Tradeoffs in planning marine protected areas for kelp forest resilience: protecting climate refugia is not always the best solution

22 Marine protected areas (MPAs) are increasingly promoted as climate mitigation tools, yet 23 guidance on their placement to maximize resilience against climate stressors like marine 24 heatwaves remains limited. Here, we develop MPA placement guidelines that explicitly 25 consider a mechanistic pathway through which MPAs could enhance kelp forest resilience to 26 heatwaves: protecting fishery-targeted urchin predators to prevent kelp overgrazing. Using 27 a spatially explicit, tri-trophic model of California kelp forests, we evaluate alternative MPA 28 configurations across a hypothetical coastline where half the habitat experiences an 29 increased probability of experiencing heatwaves. We found that effective MPA placement 30 depends on whether MPAs are being newly established or reconfigured within an existing 31 network, and that among-patch connectivity and spillover played vital roles in the relative 32 effectiveness of different MPA configurations. Changes in resilience occurred primarily at 33 the patch scale, with trade-offs between increased within-MPA resilience and decreased 34 resilience in some fished areas, resulting in minimal coastwide population effects. For 35 example, for new MPAs, large single MPAs within heatwave-prone areas maximized within-36 MPA resilience gains, while multiple small MPAs in heatwave refugia best supported whole-37 coast resilience. When reconfiguring established networks, expanding existing MPAs in 38 refugia areas was most effective. We also demonstrate the importance of considering MPA 39 recovery timescales: for example, relocating old MPAs to heatwave refugia yielded minimal 40 short-term benefits due to the loss of rebuilt, previously fished, predator biomass. Our 41 findings demonstrate that climate-adaptive marine planning should explicitly consider the 42 spatiotemporal implications of trophic cascades, connectivity, and transient population 43 dynamics to support ecosystem resilience. 44 45 46 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint

47 The accelerating pace of climate change underscores the urgent need to identify local 48 actions that may mitigate its impacts. A key challenge is the increasing frequency, duration, 49 and severity of marine heatwaves (hereafter ‘heatwave(s)’), which pose significant risks to 50 marine ecosystems and the services they provide (Frölicher et al., 2018; K. E. Smith et al., 51 2021, 2023). Ultimately, the remedies for climate change are global, but there is also a 52 pressing need for local mitigation strategies. For example, there is currently much interest in 53 the role marine protected areas (MPAs) may play in mitigating climate impacts like 54 heatwaves (Jacquemont et al., 2022; Roberts et al., 2017; J. G. Smith et al., 2023; J. W. 55 White et al., 2025), even though MPAs have historically been established to protect marine 56 ecosystems from different local stressors, such as fishing (Lubchenco & Palumbi, 2003). 57 A plausible mechanism through which MPAs – especially no-take MPAs, which is our focus – 58 may mitigate heatwave impacts is by rebuilding and protecting higher-trophic species 59 targeted by fisheries, leading to increased ecosystem resistance and resilience through 60 trophic cascades (Jacquemont et al., 2022; Kumagai et al., 2024; J. W. White et al., 2025). 61 For example, in many kelp forests, abundant predators can keep populations of herbivorous 62 sea urchins in check, reducing the likelihood of shifting from a productive kelp-dominated 63 state to a less desirable, overgrazed urchin barren (Eisaguirre et al., 2020; Hamilton et al., 64 2023). Heatwaves can affect this dynamic by reducing kelp growth and recruitment 65 (Hollarsmith et al., 2020; Michaud et al., 2022; Zimmerman & Kremer, 1986), leading to a 66 reduced production of detached kelp fronds ('drift kelp'), a preferred food source for 67 urchins. This loss of drift kelp, combined with heatwave-induced increased grazing rates, can 68 trigger urchins to switch from feeding cryptically on drift in rocky crevices to roaming 69 exposed on the reef, targeting standing kelp stipes (Kriegisch et al., 2019; Rennick et al., 70 2022; J. G. Smith & Tinker, 2022) (Fig 1). Because kelp forest resilience to heatwaves is a fine 71 balance of consumer-resource dynamics, using MPAs to protect and rebuild fishery-targeted 72 urchin predator populations may buffer against heatwave-induced kelp loss (Hopf et al., 73 2025). This benefit of MPAs was documented empirically in California, USA, where MPAs 74 that protected urchin predators were less likely than fished areas to lose kelp biomass 75 during a heatwave (Eisaguirre et al., 2020; Kumagai et al., 2024). 76 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint As MPA networks have typically been designed using biophysical guidelines (e.g., habitat 77 representation, connectivity) that assume stationary climate conditions, climate change 78 impacts and the mechanisms of climate resilience have not been explicitly considered in 79 traditional MPA planning (Lopazanski et al., 2023). In contrast, emerging climate-focused 80 conservation planning (often termed “climate-smart” or climate-adaptive planning) aims to 81 incorporate projected spatial and temporal patterns of climate change into MPA design 82 (e.g., Arafeh-Dalmau et al., 2023; Brito-Morales et al., 2022; Buenafe et al., 2025). These 83 approaches typically prioritize areas of reduced climate exposure, climate refugia, or high 84 connectivity under future scenarios, and are increasingly promoted in international policy, 85 including the post-2020 Global Biodiversity Framework (COP15 2022). However, while 86 climate-smart planning could provide important guidance on where MPAs might best be 87 placed under climate change, this approach has generally focused on metrics of exposure to 88 perceived climate stressors (e.g., changes in the mean or variance of sea surface 89 temperature; Brito-Morales et al., 2022) rather than explicitly considering the ecological 90 mechanisms that may promote resilience to specific climate stressors such as heatwaves 91 (but see White et al., 2025). 