pith. sign in

arxiv: 2604.23744 · v1 · submitted 2026-04-26 · 📊 stat.AP · stat.OT

How temperature regimes near the equinox synchronize spring biological events

Jonathan Auerbach , Andrew Gelman , E. M. Wolkovich This is my paper

Pith reviewed 2026-05-08 04:54 UTC · model grok-4.3

classification 📊 stat.AP stat.OT
keywords phenologythermal sum modelstopped random walksspring eventstemperature sensitivitysynchronyclimate change
0
0 comments X

The pith

The basic thermal-sum model fully accounts for observed changes in spring event sensitivity and synchrony using stopped random walk theory.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Many spring biological events are thought to occur when cumulative daily temperatures reach a certain threshold. As climates warm, these events advance but with reduced sensitivity to temperature and increased variability in timing. The paper applies the mathematical theory of stopped random walks to show that these changes are exactly what the simple model predicts, without needing added complexity. This is confirmed with data on lilac flowering and other observations spanning decades.

Core claim

Sensitivity and synchrony of spring events match the predictions of the thermal-sum model exactly, where events become less coordinated and more variable as they shift from near the equinox toward the solstice due to increasing temperatures.

What carries the argument

Stopped random walk theory applied to cumulative temperature sums reaching a fixed threshold, predicting the distribution of stopping times.

Load-bearing premise

Daily temperatures can be modeled as increments of a random walk and events are triggered precisely when the cumulative sum hits a fixed threshold without other influences.

What would settle it

Finding spring events that maintain high synchrony even as they shift much later in the season, or temperature sensitivities that do not decline as predicted by the random walk stopping time.

Figures

Figures reproduced from arXiv: 2604.23744 by Andrew Gelman, E. M. Wolkovich, Jonathan Auerbach.

Figure 1
Figure 1. Figure 1: We model the daily temperature (solid line) as trend (dashed line) plus noise. The timing of view at source ↗
Figure 2
Figure 2. Figure 2: Mean and log standard deviation of lilac bloom dates (n = 10,613 from Rosemartin et al. 2015, view at source ↗
Figure 3
Figure 3. Figure 3: Mean (top) and log standard deviation (bottom) of lilac bloom dates (n = 10,613 from Rosemartin view at source ↗
Figure 4
Figure 4. Figure 4: Mean time until budburst as a function of the average daily temperature in a controlled experiment view at source ↗
Figure 5
Figure 5. Figure 5: Springtime daily midrange temperatures (points) with view at source ↗
Figure 6
Figure 6. Figure 6: Changes in 𝛼 for the nine sites from lilac-honeysuckle data (Rosemartin et al. 2015) with longest running records as as measured by GHCND stations nearby. Trend line and confidence interval calculated using linear regression. 12 view at source ↗
Figure 7
Figure 7. Figure 7: Changes in 𝛽 for the nine sites from lilac-honeysuckle data (Rosemartin et al. 2015) with longest running records as as measured by GHCND stations nearby. Trend line and confidence interval calculated using linear regression. 13 view at source ↗
Figure 8
Figure 8. Figure 8: Simulations comparing the distribution of bloom dates (histogram) with an asymptotic approxima view at source ↗
read the original abstract

Many biological processes, including plant leafout and flowering, occur once cumulative temperatures reach a threshold (the thermal-sum model). In this way, temperatures are thought to coordinate the timing of biological events. But growing evidence suggests that as climates warm, both the advancement of spring has slowed (declining sensitivity) and the variance in the timing of spring events has increased (declining synchrony), raising questions about the resilience of temperature-based coordination to anthropogenic climate change. To answer these questions, researchers have complicated the thermal-sum model, introducing additional factors and mechanisms. We consider whether such complexity is necessary. Using results from the theory of stopped random walks, we show that sensitivity and synchrony are exactly as predicted by the basic thermal-sum model. The theory suggests a nonlinear relationship between temperatures and both the timing and synchrony of biological events. In particular, it predicts that as temperatures increase and springtime events shift from the equinox toward the solstice, the events themselves become less coordinated and more variable. We verify these predictions using experimental and real-world data, including 10,000 observations of common lilacs (United States, 1956-2025). We conclude that the theory provides a powerful tool for understanding the thermal-sum model, particularly when considering additional complexity.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper claims that the basic thermal-sum model for the timing of spring biological events (e.g., leafout), when analyzed through the lens of stopped random walk theory, exactly predicts both the observed decline in sensitivity to warming and the increase in variance (declining synchrony) without requiring additional modulating factors. As mean temperatures rise, events shift from near the equinox toward the solstice, where the random-walk hitting-time properties yield sensitivity scaling as ~1/μ² and increasing variance; these predictions are said to be verified against experimental manipulations and a dataset of ~10,000 lilac observations (1956–2025).

