Evaluating the effectiveness of neutral landscape models to represent a real landscape

Neutral landscape models are often employed to represent real landscapes as the null hypothesis. They usually have statistical characteristics similar to real ones. But the spatial characteristics of the real and generated maps are seldom compared. In this study, the neutral landscape models generated by Rule and SimMap are tested against a real forest landscape in Northeastern China. A set of landscape metrics is used for the comparison. Values of some metrics (total number of patches, total perimeter, and aggregation index) suggest that some level of agreement between the maps generated by neutral landscape models and the real landscape do exist at landscape and class levels. But there are also metrics that do not show any agreement between generated maps and the real landscape. Neutral models tend to over-aggregate small classes at higher aggregation levels. Each neutral model has its own strength in representing the real landscape, though neither is perfect. Some metrics, for example, double-logged fractal dimension, are found to have limited capabilities in differentiating landscape structures.