Global effects of marine protected areas on food security are unknown

Abstract

Marine protected areas (MPAs) can be a powerful conservation tool and can positively or negatively affect food security. Sala et al. (2020) estimate the effects of a global network of MPAs designed for biodiversity, carbon sequestration and food security. However, the model used to project these benefits depends on a series of unrealistic and insufficiently tested assumptions that are inconsistent with its source material; using a more realistic model markedly changes the map of priority MPAs and reduces potential food benefits by 62%. This extreme sensitivity in the outcomes of MPA networks to highly uncertain parameters and modelling assumptions means that the true global effects of MPAs on food security remain unknown. We agree with Sala et al.1 that MPAs can have an important role in managing and conserving marine ecosystems. But we are concerned that the model used by the authors does not present a reliable assessment of the effect of MPAs on the yields of fisheries and, by extension, that it is not a reliable foundation for the broader assessment of the role of MPAs in achieving multiple objectives of marine conservation, food security and climate action1. The results of Sala et al.1 depend on the same model as those of a previously published study2 (see also ref. 3), which assumes that density dependence is a function of total pooled population size, independent of how fish are distributed in space, and that unassessed fish stocks (that is, stocks not included in the RAM Legacy Stock Assessment Database) of a given species are a single global interconnected population. These two assumptions generate results that are neither consistent with their source material4 nor ecologically reasonable. The global distribution assumed for unassessed stocks implies that MPAs around Australia can increase catches along the shores of North America3, or that a single fish population can be affected both by MPAs in the Caribbean and in the waters off of China (Supplementary Fig. 2). When movement rates are low under their assumption of pooled density dependence, fishing more outside an MPA can produce higher biomass inside the MPA than there would have been in the absence of any fishing at all (Supplementary Fig. 5). The food projections made by Sala et al.1 are based on estimates of fishing mortality rates and life history values provided by a previous study4. In that study, a Pella–Tomlinson5 population model was used and it was assumed that separate stock units exist inside the waters of a specific country within a major statistical area designated by the Food and Agriculture Organization (FAO) for each unassessed taxonomic group, except for highly mobile unassessed stocks, which are assumed to be well-mixed within FAO major statistical areas. Sala et al.1 aggregated all the individual unassessed stocks assumed by the previous study4 into one global stock per species and converted the underlying population-dynamics model to a logistic growth equation. We call these assumptions made by Sala et al.1 the ‘global’ scenario. To assess the effect of these strong choices, we ran a version of the analysis by Sala et al.1 changing three key assumptions to be consistent with those of the previously published study4: the spatial resolution of the simulated populations, the population-dynamics model used and the nature of the density dependence. For our base results, we assume that density dependence (such as the competition for food or habitat) occurs at a local scale, with MPAs providing a spill-over of fish biomass to fished areas through the movement dynamics in the model. We call this alternative group of assumptions the ‘regional’ scenario. Under the global assumptions, global food production is maximized with an MPA network covering 22% of the carrying capacity, which can be achieved by protecting 24% of the ocean surface. Under the regional assumptions, the maximum yield benefits were much lower; 38% of the maximum benefits of the global assumptions could be achieved by protecting 14% of the carrying capacity (29% of ocean surface) (Fig. 1). The flatter form of the curve for the regional model in Fig. 1a suggests that a greater portion of carrying capacity could be protected without substantially reducing global fishery catches. The global results place much of the west coast of North America in the top 30% of areas for protection, but omit much of the coastal Indian Ocean and the Coral Triangle. These results are flipped under our regional assumptions. The global assumptions of Sala et al.1 suggest that 46% of the exclusive economic zone of the USA could be placed in MPAs while increasing or maintaining food production, whereas under our regional assumptions that number drops to 13% (Fig. 2). The assumption that density dependence occurs at the local scales used in our regional results is common in the MPA modelling literature, including in studies6–11 authored by authors of the study by Sala et al.1. We tested the sensitivity of our regional results to using the same approximation of larvae commonly dispersing outside the MPA to fished areas as Sala et al.1 did; the stark contrast in both the magnitude and design of a global MPA network for food provision remains (see the pooled assumption results in Supplementary Figs. 3 and 4). Fish often disperse vast distances at one or more phases of their life cycle. However, even for the most mobile species, dispersal and complete mixing across entire ocean or planetary scales is rare12. Sala et al.1 used the spatial stock structure described previously13 for the assessed fisheries; the footprints of these stocks are generally much smaller than the entire exclusive economic zone of a country, and of the unassessed fisheries (Supplementary Fig. 6). It is inconsistent to use the smaller footprints13 for the assessed stocks, as Sala et al.1 have done, but then skip past the regional stock structure to the much larger single global stock distribution for unassessed species, as assumed in the global results. The alternative assumption made previously4 that stocks of species that are not highly mobile are contained within country borders is not perfect, but it is more in line with the best available evidence of stock sizes13. We are not suggesting that the regional results are the ‘right’ findings. Instead, we are demonstrating that the central results of Sala et al.1 are not robust to changes to their core assumptions. Other shortcomings remain in both the global and regional scenarios. The spatial complexity of MPAs is simplified to a two-patch surplus production model. The models assume that displacing fishing effort for one species outside an MPA has no effect on other species or habitats in the remaining fished area; these dynamics must be taken into consideration when assessing not only the yield but also the biodiversity and carbon impacts of MPAs. There are places on Earth where MPAs can benefit food production, particularly where stocks are heavily overfished. However, these locations and the resulting effects on food provision cannot be reliably identified using the global-scale model and data used by Sala et al.1. Refinements to their assumptions, in accordance with their own references, do not just alter the results at the margin, but fundamentally change their conclusions at multiple scales. Assessments of the role of MPAs in food provision should be wary of these issues, and clearly evaluate and communicate key sensitivities and potential trade-offs between conservation and food provision arising from alternative sets of plausible assumptions, so that communities can make decisions on MPAs with the full knowledge of both the potential and uncertainty of the effects of MPAs on food security.