{"id":"fe3f008f-2276-4827-801d-0ba4dac55a9a","arxiv_id":"2411.13228","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"For populations with density-dependent growth, the probability of extinction within 100 years is well described by a modified Gompertz function of carrying capacity.","lead":"This paper finds that the chance a population goes extinct rises along the same S-shaped curve as its habitat's carrying capacity falls. The result comes from billions of population simulations and could help conservationists estimate extinction risk from simple life-history traits.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Appendix F's OU derivation is internally inconsistent: the FPE and near-K rescaling do not follow from Eq. F1, so the claimed analytical universality of the modified Gompertz law is unsupported.","rationale":"The empirical half of the paper is strong: ~5 billion simulations, deposited code and data, r²>0.995 across >1700 parameter sets, and three external empirical fits. I credit that as real evidence that a three-parameter modified Gompertz describes the simulated models and the limited empirical cases. But the abstract's and Discussion's stronger claim—that few assumptions suffice for this relationship to hold for any extant population subject to density-dependent growth—is a mathematical claim, and it is supported only by Appendix F. That appendix is the load-bearing wall. As printed, it contains a sign error in the Fokker-Planck drift, a diffusion term inconsistent with both F1 and the stationary distribution F4, and an OU approximation that changes both the noise scale and the drift coefficient. The K-dependence that produces the Gompertz exponent disappears if the actual multiplicative noise σX is retained. These are not cosmetic typos: the main theoretical result F15 is obtained from the inconsistent rescaling. The paper also concedes that the modified Gompertz shape parameter γ is not predicted (the over-identification problem) and that the standard Gompertz systematically misfits, so the exact curve being claimed universal has no theoretical derivation. The reader's weakest-assumption identification is the same: Appendix F's approximate OU step with unverified bounds and inconsistencies. I agree. A clean re-derivation and direct SDE test would settle it. Absent that, the verdict should remain conditional: accept the empirical phenomenological result, but do not accept the mathematical universality claim.","tokens_in":24960,"tokens_out":18200,"duration_ms":204783,"concrete_test":"Re-derive Appendix F symbolically from Eq. F1: write the exact FPE and scale/speed functions, then perform the large-K OU reduction keeping the true diffusion coefficient σX and absorbing boundary X=1; compute the first-passage survival PS(T;K) and its asymptotic form as K→∞. Then Monte Carlo the continuous SDE F1 directly (Euler-Maruyama with X_t=1 as absorbing) for r=0.158, σ=0.11, T=100, K ranging over at least 10^1–10^4, and compare PS(T;K) with (i) Eq. F15, (ii) the standard Gompertz, and (iii) a modified Gompertz fit. If the corrected OU has no K dependence, or if the exact SDE's PS(T;K) deviates systematically from the Gompertz family, the claimed analytical universality is not established. Also fit the same test to a theta-logistic variant (e.g., θ=0.5) to check the assertion that any negative density dependence yields the same curve.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that 'few assumptions suffice' for a universal modified Gompertz relationship rests entirely on Appendix F. As printed, that derivation does not hang together. Eq. F3 writes the Fokker-Planck drift as -∂x[r x(1 + K⁻¹x)P] and the diffusion as (σ²/2)∂xx[xP]; the SDE F1 has drift r x(1 - K⁻¹x) and diffusion σx, so the FPE should be -∂x[r x(1 - K⁻¹x)P] + (σ²/2)∂xx[x²P]. The Gamma stationary law F4 corresponds to the x² diffusion, not to the printed FPE, so the transition density used for the first-passage calculation is not well defined. The subsequent 'Lamperti transform' is also not a transform of F1: for X near K, Y = X-K satisfies dY ≈ -rYdt + σK dW, not -(r/K)Ydt + σdW. With the correct noise amplitude, standardizing by σK puts the absorbing boundary at -(K-1)/(σK) ≈ -1/σ, independent of K, so the OU approximation as written produces no K-dependence in survival over a fixed horizon. The K-dependence in t₁(K) and Eq. F15 is an artifact of replacing σK by σ while keeping the boundary at K-1. No error bound is given for keeping only the first eigenvalue, and F15 has dimensional inconsistencies. Finally, even if the derivation were repaired, it predicts the ordinary Gompertz (γ=1); the empirically crucial modified Gompertz with γ<1 is never derived, and γ is fitted rather than predicted. Thus the theoretical bridge from SDE F1 to the universal modified-Gompertz claim is