area density, forest connectivity, and land cover: Forest area density (FDEN81) and forest connectivity (FRAG81) were obtained from the NLCPD in binary format. Each pixel value represents a continuous [0,1] variable with byte values in [1,255] with zero (0) representing a missing value. Larger values represent more forest and higher connectivity, respectively. Note that the two maps were derived from the MRLC land cover map, and that "woody wetland", and "upland forest" were considered to be "forest" when computing the index values in Edition 4 of the NLCPD. Also note that the MRLC types "water", "ice", and "bare rock" were considered as missing values for the density and connectivity measurements. This means that fragmentation was not "caused" by those land cover types in this analysis; they were ignored in calculations. Land cover was obtained from the MRLC in binary format. Each pixel value represents one of 21 land-cover types in [1,92] with zero (0) representing a missing value. The three images were overlaid, pixel by pixel, using an in-house software tool named FRAGCL.C to assign a fragmentation index value. The classification model identifies twelve fragmentation categories, two missing value categories, and one water/rock category: >(0) Missing_1, if MRLC is missing (0). >(1) Edge_1, if FRAG81 >= FDEN81, and if 153 < FDEN81 < 230. >(2) Perforated_1, if FRAG81 < FDEN81, and if 153 < FDEN81 < 230. >(3) Transitional_R, if 101 < FDEN81 < 154, and > if FRAG81 >= FDEN81. >(4) Transitional_L, if 101 < FDEN81 < 154, and > if FRAG81 < FDEN81. >(5) Patch_R, if FDEN81 < 103 and if FRAG81 >= FDEN81. >(6) Patch_L, if FDEN81 < 103 and if FRAG81 < FDEN81. >(7) Interior, if FDEN81 = 255. >(8) Non-forest, if FDEN81 = 1. >(9) Water/rock, if MRLC is either water(11), ice(12), or bare rock(31). >(10) Missing_2, if FRAG81 is missing (0) and FDEN81 is > not missing (> 0). >(11) Edge_2, if FRAG81 >= FDEN81, and if 242 < FDEN81 < 255. >(12) Perforated_2, if FRAG81 < FDEN81, and if 242 < FDEN81 < 255. >(13) Edge_3, if FRAG81 >= FDEN81, and if 229 < FDEN81 < 243. >(14) Perforated_3, if FRAG81 < FDEN81, and if 229 < FDEN81 < 243. For pixels that are have defined values for land cover, forest area density, and forest connectivity, the classification model partitions the two-dimensional space defined by FRAG81 and FDEN81 into twelve fragmentation categories. The category is then changed to water/rock if the MRLC land-cover map is water, ice, or bare rock. Note that the critical values 102 and 153 represent the proportions 0.4 and 0.6, respectively, before the FDEN81 and FRAG81 values were converted from continuous [0,1] values to integer [1,255] values. The "undetermined" category in Riitters et al. (2000) is considered to be "edge". The "header file" information for the derived map is as follows. This information is needed for some image import filters. Ulxmap and Ulymap refer to the center of the upper-left pixel.
Data from De, Md, Nj, Ny and Pa were merged and cut to a window that encompasses the Delware River Basin and reprojected to Albers.
11 Campus Boulevard, Suite 200
The value is a coarse-scale measure of forest fragmentation in the area surrounding the grid cell. Fragmentation is described by twelve classes and there are two missing value classes and one water/rock class.
U.S. Forest Service
Missing_1
U.S. Geological Survey
Edge_1 - 60-90 percent of the cells in the surrounding area are forested and this cell appears to be part of the outside edge of a forest patch.
U.S.Forest Service
Perforated_1 - 60-90 percent of the cells in the surrounding area are forested and this cell appears to be part of an inside edge of a forest patch. In other words, this cell is near a non-forest inclusion within a forest patch.
U.S. Forest Service
Transitional_R - about half of the cells in the surrounding area are forested and this cell may appear to be part of a patch, edge, or perforation depending on the local forest pattern. Forest connectivity is greater in comparison to the Transitional_L class.
U.S. Forest Service
Transitional_L - about half of the cells in the surrounding area are forested and this cell may appear to be part of a patch, edge, or perforation depending on the local forest pattern. Forest connectivity is lower in comparison to the Transitional_R class.
U.S. Forest Service
Patch_R - most of the cells in the surrounding area are not forested and this cell is part of a forest inclusion or patch of forest on a non-forest background. Patches tend to be larger in comparison to the Patch_L class.
U.S. Forest Service
Patch_L - most of the cells in the surrounding area are not forested and this cell is part of a forest inclusion or patch of forest on a non-forest background. Patches tend to be smaller in comparison to the Patch_R class.
U.S. Forest Service
Interior - all of the cells in the surrounding area are labeled as forest in the land-cover map.
U.S. Forest Service
Non-forest - none of the cells in the surrounding area are labeled as forest in the land-cover map.
U.S. Forest Service
Water/rock - the cell was water, ice, or bare rock on the land-cover map.
U.S. Geological Survey
Missing_2 - the forest connectivity index was missing and the forest area density index was not missing.
U.S. Forest Service
Edge_2 - >95 percent of the cells in the surrounding area are forested and this cell appears to be part of the outside edge of a forest patch.
U.S.Forest Service
Perforated_2 - >95 percent of the cells in the surrounding area are forested and this cell appears to be part of an inside edge of a forest patch. In other words, this cell is near a non-forest inclusion within a forest patch.
U.S. Forest Service
Edge_3 - 90-95 percent of the cells in the surrounding area are forested and this cell appears to be part of the outside edge of a forest patch.
U.S.Forest Service
Perforated_3 - 90-95 percent of the cells in the surrounding area are forested and this cell appears to be part of an inside edge of a forest patch. In other words, this cell is near a non-forest inclusion within a forest patch.
U.S. Forest Service
11 Campus Boulevard, Suite 200
Although these data have been processed successfully on a computer system at the U.S. Forest Service, no warranty expressed or implied is made by the USFS regarding the utility of the data on any other system, nor shall the act of distribution constitute any such warranty.
Box 12254 3041 Cornwallis Road