We present a new approach to spatial prioritization in systematic conservation planning that integrates adequacy and representativeness into the design of protected area networks. Adequacy refers to the persistence of biodiversity, while representativeness measures how well a network samples (surrogates of) biodiversity. Commonly, optimization methods are used to identify representative networks, with modifications to size, intactness and connectivity designed to enhance adequacy. Our approach is based on operationalizing the concept of ecological benchmark, by constructing candidate reserves from connected sets of hydrological catchments that satisfy adequacy criteria. The results are then assembled into network solutions that satisfy total area targets and that achieve high representation globally Representation is assessed by multivariate distance measures on distributions of environmental surrogates within a network solution and a chosen reference area. Large random samples of networks are generated and ranked by representation, the most representative can be considered as set of equivalently good possible solutions for input to the full planning process. We also developed a new index of site (catchment) irreplaceability based on the Harmonic number,suitable to this sampling procedure.
We demonstrate our approach in the context of Quebec’s Moyen Nord, a 688 000 km2 area of boreal forest in Canada, where the provincial government plans to set aside 50% of the region for conservation purposes, with 20% strictly protected. At present just over 8% of the region is protected. We contrasted two sets of designs: de novo (S1), and designs that augmented the existing network by adding new benchmarks to reach the areal target (S2). In all cases, the benchmark construction was constrained to exclude catchments with any mapped human disturbances e.g. by roads, hydroelectric installations or forest harvesting. S2, which included existing protected areas, achieved the best representation, while including important hotspots of irreplaceability revealed under S1. The reason for S2's better performance was the presence latitudinal gradients in both climate and vegetation, but also in the degree of human disturbance. Intensive forest harvesting in the precluded benchmark construction, but the existing network of small protected areas did enhance overall ecological representation. Despite these two gradients, both S1 and S2 produced solutions more representative of the region than is the existing protected areas network. The advantage of our method is that, by construction, it achieves representation simultaneneously with other important characteristics such as connectivity, intactness and large size.