missing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that the probability of extinction of a population within 100 years, as a function of carrying capacity K, is universally well described by a modified Gompertz curve, PS(K) = exp(-exp(a + b K^γ)). This claim is supported by a very large simulation campaign across four population models of increasing demographic complexity, by fits to three empirical/observational datasets, and by an analytical argument in Supplementary Appendix F intended to show that a standard Gompertz curve follows from a logistic stochastic differential equation under two broad assumptions. The authors further use the fitted curve to discuss conservation triage, focusing on the inflection point and on the carrying capacity K10 corresponding to 10% extinction risk.","tokens_in":25348,"tokens_out":4284,"duration_ms":45902,"significance":"If the empirical relationship holds as a general pattern, this is a valuable contribution to conservation science: it provides a compact quantitative description of extinction risk versus carrying capacity that could inform status assessments and priority setting. The simulation effort is extensive and the data and code are made publicly available, which is a strength of the paper. The empirical fits are consistently very good (R² > 0.99 for simulated curves, and strong fits for the three empirical examples). However, the analytical derivation in Appendix F, which is the basis for the claim that the relationship is universal and follows from 'few assumptions', contains serious mathematical inconsistencies. As printed, the derivation does not bridge the gap from the stochastic logistic equation to the modified Gompertz law, and it does not predict the fitted shape parameter γ. The empirical law may still be a useful descriptive generalization, but the theoretical universality claim is not supported by the material presented.","major_comments":[{"comment":"The Fokker-Planck equation does not match the stochastic differential equation (F1). For dX = rX(1 - K^{-1}X)dt + σX dW, the Kolmogorov forward equation should have drift -∂x[r x(1 - x/K)P] and diffusion (σ²/2)∂xx[x² P]. Equation (F3) instead has a plus sign before the K^{-1} term and a diffusion coefficient proportional to x rather than x². As a result, the stationary distribution (F4) is not the stationary distribution of (F1), and the transition density used in the subsequent first-passage calculation is not well defined.","section":"Appendix F, Eq. (F3)"},{"comment":"The Ornstein-Uhlenbeck approximation is internally inconsistent. For X near K, with Y = X - K, the correct linearization of (F1) is dY ≈ -rY dt + σK dW, not dY = -ρY dt + σ dW with ρ = r/K. With the correct noise amplitude σK, the absorbing boundary at Y = -(K-1) is at a standardized distance approximately 1/σ, independent of K. Consequently the first-passage properties of the OU process do not depend on K, and the K-dependence in Eqs. (F14) and (F15) is an artifact of the inconsistent scaling. The claimed Gompertz dependence of survival on K is therefore not obtained from this approximation.","section":"Appendix F, OU approximation after Eq. (F13)"},{"comment":"The analytical derivation, even if corrected, yields at most a standard Gompertz curve (γ = 1), not the modified Gompertz curve with γ < 1 that is fitted to the simulations. The shape parameter γ is estimated from the same simulation data used to establish the law, and the paper provides no theoretical mechanism predicting its value. Thus the statement that 'few assumptions suffice to show that the relationship holds' for the modified Gompertz law is not supported by the derivation; the gap between the derived standard Gompertz and the empirically fitted modified Gompertz is acknowledged in the text but not resolved.","section":"Main text, 'The theoretical basis for the modified Gompertz' and Eq. (1)"},{"comment":"No error bound or quantified domain of validity is provided for the first-eigenvalue approximation, and Eq. (F15) has dimensional inconsistencies (the exponent contains products of r, K, and σ with units that do not cancel). The paper repeatedly invokes 'large K' and 'large t' without specifying how large is required, which is especially important because the simulations use a fixed 100-year horizon and K values from 1 to 3 million.","section":"Appendix F, Eqs. (F14)-(F15) and 'large K' claim"}],"minor_comments":[{"comment":"There is a typographical issue in the sentence 'Wolff et al. (2023) re-analysed Hilbers et al. (2016) data': the year should refer to the Hilbers et al. (2016) study consistently.","section":"Main text, Introduction"},{"comment":"The text refers to 'Figure 4' and 'Figure 1(b)' in several places, but the actual figure numbers in the appendix are E1, E2, E3, and E4; these cross-references should be corrected.","section":"Supplementary