Climate change reduces long‐term population benefits from no‐take marine protected areas through selective pressures on species movement

Marine protected areas (MPAs) are important conservation tools that confer ecosystem benefits by removing fishing within their borders to allow stocks to rebuild. Fishing mortality outside a traditionally fixed MPA can exert selective pressure for low movement alleles, resulting in enhanced protection. While evolving to move less may be useful for conservation presently, it could be detrimental in the face of climate change for species that need to move to track their thermal optimum. Here, we build a spatially explicit simulation model to assess the impact of movement evolution in and around static MPAs resulting from both fishing mortality and temperature‐dependent natural mortality on conservation benefits across five climate scenarios: (i) linear mean temperature shift, (ii) El Niño/La Niña conditions, (iii) heat waves, (iv) heatwaves with a mean temperature shift, and (v) no climate change. While movement evolution allows populations within MPAs to survive longer, we find that over time, climate change degrades the benefits by selecting for higher movement genotypes. Resulting population declines within MPAs are faster than expected based on climate mortality alone, even within the largest MPAs. Our findings suggest that while static MPAs may conserve species for a time, other strategies, such as dynamic MPA networks or assisted migration, may also be required to effectively incorporate climate change into conservation planning.

small areas-less than half the total movement extent of an individual-are effective for ensuring population persistence (Green et al., 2015).While under certain circumstances MPAs can promote climate change adaptation and enhance biodiversity protection in a changing seascape (Jacquemont et al., 2022;Roberts et al., 2017), climate change is overwhelming and will continue to threaten the effectiveness of MPAs (Bruno et al., 2018;Hannah et al., 2007), specifically through impacts on species distributions (Fredston-Hermann et al., 2018;Pecl et al., 2017;Thuiller, 2004).Most of the world's MPAs are set to be established in the next decade through calls to protect 30% of the ocean by 2030 (Convention on Biological Diversity, 2021).With up to 76% of existing MPAs expected to experience novel climatic conditions by the end of the century (Johnson & Watson, 2021), understanding how population persistence within MPAs interacts with climate change is a global imperative to advance biodiversity conservation.
Protecting fish from fishing using MPAs can result in the selection of low movement traits in fish populations-thereby decreasing genetic diversity-particularly when fishing mortality outside the MPA is high (Mee et al., 2017b;Pinsky & Palumbi, 2014).As fishing mortality increases, so too does selective pressure to move less within MPAs (Mee et al., 2017b), resulting in increased residency (Parsons et al., 2010) and decreased home range sizes (Thorbjørnsen et al., 2021) within the MPA.While evolving to move less is useful for conservation efforts in the present (Jiao et al., 2018;Mee et al., 2017b), it could be strongly detrimental in the face of climate change, which will require species to move potentially quickly to track their thermal tolerances.The resulting "protection paradox" (Bates et al., 2019)-that is, when one pressure selects for traits that are deleterious in the face of other pressures-underscores the need for climate change specific solutions to MPA planning and design.
Marine species are currently expected to track climate change well because of their average high mobility relative to species on land (Lenoir et al., 2020).However, strong protection within MPAs may select for reduced movement that protects individuals from enhanced fishing mortality in nearby areas outside the protected area, but these benefits from restricted movement may simultaneously compromise the species' ability to keep pace with shifting climate regimes.While increased dispersal rates are predicted in some cases (Kokko & Lopez-Sepulcre, 2006), particularly near range edges (Nadeau & Urban, 2019), as an evolutionary response to climate change (Catullo et al., 2019;Palumbi et al., 2014;Stockwell et al., 2003), understanding movement evolution dynamics within MPAs facing rapid climate change is critical to advancing effective climate conscious MPA design.To advance conservation planning in a world affected by climate change, new frameworks for MPA design that incorporate climate change adaptation have been developed (Grorud-Colvert et al., 2021;Tittensor et al., 2019;Wilson et al., 2020).Current approaches for "climate smart" MPAs include either selecting areas with low rates of climate change, considered "climate refugia" (e.g., Arafeh-Dalmau et al., 2021;Brito-Morales et al., 2022;Buenafe et al., 2023), or protecting areas with high historical sea surface temperature variability (e.g., Green et al., 2014).
