SecSel, a new software tool for conservation prioritization that is applicable to ordinal-scale data for multiple biodiversity features

are not necessarily among the sites that most effectively increase the number of features that fulfill the top-N included criteria. The number of candidate sites, including the less efficient but less costly sites, reflects the emphasis on cost reduction. The larger the number of candi- date sites, the greater the odds that a less efficient site may be selected because of its small cost. This process may increase the number of sites required to achieve the targets of all fea- tures. The final set of sites is not necessarily the set of sites that minimizes the total cost irre- spective of the number of candidate sites. This result is an inevitable consequence of a heuristic algorithm. Selecting less-scattered sites. A less-scattered set of sites can be selected in SecSel. SecSel implements two methods for this purpose. One of the methods considers boundary length. During the process of site selection, the total boundary length of the set of selected sites changes. It typically, but not neces- sarily, increases. The change in length by addition of a site depends on the geometrical relationship of the additional and already-selected sites. Addition of a site that shared a boundary with already-selected sites would increase the total boundary length less than addition of an isolated site. Thus, by considering the change of the total boundary length, sites adjacent to already selected sites are more likely to be selected. To break ties in the selection of sites based on the number of features newly fulfilled in the top-N included

criterion, the site that causes the smallest increment to the total boundary length is selected. The other method considers the distance of a site to the nearest site among the other selected sites In this method, the distance to the nearest selected site is used to break ties in the same way that the boundary length method is used in the site-selection process. The site that lies the shortest distance from the selected site is chosen. The merit of this method is that it is applicable in a case when the sites naturally do not share a common boundary (e.g., each site is a different pond or a fragmented patch of forest). This method helps to avoid selecting isolated sites that are distant from already-selected sites, and it is expected to reduce the maximum of “the distance to the nearest sites” within the conservation area. Consideration of both cost and scattering. In the consideration of both the cost and scat- tering of sites, scattering is the first criterion used to break ties at each step of site selection. If there are still ties after consideration of scattering, then cost is considered. In the early stages of the site-selection processes, sites that are highly efficient with respect to fulfilling the top-N inclusion requirement are not likely to be adjacent to each other as long as there is spatial auto- correlation in the distribution of each feature. Neighboring sites are likely to have more similar compositions than randomly selected sites, and they are less likely to complement each other.

Scattering is therefore not likely to determine which of the tied sites should be selected. If there are differences in costs among sites, those differences could decide which site is selected. As more and more sites are selected, it becomes increasingly likely that candidate sites will share a boundary with or close to already-selected sites. The protected area thus expands by adding adjacent sites to low-cost, previously selected sites. If there is spatial autocorrelation of costs, the costs of the sites added to the low-cost sites are also expected to be low. The protected area is therefore likely to start from one or a few core areas in a low-cost region; adjacent sites are then added to expand the core areas. The final result is a less-scattered set of selected sites in low-cost regions. Consideration of costs and scattering may lead to a protected area with a relatively large number of sites after the heuristic selection process. The number of sites varies between cases. If there are relatively large numbers of highly ranked local units with numerous features, there is more flexibility in site selection to achieve targets, and consideration of costs and scattering would have less of an effect on the final size of the protected area. PLOS ONE SecSel, a new software tool for conservation prioritization applicable to ordinal-scale data PLOS ONE | https://doi.org/10.1371/journal.pone.0247737 July 23, 2021 7 / 19