Body size predicts how long ant workers live - but not how they age or how they die from heat

RESEARCH ARTICLE Physiological Ecology

Body size predicts how long ant workers live — but not how they age or how they die from heat

arXiv:2608.14245v1 [q-bio.PE] 14 Aug 2026

Alana Moscardi1

|

Rafael da Silva2

1 Graduate Programme in Ecology, Instituto

Nacional de Pesquisas da Amazônia (INPA), Manaus, Brazil 2 Graduate Programme in Data Science,

|

Gleycon Silva3

1. In social insects, ‘mortality risk’ bundles three components that need not share a common predictor: duration (how long a worker lives on average), senescence trajectory (whether mortality accelerates with age,

Eastern University, St. Davids, PA, USA

captured by the Weibull shape parameter ρ), and thermal vulnerability

3 Graduate Programme in Ecology, Instituto

(how strongly mortality hazard rises with temperature).

Nacional de Pesquisas da Amazônia (INPA), Manaus, Brazil

2. We test all three axes in 18 Australian ant species via paired field– laboratory survival assays (2,363 cohort-day observations; 1,148 workers); no single predictor is detectable across all three axes at this sam-

Correspondence Alana Moscardi, Graduate Programme in Ecology, Instituto Nacional de Pesquisas da Amazônia (INPA), Manaus, AM, Brazil Email: alanamoscardi1@gmail.com

ple size. 3. Body size predicts duration (Cox HR = 0.67, p = 0.002); colony size (p = 0.60) and the size × temperature interaction (p = 0.72) show no detectable moderating effect. The body-size × foraging-rate interaction is weak but significant (LRT p = 0.014). By exclusion of

Funding information [To be completed before submission]

the two null extrinsic moderators, intrinsic physiology remains the most parsimonious explanation for the main size–longevity pattern, though the foraging-rate interaction warrants cautious interpretation (VIF ≈ 4) and unmeasured confounders cannot be ruled out at N = 18 species. 4. Senescence trajectory is associated with circadian niche, not size: steepest in matinal species (Kruskal–Wallis p = 0.009; matinal vs. crepuscular p = 0.002), and uncorrelated with body mass (Spearman p = 0.32). 5. Thermal hazard plateaus above T ∗ = 20 °C (∆AIC = −38; p < 0.001); elevated thermal sensitivity observed in the Rhytidoponera lineage (Ectatomminae) above the plateau (5 %/°C, p = 0.015). 6. Circadian regime and lineage identity — not body size — emerge as the climate-relevant axes, although they are strongly collinear in this assemblage (Cramér’s V = 0.85) and cannot be fully disentangled here; they converge on matinal Rhytidoponera (Ectatomminae) as the can-

Abbreviations: KM, Kaplan–Meier; PH, proportional hazards; HR, hazard ratio; CI, confidence interval; AIC, Akaike information criterion; LRT, likelihood-ratio test; IQR, interquartile range; ROS, reactive oxygen species

1

2

Moscardi et al.

didate most-vulnerable lineage. Because hazard plateaus near the assemblage mean, the operative climate variable is the frequency of extreme-heat days rather than mean warming. The practical implication is concrete and testable: size-based vulnerability indices, which capture only the duration axis, will systematically misrank the most exposed taxa in assemblages where circadian niche and thermal sensitivity are decoupled from body size — a pattern documented here and expected wherever matinal lineages co-occur with larger, thermally buffered species. KEYWORDS

circadian niche, climate vulnerability, Formicidae, thermal mortality, Weibull senescence, worker longevity

1

|

INTRODUCTION

Body size is among the most intensively studied life-history traits in insects, and across social insects larger workers generally live longer — a pattern documented directly in ants (Riskas et al., 2026). This robust size–longevity relationship is so well established that body size is often used as a one-dimensional proxy for mortality risk in comparative and applied work. But ‘mortality risk’ bundles together properties that need not share the same predictor: how long a worker lives on average, how steeply its risk of death accelerates with age, and how vulnerable it is to environmental extremes such as heat. If these components have distinct drivers, then a size-centric framework will systematically mispredict which workers — and which lineages — are most at risk under environmental change. Riskas et al. (2026) documented a strong size–longevity effect in this same assemblage of 18 Australian ant species, but did not partition mortality risk into its component axes, test whether size buffers thermal mortality, or examine the trajectory of senescence. The present study extends that work by addressing three questions that are jointly sufficient to test whether body size is a strong comparative predictor of ant mortality risk in the broad sense, or only its most commonly measured component. We organise the study around three questions, each resolved by a falsifiable claim (Table 1). Q1 — What kind of mechanism links body size to longevity? Three non-exclusive hypotheses dominate the literature: (i) an intrinsic-physiology mechanism, in which larger workers have lower mass-specific metabolic rates (West et al., 2001; Brown et al., 2004) and plausibly reduced oxidative stress (Kramer et al., 2021); (ii) an extrinsic-buffering mechanism, in which greater mass buffers workers against desiccation (Lighton et al., 1994; Bujan et al., 2016) and heat stress (Baudier and O’Donnell, 2018); and (iii) a colony-level dilution hypothesis: that larger colonies weaken the perworker size benefit. Discriminating among these mechanisms requires testing whether the size benefit is modulated by colony size, foraging context, or temperature — questions unaddressed by prior work on this assemblage. Q2 — What shapes the trajectory of mortality? Unlike mean lifespan, the Weibull shape parameter ρ captures whether mortality accelerates with age (senescence, ρ > 1) or decelerates (frailty selection, ρ < 1) — a property that is, at least in principle, independent of average longevity (Ricklefs, 2010; Dahlgren et al., 2016; Lemaitre et al., 2024). Because trajectory integrates an organism’s physiological and ecological context, we ask whether it tracks body size — as a naïve extension of the size–longevity rule would predict — or instead is associated with circadian activity regime, i.e. the timing of foraging relative to the daily temperature cycle. Q3 — How does temperature act, and is thermal risk taxon-specific? Climate change is increasing global mean

