Knowledge-informed simulated annealing for generating prescribed spatial patterns in resource allocation

There has been considerable recent interest in developing geographic information systems (GIS) applications that incorporate landscape pattern criteria in spatial decision-making. Landscape patterns influence the dynamics of ecosystems, and as such, merit consideration to reconcile other goals and constraints in spatial decision-making. GIS tools capable of analyzing landscape patterns and incorporating pattern information into a spatial decision-support framework for allocating landscape entities have great potential to facilitate achieving pattern goals in land allocation. In this research, I designed, developed, and evaluated a spatial pattern optimization technique called knowledge-informed simulated annealing (KISA) for generating prescribed landscape patterns in single- and multi-objective land-allocation problems and demonstrated the use of such a technique to examine the effect of allocation features on the resulting pattern characteristics.; Two KISA rules, the compactness and the contiguity rule, were developed. They encourage the generation of the prescribed landscape patterns at individual locations through uncoordinated discrete steps and reduce the redundancy in the conventional simulated annealing (SA) algorithm. In single-objective problems, the performance of KISA in solving four problems, each using a distinct pattern metric as the optimization objective to achieve the least fragmented landscape, was examined. These metrics are coarse characterizations of shape compactness, connectivity, and interspersion of distinct homogenous areas on a landscape map. In multi-objective problems, I examined the performance of two approaches, (1) knowledge-informed Pareto simulated annealing and (2) Pareto simulated annealing with partially optimized initial solutions, to generating solutions that approximate the multi-objective Pareto front. Four problems were constructed, each with two objectives, representing all combinations of cases in which there are (1) conflicting or concordant objectives and (2) objectives with similar or different degrees of difficulty. The results of multiple-realization experiments were tested for the variation of performance and the effect of spatial constraints.; KISA improved the performance of SA in solving spatial optimization problems and illustrated the use of spatial optimization techniques for assessing the effect of design features on landscape patterns. The maps generated by the KISA approaches can be used as tools for initiating discourses on subjective goals among planners, decision makers, and publics.