92 Here, we develop MPA placement guidelines for future marine heatwaves that explicitly 93 consider the mechanistic pathway through which MPAs may enhance resilience to 94 heatwave-driven kelp forest collapse by protecting fishery-targeted urchin predators. We 95 focus on the near-term (20-year) consequences of alternative spatial designs for new and 96 established MPA networks, accounting for transient dynamics following MPA establishment. 97 Because the biomass and abundance of previously fished predator populations typically 98 recover over years to decades (Hopf et al., 2016; J. W. White et al., 2013), resilience benefits 99 arising through trophic pathways may also be delayed. These lags are particularly important 100 when 1) setting expectations for the assessment of MPA resilience outcomes, and 2) 101 considering change to MPA network design (CDFW, 2022). In the case of MPA relocation, 102 there may be trade-offs between reduced future climate exposure (if, for example, shifting 103 to a climate refugia), the time required to rebuild predator populations in the new location, 104 and the loss of conservation benefits from opening established MPAs to fishing. Specifically, 105 we ask, where should new MPAs be placed relative to regions of higher/lower heatwave 106 probability, as well as other existing MPAs, to maximize kelp forest resilience? 107 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint To answer this question, we use a three-species, multi-patch (along an idealized linear 108 coastline) population model that captures the key species and trophic interactions in 109 southern California kelp forests: giant kelp (Macrocystis pyrifera), herbivorous purple 110 urchins (Strongylocentrotus purpuratus), and a fishery-targeted predatory fish (California 111 Sheephead; Bodianus pulcher, henceforth “Sheephead”). This is a spatial extension of a 112 model previously used to investigate single-population heatwave impacts (Hopf et al., 113 2025). Here, we compared the short-term resilience consequences of alternative no-take 114 MPA placements in response to an increased incidence of heatwaves. We consider multiple 115 spatial configurations, including scenarios without previously established MPAs in the region 116 and scenarios with an established MPA network comprised of two individual MPA patches 117 covering 12.5% of the kelp-forest habitat (similar to the situation in California, Fig 2). We 118 considered a scenario in which half of the coastline has an increased chance of experiencing 119 a heatwave (25% chance each year), while the other half is a refugium habitat that 120 experiences historical levels of environmental variability. 121 122 123 Figure 1: Three-species model overview. Key patch-level processes, including those affected 124 by heatwaves (orange symbols), captured in our kelp-urchin-predator population model. 125 Purple and green text show processes relevant to the kelp-urchin sub-model, and blue text 126 indicates processes relevant to the predator sub-model. Dashed lines indicate export/loss, 127 and solid lines indicate import/growth process. Non-italicized and italicized text indicate 128 within- and among-patch processes, respectively. All artwork by Jess K. Hopf, except giant 129 kelp icons, which are modified from artwork by Jane Thomas sourced from 130 ian.umces.edu/media-library under the CC BY-SA 4.0 license. 131 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint 132 Figure 2: MPA scenarios. Schematic of the modelled, linear coastline with the MPA 133 scenarios considered in this study, including establishing new MPAs (light red squares) along 134 a coastline without any existing MPA network, and adding or relocating MPA patches in a 135 network with established MPAs (dark red and hatched squares). Each square represents a 136 single kelp forest patch covering 6.25% of the total kelp forest area. Patches are separated 137 by sandy substrate along a infinite coastline (so the leftmost and rightmost patches depicted 138 here are also neighbors). Orange regions indicate patches impacted by marine heatwaves, 139 while the non-highlighted region is a heatwave refugium. 140 141 142 143 144 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint

145 We quantified the ‘resilience’ of a given kelp forest habitat patch as the proportion of time 146 kelp biomass was above a minimum threshold (1% of no-heatwave, kelp-forested state) 147 during the first 20 years after the MPA configuration was changed (and the heatwave 148 probability had increased for half of the coastline). This definition encompasses both the 149 resistance of kelp-forest patches to a state change and the recovery of degraded patches to 150 a kelp-dominated state (primarily through kelp spore and drift imports). As our model 151 includes natural variability in larval recruitment (and the incidence of heatwaves; see 152 Methods), there is a chance of kelp forest decline independent of heatwave perturbation. 153 Therefore, our definition of resilience includes overall resilience to change, not just 154 resilience to disturbances. Distributions of resilience across the 5,000 simulation runs were 155 bimodal, reflecting phase-shifting system behavior between stable kelp-dominated forests 156 and persistent urchin barrens (SI Figs). To address our research questions, we focus on the 157 absolute difference between the median resilience values of the baseline and alternative 158 MPA configurations, on both patch-by-patch and whole-coastline scales (Figs 3-4). 