Limitations.–All measures based on the adjacency matrix (i.e., the number of adjacencies between each pair of patch types) that include like-adjacencies (i.e., percentage of like adjacencies, clumpiness index, aggregation index, and contagion) are strongly affected by the grain size or resolution of the image. Given a particular patch mosaic, a smaller grain size will result in a proportional increase in like adjacencies. Given this scale dependency, these metrics are best used if the scale is held constant. Note, interspersion is not affected by resolution directly because only patch edges are considered. In addition, there are alternative ways to consider cell adjacencies. Adjacencies may include only the 4 cells sharing a side with the focal cell, or they may include the diagonal neighbors as well. FRAGSTATS uses the 4-neighbor approach for the purpose of calculating these metrics. Further, there are at least two basic approaches for counting cell adjacencies, referred to as the single count and double count methods. As noted above, FRAGSTATS adopts the double count method in which pixel order is preserved. In this method, all non-background cells inside the landscape (i.e., positively-valued cells ) are visited and the four sides of each cell are tallied in the adjacency matrix. As a result, all cell sides involving non-background classes inside the landscape are tallied twice (hence the term double count), but all cell sides involving background or landscape border (i.e., negatively-valued cells) are only counted once, as those cells are not themselves visited. Finally, mass fractal dimension and lacunarity involve the use of moving windows of many sizes; these can be computationally demanding and for large landscapes may take a very long time to compute.

Metrics based on fractal geometry such as mass fractal dimension are subject to several limitations (as discussed by Hargis et al. 1998). First, the simplifications of landscape pattern produced during the mapping process yield images that are not truly fractal, and the application of fractal measures in the strictest sense is therefore questionable. Fractal geometry assumes that the quantity being measured and the ruler length of measurement have a linear relationship when both are logarithmically transformed (Voss 1988). If this condition is not met, the object (in this case, the landscape) may not be fractal. Any smoothing or renormalization of landscape patterns during the mapping process may cause this relationship to deviate from linearity, and the use of fractal dimension may be questionable. Second, landscape extent and grain affect the ability to derive an accurate scaling relationship between box size and mass. A large ratio of extent to grain is needed to produce a reasonable range of box sizes for generating an accurate slope. Finally, the relationship between grain and average patch size can affect the accuracy and meaningfulness of the derived fractal dimension. If map resolution allows only two or three box sizes before the sample box exceeds the size of the average patch, the slope derived from these points is questionable. In this case, when larger box sizes are added to the regression line, any interesting differences are lost in the averaging process.

The subdivision metrics based on the cumulative patch size distribution are essentially free of any known limitations. Perhaps the greatest limitation is that they have not yet been used extensively by landscape ecologists, so their behavior under various conditions has not been fully explored. In addition, it is important to keep in mind the similarities and differences between these metrics and the closely related concept of contagion.