Other strategies advocate for protecting a representative portion of habitat and/or functional groups (e.g., Colton et al., 2022;McLeod et al., 2009).Collectively, current climate change specific MPA guidance considers broad habitat and community level responses to changing ocean conditions; however, there is a lack of understanding of the many ways climate change and other factors can interact to influence MPA effectiveness.For example, studies allowing for thermal tolerance evolution suggest that targeting only refugia may not fully protect thermally adapted individuals (Colton et al., 2022;Walsworth et al., 2019).More research is required, especially regarding eco-evolutionary feedbacks that can affect both the pace and magnitude of responses (Gil et al., 2020), to fully understand the viability of MPA design under climate change.
This study aims to understand how movement evolution in fish impacts MPA effectiveness under climate change.While MPAs are expected to reduce movement extents in fish experiencing relatively high fishing mortality outside the MPA (Mee et al., 2017b), climate change can increase dispersal rates on species range edges (Kokko & Lopez-Sepulcre, 2006;Miller et al., 2020;Nadeau & Urban, 2019;Ochocki & Miller, 2017).Our goal is to evaluate how these seemingly conflicting selection pressures-that of fishing and climate change within the context of MPAs-compete to determine species' movement extents and consequences for population persistence (Figure 1; Table S1; Methods).Our findings can be used to inform design and evaluation strategies for MPAs in climate change and identify other relevant conservation interventions.
We explore movement evolution dynamics by extending a framework developed by Mee et al. (2017aMee et al. ( , 2017b)), which presents a spatially explicit model to assess movement evolution within MPAs.Mee et al. (2017b) focused on highly mobile species and found that skipjack tuna, bluefin tuna, and dogfish experience low movement evolution within an MPA when fishing mortality is high.Here, we assess population changes and evolutionary dynamics within the MPA for species that move shorter distances (2-6 km year −1 ) and are faced with climate change impacts, modeled as temperature based natural mortality (adapted from Norberg et al., 2012;Figure S1) under five climate scenarios: (i) linear mean temperature shift (Mean Shift), (ii) El Niño/La Niña conditions with cyclical variation between temperature in years (El Niño/La Niña), (iii) heat waves at random intervals and intensities (Shocks), (iv) heatwaves with a mean temperature shift (Shocks with Mean Shift), and (v) no climate change (Null).We examine multiple MPA sizes to contextualize our results and determine how climate change may impact the effectiveness of various MPA designs when considering movement evolution.

| Simulation overview
We build off an approach presented by Mee et al. (2017aMee et al. ( , 2017b) ) to track eco-evolutionary changes in species movement resulting from MPA implementation and extend their work to further understand the interacting effect of climate change on movement evolution.
Our simulation consists of a 100 ("latitude") × 20 (longitude) grid, where each cell represents 2 km × 2 km.Our grid is larger in the vertical direction to capture latitudinal shifts in species space use due to climate change; expanding the grid horizontally (i.e., longitudinally) would not affect results, but would increase simulation run  S1.
time and compute power required for the analysis.Each simulation consists of 10 replicates (Figure S10-15) for each combination of climate scenario (including no climate change), MPA size, and with and without movement evolution (Figure 1).We model three age classes representing larvae, juveniles, and adults and both sexes.The simulation is initialized with 600 fish (100 per sex and age class), with genotypes distributed using Hardy-Weinberg equilibrium with the initial frequency of the low movement (a) allele set to 0.3.We use a higher initial frequency of the low movement allele (a) compared to Mee et al. (2017b), which set the initial frequency equal to 0.1, because our analysis is designed to specifically track evolutionary changes resulting from climate change on the low movement allele in both directions, as opposed to the increases due to protection that were the focus of Mee et al. (2017b).The selection and deselection of alleles is driven by differential mortality for both fishing and temperature-dependent natural mortality for different genotypes based on where they move within the simulation.The simulation is run in three phases, with subsequent phases initiated once equilibrium is reached in the prior phase: burn in prior to fishing and MPA implementation (Phase 1; 10 years to equilibrium), fishing only (Phase 2; 15 years to equilibrium), and post MPA implementation (Phase 3; 150 years to capture long-term climate change).All work was performed in R (version 4.1.3).