159 Habitat patches with newly placed MPAs had varied resilience gains depending on MPA 160 coverage, configuration (single large or several small), and placement, as well as adjacency 161 to an existing MPA patch or high heatwave likelihood (red bars in Figs 3-4). Notably, these 162 gains within the MPA came at a cost to resilience in some, but not all, fished patches (blue 163 bars in Figs 3-4). Typically, fished patches closer to new MPA patches experienced resilience 164 gains, while more distant fished patches had resilience losses. This is because there is 165 spillover of kelp spores and drift kelp from healthy kelp forests inside MPAs to neighboring 166 patches, but even redistribution of fishing pressure from new MPAs into the remaining 167 fished habitat. These trade-offs among patches resulted in none to minimal (<0.05 unit 168 change) increases in median resilience for the whole population after adding/moving MPAs 169 (grey bars in Figs 3-4). Within-patch and population effects were consistently greater with 170 larger MPA coverage (Figs 3-4). 171 Single large MPAs resulted in consistently greater resilience gains within MPA patches than 172 multiple small MPA patches of the same coverage and placement relative to the heatwave-173 impacted area. This was true for both establishing a new MPA network (compare red 174 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint horizontal dashed lines across columns, Fig 3) and reconfiguring an existing one (e.g., 175 compare red horizontal dashed lines between scenarios 2A and 2B or 2E and 2G in Fig 4). 176 However, single large MPAs also resulted in stable or declining resilience in fished patches, 177 on average. For example, no fished patches had reduced resilience when establishing two 178 separate, single-patch MPAs (covering 12.5% of the total area) in the non-impacted, 179 heatwave refugium area (1E in Fig 3), but over half the fished patches saw declines in 180 resilience when establishing a single large MPA covering 12.5% in the heatwave refugium 181 (1E in Fig 3). This difference arose because spillover of kelp forest benefits is shared 182 between habitat patches in the same large MPA but not between distant small MPAs. 183 When considering placement relative to the heatwave-impacted region, we found that 184 establishing new MPAs (without existing MPAs) in the heatwave-impacted region resulted in 185 greater within-MPA gains than establishing them in the non-impacted refugium area, 186 regardless of configuration (e.g., compare 1A, 1C, 1E, and 1G in Fig 3). Conversely, in the 187 scenario with existing MPAs, adding new MPAs or moving existing ones to the heatwave 188 refugium area resulted in markedly greater resilience gains than focusing MPA patches in 189 the impacted region (e.g., compare 2C to 2D, and 2E to 2F in Fig 4). 190 The effects on fished-patch resilience of MPA placement relative to the heatwave region 191 interacted with MPA configuration: resilience losses occurred and were greatest with single 192 large MPAs in refugia areas, regardless of whether there were previously established MPAs 193 (1A, 1B in Fig 3, and 2A, 2E in Fig 4). Critically, heatwave-impacted patches that were 194 originally MPAs but then opened to fishing saw the greatest decreases in resilience (2E, 2G 195 in Fig 4), while refugia MPA-to-fished patches saw no change (2F, 2H in Fig 4). 196 Gains in median resilience at the population scale were likewise influenced by an interaction 197 between MPA configuration (single large, or multiple small) and placement relative to the 198 heatwave-impacted region (grey bars Figs 3-4). When establishing new MPAs, the greatest 199 increases in resilience occurred with larger coverages (25%) either with a single MPA in the 200 impacted area, or multiple MPAs in the non-impacted area (1F, 1D in Fig 3). When 201 reconfiguring or adding new MPAs to the established MPA network, we found that 202 expanding existing MPAs (adding new or moving old MPA patches) had the largest resilience 203 gains across the whole population, but only when at least half of the final coverage was in 204 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint the heatwave refugium (2A, 2C, 2E in Fig 4). Moving established MPAs had little benefit (2E-205 2H in Fig 4). 206 207 Figure 3: Resilience changes when establishing new MPAs. The absolute change in median 208 resilience (proportion of years kelp-forest patches spent above the kelp biomass threshold, 209 which is 1% of the no-heatwave, kelp-forested state), 20 years post-MPA implementation. 210 Values are relative to the baseline scenario of no MPAs. Each blue/red bar is an individual 211 patch (6.25% of the total kelp forest area), ordered along an infinite modelled coastline that 212 is partially affected by marine heatwaves (orange box indicates impacted patches). Grey 213 bars and text are changes in median resilience at the population scale. Horizontal dashed 214 lines represent mean changes in patch resilience for MPA (red) and fished (blue) patches 215 (i.e., means of bar heights). Letters relate to scenarios outlined in Figure 2. See also 216 Supplementary figure x. 217 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint 218 Figure 4: Resilience effects of reconfiguring an established MPA network. The absolute 219 change in median resilience (proportion of years kelp-forest patches spent above the kelp 220 biomass threshold, which is 1% of the no-heatwave, kelp-forested state), 20 years post-MPA 221 implementation. Values are relative to the baseline scenario of two old MPAs. Each blue/red 222 bar is an individual patch (6.25% of the total kelp forest area), ordered along an infinite 223 modelled coastline that is partially affected by marine heatwaves (orange box indicates 224 impacted patches). Grey bars and text are changes in median resilience at the population 225 scale. Horizontal dashed lines represent mean changes in patch resilience for MPA (red) and 226 fished (blue) patches (i.e., means of bar heights). Letters relate to scenarios outlined in 227 Figure 2. See also Supplementary figure x. 228 229 230 231 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint

