Modelling spatial variation in an exploited species across marine reserve boundaries

Abstract

Abstract An increase in exploited species within marine reserves is often expected to benefit adjacent fished areas through the movement or ‘spill‐over’ of exploited species across reserve boundaries. However, harvest at and near reserve boundaries can also decrease abundances inside reserves, forming ‘edge‐effects’. Spatial analysis of gradients in abundance across reserve boundaries have been used to interpret the extents of edge‐effects and spill‐over. While the shape of these gradients will vary depending on species mobility and fishing intensity on reserve boundaries, they have largely been assumed to be gradual and quantified using continuous modelling strategies. This study shows that continuous modelling strategies can lead to overestimations of spill‐over and edge‐effects by underfitting the rapid transition in abundance that can occur from reserve to fished areas. The abundances of spiny lobster (Jasus edwardsii) across the boundaries of two north‐eastern New Zealand marine reserves were quantified using catch survey data collected over 2 years. Several modelling approaches were used, and the predictive performance of each approach was compared. Abrupt changes in lobster abundances were found across reserve boundaries and were best represented with models that incorporated a stepwise function across boundaries or where data from reserve and fished areas were modelled independently. Despite the mobile nature of this species, we found little evidence of edge‐effects and spill‐over within and adjacent to the surveyed marine reserves, which may reflect the low population status within these reserves. In contrast, standard continuous modelling approaches overestimated the extent of edge‐effects and spill‐over in both marine reserves. Policy implications. This study shows how accurate assessments of edge‐effects and spill‐over can be used to evaluate the effectiveness of existing marine reserves in protecting exploited species. These findings emphasise the importance comparing several different modelling strategies when evaluating edge‐effects and spill‐over to ensure appropriate model fits. Although modelling approaches such as generalised additive models present a simple solution to exploring spatial variation in ecological data, we emphasise the importance of questioning the applicability of such strategies when assessing variation across areas of rapid change such as protected area boundaries.