For each phase, the simulations proceed in four steps, each of which represents 1 year.Extended details can be found in Mee et al. (2017b) and in the "Extended methods" (including model equations, Data S1).First, the fish spawn.The number of eggs produced in each patch is drawn from the Poisson distribution with a mean of fecundity times the number of females in the patch.If a male is present in the patch, all eggs are assumed to be fertilized and distributed to genotypes first using a binomial distribution with a mean based on the allele frequencies within the males of that patch then between sexes with a binomial distribution with probability 0.5.Second, juveniles recruit to the adult age class and mortality is calculated.A binomial distribution with probability 0.33 (1/maturity.age; Table S1) determines the number of juveniles that become adults.The natural mortality of all life stages (s for juveniles and adults, s b for larvae/fry) is determined using Equation (1), which is temperature dependent and modified from Norberg et al. (2012) to take the minimum of the temperature based mortality and the natural mortality (Figure S6).
After experiencing natural mortality, fishing mortality is calculated and applied for the climate change scenarios.After MPA implementation, fishing pressure is redistributed to areas outside the MPA based on the MPA area.A patch's fishing mortality is dependent on the patch's population and based on a hyperbolic function that distributes fishing pressure such that patches with higher populations are fished more heavily.The probability of being fished (p) is determined by that patch's individual fishing mortality and uses a binomial distribution to determine the number of fish harvested from each patch.Lastly, fish move around the simulation grid.Movement extents for each fish are drawn from a negative binomial distribution with mean distance based on the genotype.The angle of movement is drawn from the uniform distribution, and genotype-informed distances are converted to number of grid cells, assuming each fish starts in the middle of its grid cell, to determine the new location of each fish.We do not use an edge correction, and instead fish "hitting" the edge of the grid bounce back into the simulation grid along the same trajectory.

| Model parameterization
Life history parameters (Table S1) for 47 fish species with movement extents between ~2 and 6 km, which was determined reasonable for the simulation grid, were obtained from FishBase (Froese & Pauley, 2022; Tables S2 and S3).For our simulations, we use the median value across all species for each life history parameter as a representative statistic.Example species with similar life history values include the bonefish (Albula vulpes), the sky emperor (Lethrinus mahsena), and the great barracuda (Sphyraena barracuda).Sensitivity analyses were used to set representative values for density dependence, movement pattern, and fishing mortality and to assess the effects of variation within other life history and climate parameters (Data S1; Figures S7-S13; Table S1).Climate values were chosen from primary literature and data sources.Further details are provided in the "Extended methods" (Data S1).

| Climate scenarios
The initial temperature for the grid wass set between the thermal optimum at high (cold) latitudes and the thermal optimum +2°C in low (hot) latitudes in all scenarios.Spatially, all climate scenarios use a linear temperature gradient with a 0.02°C increase in temperature per grid row (across latitude).Temporally, the climate is held constant for the first 25 years until MPA implementation for all climate scenarios.The climate scenarios were defined as follows: (i) Mean Shift-consistent temperature increase by 0.033°C (RCP 8.5; Bruno et al., 2018) in each grid cell and year (Figure S14), (ii) El Niño/ La Niña-simulated via a sinusoidal wave representing the change in temperature from year to year, with a cyclical deviation of 0.5°C from the mean and a consistent temperature increase of 0.033°C (Bruno et al., 2018) in each grid cell and year (Equation 2; Figure S14), (iii) Shocks-heatwaves are simulated as single year occurrences with varying probability of occurrence and intensity (departure from the mean), (iv) Shocks with Mean Shift-the shocks scenario with an addition of a yearly 0.033°C (Bruno et al., 2018)  (2) dT dt = 0.5 × sin(t) + 0.033 the next 100 years, if p is less than 0.35 (a 35% chance) a shock occurs.
Intensity, the increase over the mean, is determined by a random number generated between 1 and 4 (Frölicher et al., 2018; Figure S14).An additional analysis was run with a yearly addition of 0.013°C throughout the simulation instead of 0.033°C to the mean shift, El Niño/La Niña, and shocks with mean shift to explore dynamics when climate proceeds at a slower rate.

