{"id":"fe981113-aa67-4479-b73c-c5647388f2f2","arxiv_id":"2506.03346","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"In 11 coral species from the Great Barrier Reef, all five measured storage effects are weak and cannot overcome fitness differences, making temporal niche partitioning only a minor contributor to coexistence.","lead":"Weak storage effects: using 5 years of demographic data and size-structured models for 11 Great Barrier Reef corals, the authors find that environmentally driven coexistence mechanisms are far weaker than average fitness differences, so temporal niche partitioning contributes little to coral diversity. The result adds a data-rich test to the long-running debate on whether the storage effect maintains biodiversity, suggesting spatial processes may matter more.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Five-year window yields rank-deficient between-species correlation matrices; simulating year effects from the resulting singular MVN confines environmental responses to a low-dimensional subspace, potentially biasing storage-effect estimates downward.","rationale":"The reader's weakest_assumption—that five-year between-species correlations may not represent long-term responses—is a legitimate concern, and the paper's own Appendix B.5 acknowledges the lack of temporal replication. However, the most load-bearing aspect is more specific and mechanistic: the empirical correlation matrix computed from 5–6 annual estimates is rank-deficient, so the multivariate normal year effects used in simulations are sampled from a singular covariance. This confines all species' environmental responses to a low-dimensional subspace, directly limiting the species-specific response variation that the storage effect requires. This is not simply a question of the mean correlation being biased upward; even if the average correlation were unbiased, the rank constraint caps the dimensionality of environmental variation and can systematically deflate Δ(EC). The paper conducts extensive robustness checks, and the reader's CONDITIONAL verdict already reflects appropriate caution about the short time series. The proposed test—using a full-rank regularized or LKJ-based correlation matrix—would isolate whether the low-rank constraint materially changes coexistence outcomes. If it does not, the central claim is strengthened; if it does, the estimation of storage effects needs revision. Given the paper's thoroughness and the fact that the concern is currently unresolved rather than demonstrated, keeping the CONDITIONAL verdict is appropriate.","tokens_in":46228,"tokens_out":10221,"duration_ms":129756,"concrete_test":"Re-run the full-community module replacing the empirical correlation matrices R_G, R_F1, R_F2 with full-rank shrinkage estimates (e.g., Ledoit-Wolf shrinkage toward the identity, or an LKJ(η) prior on the correlation matrix fitted jointly with the year effects) while keeping all other model structure identical. Recompute Δ(EC) and the posterior probability of ≥2 species coexisting. If coexistence probability or Δ(EC) rises materially (e.g., by >20%), the singular-covariance constraint is inflating the apparent weakness of storage effects; if results are unchanged, the concern does not land.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that storage effects are weak rests on the simulated environment containing sufficient species-specific, multidimensional variation. In Section 2.3 and Appendix A.3, the authors compute an 11×11 empirical correlation matrix of species-specific year effects from only 5–6 annual estimates. For any posterior draw, this matrix has rank at most the number of years minus one (≤4 for growth, ≤5 for fecundity), so the covariance matrices Σ_G, Σ_F1, Σ_F2 (Eqs. A.15, A.22, A.23) are singular. IPM simulations then draw year-effect vectors from MVN(0, Σ) (Eqs. A.16, A.24, A.25), meaning all 11 species' environmental responses are confined to a ≤4–5 dimensional linear subspace in every simulated year. This artificially limits environmental dimensionality: species cannot exhibit the independent, species-specific responses that constitute the first ingredient of the storage effect. The result is a systematic downward bias in the environment–competition covariance Δ(EC) and in coexistence probabilities, independent of whether the mean correlation is upwardly or downwardly biased. The reader's concern about short-term correlations is valid, but the rank deficiency is a concrete mechanical pathway by which the five-year window can make storage effects appear negligible. Appendix B.5's diffuse-correlation robustness test uses a compound-symmetry matrix with a single shared ρ, which is full rank and therefore tests a different constraint, not the low-rank issue.