Exact integer linear programming solvers outperform simulated annealing for solving conservation planning problems

Richard Schuster; Jeffrey O Hanson; Matthew Strimas-Mackey; Joseph R Bennett

doi:10.7717/peerj.9258

Exact integer linear programming solvers outperform simulated annealing for solving conservation planning problems

PeerJ. 2020 May 27:8:e9258. doi: 10.7717/peerj.9258. eCollection 2020.

Authors

Richard Schuster^{1

2}, Jeffrey O Hanson³, Matthew Strimas-Mackey⁴, Joseph R Bennett¹

Affiliations

¹ Department of Biology, Carleton University, Ottawa, ON, Canada.
² Ecosystem Science and Management Program, University of Northern British Columbia, Prince George, BC, Canada.
³ School of Biological Sciences, University of Queensland, Brisbane, QLD, Australia.
⁴ Cornell Lab of Ornithology, Cornell University, Ithaca, NY, USA.

Abstract

The resources available for conserving biodiversity are limited, and so protected areas need to be established in places that will achieve objectives for minimal cost. Two of the main algorithms for solving systematic conservation planning problems are Simulated Annealing (SA) and exact integer linear programing (EILP) solvers. Using a case study in BC, Canada, we compare the cost-effectiveness and processing times of SA used in Marxan versus EILP using both commercial and open-source algorithms. Plans for expanding protected area systems based on EILP algorithms were 12-30% cheaper than plans using SA, due to EILP's ability to find optimal solutions as opposed to approximations. The best EILP solver we examined was on average 1,071 times faster than the SA algorithm tested. The performance advantages of EILP solvers were also observed when we aimed for spatially compact solutions by including a boundary penalty. One practical advantage of using EILP over SA is that the analysis does not require calibration, saving even more time. Given the performance of EILP solvers, they can be used to generate conservation plans in real-time during stakeholder meetings and can facilitate rapid sensitivity analysis, and contribute to a more transparent, inclusive, and defensible decision-making process.

Keywords: Conservation planning; Integer linear programming; Marxan; Optimization; Prioritization; Prioritizr.

Grants and funding

Richard Schuster is supported by a Liber Ero Fellowship and Environment and Climate Change Canada (ECCC), Jeffrey O Hanson by ECCC, Matthew Strimas-Mackey by endowments at the Cornell Lab of Ornithology, and Joseph R. Bennett by Natural Sciences and Engineering Research Council of Canada and ECCC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.