Moscardi et al.

3

temperatures and the frequency of extreme heat events; in south-eastern Australia specifically, these trends are welldocumented (Herold et al., 2021; Bureau of Meteorology, 2025; IPCC, 2021; Perkins-Kirkpatrick and Lewis, 2020). Whether thermal mortality scales linearly with temperature, and whether all ant lineages are equally vulnerable or vulnerability is structured by lineage identity, is central to predicting community-level responses to warming yet is rarely quantified at the subfamily level; existing thermal surveys operate at the species level (Diamond and Chick, 2017; Kaspari et al., 2015). Body size is a strong comparative predictor of worker longevity, but its explanatory scope across other axes of mortality risk remains uncertain. By answering these three questions within a single, unified framework on one paired field–laboratory system, we are able to test not only each component in isolation but also whether they share a common predictor — and we find no shared predictor detectable at N = 18 species. Figure 1 summarises the conceptual framework.

F I G U R E 1 Conceptual framework: three decoupled axes of worker mortality risk. Body size predicts only the duration of life (HR = 0.67); it is statistically decoupled from the senescence trajectory (ρ–mass Spearman p = 0.32) and from thermal vulnerability (size × temperature p = 0.72). The trajectory axis is associated with circadian activity regime (matinal species senesce fastest; Kruskal–Wallis p = 0.009), and the thermal axis is lineage-patterned (elevated thermal sensitivity in the Rhytidoponera lineage (Ectatomminae); PGLS p = 0.081, corroborated by individual-level Cox p = 0.015). Circadian regime and lineage identity are strongly collinear in this assemblage (Cramér’s V = 0.85) and cannot be fully separated at N = 18 species; they are therefore drawn as complementary, not independent, predictors. The three climate-relevant signals converge on matinal Rhytidoponera (Ectatomminae). Each axis is labelled in text on the schematic so the diagram is interpretable independently of colour. Abbreviations: HR, hazard ratio (Cox proportional-hazards model); ρ, Weibull shape parameter (senescence rate); KW, Kruskal–Wallis test; PGLS, phylogenetic generalised least squares; Cramér’s V , association statistic between two categorical variables.

2

|

MATERIAL AND METHODS

Each of the three claims is evaluated through a set of falsifiable sub-tests (H1–H9), described in full below and summarised, with outcomes, in Table 1. In brief, Claim 1 (§2.3) asks whether the size–longevity benefit is genuine and, if so, whether it is modulated by colony size, foraging rate, or temperature — the pattern that distinguishes an intrinsicphysiology mechanism from extrinsic buffering or colony-level dilution. Claim 2 (§2.4) asks whether the shape of

4

Moscardi et al.

the mortality trajectory (the Weibull senescence rate ρ) is associated with circadian activity regime rather than body size. Claim 3 (§2.5) characterises the temperature–hazard function — its shape, and whether thermal vulnerability is concentrated in particular lineages. Throughout, the internal labels H1–H9 cross-reference the methods, tables, supplementary material, and analysis code.