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper parameterizes integral projection models for 11 coral species at Lizard Island using five years of demographic data, simulates community dynamics, and applies simulation-based Modern Coexistence Theory to quantify five storage effects arising from survival, growth, and fecundity fluctuations interacting with larval and space competition. The central claim is that storage effects operate but are weak: they are generally smaller than fluctuation-free fitness differences, coexistence is uncommon in simulations, and coexistence becomes probable only under greatly exaggerated environmental variability or strongly negative interspecific correlations. The authors support this with extensive robustness checks, posterior uncertainty propagation, and comparisons of fecundity fluctuations to literature values.","tokens_in":46475,"tokens_out":4891,"duration_ms":66261,"significance":"If the central claim holds, this is a valuable empirical test of a prominent coexistence mechanism in a system thought to favor strong storage effects, and it would reinforce a growing literature suggesting that temporal storage effects are generally weak. The study's strengths include explicit propagation of parameter uncertainty into coexistence outcomes, quantification of five distinct storage-effect pathways, and a broad set of alternative-model and parameter-sensitivity analyses. The concern raised below about the low-rank covariance structure of simulated year effects is load-bearing for the weak-storage-effect conclusion, because the simulated environmental responses are generated from covariance matrices that cannot represent the full dimensionality of species-specific environmental responses and may thereby bias the storage-effect estimates downward.","major_comments":[{"comment":"The 11x11 empirical correlation matrices RG, RF1, and RF2 are computed from only 5-6 annual species-specific year-effect estimates, so for any posterior draw each matrix has rank at most 4 (growth) or 5 (fecundity). The covariance matrices constructed in Eqs. (A.15), (A.22), and (A.23) are therefore singular, and the MVN draws in Eqs. (A.16), (A.24), and (A.25) confine all 11 species' environmental responses to a 4-5 dimensional linear subspace. This rank constraint mechanically limits the diversity of species-specific responses, which is the first ingredient of the storage effect, and it imposes a lower bound on average squared pairwise correlations among the 11 simulated response vectors; this is a distinct pathway by which the five-year window can bias Delta(EC) and coexistence probabilities downward, independent of whether the mean correlations are themselves biased. The diffuse-correlation robustness check in Appendix B.5 uses a full-rank compound-symmetry matrix with a single shared rho and therefore does not test the low-rank constraint. Please add a robustness analysis that breaks the low-rank constraint, for example by using an LKJ or factor-model prior on the full covariance matrix, or by adding species-specific independent noise to the simulated year effects, and report the resulting coexistence probabilities and Delta(EC) values. If the weak-storage-effect conclusion persists under that alternative, it would substantially strengthen the paper.","section":"Section 2.3 and Appendix A.3/A.4, Eqs. (A.15)-(A.16) and (A.22)-(A.25)"},{"comment":"The model predicts that Goniastrea pectinata dominates in 88-91% of no-coexistence simulations, yet this species is neither the most abundant nor the most abundant within its morphological group in the observed community. The text offers two plausible explanations, but the possibility that the five-year demographic rates are unrepresentative is also directly relevant to the storage-effect estimates: the study period ended with Tropical Cyclone Nathan and included thermal stress that disproportionately affected Acropora, so the estimated year effects and their between-species correlations may be dominated by a common disturbance response. Because the central claim concerns the strength of environmental covariance, please test how the estimated year-effect covariance structure and the MCT storage-effect estimates change when the cyclone-affected year is excluded or when year effects are estimated under a longer-term disturbance regime, and report whether the qualitative conclusion remains.","section":"Section 3 and Fig. 4"}],"minor_comments":[{"comment":"The phrase 'the probability of three or more species coexisting was only than 12%' appears to contain a typo; it should likely read 'only 12%'.","section":"Results, Section 3"},{"comment":"The phrase 'Combining this with wave-disturbance morality (sub-model A.1)' should read 'mortality' rather than 'morality'.","section":"Appendix A.2"},{"comment":"The