| Analysis
Population size was averaged across replicate, genotype, and sex for all adult fish within the simulation, and then summed within the MPA grid cells to get a within-MPA population abundance.Additionally, population size was averaged similarly for index 90 (high latitude), which represents a colder area, and index 10 (low latitude), which

| Evolutionary dynamics
For scenarios with a mean increase in temperature, the low movement genotype (aa) initially increased before rapidly decreasing in frequency as temperatures begin to exceed the species thermal optimum (Figure 2).In contrast, both the heterozygous (Aa) and high movement (AA) genotypes initially decreased in frequency; however, over time, the heterozygous genotype (Aa) increased with climate change, while the high movement genotype (AA) continued to decrease (Figure 2).The low movement genotype (aa) peaked at frequencies greater than 0.99 within the MPA, typically after the population became largely confined to the MPA (i.e., local extinction outside the MPA; Figure 2; Table S4).The eventual decline in the low movement genotype (aa) corresponded to temperatures between 28.16 and 29.93°C for the mean shift scenario, 28.28 and 29.86°C for the El Niño/La Niña scenario, and 27.23 and 30.32°C for the shocks with mean shift scenario (Table S4), equating to survival rates between 13%-36%, 13%-34%, and 10%-52%, respectively.Results were consistent across climate rates (see Data S1; Figure S15).The area outside the MPA experienced some spillover of genotypes from within the MPA, with an initial increase in the low movement genotype (aa) and a significant decline of the high movement genotype (AA) (Figure 2).These evolutionary patterns were not observed in the shocks scenario (Figure 2).

| Population dynamics within the MPA
Accounting for climate change consistently lowered the population size within the MPA relative to the no climate scenario (Null) regardless of whether the population experienced low movement evolution (Figure 3).Within a medium MPA (6 × 6), the maximum population size across climate scenarios was 98.21%, 97.78%, 92.89%, and 88.31% of the maximum population size when not accounting for climate change (Table S5).The mean shift scenario resulted in a steady population decline, while populations subject to El Niño/La Niña and shocks with mean shift experienced cyclical dynamics of decline and recovery, but ultimately declined overall (Figure 3).At year 100, populations in the climate scenarios with mean shifts in temperature were between 5.47% and 27.02% the size of the no climate change scenario (Table S5).The shocks scenario resulted in the same decline-recovery pattern as shocks with mean shift, but, absent the mean increase in temperature, the population was sustained at levels similar to the no climate scenario between heatwaves (shocks population was 82.40% the size of the no climate change scenario Changes in the frequency of the high movement (AA), heterozygous (Aa), and low movement (aa) genotypes through time for two MPA sizes (no MPA, small [3 × 3], and large [10 × 10]) and two areas outside of the MPA: a high latitude (cooler temperatures) simulation area and a low latitude (higher temperatures) within the simulation area.The point at which the population for a specific genotype hits 0 on the y-axis is when that genotype has been removed from the population (as indicated by truncated lines).Time starts at the beginning of the simulation and the first dark grey line indicates the beginning of fishing and the second represents MPA implementation.MPA, marine protected area.
at year 100; Table S5).Patterns of population decline, recovery, persistence, and collapse were consistent across all MPA sizes assessed (Figure S16), when fishing mortality was greater than 0.5 (Figure S17), and outside of the MPA in the fished areas (Figure S18), whether or not movement evolution was considered.
Regardless of climate, low movement evolution led to higher equilibrium populations within the MPA than without movement evolution (Figure 3; Table S5).For all climate change scenarios, low movement evolution led to the largest population increase over no movement evolution in the small (3 × 3) and medium (6 × 6) MPAs, with populations 207.22%-218.43%and 172.96%-181.08% the size of the no movement evolution population, respectively, depending on climate scenario.However, unlike the no climate change scenario in which the population benefit derived from low movement evolution was sustained throughout the simulation, climate scenarios with a mean increase in temperature (mean shift, El Niño/La Niña, and shocks with mean shift) degraded the relative population benefit of evolving to move less through time (Figure 4).For example, within the medium (6 × 6) MPA, the average population size for low movement evolution relative to no movement evolution went from a factor of 1.96 times larger in year 75, to 1.76 times larger in year 100, and 1.68 times larger in year 125 (Table S6).Similar declines in the relative benefit of low movement evolution were observed for the El Niño/La Niña and shocks with mean shift scenarios and across MPA sizes (Table S6).The shocks scenario did not result in the same degradation of benefits due to climate change, with population level effects instead resembling the no climate change scenario (Figure 3).