Appendix E, Section E"},{"comment":"The captions of Figures G2-G5 mention 'the distribution of differences between the R2 values', but the main text does not state which baseline R² (modified vs. standard) is being compared; clarifying the definition of 'difference' would improve readability.","section":"Supplementary Appendix G, Figures G2-G5"},{"comment":"Equation (A10) appears to be a rearrangement of Equation 17 of Niel and Lebreton (2005), but the relation between the annual survival rate S_a and the age at first breeding B is written without a derivation; providing the original equation or a reference would help the reader verify the formula.","section":"Supplementary Appendix A, Eq. (A10)"},{"comment":"The caption of Figure 3 states that the power-law curve 'takes the form 1 - PE ∝ (K + X0)^0.25', but the notation is not standard; the authors should specify that this is a fitted proportional relationship and define the horizontal offset X0 more explicitly.","section":"Main text, Figure 3 caption"}],"recommendation":"major_revision","confidential_remarks":"The empirical results and the scale of the simulation study are impressive and likely of interest to the conservation community. However, the theoretical appendix is not merely sloppy: the central analytical claim is unsupported by the mathematics as written, and the fitted shape parameter γ is not derived. The authors could revise the manuscript by (i) substantially reworking or removing the theoretical universality claim, (ii) presenting the modified Gompertz as an empirical law with clearly stated scope (the four model families and the fitted parameters), and (iii) either correcting the OU derivation or stating that the K-dependence in the survival probability arises from the nonlinear logistic term rather than from the linear-noise approximation. If the authors insist on keeping the universality claim, they need a mathematically valid derivation; as it stands, the gap between the derivation and the conclusion is too large to be fixed by local edits."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Nick, the short version: the simulation work is substantial and the empirical result is worth knowing, but the theoretical appendix is not in a shape to support the universality claim. As it stands, the paper does not establish a general law.\n\nWhat is actually new is a coordinated simulation campaign across four progressively realistic population models—billions of runs—all pointing to a modified Gompertz relationship PE(K)=1−exp(−exp(a+bK^γ)), with fitted γ consistently below 1. That systematic deviation from the standard Gompertz (Wolff et al. 2023) is a real finding, and the comparison against power-law and logistic alternatives is clear and useful. The practical regressions for K10 and related thresholds are a nice bonus, and the analyses of non-equilibrium initial sizes and Allee effects widen the scope. Code and data are deposited, which is exactly the kind of reproducibility that makes the empirical core credible.\n\nThe soft spot is Appendix F. It does not hold together as printed: the Fokker–Planck drift has a sign error, the diffusion term is written with x rather than x², the stated stationary Gamma distribution does not match the printed FPE, and the later 'Lamperti transform' is not a transform of the original SDE. With the correct noise amplitude, the OU approximation produces no K-dependence in survival over a fixed horizon; the K-dependence in t1(K) and equation (F15) appears to be an artifact of the rescaling. There is also no error bound for keeping only the first eigenvalue. More important, even a repaired derivation would at best yield the standard Gompertz (γ=1), not the modified form that the paper's empirical discovery hinges on. So the abstract's 'few assumptions suffice' claim overreaches. The authors admit γ is fitted and that they could not predict the modified Gompertz from life-history parameters, which is honest, but the framing in the introduction and abstract does not reflect that limitation.\n\nThe empirical content deserves a serious referee; the theory needs major repair or an explicit demotion to a provisional sketch. I would send it to review, but with a directive to fix Appendix F, provide bounds or drop the analytical claim, and reframe the paper as a well-supported empirical generalization with a tentative theoretical note.","headline":"The simulations are impressive and the empirical modified-Gompertz pattern is credible, but Appendix F's theoretical derivation is internally inconsistent and does not support the claimed universality.","tokens_in":25868,"tokens_out":2773,"would_cite":false,"duration_ms":31256,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D25","92D40","60J70"],"pacs":[],"model":"deepseek-v4-flash","headline":"Extinction risk versus carrying capacity collapses