2.1

|

Levels of inference

Analyses operate at three nested levels, and results from different levels are not treated as commensurate without qualification. Cohort-day (individual survival process) (N = 2,363 observations; 1,826 events; 106 cohorts): Cox models M0/M1/M2, the piecewise temperature spline, and subfamily × temperature interactions. Colony (N = 39): OLS tests of the body-size × colony-size moderator (H2a). Species (N = 18, or N = 14–15 after phylogenetic or cohort filters): Weibull ρ, Kruskal–Wallis tests of circadian niche, Spearman correlations, PGLS, and variance partitioning of ρ. Each claim states its primary inferential unit; species-level tests are underpowered relative to cohort-day Cox models, and genus- or subfamily-level contrasts resting on few tip taxa are interpreted as provisional. Our approach is survival modelling — estimating how instantaneous mortality hazard changes as a function of candidate predictors — not bivariate association. Cox proportional-hazards models provide directional, time-to-event inference with censoring, clustering, and simultaneous covariates; where we report correlations (e.g. Spearman ρ), they serve as auxiliary checks, not as the primary evidence. The inferential strategy rests on mechanistic exclusion: by testing three candidate moderators of the size–longevity effect — finding colony size and temperature null, and a weak but significant foraging-rate interaction — we narrow the mechanistic window rather than simply documenting a relationship. Finally, we test for nonlinearity in the temperature–hazard function (piecewise spline with Davies permutation) and for effect heterogeneity across subfamilies (interaction terms), both of which go beyond what a correlational framework can address.

2.2

|

Study system and data

We re-analyse the field mark–recapture dataset of Riskas et al. (2026), comprising 18 Australian ant species monitored at the La Trobe Wildlife Sanctuary, a 28-ha reserve on La Trobe University’s Melbourne campus, Victoria, in temperate south-eastern Australia (∼37.7◦ S, 145.1◦ E). The site has a mesic, temperate climate (mean annual minimum and maximum air temperatures of 9.7 °C and 20.2 °C; ∼656 mm annual rainfall; Bureau of Meteorology 2024). The field dataset contains 2,363 cohort-day observations spanning 106 cohorts across 39 colonies, collected between February 2022 and May 2023, with 1,826 recorded daily mortality events. Daily maximum temperature (maxt) averaged 22.8 °C (range 13.1–40.5 °C). For laboratory comparisons (Claims 1 and 2) we use the paired laboratory dataset of 1,148 individual workers (39 colonies, 18 species) maintained under controlled temperature and humidity. Full fieldcollection protocol and laboratory rearing conditions are described in Riskas et al. (2026). Descriptive statistics for all 18 species are given in Table 2. Species were chosen to span the full range of body sizes, subfamily identities, and circadian activity regimes present in this temperate assemblage, following the sampling design of Riskas et al. (2026). Colonies were recruited opportunistically at the La Trobe Wildlife Sanctuary by active search and pitfall trapping, with inclusion conditional on at least two successfully marked cohorts per colony. The resulting sample — 18 species across five subfamilies — captures enough interspecific variation to estimate Cox and Weibull parameters, but at N = 18 species it is too small to support fully factorial inference across all three mortality axes simultaneously; we return to this constraint in the Limitations (§4.6).

Moscardi et al.

5

TA B L E 1 Analytical roadmap: questions, claims, internal hypotheses, and outcomes. Internal labels (H1–H9) crossreference the methods, code, and supplementary tables. Analysis

Outcome

Q1. What mechanism links body size to longevity? Claim 1. Body size predicts duration; intrinsic physiology inferred by exclusion. Larger workers live longer (H1)

✓ HR = 0.67, p = 0.002

Colony size does not modulate it (H2a)

◦ p = 0.60

Foraging rate modulates it weakly (H2b)

✓ LRT p = 0.014 (M2)

Body size does not buffer heat (H3)

◦ HR = 1.00, p = 0.72

Size effect stronger in field (H2c)

▷ 1.28×

Q2. What shapes the trajectory of mortality? Claim 2. Trajectory associated with circadian niche, not body size. Senescence is the rule (H4) Trajectory is associated with circadian niche (H5) Field steepens senescence (H6) Trajectory unrelated to body mass

✓ 14/18 ρ > 1 ✓ KW p = 0.009 ✓ 17/18, p < 0.001 ◦ Sp. ρ = −0.25, p = 0.32

Q3. How does temperature act? Is risk taxon-specific? Claim 3. Thermal risk nonlinear and taxonomically patterned. Temperature raises hazard (H7)

✓ HR = 1.036, p < 0.001

Hazard plateaus above 20 °C (H8)

✓ T ∗ = 20 °C, p < 0.001

Vulnerability is subfamily-specific (H9) ✓

Supported.

◦ Null result.

✓ joint p = 0.017

▷ Directional (consistent, not conclusive).

Abbreviations: HR, hazard ratio (Cox proportional-hazards model); KW, Kruskal–Wallis test; Sp., Spearman rank correlation; T ∗ , estimated temperature breakpoint; ρ, Weibull shape parameter. Fractions of the form X /18 refer to the 18 study species. Internal hypothesis labels (H1–H9) are defined in the Methods (§2.3–§2.5).