| Movement evolution increases persistence within and decreases persistence outside the MPA
Climate change affected the time to collapse both within and outside the MPA, and these climate change effects were mediated F I G U R E 3 Population size within a medium (6 × 6) MPA for all five climate scenarios (panels) with and without movement evolution (colors).Time starts at Phase 3 of the simulation procedure when an MPA is implemented, and the climate change simulation begins.Fishing mortality outside the MPA is set to 0.7 throughout the simulation.MPA, marine protected area.
by whether populations experienced low movement evolution.
Low movement evolution led to extended time to population collapse within the MPA (1% collapse point-Figure 5; Figure S19; 5% collapse point-Table S7), except within the smallest (1 × 1) MPA, which was too small to sustain the population even without climate change.However, all populations in sufficiently sized MPAs eventually collapsed in scenarios with a mean increase in temperature (mean shift, El Niño/La Niña, and shocks with mean shift).When comparing the time to population collapse for populations with movement evolution to those without, the greatest difference was in the small (3 × 3) MPA (18 years for mean shift, 19 years for El Niño/La Niña, and 9 years for shocks with mean shift; Figure 5).
Outside the MPA, in the hottest and coolest area of the species range, low movement evolution had the opposite effect, leading to faster times to population collapse with low movement evolution than without in many of the climate scenarios and across MPA sizes (Figure S20).Larger MPAs exacerbate this effect, with a large (10 × 10) MPA in the El Niño/La Niña and a population experiencing movement evolution collapsing 6 years before a no movement evolution population in the hotter zone and 1 year before a no movement evolution population in the cooler zone (Figure 5).Additionally, when an MPA was in place, particularly a large MPA, populations outside the MPA collapsed earlier than they would have without the MPA with movement evolution (Figure 5).

| Movement evolution increases rates of decline
Movement evolution led to faster rates of population decline in all climate scenarios with a mean increase in temperature compared to no movement evolution (1.32-2.27times faster, depending on the climate scenario and excluding the smallest [1 × 1] MPA; Figure 6; Figure S21; Table S8).Relative decline rates between movement evolution and no movement evolution were also higher (decline rates up to 1.81 times faster with movement evolution for mean shift and decline rates up to 6.5 times faster with movement evolution for El Niño/La Niña, excluding the smallest [1 × 1] MPA; Table S9), removing the possibility that the higher rate of population decline was driven by the initial larger population size of the low movers.For the shocks scenario, the population also declined faster with low movement evolution than without following a heat wave, but subsequently recovered faster after the heatwave subsided (Figure S21; Table S9).
When comparing a higher climate rate (0.033°C) to a lower climate rate (0.013°C) with movement evolution, the lower rate led to decline rates 1.97 time slower in the mean shift scenario, 1.55 times slower in the El Nino/La Nina Scenario, and 1.75 times slower in the shocks with mean shift scenario than the higher rate (excluding the smallest [1 × 1] MPA; Figure S22).This resulted in longer persistence under the lower climate rate (average 94.6 years longer).

| DISCUSS ION
This study demonstrates the interaction between climate change and movement evolution in the context of MPAs.We show that even though low movement evolution leads to larger populations that survive longer inside the MPA (sensu Mee et al., 2017b), it also leads to several sub-optimal results under large-scale temperature shifts.
Populations outside the MPA that experience movement evolution often collapse more quickly than without movement evolution, leaving an ecologically trapped population within the MPA.By the time climate change begins to increase movement rates selected for by mean increases in temperature, populations are prevented from regaining their highest movement extents.This results in a decrease in the expected population benefits from low movement evolution and faster than expected rates of population decline within the MPA.Without a mean increase in temperature, shocks was the only climate scenario that did not lead to collapse, but recovery potential rapidly declined when temperature increased alongside heatwaves.It is likely that environmental stochasticity increases the rate of adaptation in movement traits, analogous to environmental variability increasing the rate of thermal tolerance adaptation in corals (Palumbi et al., 2014).Theoretical and empirical evidence suggests that climate change results in increased dispersal distances through multiple mechanisms (Boeye et al., 2013;Miller et al., 2020;Nadeau & Urban, 2019).