paragraph beginning 'The early life history of corals involves two critical transitions...' is duplicated verbatim; one copy should be removed.","section":"Appendix C"},{"comment":"The diffuse-correlation robustness model is described qualitatively but its quantitative results are not reported in Table B.1 or elsewhere; please add the coexistence probabilities for this scenario so that readers can compare it with the baseline and other alternatives.","section":"Appendix B.5 and Table B.1"},{"comment":"The reference list contains entries cited in the text that appear with inconsistent formatting (e.g., some entries have missing journal names or incomplete page ranges); a careful copyedit of the reference list is recommended.","section":"Section 5 / References"}],"recommendation":"major_revision","confidential_remarks":"The rank-deficiency issue is the main technical obstacle to acceptance. The paper's current robustness checks do not address it, but the proposed fix is straightforward and within the scope of the manuscript. The G. pectinata dominance mismatch is also worth a sensitivity analysis, though it is secondary to the rank-deficiency concern. The authors should be encouraged to make the code and posterior draws for the proposed new analyses available, given that the paper already promises supplementary files and reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a real step forward in quantifying storage effects, and the authors are honest about their model's failures. But I think the rank-deficient covariance issue is a genuine soft spot that the paper doesn't fully address.\n\nWhat's new: they run 11-species size-structured IPMs with Bayesian parameter uncertainty, decompose coexistence into five process-based storage effects, and show that only the fecundity-larvae term matters, and even that is usually too weak to overcome fitness differences. The sensitivity analyses are extensive, and the fecundity comparison with literature is a nice touch. Credit where due: they propagate uncertainty properly and they flag the G. pectinata dominance problem themselves.\n\nThe soft spot: they estimate 11x11 between-species correlation matrices for growth and fecundity year effects from five or six annual data points. The resulting covariance matrices are singular; when they simulate year effects from MVN(0,Sigma), all species' environmental responses are confined to a four- or five-dimensional subspace. That's not just a statistical detail--it removes the species-specific environmental dimensions that storage effects need. Their diffuse-correlation robustness check uses a full-rank compound-symmetry matrix, so it doesn't test this low-rank constraint. I think this could bias the storage-effect estimates downward, and I'd want it addressed before taking the 'negligible storage effect' conclusion at face value.\n\nThe G. pectinata issue is a separate misspecification signal, but they acknowledge it and the conclusion is robust across community modules, so I see it as a moderate concern, not a killer.\n\nWho should read this: anyone interested in coexistence theory, and coral ecologists. It's a serious empirical data point. I'd send it to peer review, but I'd ask referees to look hard at the covariance construction and, ideally, require a regularized full-rank robustness analysis. As is, conditional acceptance.","headline":"A serious empirical test of the storage effect in corals, but the five-year window makes me wary of the covariance structure that drives the central conclusion.","tokens_in":47004,"tokens_out":4746,"would_cite":true,"duration_ms":55747,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D25","92D40"],"pacs":[],"model":"deepseek-v4-flash","headline":"A simulation-based test on 11 Great Barrier Reef coral species finds that the storage effect—coexistence through temporal niche partitioning—operates but is too weak to maintain coral biodiversity.","keywords":["storage effect","coral reef","coexistence","integral projection model","modern coexistence theory","environmental fluctuations","Great Barrier Reef","temporal niche partitioning"],"falsifier":"A multi-decadal demographic time series from the same reef that reveals strongly negative correlations among species' fecundity year effects together with high interannual variance would place the community inside the coexistence regime identified in the paper's Figure 6 and undermine its central conclusion; the same measurement showing persistently high positive correlations with modest variance would corroborate the finding.","tokens_in":45962,"feed_emoji":"🪸","tokens_out":8325,"duration_ms":86626,"temperature":0.7,"pith_summary":"The paper sets out to test whether the storage effect—the classic theory that species can coexist by specializing on different states of