232 MPAs can enhance resilience against heatwave-driven kelp forest collapse by protecting and 233 rebuilding fishery-targeted urchin predators (Eisaguirre et al., 2020; Jacquemont et al., 234 2022; Kumagai et al., 2024). By explicitly modelling this mechanistic pathway and evaluating 235 alternative MPA configurations, we show that the extent of the climate mitigation benefit 236 that could be expected depends not only on where MPAs are placed relative to climate 237 refugia but also on spillover1 and connectivity processes, transient dynamics, and existing 238 MPA protection. Critically, resilience responses were highly spatially heterogeneous and 239 emerged primarily at the patch level, rather than the entire population, with trade-offs 240 between increased resilience within-MPAs and decreased resilience in some (but not all) 241 fished patches leading to minimal overall changes in resilience at the whole-coast scale. This 242 lack of whole-coast resilience benefits from MPAs reflects current understanding of 243 population dynamic responses to MPAs in systems without severe overfishing: there should 244 be local within-MPA increases in biomass but not substantial whole-population biomass 245 increases (e.g., Ovando et al., 2021; White et al., 2025). 246 The management implications of our findings will depend on planning objectives; for 247 example, if a climate-adaptation goal of a new MPA network is to maximize resilience 248 against heatwave-driven kelp forest collapse primarily within MPA areas, then establishing a 249 large single MPA in heatwave-vulnerable areas may be best (Fig 5). In the fished area, 250 however, this approach will have minimal benefits, which will diminish further with larger 251 MPA coverages. Conversely, if the goal is to use MPAs to support more uniform resilience 252 across the whole coastline, then multiple small MPAs in the heatwave refugia area (or less 253 so, the impacted area) may be more suitable (Fig 5). This work provides starting foundations 254 for such considerations to be folded into climate-adaptive planning alongside climate-255 refugia approaches (e.g., Brito-Morales et al., 2022), risk-spreading strategies (e.g., Allison et 256 al., 2003), dynamic MPAs (e.g., Tittensor et al., 2019), and other mitigation strategies 257 (Arafeh-Dalmau et al., 2023; Buenafe et al., 2025). 258 1 Note that we are referring to the ‘spillover’ of kelp spores and drift, rather than the traditional MPA spillover of fishery-targeted fish. In our model, adult fish remain within their recruited patch. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint Among-patch connectivity and spillover played a vital role in the relative effectiveness of 259 different MPA configurations in our analysis. Whether changes in the MPA configuration 260 increased or decreased the resilience of fished patches reflected a balance between 261 exported benefits from MPAs (increased kelp spore supply and drift kelp biomass, which 262 primarily affected neighboring patches) and the even redistribution of fishing pressure from 263 newly closed areas to the remaining open patches. As such, fished patches adjacent to 264 newly established and long-standing MPAs typically exhibited increased resilience, whereas 265 patches further away often experienced decreases. To our knowledge, this is the first study 266 to explicitly call attention to potential negative resilience outcomes associated with MPA 267 configurations under climate stress. Such outcomes are consistent with the expected 268 implications of MPAs on fisheries (Botsford et al., 2009; White et al., 2014), in which 269 redistributed fishing pressure can temporarily or permanently reduce expected fishery 270 yields. The spillover of MPA benefits (kelp spores and larvae) between patches is also a 271 primary driver behind our finding that expanding established MPAs was notably more 272 effective at increasing resilience within those MPAs, than placing distant MPAs. 273 Our results demonstrate the importance of accounting for ecological recovery timescales 274 when evaluating the climate-mitigation potential of MPAs. The mechanistic pathway for 275 resilience we focused on requires sufficient biomass of fished predators to reduce urchin 276 numbers below thresholds (Ling et al., 2015). While the time required to reach these 277 biomasses is unknown (and will depend on pre-protection fishing levels; Nickols et al., 278 2019), MPA literature indicates that it may take years to decades to rebuild previously 279 fished populations to pre-exploitation levels (Claudet et al., 2008; Hopf et al., 2016; J. W. 280 White et al., 2013). Indeed, older MPAs have demonstrated greater resilience to kelp loss 281 than younger ones (Eisaguirre et al., 2020). In our model, we assume that old MPAs had fully 282 rebuilt Sheephead biomass and population structure before increasing heatwave frequency. 283 Under these conditions, we found that offsetting the cost of relocating these old MPAs (loss 284 of established biomass) may occur only in limited situations: specifically, if there is an 285 existing MPA in the non-impacted area that can be expanded to maintain overall coverage 286 (scenario 2E in Fig 4). Notably, we found that moving an established MPA to a refugia area, 287 but distant from other established MPAs, was unlikely to have notable benefits (scenario 2F 288 in Fig 4). In that case, the cost of no longer protecting established biomass was greater than 289 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint the benefits of moving away from heatwaves. This suggests caution when relocating MPAs 290 under climate-adaptive planning: if the goal is promoting resilience against heatwaves 291 through protecting trophic cascades, moving MPAs to less-impacted areas is not always 292 optimal. While placing MPAs in climate refugia may be better long-term, relocation could 293 incur prohibitive short-term costs (including ecological and non-ecological; Green et al., 294 2014). If moving MPAs is a key objective, maintaining old MPAs until newer ones have 295 rebuilt biomass may mitigate the short-term costs. 