This study focuses on selection (Travis et al., 2013), specifically of adult dispersal distance (Clobert et al., 2009), which is expected to experience heterogeneous pressures across a species' range, particularly at leading and trailing edges (Ochocki & Miller, 2017;Szűcs et al., 2017;Weiss-Lehman et al., 2017;Williams et al., 2016).We exclude these edge dynamics, instead modeling movement as a uniform random walk.As a result, high movement genotype (AA) fish are as likely to move into hotter regions of the simulation as into cooler regions to effectively escape the effects of climate change (i.e., movement was not temperature dependent).While random walks should allow for spatial sorting (Shine et al., 2011), which would lead the high movement genotype (AA) to dominate on the leading range edge, we found limited evidence of this, with the high F I G U R E 6 Rate of change within a medium (6 × 6) MPA for each climate scenario (panels) and evolution (colors) scenario.For shocks with mean shift, shocks, and El Nino/La Nina, movement evolution leads to higher rates of decline when temperatures increase within the MPA, but also quicker rates of recovery with subsequent declining temperature.This pattern is most pronounced in the mean shifts scenario, as seen by the clear separation of the with (purple) and without (blue) movement evolution lines.MPA, marine protected area.
latitude fish containing only marginally higher frequencies of the high movement genotype (AA) than the low latitude fish, possibly due to either the relatively small size of our simulation, spillover of ical studies (Gienapp et al., 2008).Here, we were able to isolate climate change as the driving force of both phenotypic and genotypic changes in fish species over relatively short time scales.Although we did not model phenotypic plasticity, we expect this mechanism to alter movement extents (Bowler & Benton, 2005) and predict it would lead to similar population dynamics to those shown here, possibly on shorter time scales if climate change does not disrupt cues (Bonamour et al., 2019;DeWitt et al., 1998).It is, however, important to explore the extent to which movement is plastic versus evolved, as plastic responses are quicker to reverse and may avoid some of the maladaptive responses associated with evolution highlighted here.
Overall, this study suggests that single, static MPAs may be un- within MPAs (Bruno et al., 2019).
Generally, species undergoing range shifts are expected to move out of MPAs (Gilmour et al., 2022;Weinert et al., 2021), therefore, as the world moves towards implementing 30 × 30 commitments (to protect 30% of the planet by 2030; Jones et al., 2020;O'Leary et al., 2016), carefully designed MPA networks that can serve as stepping-stones, especially for lower movement extent fish, may be required.While many MPA networks are currently in place, most (86%) were not designed nor are currently managed for climate change (Lopazanski et al., 2023), and many are known to have a high likelihood of climate change impacts (Kyprioti et al., 2021).Due to uncertainty in climate change effects on MPAs, designing MPA networks for both connectivity and a portfolio effect with both diversity and replication of habitats may be the best way to ensure species are protected now and in the future (Hopkins et al., 2016;McLeod et al., 2009).To accomplish this, we must improve our understanding of species connectivity among MPAs via both larval dispersal (Gaines et al., 2010;Rassweiler et al., 2020) and adult movement and explicitly incorporate this information into conservation planning efforts.
Climate change is predicted to alter fish movement extents, with larval dispersal distances, in particular, expected to decrease (O'Connor et al., 2007).Our findings suggest that multiple MPAs may also affect movement dynamics via evolution in fish; however, we do not explore potential consequences on larval movement.
Therefore, future research is needed to investigate how adult movement, larval dispersal, and their evolutionary dynamics interact in and extinction (Hewitt et al., 2011;Ricciardi & Simberloff, 2009), and species with small thermal ranges may not be good candidates (Backus & Baskett, 2021), moving a large enough fraction (50%-60%) of a population can serve to successfully overcome extinctions associated with climate change (Backus & Baskett, 2021).While research is largely lacking on assisted migration between protected areas in the sea (Twardek et al., 2023), lower dispersal rates, e.g., resulting from low movement evolution, could increase translocation success (Backus & Baskett, 2021), suggesting that MPAs in a species' historical range where low movement has evolved could effectively serve as donor reserves.It will likely be important to time translocations to target species that have evolved maximum low movement extents within the MPA, but that have not yet collapsed or increased movement due to climate change.We expect these individuals to have higher establishment success when moved into a new MPA due to high protection levels within their new habitat.More research is needed to determine the feasibility and necessity of this intervention, including identifying key challenges such as whether competitio or other factors, would lead to failed introductions (Backus & Baskett, 2021).