†

Moscardi et al. 6

61 6

8

7

64.5

94.5

76.5

44.0

81.0

0.73

0.86

0.93

0.74

0.74

23.5

24.4

26.0

22.4

23.3

mean (°C)

0.96

0.89

1.15

1.08

1.70

1.99

1.38

ρlab

5.15

5.65†

4.89

2.85

3.59

2.94

0.54

ρfield

TA B L E 2 Species-level descriptive statistics for the 18 study species, grouped by subfamily and sorted by log10 body mass within subfamily. Columns: log10 mass (mg), mean worker wet mass; Colony size, mean number of workers per colony; Cohorts (N ), number of field mark–recapture cohorts; Median obs (d), median per-worker observation window in days; Foraging rate, mean workers leaving the nest per minute; maxt mean (°C), mean daily maximum air temperature experienced; ρlab , Weibull shape parameter from individual laboratory survival; ρfield , Weibull shape parameter from field cohort survival. For all ρ, values > 1 indicate senescence, ≈ 1 constant hazard, and < 1 frailty selection. Morphospecies designators (e.g. Camponotus claripes A, Papyrius A) follow the genus name in roman type.

194

8

23.7

maxt

0.66 257

4

26.7

rate

Subfamily

0.37

192

0.70

Foraging

Ectatomminae −0.18

161

1.81

obs (d)

Species

Ectatomminae

0.94

61.0

61.0

Median

Rhytidoponera metallica Ectatomminae

0.33

2

(N )

Rhytidoponera tasmaniensis

Formicinae

8

Cohorts

Rhytidoponera victoriae

Formicinae

871

295

size

Camponotus consobrinus

−0.81

−0.09

Colony

Camponotus claripes A Formicinae

Formicinae

log10

Camponotus nigroaeneus

mass (mg)

Melophorus turneri

1.46

1.82

1.85

2.12

2.60

1.34

5.65†

3.28

1.25

24.8

1.47

4.69

4.47

25.1

25.9

1.56

0.83

0.87

26.1

1.07

0.81 96.0

1.06

2.64

22.3

24.6

75.5

0.84

25.3

20.6 8

83.5

7.25

0.64

2

123.0

8.54

3.79

3.99

0.66 179

10

89.5

1.12

91.0 0.09

472

2123

5

50.0

25.2

2.51

1.01

121.5

−0.45

6

2.10

21.2

25.9

3

−0.68

142

8

57.0

1.58

5.58

6

Myrmicinae

−0.71

7964

16666

73.0

150

Meranoplus fenestratus Myrmicinae

Myrmicinae

−0.33

5

70.0

127

Crematogaster laeviceps Myrmicinae

−0.59

3155

2

8

0.98

Chelaner cinctum

Dolichoderinae

5711

1.21

Meranoplus diversus

Dolichoderinae

−0.61

1006

Myrmeciinae

Anonychomyrma nitidiceps

−0.89

Myrmeciinae

Iridomyrmex septentrionalis

Dolichoderinae

−0.95

Myrmecia pyriformis

Iridomyrmex splendens

Dolichoderinae

Dolichoderinae

Myrmecia piliventris

Papyrius A

5.65†

Iridomyrmex notialis

Field ρ estimated from pooled assemblage AFT fallback (fewer than 3 cohorts available for separate Weibull fit).

Moscardi et al.

7

F I G U R E 2 Representative study organism: Rhytidoponera sp. (Ectatomminae). A worker of Rhytidoponera sp., one of three ectatomminine species included in the field mark–recapture dataset (Ectatomminae: R. metallica, R. tasmaniensis, R. victoriae). This subfamily is distinguished by notably low colony size (mean ≈170 workers) relative to myrmicine and dolichoderine species, the steepest actuarial senescence trajectories in laboratory survival assays (ρlab ≈ 1.4–2.0), and elevated thermal hazard sensitivity in the field. Photo: Rhytidoponera sp. by P. Kavanagh (https://www.flickr.com/photos/patrick_k59/49413794622), CC BY 2.0.

|

Replication Statement

Replication Statement (Functional Ecology requirement) Scale of inference

Factor(s) applied

Number of replicates

Species (N = 18)

Body mass (log10 , centred); circadian niche (3 groups); subfamily (5

18 species

groups) Cohort

Daily mortality event; observation day

106 cohorts, 39 colonies

Assemblage

Daily maximum air temperature (maxt, °C)

2,363 cohort-day observations

Individual (lab)

Body mass; colony identity

1,148 workers, 39 colonies

Each of the three claims maps to a single pre-specified primary endpoint (Table 1): the Cox M0 body-size hazard ratio for duration (H1); the laboratory Weibull shape ρlab , tested across circadian niches by Kruskal–Wallis, for senescence trajectory (H5); and a piecewise Cox model selected by AIC, with breakpoint significance from a Davies permutation test, for the thermal plateau (H8). We control confounding through three complementary devices rather than a single adjustment: cluster-robust standard errors (cluster = cohort) absorb within-cohort autocorrelation; subfamily fixed effects absorb shared phylogenetic baseline hazard; and a suite of sensitivity analyses (species fixed effects, stratified Cox, and exclusion of pooled-Weibull fallback species) probes robustness to the strongest assumptions. Because the three claims address orthogonal null hypotheses by design, we do not apply a multiple-comparison correction across them.