296 A key insight from our findings is that when establishing new MPAs, focusing protection on 297 patches within heatwave-prone areas was the best strategy (for within-MPA effects), but 298 when reconfiguring an established MPA network, protecting heatwave refugia was more 299 effective. This difference arises because in the case of existing MPAs, there was a large 300 difference in resilience inside versus outside MPAs (i.e., the median inside was 100%; Fig 301 S3). When an old MPA was expanded in the non-impacted refugium habitat, existing kelp 302 drift and spore spillover from that MPA patch was able to support its new MPA neighbor to 303 dramatically increase resilience. This spillover, however, was less effective from an existing 304 MPA in the heatwave-impacted habitat. By contrast, in the case of adding new MPAs where 305 none previously existed, the time required to build up biomass (i.e., the transient dynamics) 306 limited the capacity for resilience benefits. What effect existed was similar in magnitude in 307 both the impacted and refuge parts of the coastline, but because the baseline scenario (no 308 MPAs) had lower resilience in the impacted zone, the relative benefit of MPAs was greater 309 there (Fig S2). In other words, when expanding MPAs, doing so in the non-impacted refuge 310 area was more effective both in absolute terms (higher resilience within MPA boundaries) 311 and relative to the baseline scenario. But when adding MPAs, there was no spatial 312 difference in absolute terms, but adding MPAs to the heatwave-impacted area was better 313 relative to the no-action baseline. 314 It is important to recognize that our results regarding MPA planning apply to a specific 315 ecological context: MPAs that protect fishery-targeted grazer (i.e. urchin) predators to 316 reduce the likelihood of heatwave-driven kelp forest collapse through overgrazing. The key 317 elements driving our findings are that the urchin predator is fished (which is not the case in 318 all kelp forests; Kumagai et al., 2024; Spiecker et al., 2025), that the dispersal of kelp spores 319 and drift kelp is highly local, relative to the connectivity of fish and urchin larvae, and that 320 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint heatwaves are the primary climate stressor. Heatwaves have emerged as a key threat to 321 kelp forest resilience (Parnell et al., 2026), but there are other climate stressors in 322 temperate systems that do not necessarily covary with temperature (Hamilton et al., 2023). 323 2023). As minimizing exposure to one stressor may increase exposure to another (Bruno et 324 al., 2018), our findings need to be considered in a broader planning approach that balances 325 MPA configurations to achieve clear goals. 326 The general tri-trophic framework we have developed could also apply to some coral reef 327 systems where protecting a functionally diverse herbivore community supports coral 328 persistence (Rasher et al., 2013). System-specificity aside, an important general lesson from 329 our analysis is the value of an ecologically mechanistic modeling approach for climate-330 informed conservation planning. By explicitly modeling trophic dynamics and transient 331 population dynamics, we revealed subtle and counterintuitive guidance for MPA placement 332 (e.g., the trade-offs involved in expanding existing MPAs in refuges versus adding new MPAs 333 to heatwave-impacted habitats) that are not possible from only examining distributions of 334 forecasted physical ocean statistics (e.g., sea surface temperatures) and choosing to protect 335 'refuge' habitat (Green et al., 2014). As this study demonstrates, it is important to explicitly 336 account for the spatiotemporal constraints of connectivity and transient population 337 dynamics when considering climate-informed spatial management. 338 339 340 Figure 5: Summary of guidelines for resilient kelp-forest MPAs. Overarching guidelines for 341 establishing new MPA networks and reconfiguring (expanding or relocating patches) existing 342 MPA networks based on changes to resilience (proportion of years kelp-forest patches spent 343 above the kelp biomass threshold) within kelp-forest patches and across a linear coastline. 344 ‘Non-impacted’ and ‘impacted’ refer to where the newly established MPA patches are 345 focused relative to heatwaves. Symbols qualitatively indicated the effectiveness of a given 346 strategy relative to the baseline and other scenarios, with larger arrows indicating greater 347 relative change. 348 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint

349 Three-species, multi-patch population model 350 Here we extend the three-species, single-patch model from Hopf et al. (2025) – which 351 captures the interactions between giant kelp, herbivorous purple urchins, and a fishery-352 targeted urchin-eating predatory fish (California Sheephead) – to a 16-patch along-coast 353 population model with connectivity between all patches. To account for seasonal variations 354 in ecological processes (recruitment, kelp growth, urchin grazing, and heatwave effects), we 355 use seasonal (3-month) time steps. The model captured the characteristic kelp forest 356 dynamics in which local overgrazing by urchins leads to urchin barrens at the patch scale. 357 Kelp spore recruitment and external drift supply can then rescue extinct patches under low 358 urchin densities. Here we give an overview, but details are provided in the Supplemental 359 Information, and code is available at DOI: [[TBA]]. 