Using Stats to Protect Oceans

Marine protected areas (MPAs) are areas in the ocean where fishing is either prohibited or tightly controlled to – hopefully – conserve ocean resources and protect species and habitats.

But do MPAs work? And does one type of MPA work better than another? A recent journal article in PNAS addressed that question and found that both no-take (where fishing is prohibited) and multiple-use (fished) MPAs see increases in the numbers of fish.

Shu Yang and Brian Reich are NC State statisticians who worked on this project, which was led by the Duke Marine Laboratory. They responded in writing to questions from The Abstract about the research and the role that statistics play in important projects like this one.

The Abstract (TA): The researchers found that both managed (including some regulated fishing) and protected (fishing prohibited) marine areas can see increases in fish populations over time. Was this result surprising to you?

Yang and Reich: Most existing literature established that Marine Protected Areas (MPAs) with a strict no-take policy would outperform those with less restrictive, multi-use policies, to a large extent, in terms of conservation. However, by applying causal inference methods to adjust for confounding biases, the research demonstrated that well-managed and well-staffed multi-use MPAs could potentially achieve conservation outcomes comparable to no-take MPAs, even in areas likely to be more stressed by human activity. The findings suggest that policymakers aiming to optimize conservation results while also considering socioeconomic factors should focus more on the effective management of MPAs and the local context than the type of MPA.

TA: As statisticians, what was your role in the work?

Yang and Reich: Our role was to assist the team implementing and interpreting a causal inference analysis. The analysis would have been far simpler if parcels of the ocean could be randomized to different policies, but for obvious reasons this is impractical. In reality, policies are determined by practical considerations that may bias simple comparisons. A causal inference tries to address these biases to isolate the true effects of policies.

TA: When you created this analysis, what variables did you take into account?

Yang and Reich: To isolate the effects of the management policies, we tried to remove bias by adjusting for factors that lead to differences in fish populations and where decision makers place MPAs. This includes habitat type, ocean conditions, likelihood of fishing and more.

TA: Did this work raise additional questions that you want to explore; i.e., what would the next steps be in a project like this one?

Yang and Reich: There are many exciting directions to follow. We have studied the effect of a policy in the area where it is implemented, but of course, fish swim! Determining the effects of a management policy in one location on the biodiversity in another is challenging but critical for policy making. We have also established that different types of regions benefit more from conservation, and so using this information to allocate resources efficiently under budget constraints is an area of future work.

This post was originally published in NC State News.