8

Moscardi et al.

2.3

|

Duration models — Cox proportional hazards (Claim 1)

We fit Cox proportional-hazards (PH) models to the cohort-day dataset with cluster-robust standard errors (cluster = cohort_id) to account for within-cohort temporal autocorrelation. The event is the daily mortality indicator (event_wd), the time variable is days since cohort formation (Day), and the focal predictor is log10 worker body mass (logw_mass, centred as logw_c). Subfamily (SF) is a fixed effect with Dolichoderinae as reference. The proportional hazards assumption was verified using Schoenfeld residuals (lifelines proportional_hazard_test, rank transform): no covariate showed a significant time-trend (p ≥ 0.17 for all predictors including body size and maxt; Table S1 in supplementary materials). As a species-frailty robustness check (supplementary Table S2), we re-fitted model M0 with (i) species fixed effects (C(SPEC), 18 dummies) and (ii) stratification by species (strata = [SPEC]). Both variants yielded a directionally consistent body-size coefficient (HR = 0.77–0.71, respectively), though non-significant within species where intraspecific size variation is narrow; the hazard ratio obtained from M0 therefore reflects a predominantly interspecific size–longevity relationship. To test whether genus-level clustering further confounds the result (several genera contribute two or three species: Rhytidoponera, Iridomyrmex, Camponotus, Meranoplus), we attempted to fit M0 stratified by genus. The model failed to converge due to matrix singularity, indicating that genus and subfamily structures are too collinear to separate at N = 18; the subfamily fixed-effect parameterisation of M0 is therefore retained. The analysis variables are derived from the raw mark–recapture records as follows. The cohort-day table has one row per cohort-day combination, with the event set to event_wd (1 if at least one worker died on that calendar day, 0 otherwise) and time measured as days since cohort initiation (Day). Body mass (logw_mass) is the log10 of worker wet mass in mg, centred at the assemblage mean (logw_c = logw_mass − x̄ ). Daily maximum temperature (maxt) is taken from the nearest Bureau of Meteorology station (Bundoora, ∼1 km from the site) and, for interaction models, centred analogously (maxt_c). The three species with fewer than three field cohorts (Camponotus nigroaeneus, Crematogaster laeviceps, Iridomyrmex notialis) receive ρfield from a pooled assemblage Weibull AFT model and are flagged † throughout.

•

H1 (size → longevity): Do larger workers survive longer? Expected: HR < 1 for logw_c. Model: hazard ∼ logw_c +

•

H2a (colony-size moderator): Does colony size weaken the per-worker size benefit (colony-dilution hypothesis)? Ex-

maxt + C(SF) (M0). pected: null interaction if size acts intrinsically. OLS on cohort means (N = 39 colonies): total_days ∼ logw_c × cs_c + C(SF). •

H2b (foraging-rate moderator): Does foraging intensity modulate size-longevity? Expected: null if size acts primarily via intrinsic physiology. Model M2 (the size × foraging model): hazard ∼ logw_c × fr_c + maxt_c + C(SF) (Table 5).

•

H2c (field amplification of size): Is the size benefit stronger in the field than in the lab? Compares | βˆlogw | between

•

H3 (thermal-buffer null): Does body size buffer thermal mortality (extrinsic-buffering hypothesis)? Expected: null in-

field and laboratory Cox models, plus a species-level Spearman correlation of survival indices (Φ; Fig. 3). teraction if size acts intrinsically. Model M1 (the size × temperature model): hazard ∼ logw_c × maxt_c + C(SF).

Nested models are compared by partial likelihood-ratio test (LRT) and AIC.

Moscardi et al.

2.4

|

9

Trajectory models — Weibull survival (Claim 2)

We fit Weibull survival models per species to estimate the shape parameter ρ (senescence rate) and scale λ. Laboratory ρlab is estimated from individual-level data (one row per worker; event = death; time = days to event or rightcensoring) using maximum-likelihood Weibull regression per species. Field ρfield is estimated from the cohort-day counting-process data (one row per worker-day; event = daily mortality indicator; time window = entry/exit day for that worker-day record). For each species with three or more cohorts, a separate Weibull proportional-hazards model is fitted to that species’ cohort-day records, extracting ρfield as the estimated shape. For three species with fewer than three cohorts (Camponotus nigroaeneus, Crematogaster laeviceps, Iridomyrmex notialis), a pooled assemblage AFT model supplies a fallback ρfield estimate; these fallback cases are flagged with † in all tables and figures. To quantify the collinearity between circadian niche and subfamily as explanatory variables of ρlab , we performed a variance partitioning analysis (Legendre and Legendre, 2012): three OLS models (niche-only, subfamily-only, and combined) were fitted on the 18 species-level ρlab estimates, and the unique and shared fractions of adjusted R 2 were computed following Legendre and Legendre (2012).