360 At the patch level (~1 km2), our model is comprised of two linked sub-models: (1) a stage-361 based kelp-urchin component tracking biomass of juvenile/adult kelp (standing and drift) 362 and juvenile/adult urchins (exposed and hiding), and (2) a predator integral projection 363 model (IPM) tracking Sheephead length-abundance converted to biomass. Our model 364 incorporates behavioral switching where urchins passively consume drift kelp while hiding 365 but emerge to graze standing kelp when drift becomes scarce, with exposure modeled as a 366 declining function of drift availability (Randell, 2022; Rennick et al., 2022). We used a Type-367 II functional response for urchin grazing and a Type-I response for Sheephead predation on 368 adult urchins, with higher predation mortality on exposed versus cryptic urchins (Nichols et 369 al., 2015). When kelp was absent for >3 months, Sheephead ceased consuming nutritionally-370 depleted urchins (Liebergesell, 2022). 371 We modelled an effectively infinite linear coastline (the first patch neighbors the last patch), 372 representing the Southern California coast, where kelp forest patches are interspersed with 373 sandy habitat. Kelp forest patches were demographically connected through kelp spore and 374 drift dispersal, and purple urchin and Sheephead larval dispersal. We used a 1km2 patch 375 size, which encompasses adult Sheephead (Topping et al., 2006) and urchin (Rogers-Bennett 376 & Okamoto, 2020) home ranges, and captures the majority (~65%; Gaylord et al., 2002), but 377 not all, of kelp spore dispersal. Kelp spore dispersal between patches followed a data-378 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint parameterized negative exponential function (Gaylord et al., 2002), and we assumed that 379 kelp drift only dispersed between nearest-neighbor patches (Figurski, 2010; Hobday, 2000). 380 Urchin and Sheephead larval dispersal were assumed to be well-mixed among patches, 381 reflecting the long larval durations and distances of both species (Alonzo et al., 2004; Kinlan 382 & Gaines, 2003). Stochasticity was captured through year-to-year variation in recruitment 383 for all species, with recruitment each year drawn randomly from data-parameterized 384 normal distributions. Where possible, the model is parameterized using published data from 385 the Channel Islands, southern California, otherwise data from comparable regions are used 386 (see SI for more details). 387 388 Modelling heatwaves & MPA strategies 389 Reflecting the known effects of heatwaves on kelp and urchins, we implemented heatwaves 390 as a period of reduced kelp recruitment (Hollarsmith et al., 2020) and growth (Zimmerman 391 & Kremer, 1986), and increased urchin grazing rates (Spindel, 2023). As an illustrative 392 scenario, we assumed that half of the modelled coastline experienced heatwaves, and that 393 there was a 0.25 probability that the heatwave would occur each year, reflecting the 394 expected century-scale increase in the frequency of heatwaves in the California Current (Shi 395 et al., 2021). 396 To establish initial conditions, we ran the two modelled baseline scenarios (without MPAs 397 and with 12.5% MPA coverage, a value chosen because it is approximately the coverage of 398 MPAs in California waters) for 40 years without the increased heatwave effect. Given that 399 we parameterized the model (not including the modelled heatwave effects) using data 400 primarily collected before the 2014-2016 Californian heatwave, we implicitly assume that 401 the initial conditions reflect baseline climate levels of heatwaves. We then ran the model for 402 20 years with the increased heatwave effects and intensity under a range of MPA scenarios 403 (Fig 2), simulating 5,000 replicates of each scenario. In scenarios with increased in MPA 404 coverage, Sheephead fishing pressure was evenly redistributed to the patches that 405 remained open to fishing. 406 407 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint Resilience calculations 408 We used the proportion of time kelp biomass was above a minimum threshold (1% of no-409 heatwave, kelp-forested state) during the simulation run (20 years) as our resilience 410 response variable. We plotted the distributions of resilience at the patch and whole coast 411 level (SI Figs) and focused on the change in median resilience between the baseline and 412 alternative scenarios for both the with and without old MPA cases. 413 414

415 This work was funded by the David and Lucile Packard Foundation (award 2022-74730), the 416 California Ocean Protection Council (agreements C0874012 and C0874015), Oregon 417 Department of Fish and Wildlife (contract IGA 251-23) and in part by the Oregon 418 Agricultural Experiment Station with funding from the Hatch Act capacity funding program, 419 award numbers NI25HFPXXXXXG022 and/or NI25HMFPXXXXG029, from the USDA 420 National Institute of Food and Agriculture. This is publication XX of the Partnership for 421 Interdisciplinary Study of Coastal Oceans, funded primarily by the David and Lucile Packard 422 Foundation 423 424

425 Allison, G., Gaines, S., Lubchenco, J., & Possingham, H. P. (2003). Ensuring persistence of 426 marine reserves: Catastrophes require adopting an insurance factor. Ecological 427 Applications, 13(1), 8–24. 428 Alonzo, S. H., Key, M., Ish, T., & MacCall, A. (2004). Status of the California sheephead 429 (Semicossyphus pulcher) stock (2004). Center for Stock Assessment Research, 430 California Department of Fish and Game, University of California Santa Cruz, and 431 National Marine Fisheries Service, Santa Cruz, California, USA. 432 Arafeh-Dalmau, N., Munguia-Vega, A., Micheli, F., Vilalta-Navas, A., Villaseñor-Derbez, J. C., 433 Précoma-de La Mora, M., Schoeman, D. S., Medellín-Ortíz, A., Cavanaugh, K. C., Sosa-434 Nishizaki, O., Burnham, T. L. U., Knight, C. J., Woodson, C. B., Abas, M., Abadía-435 Cardoso, A., Aburto-Oropeza, O., Esgro, M. W., Espinosa-Andrade, N., Beas-Luna, R., 436 … Possingham, H. P. (2023). Integrating climate adaptation and transboundary 437 management: Guidelines for designing climate-smart marine protected areas. One 438 Earth, 6(11), 1523–1541. https://doi.org/10.1016/j.oneear.2023.10.002 439 Brito-Morales, I., Schoeman, D. S., Everett, J. D., Klein, C. J., Dunn, D. C., García Molinos, J., 440 Burrows, M. T., Buenafe, K. C. V., Dominguez, R. M., Possingham, H. P., & Richardson, 441 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint A. J. (2022). Towards climate-smart, three-dimensional protected areas for 442 biodiversity conservation in the high seas. Nature Climate Change, 12(4), 402–407. 