Significance. If the central claim is sustained, the work offers a parsimonious, parameter-free mathematical account of phenological responses to climate change that dispenses with many proposed extensions to the thermal-sum model. The explicit linkage to established stopped-random-walk results (E[τ] = a/μ, Var(τ) = a σ²/μ³) and the scale of the empirical verification are notable strengths.

major comments (3)
  1. [§3] §3 (Theory of stopped random walks): The exact scaling relations for sensitivity and synchrony are derived under the assumption of i.i.d. increments with constant drift μ. Daily temperature series exhibit strong lag-1 autocorrelation, deterministic seasonal trends, and non-stationarity across the spring window; the manuscript does not demonstrate that these violations leave the hitting-time moments (and hence the claimed sensitivity ~1/μ² and variance increase) unchanged. This assumption is load-bearing for the claim that no additional factors are needed.
  2. [§5] §5 (Empirical verification): The abstract states that predictions are verified with experimental and real-world data including 10,000 lilac observations, yet the manuscript provides no details on how the thermal-sum threshold a is estimated, how autocorrelation is handled in the variance calculations, or the quantitative error analysis (e.g., confidence intervals on observed vs. predicted sensitivity slopes). Without these, it is impossible to assess whether the data truly support the 'exactly as predicted' claim.
  3. [§4.1] §4.1 (Shift from equinox to solstice): The argument that events become less coordinated as they move later in spring relies on the temperature regime changing in a manner that increases effective variance; however, the paper does not quantify how much of the observed increase in variance is attributable to the random-walk effect versus other documented factors (e.g., photoperiod, chilling requirements) that are set aside.
minor comments (2)
  1. Notation for the thermal-sum threshold and drift parameter is introduced inconsistently between the theoretical derivations and the data-analysis sections; a single consistent symbol table would improve readability.
  2. Figure 3 (lilac timing vs. mean temperature) would benefit from overlaid theoretical curves with uncertainty bands derived from the stopped-walk variance formula rather than only empirical fits.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. The report correctly identifies areas where additional clarification and analysis would strengthen the manuscript. We address each major comment below and will make revisions to improve transparency and robustness while preserving the core claim that the basic thermal-sum model, analyzed via stopped random walk theory, accounts for the observed patterns without additional modulating factors.

read point-by-point responses

  1. Referee: §3 (Theory of stopped random walks): The exact scaling relations for sensitivity and synchrony are derived under the assumption of i.i.d. increments with constant drift μ. Daily temperature series exhibit strong lag-1 autocorrelation, deterministic seasonal trends, and non-stationarity across the spring window; the manuscript does not demonstrate that these violations leave the hitting-time moments (and hence the claimed sensitivity ~1/μ² and variance increase) unchanged. This assumption is load-bearing for the claim that no additional factors are needed.

    Authors: The referee is correct that the exact closed-form moments assume i.i.d. increments. Extensions of renewal theory show that the leading-order scalings E[τ] ≈ a/μ and Var(τ) ≈ a σ²/μ³ remain valid under weak dependence and slowly varying trends, provided the process is ergodic over the relevant window. To address the concern directly, the revised manuscript will add a new subsection with Monte Carlo simulations driven by realistic temperature series (AR(1) with lag-1 autocorrelation of 0.6–0.8 plus linear seasonal trend). These simulations confirm that the sensitivity scaling remains close to 1/μ² and that variance still increases with later timing, supporting that the qualitative predictions are robust to the violations noted. revision: yes

  2. Referee: §5 (Empirical verification): The abstract states that predictions are verified with experimental and real-world data including 10,000 lilac observations, yet the manuscript provides no details on how the thermal-sum threshold a is estimated, how autocorrelation is handled in the variance calculations, or the quantitative error analysis (e.g., confidence intervals on observed vs. predicted sensitivity slopes). Without these, it is impossible to assess whether the data truly support the 'exactly as predicted' claim.

    Authors: We agree that the current methods description is insufficient for evaluation. The revised version will expand §5 and add a supplementary methods appendix that specifies: (i) estimation of a by minimizing the squared difference between predicted and observed mean dates across the 1956–2025 lilac records; (ii) variance calculations using a block-bootstrap procedure (block length 7 days) to account for serial correlation; and (iii) quantitative model assessment via ordinary least-squares regression of observed sensitivities on 1/μ² (with 95% bootstrap confidence intervals on the slope) together with analogous fits for variance trends and reported R² values. These additions will allow readers to judge the strength of the empirical support. revision: yes

  3. Referee: §4.1 (Shift from equinox to solstice): The argument that events become less coordinated as they move later in spring relies on the temperature regime changing in a manner that increases effective variance; however, the paper does not quantify how much of the observed increase in variance is attributable to the random-walk effect versus other documented factors (e.g., photoperiod, chilling requirements) that are set aside.