onto a single modified Gompertz curve.","keywords":["extinction risk","carrying capacity","modified Gompertz curve","population viability analysis","density dependence","stochastic population dynamics","Ornstein-Uhlenbeck approximation","conservation prioritization"],"falsifier":"For a viable parameter set with $r/\\sigma^2$ just above the extinction threshold and $K$ around 100, compute the exact first-passage-time survival probability of the logistic stochastic differential equation by summing the full spectral expansion rather than keeping only the first eigenvalue; if the exact curve deviates from the fitted modified Gompertz by more than the Monte Carlo error of 10,000 replicates, the claimed universality fails in that regime.","tokens_in":24762,"feed_emoji":"📉","tokens_out":7538,"duration_ms":67083,"temperature":0.7,"pith_summary":"This paper sets out to establish a universal quantitative law: for any population whose growth is limited by density dependence, the probability of extinction within a fixed horizon is a modified Gompertz function of carrying capacity, $P_E(K) = 1 - \\exp(-\\exp(a + b K^{\\gamma}))$. The claim is supported by roughly five billion individual-based simulations across four population models of increasing realism, with fits exceeding $r^2 = 0.995$ for every viable parameter combination, and by an analytic argument that starts from the logistic stochastic differential equation and reduces it near carrying capacity to an Ornstein-Uhlenbeck process. If the law holds, then the extinction-risk consequences of habitat loss can be read off a three-parameter curve, and conservation triage can be guided by where a population sits on that curve. The paper also shows that the standard Gompertz curve ($\\gamma = 1$) fits well but systematically underestimates risk at high carrying capacity, and that the modified curve's inflection point consistently lies at extinction probabilities above 0.5, around 0.65 to 0.85.","feed_headline":"Extinction risk follows one curve as habitat shrinks","feed_subtitle":"Billions of simulations plus a stochastic growth theory collapse extinction risk versus carrying capacity onto one curve.","key_machinery":"The load-bearing object is the modified Gompertz function $P_S(K) = \\exp(-\\exp(a + b K^{\\gamma}))$, where $\\gamma$ is a shape parameter that skews the curve and is consistently estimated below one across all models and most empirical datasets. The analytic machinery is the reduction of the logistic stochastic differential equation near carrying capacity to an Ornstein-Uhlenbeck process, followed by a first-eigenvalue approximation of the first-passage-time distribution to the extinction boundary; this turns an exponential decay in time into a double-exponential (Gompertz) dependence on $K$. The shape parameter $\\gamma$ corrects the systematic deviation of the standard Gompertz at high $K$, and the paper argues this improvement is systematic rather than a dimensionality artifact because $\\gamma$ clusters below one.","core_discovery":"The central discovery is that survival probability as a function of carrying capacity takes the form $P_S(K) = \\exp(-\\exp(a + b K^{\\gamma}))$, and therefore extinction probability follows a modified (asymmetric) Gompertz sigmoid. This pattern held across more than 1700 simulation settings spanning two-sex demography, biparental care, delayed maturation, and temporally autocorrelated environmental noise, and it also held when populations started away from carrying capacity or experienced Allee effects. The paper argues that the form is not empirical coincidence: linearizing the logistic stochastic differential equation around carrying capacity gives an Ornstein-Uhlenbeck process, and keeping only the slowest mode of its first-passage-time spectrum yields a Gompertz survival curve in $K$ under just two assumptions, negative density dependence and a viable steady state at large $K$. This is a broad universality claim: it does not depend on the specific mechanism of density dependence, and it implies that the extinction-risk-versus-capacity relationship for any extant population can be summarized by three parameters.","pith_inferences":["If the modified Gompertz is truly universal, then comparing species reduces to comparing three parameters, so a promising next step is to model $a$, $b$, and $\\gamma$ as functions of measurable life-history traits and phylogenetic relatedness; the paper only predicts the standard Gompertz parameters from traits.","The consistent $\\gamma < 1$ across models and empirical datasets hints that extinction risk is a stretched-exponential function of $K$; whether $\\gamma$ clusters by taxon or environmental regime is a testable extension not pursued here.","The analytic approximation should degrade for small carrying capacities or for parameter sets near the viability