443 https://doi.org/10.1038/s41558-022-01323-7 444 Bruno, J. F., Bates, A. E., Cacciapaglia, C., Pike, E. P., Amstrup, S. C., Van Hooidonk, R., 445 Henson, S. A., & Aronson, R. B. (2018). Climate change threatens the world’s marine 446 protected areas. Nature Climate Change, 8(6), 499–503. 447 https://doi.org/10.1038/s41558-018-0149-2 448 Buenafe, K. C. V., Dunn, D. C., Metaxas, A., Schoeman, D. S., Everett, J. D., Pidd, A., Hanson, 449 J. O., Bentley, L. K., Kim, S. W., Neubert, S., Scales, K. L., Dabalà, A., Brito-Morales, I., 450 & Richardson, A. J. (2025). Current approaches and future opportunities for climate-451 smart protected areas. Nature Reviews Biodiversity, 1(5), 284–297. 452 https://doi.org/10.1038/s44358-025-00041-0 453 Claudet, J., Osenberg, C. W., Benedetti-Cecchi, L., Domenici, P., García-Charton, J.-A., Pérez-454 Ruzafa, Á., Badalamenti, F., Bayle-Sempere, J., Brito, A., Bulleri, F., Culioli, J.-M., 455 Dimech, M., Falcón, J. M., Guala, I., Milazzo, M., Sánchez-Meca, J., Somerfield, P. J., 456 Stobart, B., Vandeperre, F., … Planes, S. (2008). Marine reserves: Size and age do 457 matter. Ecology Letters, 11(5), 481–489. https://doi.org/10.1111/j.1461-458 0248.2008.01166.x 459 Eisaguirre, J. H., Eisaguirre, J. M., Davis, K., Carlson, P. M., Gaines, S. D., & Caselle, J. E. 460 (2020). Trophic redundancy and predator size class structure drive differences in kelp 461 forest ecosystem dynamics. Ecology, 101(5), 1–11. https://doi.org/10.1002/ecy.2993 462 Figurski, J. D. (2010). Patterns and sources of variation in drift algae and the ecological 463 consequences for kelp forests. University of California Santa Cruz. 464 Frölicher, T. L., Fischer, E. M., & Gruber, N. (2018). Marine heatwaves under global warming. 465 Nature, 560, 360–364. 466 Gaylord, B., Reed, D. C., Raimondi, P. T., Washburn, L., & McLean, S. R. (2002). A physically 467 based model of macroalgal spore dispersal in the wave and current-d ominated 468 nearshore. Ecology, 83(5), 1239–1251. https://doi.org/10.1890/0012-469 9658(2002)083%5B1239:APBMOM%5D2.0.CO;2 470 Hamilton, S. L., Kennedy, E. G., Zulian, M., Hill, T. M., Gaylord, B., Sanford, E., Ricart, A. M., 471 Ward, M., Spalding, A. K., & Kroeker, K. (2023). Variable exposure to multiple climate 472 stressors across the California marine protected area network and policy 473 implications. ICES Journal of Marine Science, 80(7), 1923–1935. 474 https://doi.org/10.1093/icesjms/fsad120 475 Hobday, A. (2000). Abundance and dispersal of drifting kelp Macrocystis pyrifera rafts in the 476 Southern California Bight. Marine Ecology Progress Series, 195, 101–116. 477 https://doi.org/10.3354/meps195101 478 Hollarsmith, J. A., Buschmann, A. H., Camus, C., & Grosholz, E. D. (2020). Varying 479 reproductive success under ocean warming and acidification across giant kelp 480 (Macrocystis pyrifera) populations. Journal of Experimental Marine Biology and 481 Ecology, 522, 151247. https://doi.org/10.1016/j.jembe.2019.151247 482 H opf, J. K., Giraldo-O spina, A., Caselle, J. E., Kroeker, K. J., Carr, M. H., Botsford, L. W., 483 Hastings, A., & White, J. W. (2025). Short -T erm Management of Kelp Forests for 484 Marine Heatwaves Requires Planning. Conservation Letters, 18(4), e13130. 485 https://doi.org/10.1111/conl.13130 486 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint Hopf, J. K., Jones, G. P., Williamson, D. H., & Connolly, S. R. (2016). Fishery consequences of 487 marine reserves: Short-term pain for longer-term gain. Ecological Applications, 26(3), 488 818–829. https://doi.org/10.1890/15-0348 489 Jacquemont, J., Blasiak, R., Le Cam, C., Le Gouellec, M., & Claudet, J. (2022). Ocean 490 conservation boosts climate change mitigation and adaptation. One Earth, 5(10), 491 1126–1138. https://doi.org/10.1016/j.oneear.2022.09.002 492 Kinlan, B. P., & Gaines, S. D. (2003). Propagule dispersal in marine and terrestrial 493 environments: A community perspective. Ecology, 84(8), 2007–2020. 494 https://doi.org/10.1890/01-0622 495 Kriegisch, N., Reeves, S. E., Flukes, E. B., Johnson, C. R., & Ling, S. D. (2019). Drift-kelp 496 suppresses foraging movement of overgrazing sea urchins. Oecologia, 190(3), 665–497 677. https://doi.org/10.1007/s00442-019-04445-6 498 K umagai, J. A., Goodman, M. C., Villaseñor -D erbez, J. C., Schoeman, D. S., Cavanuagh, K. C., 499 Bell, T. W., Micheli, F., De Leo, G., & Arafeh-D almau, N. (2024). Marine Protected 500 Areas That Preserve Trophic Cascades Promote Resilience of Kelp Forests to Marine 501 Heatwaves. Global Change Biology, 30(12), e17620. 502 https://doi.org/10.1111/gcb.17620 503 Liebergesell, M. W. (2022). Evaluating mechanisms of prey selection by a fish predator and 504 their sea urchin prey linked to habitat-specific differences in kelp forests and urchin 505 barrens.pdf [Masters, San Diego State University]. 506 https://www.proquest.com/openview/2be1d752005d4f83f7a0c985cf341dd2/1?cbl=507 18750&diss=y&parentSessionId=9%2BVGAEWoIcHWaDJOsw8Ed3yhOEt8bByTF%2B8508 KITfNoG4%3D&pq-origsite=gscholar&accountid=13013 509 Ling, S. D., Scheibling, R. E., Rassweiler, A., Johnson, C. R., Shears, N., Connell, S. D., Salomon, 510 A. K., Norderhaug, K. M., Pérez-Matus, A., Hernández, J. C., Clemente, S., Blamey, L. 511 K., Hereu, B., Ballesteros, E., Sala, E., Garrabou, J., Cebrian, E., Zabala, M., Fujita, D., 512 & Johnson, L. E. (2015). Global regime shift dynamics of catastrophic sea urchin 513 overgrazing. Philosophical Transactions of the Royal Society B: Biological Sciences, 514 370(1659), 20130269. https://doi.org/10.1098/rstb.2013.0269 515 Lopazanski, C., Foshay, B., Couture, J. L., Wagner, D., Hannah, L., Pidgeon, E., & Bradley, D. 516 (2023). Principles for climate resilience are prevalent in marine protected area 517 management plans. Conservation Letters, 16(5), e12972. 