This study has several key limitations (discussed in more depth in the "Model Assumptions & Limitations", Data S1).Many of the key assumptions are borrowed from Mee et al. (2017b), including modeling of movement evolution as a single locus with only two alleles.Despite this simplification, the results of this model are likely similar if not conservative compared to species with more complex underlying genetics controlling movement (Mee et al., 2017b).While there is some evidence of reductions of movement in MPAs (Parsons et al., 2010;Thorbjørnsen et al., 2021), one long-term study on cod in an MPA found no evidence of reductions or changes to home range sizes after MPA implementation (Villegas-Ríos et al., 2022), suggesting that not all species have the potential to evolve movement at fast rates.We also assume that species do not have the capacity to range expand or migrate, instead modeling movement as a random walk.The inclusion of edge dynamics and related movement strategies could change the overall outlook for population survival and represents an important next step toward understanding movement evolution under climate change within MPAs.We also did not account for possible physiological factors that may lead to decreased metabolic activity and therefore lower movement extents under climate change, as evidence linking these factors is weak (Peck et al., 2009).

F
Model schematic detailing the different phases of the simulation model representing the entire species range.(a) Phase 1 includes four model steps: spawning, recruitment, natural mortality, and movement.(b) During Phase 2 fishing begins, with fish experiencing fishing mortality after natural mortality and before movement.(c) In Phase 3, the marine protected area is implemented, and climate change is simulated, which increases linearly across latitude, with cooler temperatures in the "south" (blue area) and warmer temperatures in the "north" (pink area).Natural mortality is now based on temperature drawn from either the no climate or one of the four climate change scenarios.Parameter information is shown in Table temperature increase, and (v) Null-no climate change.For shocks and shocks with mean shift, shock years are determined by randomly generating probabilities (p) of a shock based on the uniform distribution.During the first 50 years, if p is less than 0.1 (a 10% chance), a shock occurs in that year.During (1) min s, e (−SST−opt.temp) 2 temp.range 2 represents a hotter area, to determine the population dynamics across the fishs full range.The point of population collapse, at which the species cannot recover and will go extinct, was defined as either 5% (when considering only individuals inside the MPA) or 1% (when considering individuals both inside and outside the MPA) of emerging biomass.This population collapse point was used as a reference point to identify the time to population collapse across different climate scenarios and between simulations with and without movement evolution.These 1% and 5% collapse points only inside the MPA and across the entire simulation, respectively, were chosen to maximize interpretability, as populations outside the MPA declined below 5% of equilibrium population sizes too quickly across all climate scenarios to compare the 5% collapse point.We calculated the first derivative of population size through time to compare the population rate of change for different MPA sizes and climates.We also assess the relative rate of change within the MPA (first derivative/ population) to compare between simulations.To determine how adult genotype frequencies changed over time, population size was averaged across replicates and sex for each of the following: (1) MPA grid cells, (2) cells at latitude index 90 (high latitude) and (3) cells at index 10 (low latitude).The peak frequency of the low movement genotype (aa) was determined within the MPA and at each latitude, as well as the corresponding temperature associated with a permanent decline of the low movement genotype (aa) to a frequency <0.9 after climate change begins.Additional details are available in Data S1.

F
Relative population size with movement evolution compared to without movement evolution for a medium (6 × 6) MPA for all five climate scenarios (represented by different colors).Each curve represents the relative change in population size over time for each climate scenario.Relative population sizes were calculated up to the point in time when either evolution scenario drops below five fish to avoid capturing the effect of population extinction.MPA, marine protected area.