H4 (senescence is prevalent): Is actuarial senescence (ρ > 1) the dominant life-history pattern, rather than frailty

•

selection (ρ < 1)? Assessed as the proportion of species with ρ > 1 and 95 % CI above 1. H5 (circadian structuring of trajectory): Is senescence rate associated with circadian activity niche rather than body

•

size? ρ compared among niches (matinal/diurnal/crepuscular) by Kruskal–Wallis with pairwise follow-up (Dunn’s test with Bonferroni correction, implemented via scikit-posthocs posthoc_dunn), and correlated with body mass (Spearman). Circadian niche assigned from peak foraging hour (peak_niche; Riskas et al. 2026). H6 (field amplification of trajectory): Do field conditions steepen senescence trajectories relative to the laboratory?

•

∆ρ = ρfield − ρlab tested against zero by Wilcoxon signed-rank (∆ρ > 0 = steeper field senescence; Table 7).

2.5

|

Thermal-risk models — Cox PH and piecewise spline (Claim 3)

We model the assemblage-level temperature–hazard relationship, test it for nonlinearity, and ask whether thermal vulnerability differs among subfamilies.

•

H7 (temperature main effect): Does higher daily maximum temperature increase mortality hazard at the assemblage level? Expected: HR > 1 for maxt. Model M0: hazard ∼ logw_c + maxt + C(SF); interest is βˆmaxt (HR per °C).

•

H8 (nonlinear plateau): Is the temperature–hazard relationship nonlinear, with a plateau above some breakpoint

H7 establishes the assemblage-level baseline for H8 and H9. T ∗ consistent with behavioural buffering? Piecewise-linear spline models are fitted across T ∗ ∈ {15, . . . , 34} °C: hazard ∼ logw_c + maxt + (maxt − T ∗ )+ + C(SF); best T ∗ by AIC; significance by Davies permutation test (Table 8). •

H9 (taxon-specific vulnerability): Do subfamilies differ in thermal sensitivity, predicting that some lineages remain exposed above the plateau? Subfamily × temperature interactions added: hazard ∼ logw_c + maxt_c + Σi (maxt_c × SFi ) + C(SF); reference category for all subfamily contrasts is Dolichoderinae (the most species-rich group), encoded as the omitted dummy in C(SF). Significance assessed by joint LRT (χ 2 (4)) first; individual post-hoc comparisons (each SFi × maxt_c Wald test) reported only upon rejection of the joint null. A binary variant uses a heat-event indicator (heat30, maxt > 30 °C; Table 9).

10

Moscardi et al.

2.6

|

Software

All analyses use Python 3.13: Cox PH and Weibull models with lifelines v0.29; OLS/WLS with statsmodels v0.14; numerical routines with numpy/scipy; figures with matplotlib v3.9. A fixed random seed (42) is used throughout. All code and processed data are available at the Open Science Framework (https://osf.io/[FILL-BEFORESUBMISSION]): the Python pipeline, processed CSVs, and a README with step-by-step reproduction instructions.

3

|

RESULTS

3.1 | Claim 1 — Body size predicts duration; colony size and temperature are null moderators, with a weak foraging-rate interaction Body size has a strong, largely intrinsic (context-independent) effect on longevity that is insensitive to colony size and thermal context. In Cox proportional-hazards model M0 (logw_c + maxt + C(SF), cluster-robust, N = 2,363 cohort-days), each log10 unit increase in body mass reduces daily hazard by 33 % (HR = 0.67, 95 % CI [0.52–0.87], p = 0.002; Table 3), consistent with the foundational size–longevity relationship established in Riskas et al. (2026). The main-effects model has moderate discrimination (concordance = 0.637; AIC = 24,542), and among subfamilies Myrmicinae (HR = 0.31, p < 0.001) and Ectatomminae (HR = 0.55, p = 0.004) carry markedly lower baseline hazard than the Dolichoderinae reference (Fig. 3). The size–longevity relationship is a between-species signal. When species identity was absorbed as fixed effects (18 dummies) or as Cox strata, the body-size coefficient remained directionally consistent but became non-significant (HR = 0.77–0.71, p = 0.47–0.35; Table S2), reflecting narrow intraspecific body-size variation rather than an absence of effect. The HR = 0.67 reported throughout therefore quantifies a between-species comparative pattern — a distinction that matters for applying size-based vulnerability indices to species comparisons versus within-colony trait variation. Phylogenetic generalised least squares (PGLS) confirmed that this size–longevity signal carries negligible phylogenetic structure: Pagel’s λ̂ = 0.06 (LRT χ 2 (1) = 0.02, p = 0.90), indicating that species-level body-size log-hazard estimates show no detectable covariation with shared evolutionary history at this phylogenetic depth. The nonphylogenetic Cox model (M0) is therefore retained as the primary model (Supplementary Table S4). Of the three contexts we tested, two did not moderate the size effect and one showed a weak interaction. First, colony size: the body-size × colony-size interaction was null (OLS interaction p = 0.60, 95 % CI [−58, +77] days, N = 39 colonies; Table 4). Second, foraging rate: the body-size × foraging-rate interaction was weak but significant (M2 vs. M0fr : LRT χ 2 (1) = 6.08, p = 0.014, ∆AIC = −4.1; Table 5). Thus only colony size and temperature remain clearly null moderators; the foraging interaction qualifies the otherwise intrinsic reading and should be interpreted cautiously given moderate collinearity (VIF ≈ 4; §4). Third, temperature: in Cox model M1temp (M0 extended with the body size × temperature product), the interaction term was essentially absent (HR = 1.00, 95 % CI [0.98–1.03], p = 0.72; ∆AIC = +1.45, LRT p = 0.46 vs. M0; Fig. 5B), and hazard curves across body-size quartiles remained parallel as temperature rose (Fig. 4). The interaction CI [0.98, 1.03] corresponds to an upper bound of ≈ 3 % per °C on any thermal moderation of the size benefit, against a main temperature effect of HR = 1.04 (≈ 3.6 %/°C). The one signal of an environmental role is directional rather than conclusive. The protective effect of body size is stronger in the field (HR = 0.67) than in the laboratory (HR = 0.73, 95 % CI [0.45–1.21], p = 0.22) — a 1.28-fold difference on the log-hazard scale (| βˆ |: 0.40 vs. 0.31) — and field and laboratory survival indices are positively but not significantly correlated across species (Spearman ρ = 0.41, p = 0.088, N = 18). Taken together — a robust main effect, two null moderators (colony size, temperature), a weak but significant foraging-rate interaction, and a null thermal

