Optimization methodology for land use patterns—evaluation based on multiscale habitat pattern comparison

In this paper, the methodological concept of landscape optimization presented by Seppelt and Voinov [Ecol. Model. 151 (2/3) (2002) 125] is analyzed. Two aspects are chosen for detailed study. First, we generalize the performance criterion to assess a vector of ecosystem functions. This leads to a multidimensional analysis of the results of landscape pattern optimization. Second, we consider how these results relate to an existing landscape presented by some current land use maps. We apply an algorithm of pattern matching, which allows us to compare landscape patterns. With this tool we identify patterns in the study area, which are invariant during optimization and identify the relationships to recent landscape patterns known from data. As expected, the local optimization algorithm does not work well in certain situations that involve strong spatial interactions. We are presenting an example when global optimization cannot be efficiently performed in terms of local variables.