F
Population collapse points (1% of emerging biomass) for each climate scenario (colors) across all five MPA sizes (no MPA, smallest [1 × 1], small [3 × 3], medium [6 × 6], and large [10 × 10]) with (triangles) and without (circles) movement evolution.The top panel shows the collapse points outside the MPA at the hot extreme (high temperature), the middle panel shows the collapse points within the MPA, and the bottom panel shows the collapse points outside the MPA at the cool extreme (low temperature) of the species range.While movement evolution leads to longer time to collapse within the MPA, as shown by most triangles of one color being to the right of the corresponding circles, it is detrimental outside the MPA, leading the population to collapse sooner than when there is no movement evolution, as shown by circles to the right of the triangles of the same color in many cases.MPA, marine protected area.Climate change increased species movement extents by selecting for the heterozygous genotype (Aa), not the high (AA) genotype as might be expected.Instead, both the high (AA) and low (aa) movement genotypes decreased with climate change.While the increased frequency of the heterozygote led to an overall increase in movement extent with climate change, this occurred when species were locally extinct outside the MPA due to overfishing.These findings suggest that, for fished stocks, MPAs will inhibit species from reaching their full movement potential for tracking climate change, with larger MPAs exacerbating this maladaptive response-as shown by decreased persistence outside the MPA-by further increasing the frequency of the low movement genotype (aa).Even without population collapse due to a mean increase in temperature, a complete lack of selection for increased movement extents with heatwaves alone could harm species if heatwave intensity increases as is predicted (Frölicher et al., 2018).
alleles from the MPA, low genetic diversity within the fragmented MPA, or our rebounding effect within the simulation.Despite the assumption of no range expansion capacities, spatial sorting would likely lead to a similarly isolated, low movement population within MPAs.Future work should focus on the complexity and directionality of dispersal evolution and its effects on the evolution of specific movement genotypes in climate change.Linking movement adaptation to climate change in simulation is an important first step in understanding how evolution may impact a species' ability to track climate change.It is difficult to prove adaptation over other mechanisms, such as phenotypic plasticity, or that climate change is the specific driver of genotypic changes in empir- able to effectively protect species long-term under climate change, although they can preserve a small population within MPAs for an extended period.Our results emphasize a need to carefully match managers' goals with conservation interventions to reach a desired outcome in climate change.If local persistence is the primary objective, then single MPAs may effectively protect and extend population survival in a small region, at least for a time.However, if overall population persistence is the primary goal, single, even large MPAs, fail to reach this objective under climate change, consistent with an empirical review showing limited evidence of climate resilience MPA networks facing climate change to determine their success as a climate conscious strategy and how to best incorporate connectivity into their design.Beyond designing and managing MPAs to account for the role of movement evolution under climate change, alternative conservation interventions may become essential to ensure species persistence in the future.Assisted migration (McLachlan et al., 2007), whereby populations are moved outside their historical range and beyond areas accessible through natural dispersal, has been proposed as a solution to overcome climate constraints.While assisted migration has associated risks, including ecosystem disruption, invasion, Finally, although we model multiple climate scenarios, each individual scenario is relatively simplistic and ignores co-occurring climate patterns.For example, although El Nino/La Nina and heat waves are known to often occur simultaneously, our simulations model these as independent scenarios to isolate climate effects resulting from mean shifts from random shocks and predictable cyclical patterns.This paper is the first to determine the dynamics of movement evolution within MPAs under climate change.Our study reveals the importance of considering the evolutionary consequences of MPAs and climate change and how these may conflict and interact, leading to both expected and unexpected consequences for MPAs.Such interactions have not been well explored, despite their demonstrated importance (Colton et al., 2022; Palumbi et al., 2014) in establishing effective MPAs by 2030.We show that climate change selects for higher movement extents, which is in direct conflict with selection for lower movement via protection.While the time to collapse is extended within the MPA, evolved populations decline more rapidly than non-evolving populations.Single, static MPAs alone may not be the best tool for species conservation under climate change.Depending on manager goals, the evolution of low movement extents within MPAs could serve to secure populations when implemented in tandem with other conservation tools, such as assisted migration.As the full impacts of climate change on species is not yet understood, it is imperative to put more thought into how to best incorporate climate change in the design of conservation strategies, particularly when discussing static efforts such as MPAs and MPA networks.