Moscardi et al.

11

Fig. 2 Cox model M0: hazard ratios (N = 2363 cohort-days, 1826 events) (diamond = p < 0.05; cross = null predictor; circle = subfamily contrast vs Dolichoderinae)

Body size (log mass)

p = 0.002

Max temperature (°C)

p = <0.001

Foraging rate

p = 0.658

Body size × Temperature

p = 0.740

SF: Ectatomminae

p = 0.001

SF: Formicinae

p = 0.694

SF: Myrmeciinae

p = 0.231

SF: Myrmicinae

Significant (p < 0.05) Null predictor Subfamily contrast

p = <0.001

0.2

0.4

0.6

0.8

1.0

1.2

Hazard ratio (cluster-robust 95% CI)

1.4

F I G U R E 3 Forest plot of hazard ratios from the main Cox proportional-hazards model M0 (Claims 1 and 3). Each row shows a predictor with its point estimate and cluster-robust 95 % CI (cluster = cohort_id). Marker shape encodes effect type independently of colour: diamonds mark statistically significant predictors (Wald p < 0.05: body size and maximum temperature); crosses mark null predictors (body size × temperature interaction); circles mark subfamily baseline-hazard contrasts relative to the Dolichoderinae reference. Vertical dashed reference line at HR = 1. Model fitted on N = 2,363 cohort-day observations (1,826 events). M0, primary Cox model (logw_c + maxt + C(SF)); HR, hazard ratio; CI, confidence interval; SF, subfamily.

TA B L E 3 Main Cox proportional-hazards model (M0) coefficients (Claims 1 and 3, H1, H3, H7). Reference category: Dolichoderinae. HR, hazard ratio; CI, 95 % confidence interval; z , Wald test statistic. Covariate

p

β

SE

HR

95 % CI

Body size (log10 mass, centred)

−0.398

0.130

0.67

[0.52, 0.87]

0.002

Max. temperature (°C)

+0.035

0.010

1.036

[1.016, 1.057]

< 0.001

Subfamily: Ectatomminae

−0.600

0.210

0.55

[0.36, 0.83]

0.004

Subfamily: Formicinae

−0.105

0.193

0.90

[0.62, 1.32]

0.589

Subfamily: Myrmeciinae

−0.399

0.280

0.67

[0.39, 1.16]

0.155

Subfamily: Myrmicinae

−1.205

0.231

0.31

[0.19, 0.47]

< 0.001

Size × temperature interaction (H3): HR = 1.00 [0.98, 1.03], p = 0.72; ∆AIC = +1.45 vs. M0 Concordance = 0.64; AIC = 24,542; N = 2,363 cohort-day obs., 106 cohorts, 39 colonies. Cluster-robust SEs (cluster = cohort).

Moscardi et al.