a fluctuating environment—actually maintains biodiversity in a system built for it: a community of 11 corals on the Great Barrier Reef with long-lived adults, large recruitment variability, and a competitive propagule stage. Using five years of size-structured demographic data, the authors parameterize integral projection models, simulate the community, and decompose the mechanisms of coexistence with Modern Coexistence Theory. They find that all five storage effects operate but are weak, typically about five times smaller than fluctuation-free fitness differences, and that simulated coexistence occurs mainly when species are nearly equal in the absence of fluctuations, not when temporal niches are strong. The paper concludes that environmental fluctuations contribute little to coral coexistence and that spatial processes, including microhabitat partitioning and asymmetric dispersal, are more likely to explain coral biodiversity.","feed_headline":"Temporal niches barely help corals coexist, model shows","feed_subtitle":"Five-year demographic simulations of 11 reef species find storage effects five times weaker than fitness differences.","key_machinery":"The argument is carried by three coupled tools. An Integral Projection Model (IPM)—a size-structured population model that tracks colony planar area through annual survival, growth, spawning, and density-dependent recruitment—generates simulated coral communities, with environmental fluctuations entering as species-specific year effects on growth and fecundity plus a wave-disturbance mortality sub-model. Bayesian hierarchical fitting propagates parameter uncertainty into every coexistence outcome. Simulation-based Modern Coexistence Theory (MCT) decomposes each species' per-capita growth rate when rare into components for mean conditions, environmental variation, competition variation, and the environment–competition interaction, and then compares invaders to residents; a fine-grained partition splits the storage effect into five sub-effects ($\\Delta_{(SL)}$, $\\Delta_{(SA)}$, $\\Delta_{(GL)}$, $\\Delta_{(GA)}$, $\\Delta_{(FL)}$) indexed by which demographic process covaries with which competition axis. The load-bearing diagnostic is a two-parameter map of coexistence probability against the between-species fecundity-year-effect correlation and the scale of environmental variability, which places the fitted species in a high-correlation, low-variability region far from where the storage effect can stabilize coexistence.","core_discovery":"The central discovery is that the storage effect is real but negligible: it promotes coexistence yet is nowhere near sufficient for it in this coral community. In simulations, the posterior probability that two or more of the 11 species coexist is 38%, and for three or more species only about 12%; in the exclusion cases, Goniastrea pectinata typically monopolizes the community. The Modern Coexistence Theory partition shows that fluctuation-free effects $\\Delta_0$ are typically five times larger than the storage effect $\\Delta_{(EC)}$, and coexistence arises when $\\Delta_0$ is near zero rather than when the storage effect is unusually strong. Of the five process-based storage effects, only the fecundity-to-larvae effect $\\Delta_{(FL)}$—the classic lottery-model mechanism—makes a substantive contribution; growth-based and survival-based storage effects are weak because growth fluctuations lack a positive environment–competition covariance and survival fluctuations barely move equilibrium growth rates. The decisive sensitivity result is that high-probability coexistence via the storage effect requires roughly tenfold-greater environmental variability, or roughly fivefold variability combined with strongly negative between-species correlations in fecundity year effects—conditions far outside the posterior estimates for the studied species.","pith_inferences":["If the cyclone- and heat-stress years inflated the estimated between-species correlations, a multi-decadal demographic record could shift the fitted community toward the negative-correlation, high-variability regime where the storage effect stabilizes coexistence—my inference, not a claim the paper makes.","Because the model is deliberately aspatial, it does not rule out a spatial storage effect; a spatially explicit version with microhabitat patches and larval dispersal could test whether space substitutes for time in maintaining the same coral diversity.","The simulated dominance of Goniastrea pectinata, which contradicts its field abundance, implies an omitted process such as species-specific thermal sensitivity or microhabitat segregation; quantifying that omission would show how much real-world coexistence the single-habitat, fluctuation-only model misses.","Recruitment-density