518 https://doi.org/10.1111/conl.12972 519 Lubchenco, J., & Palumbi, S. (2003). Plugging a hole in the ocean: The emerging science of 520 marine reserves 1. Ecological \ldots, 13(1), 3–7. 521 Michaud, K. M., Reed, D. C., & Miller, R. J. (2022). The Blob marine heatwave transforms 522 California kelp forest ecosystems. Communications Biology, 5(1), 1143. 523 https://doi.org/10.1038/s42003-022-04107-z 524 Mumby, P. J., Hastings, A., & Edwards, H. J. (2007). Thresholds and the resilience of 525 Caribbean coral reefs. Nature, 450(7166), 98–101. 526 https://doi.org/10.1038/nature06252 527 Nichols, K. D., Segui, L., & Hovel, K. A. (2015). Effects of predators on sea urchin density and 528 habitat use in a southern California kelp forest. Marine Biology, 162(6), 1227–1237. 529 https://doi.org/10.1007/s00227-015-2664-2 530 Nickols, K. J., White, J. W., Malone, D., Carr, M. H., Marissa, M. S., Alan, L. B., & Louis, H. 531 (2019). Setting ecological expectations for adaptive management of marine 532 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint protected areas. Journal of Applied Ecology, 56(February), 2376–2385. 533 https://doi.org/10.1111/1365-2664.13463 534 Ovando, D., Caselle, J. E., Costello, C., Deschenes, O., Gaines, S. D., Hilborn, R., & Liu, O. 535 (2021). Assessing the population-level conservation effects of marine protected 536 areas. Conservation Biology, 35(6), 1861–1870. https://doi.org/10.1111/cobi.13782 537 Randell, Z. (2022). Grazer-resource Behavior, Interactions, and Feedbacks Control Kelp-forest 538 Dynamics and Stability. 539 Rennick, M., DiFiore, B. P., Curtis, J., Reed, D. C., & Stier, A. C. (2022). Detrital supply 540 suppresses deforestation to maintain healthy kelp forest ecosystems. Ecology, 541 103(5), e3673. https://doi.org/10.1002/ecy.3673 542 Roberts, C. M., O’Leary, B. C., Mccauley, D. J., Cury, P. M., Duarte, C. M., Lubchenco, J., 543 Pauly, D., Sáenz-Arroyo, A., Sumaila, U. R., Wilson, R. W., Worm, B., & Castilla, J. C. 544 (2017). Marine reserves can mitigate and promote adaptation to climate change. 545 Proceedings of the National Academy of Sciences of the United States of America, 546 114(24), 6167–6175. https://doi.org/10.1073/pnas.1701262114 547 Rogers-Bennett, L., & Okamoto, D. (2020). Mesocentrotus franciscanus and 548 Strongylocentrotus purpuratus. In Developments in Aquaculture and Fisheries 549 Science (Vol. 43, pp. 593–608). Elsevier. https://doi.org/10.1016/B978-0-12-819570-550 3.00032-9 551 S hi, H., García-R eyes, M., Jacox, M. G., Rykaczewski, R. R., Black, B. A., Bograd, S. J., & 552 Sydeman, W. J. (2021). Co-o ccurrence of California Drought and Northeast Pacific 553 Marine Heatwaves Under Climate Change. Geophysical Research Letters, 48(17), 554 e2021GL092765. https://doi.org/10.1029/2021GL092765 555 Smith, J. G., & Tinker, M. T. (2022). Alternations in the foraging behaviour of a primary 556 consumer drive patch transition dynamics in a temperate rocky reef ecosystem. 557 Ecology Letters, (April), 1–12. https://doi.org/10.1111/ele.14064 558 Smith, K. E., Burrows, M. T., Hobday, A. J., King, N. G., Moore, P. J., Sen Gupta, A., Thomsen, 559 M. S., Wernberg, T., & Smale, D. A. (2023). Biological Impacts of Marine Heatwaves. 560 Annual Review of Marine Science, 15(1), 119–145. https://doi.org/10.1146/annurev-561 marine-032122-121437 562 Smith, K. E., Burrows, M. T., Hobday, A. J., Sen Gupta, A., Moore, P. J., Thomsen, M., 563 Wernberg, T., & Smale, D. A. (2021). Socioeconomic impacts of marine heatwaves: 564 Global issues and opportunities. Science, 374(6566), eabj3593. 565 https://doi.org/10.1126/science.abj3593 566 Spindel, N. B. (2023). Ecophysiology of ectothermic ecosystem engineers: Bioenergetic 567 effects of climate and food on dominant consumers and their consequences for 568 coastal ecosystems. 569 Tittensor, D. P., Beger, M., Boerder, K., Boyce, D. G., Cavanagh, R. D., Cosandey-Godin, A., 570 Crespo, G. O., Dunn, D. C., Ghiffary, W., Grant, S. M., Hannah, L., Halpin, P. N., 571 Harfoot, M., Heaslip, S. G., Jeffery, N. W., Kingston, N., Lotze, H. K., McGowan, J., 572 McLeod, E., … Worm, B. (2019). Integrating climate adaptation and biodiversity 573 conservation in the global ocean. Science Advances, 5(11), eaay9969. 574 https://doi.org/10.1126/sciadv.aay9969 575 Topping, D. T., Lowe, C. G., & Caselle, J. E. (2006). Site fidelity and seasonal movement 576 patterns of adult California sheephead Semicossyphus pulcher (Labridae): An 577 acoustic monitoring study. Marine Ecology Progress Series, 326, 257–267. 578 https://doi.org/10.3354/meps326257 579 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint White, J. W., Botsford, L. W., Hastings, A., Baskett, M. L., Kaplan, D. M., & Barnett, L. A. K. 580 (2013). Transient responses of fished populations to marine reserve establishment. 581 Conservation Letters, 6(3), 180–191. https://doi.org/10.1111/j.1755-582 263X.2012.00295.x 583 White, J. W., Hopf, J. K., Arafeh-Dalmau, N., Ban, N. C., Bates, A. E., Claudet, J., Lopazanski, 584 C., Sunday, J. M., & Caselle, J. E. (2025). Measurements, mechanisms, and 585 management recommendations for how marine protected areas can provide climate 586 resilience. Marine Policy, 171, 106419. 587 https://doi.org/10.1016/j.marpol.2024.106419 588 Zimmerman, R., & Kremer, J. (1986). In situ growth and chemical composition of the giant 589 kelp, Macrocystis pyrifera: Response to temporal changes in ambient nutrient 590 availability. Marine Ecology Progress Series, 27, 277–285. 591 https://doi.org/10.3354/meps027277 592 593 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 4, 2026. ; https://doi.org/10.64898/2026.04.01.715997doi: bioRxiv preprint