Structural diversity is an important and difficult to measure landscape characteristic. It cannot be measured directly but is described by indices, that base upon measurable attributes such as number, size, shape and edge-length of individual areas and distance between areas.
On the basis of a sample map the study analyses four principal properties of indices: Their ability to differentiate between obviously different structures, their mutual statistical independence, their scale invariance and the possibility to estimate them from samples.
The indices LPI, LSI and MSI proved to be suitable for an analysis of spatial structure. The fractal indices MPFD and DLFD showed less ability to differentiate between different structures.
Unbiased sample-based estimates (using square windows of the entire map) are feasible only for indices which are composed of components which can be estimated without bias themselves (absolute area, length of edge). Also the fractal area/perimeter dimension gave good estimates in the sampling study.