boundary, so a targeted simulation campaign at $K$ between 1 and 100 would map the curve's domain of validity; the paper does not report such a boundary.","Because the curve is asymmetric, the extinction-risk impact of a proportional habitat loss depends on where the population starts on the curve, implying that safe operating spaces for habitat retention should be defined relative to each population's inflection point."],"forward_implications":["A power-law or logistic curve misstates how extinction risk changes with carrying capacity; for populations at low risk, a power law underestimates the rate of change, and a logistic places the maximum benefit of intervention at $P_E = 0.5$ rather than at the observed $0.65$ to $0.85$.","For a population already at risk, the largest marginal reduction in extinction probability from increasing carrying capacity occurs when current extinction probability is roughly 0.65 to 0.85, which can inform where to target habitat protection or restoration.","The carrying capacity at which a population has a 10% chance of extinction in 100 years ($K_{10}$, aligned with IUCN Criterion E) can be estimated from growth rate and environmental stochasticity via linear models, at least for the simpler population structures, offering a rapid assessment tool.","The same Gompertz relationship holds when initial population size differs from carrying capacity and when Allee effects are present, so the curve can be applied to reintroductions, newly protected habitats, and populations with mate-finding or cooperative-breeding constraints."],"supporting_citations":[{"why":"Re-analysed Hilbers et al. data and found that a Gompertz curve fits most mammals, the immediate predecessor this paper extends.","marker":"Wolff et al. (2023)"},{"why":"Derived a power-law extinction-risk relation that the paper shows deviates strongly from simulated curves.","marker":"Lande (1993)"},{"why":"Studied scaling of the stochastic logistic model and suggested sigmoidal relationships, providing a baseline to compare against.","marker":"Dushoff (2000)"},{"why":"Observational data on isolated bighorn sheep populations used as the first empirical test of the modified Gompertz.","marker":"Berger (1990)"},{"why":"Modelling study of New Zealand brown mudfish carrying capacities used as the second empirical test.","marker":"White et al. (2017)"},{"why":"Grizzly bear population viability estimates used as the third empirical test.","marker":"Shaffer and Samson (1985)"},{"why":"Supplies the large-time, first-eigenvalue survival approximation for the Ornstein-Uhlenbeck process used in the analytic derivation.","marker":"Giorgini et al. (2020)"},{"why":"Provides the spectral first-passage-time representation for the Ornstein-Uhlenbeck process that underlies the analytic Gompertz result.","marker":"Ricciardi and Sato (1988)"}],"fun_headline_variants":["Extinction risk collapses onto one universal curve","Billions of simulations reveal a single extinction law","One curve links extinction risk and carrying capacity","Universal Gompertz curve predicts extinction risk","Extinction risk has a universal shape: Gompertz"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The universal curve rests on treating a population near carrying capacity as an Ornstein-Uhlenbeck process and on assuming that, over 100 years, survival is dominated by the slowest decaying mode of that process; if that spectral approximation fails at moderate carrying capacities, the analytic claim that every density-dependent population follows a Gompertz curve is unsupported even if the simulation fits remain excellent.","fun_headline_variants_meta":{"raw":{"variants":["Extinction risk collapses onto one universal curve","Billions of simulations reveal a single extinction law","One curve links extinction risk and carrying capacity","Universal Gompertz curve predicts extinction risk","Extinction risk has a universal shape: Gompertz"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000727,"raw_usage":{"total_tokens":3224,"prompt_tokens":876,"completion_tokens":2348,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":2278}},"tokens_in":492,"tokens_out":2348,"duration_ms":17604,"temperature":1.0,"reasoning_tokens":2278,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:42:07.870453+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a viable parameter set with $r/\\sigma^2$ just above the extinction threshold and $K$ around 100, compute the exact first-passage-time survival probability of the logistic stochastic differential equation by summing the full spectral expansion rather than keeping only the first eigenvalue; if the exact curve deviates from the fitted modified Gompertz by more than the Monte Carlo error of 10,000 replicates, the claimed universality fails in that regime.","supporting_citations":[],"review_version":1}