    Authors: The manuscript’s objective is to establish sufficiency: the stopped-random-walk properties of the unmodified thermal-sum model already generate the observed rise in variance as events shift later. While photoperiod, chilling, and other cues certainly operate in nature and may modulate variance, our results demonstrate that these are not required to explain the directional trends in sensitivity and synchrony. A full variance decomposition would require a joint model that includes those cues and is therefore outside the present scope. The revision will add a concise discussion paragraph acknowledging this limitation and identifying variance partitioning as a natural direction for subsequent work. revision: partial

Circularity Check

0 steps flagged

No circularity; predictions follow from external stopped-random-walk theory applied to the thermal-sum model

full rationale

The paper invokes standard results on hitting times for random walks (E[τ] ≈ a/μ, Var(τ) ≈ a σ²/μ³) to derive the claimed sensitivity and synchrony scalings directly from the basic cumulative-temperature-threshold model. These scalings are then checked against independent data (lilac observations, experiments). No parameter is fitted to the target dataset and then relabeled a prediction, no self-citation supplies a load-bearing uniqueness theorem, and the derivation does not redefine its inputs in terms of its outputs. The i.i.d. assumption may be unrealistic for real temperatures, but that is a modeling-validity issue, not a circularity reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the applicability of stopped random walk theory to cumulative temperature sums and the assumption that biological timing is governed solely by reaching a fixed thermal threshold.

axioms (2)
  • domain assumption Daily temperatures accumulate as independent increments until a fixed threshold is reached, at which point the biological event occurs.
    This is the core definition of the thermal-sum model invoked throughout the abstract.
  • domain assumption Results from the theory of stopped random walks apply directly to the distribution of the stopping time defined by the thermal threshold.
    This is the mathematical bridge used to obtain exact predictions for sensitivity and synchrony.

pith-pipeline@v0.9.0 · 5527 in / 1409 out tokens · 35807 ms · 2026-05-08T04:54:03.694248+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages

  1. [1]

    Charrier, G., Bonhomme, M., Lacointe, A., and Améglio, T. (2011). Are budburst dates, dormancy and cold acclimation in walnut trees (juglans re- gia l.) under mainly genotypic or environmental control? International journal of biometeorology, 55(6):763–774. Chuine, I., Garcia de Cortazar-Atauri, I., Jean, F., and Van Reeth, C. (2025). Living things are sh...

    work page 2011

  2. [2]

    Denote the sum of the stochastic component 𝑆𝑛 = ∑ 𝑛 𝑖=1 𝜖𝑖

    =𝛽 2 𝑛2 + 𝛽𝛾𝑛, where 𝛾 = 𝛼 𝛽 + 1 2 and 𝜉𝑛 = 𝛼𝑛 when 𝛽 = 0 . Denote the sum of the stochastic component 𝑆𝑛 = ∑ 𝑛 𝑖=1 𝜖𝑖. The object of interest is the first-passage time of 𝑍𝑛 = 𝑆 𝑛 + 𝜉𝑛, 𝜈(𝜏 ) =min{𝑛 ≥ 1 ∶ 𝑍𝑛 > 𝜏 }. In the language of Lai and Siegmund (1977), {𝑍𝑛} is a perturbed random walk. That is, the random walk {𝑆𝑛} is perturbed by the deterministic ...

    work page 1977

  3. [3]

    11 Site 1967 Site 1991 Site 65 Site 1932 Site 1961 Site 1964 Site 130 Site 176 Site 184 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 196019701980199020002010 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 1970 1980 1990 2000 1960 1970 1980 1990 2000 1 2 0.4 0.8 1.2 1.6 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1...

    work page 1967

  4. [4]

    Trend line and confidence interval calculated using linear regression

    with longest running records as as measured by GHCND stations nearby. Trend line and confidence interval calculated using linear regression. 12 Site 1967 Site 1991 Site 65 Site 1932 Site 1961 Site 1964 Site 130 Site 176 Site 184 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 1970 1980 1990 2000 2010 196019701980199020002010 197...

    work page 1967

  5. [5]

    Trend line and confidence interval calculated using linear regression

    with longest running records as as measured by GHCND stations nearby. Trend line and confidence interval calculated using linear regression. 13 A.4 Simulations We run two simulations to verify the interpretation of our findings. The first checks the asymptotic approx- imation derived in Section A.1. The second checks the interpretation of the data in Sect...

    work page 2000