parameters were tuned to produce realistic coral cover rather than estimated from data, so direct species-level recruit measurements would sharpen the fluctuation-free fitness differences and could confirm or revise the reported fivefold gap."],"forward_implications":["If the storage effect is this weak in a coral community possessing all its theoretical prerequisites, fluctuation-driven coexistence mechanisms are likely minor in most ecosystems, and fluctuating environments should not be treated as a default explanation for biodiversity.","Coral coexistence research should shift toward spatial mechanisms—microhabitat settlement preferences, spatial fitness-density covariance, and asymmetric larval dispersal—which the paper identifies as the most plausible alternative supports.","Among storage-effect pathways, fecundity fluctuations coupled to larval competition dominate, so empirical effort on temporal coexistence should concentrate on propagule production and settlement rather than on growth- or survival-driven mechanisms.","Coexistence in this system, when it happens, reflects near-equal mean fitnesses rather than strong stabilization, so demographic equalizing processes deserve as much attention as stabilizing mechanisms."],"supporting_citations":[{"why":"Introduces the lottery-model storage effect that the paper tests, defining the fecundity-to-larvae mechanism as the classic pathway.","marker":"Chesson and Warner, 1981"},{"why":"Supplies the ingredient-list definition of the storage effect and the theoretical foundation for decomposing coexistence in variable environments.","marker":"Chesson, 1994"},{"why":"Provides the simulation-based method for quantifying temporal storage effects without Taylor-series math, which the paper's MCT analysis implements.","marker":"Ellner et al., 2016b"},{"why":"Extends modern coexistence theory for empirical applications; the paper's coarse- and fine-grained partitions follow this framework.","marker":"Ellner et al., 2019"},{"why":"Contributes the five-year, 11-species demographic dataset on survival, growth, and fecundity that parameterizes the integral projection models.","marker":"Madin et al., 2023"},{"why":"Supplies the wave-disturbance mortality sub-model and a prior coral IPM application at the same site that the simulations build upon.","marker":"Alvarez-Noriega et al., 2023"},{"why":"Provides theoretical and comparative arguments that storage effects are generally weak, which the paper's conclusion directly extends.","marker":"Stump and Vasseur, 2023"}],"fun_headline_variants":["Coral storage effects negligible: fitness differences win","Storage effect five times weaker than fitness in coral model","Fecundity storage only substantive coral storage effect","Tenfold variability needed for coral storage effect to matter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the between-species correlations in year-to-year vital rates computed from five years of data represent true long-term environmental responses; the study window included a cyclone and thermal stress, and if those events inflated the correlations, the storage effect's strength could be substantially underestimated.","fun_headline_variants_meta":{"raw":{"variants":["Coral storage effects negligible: fitness differences win","Storage effect five times weaker than fitness in coral model","Fecundity storage only substantive coral storage effect","Tenfold variability needed for coral storage effect to matter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000668,"raw_usage":{"total_tokens":3047,"prompt_tokens":944,"completion_tokens":2103,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":560,"completion_tokens_details":{"reasoning_tokens":2041}},"tokens_in":560,"tokens_out":2103,"duration_ms":18160,"temperature":1.0,"reasoning_tokens":2041,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:06:11.699445+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A multi-decadal demographic time series from the same reef that reveals strongly negative correlations among species' fecundity year effects together with high interannual variance would place the community inside the coexistence regime identified in the paper's Figure 6 and undermine its central conclusion; the same measurement showing persistently high positive correlations with modest variance would corroborate the finding.","supporting_citations":[{"cited_title":"S., Baird, A","cited_arxiv_id":null,"evidence_quote":"Contributes the five-year, 11-species demographic dataset on survival, growth, and fecundity that parameterizes the integral projection models."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides theoretical and comparative arguments that storage effects are generally weak, which the paper's conclusion directly extends."}],"review_version":1}