Survival probability

12

Fig. 1 Worker survival probability by body-size quartile (shading = 95% CI; ticks = censored observations; log-rank test) 1.0 Body-size quartile Q1 (smallest) (N = 620) Q2 (N = 563) Q3 (N = 596) 0.8 Q4 (largest) (N = 584) 0.6 0.4 0.2 0.0 0

50

100

At risk Q1 (smallest) 588 Q2 533 Q3 573 Q4 (largest) 563

185 123 239 237

105 24 70 80

150

200

250

300

350

400

81 13 19 14

68 13 0 0

62 9 0 0

29 3 0 0

15 0 0 0

0 0 0 0

Days since cohort formation

F I G U R E 4 Kaplan–Meier survival curves by body-size quartile (Claim 1, H1). Field workers are grouped into four body-size quartiles (Q1 = smallest, N = 620; Q2, N = 563; Q3, N = 596; Q4 = largest, N = 584; quartile boundaries defined from the full cohort-day field dataset). Shaded bands are pointwise 95 % confidence intervals; vertical ticks mark censored observations; the number of individuals at risk is tabulated below the time axis. The stepwise separation between quartiles illustrates the intrinsic size–longevity relationship: larger workers survive significantly longer across the full observation window (log-rank p < 0.001; Cox HR per log10 unit = 0.67, 95 % CI [0.52–0.87], model M0, Table 3).

Moscardi et al.

13

TA B L E 4 Colony-level OLS regression of total lifespan on body size and colony size (H2a). Interest is the body-size × colony-size interaction (null hypothesis: colony dilution does not moderate the size benefit). Reference: Dolichoderinae. Term

p

β

SE

t

95 % CI

Intercept

71.9

35.5

2.02

[−1.0, 144.7]

0.053

Subfamily: Ectatomminae

10.2

44.5

0.23

[−81.1, 101.6]

0.820

Subfamily: Formicinae

7.5

38.4

0.20

[−71.3, 86.3]

0.846

Subfamily: Myrmeciinae

50.6

51.0

0.99

[−54.0, 155.2]

0.329

Subfamily: Myrmicinae

51.4

46.1

1.11

[−43.3, 146.1]

0.275

Body size (centred)

13.3

18.1

0.73

[−23.8, 50.4]

0.470

Colony size (centred)

11.3

25.8

0.44

[−41.5, 64.2]

0.663

Body size × Colony size

9.4

32.8

0.29

[−57.9, 76.8]

0.776

N = 39 colonies; one row per colony. OLS fitted with statsmodels using heteroscedasticityconsistent standard errors. Variable definitions: ‘Body size (centred)’ = log10 worker mass in mg, centred at the assemblage mean (x̄ = 0 on the centred scale); ‘Colony size (centred)’ = log10 total colony worker count, centred similarly. The response is total days of colony observation (the survival window), an imperfect colony-level proxy for individual worker lifespan. Note

The body-size × colony-size interaction p = 0.78 (coefficient test, this table) and p = 0.60 (F-test vs. the reduced model) are both consistent with the null that colony size does not moderate the size–longevity benefit.

14

Moscardi et al.

B Partial effects by body-size quartile (parallel curves confirm null size x temperature interaction, H3a)

5 4 3

95% CI (pointwise) T* = 20 °C (breakpoint)

2 1 Below 20 °C: HR = 1.216 /°C Above 20 °C: HR = 1.016 /°C (plateau)

0 15

20

25

30

35

Daily maximum temperature (°C)

Log-hazard ratio (centred at assemblage mean, 22.8 °C)

Log-hazard ratio (centred at assemblage mean, 22.8 °C)

A Piecewise Cox: temperature effect on mortality hazard (H8: Delta-AIC = -38 vs linear, Davies permutation p < 0.001) 4 3

Body-size quartile (other covariates at mean) Q1 (smallest) Q2 Q3 Q4 (largest)

2 1 0

40

15

20

25

30

35

Daily maximum temperature (°C)

40

F I G U R E 5 Temperature–mortality relationship under the piecewise spline model (Claim 3, H8). Panel A: Loghazard ratio (centred at the assemblage mean daily maximum temperature, T¯max = 22.8 °C) as a function of daily maximum temperature (maxt), with the estimated breakpoint T ∗ = 20 °C marked by a dashed vertical line. The shaded band is the pointwise 95 % confidence interval derived from the cluster-robust variance matrix of the piecewise Cox model; faint background shading distinguishes the rising segment (below 20 °C; HR = 1.216/°C) from the near-plateau segment (above 20 °C; HR = 1.016/°C). The piecewise model fits substantially better than a linear model (∆AIC = −38; Davies permutation p < 0.001; Table 8). Panel B: Partial-effect hazard curves for each body-size quartile over the same temperature range, with all other covariates held at their means. The four curves are parallel, consistent with the null size × temperature interaction (H3, Table 3), and the plateau is equally present across all size classes.

interaction — the evidence remains consistent with intrinsic physiology as the most parsimonious account of the size– longevity rule (inferred by exclusion of extrinsic-buffering and colony-dilution alternatives), while acknowledging a modest foraging-context qualification; it cannot formally rule out unmeasured confounders at N = 18 species. TA B L E 5 Cox proportional-hazards model comparison (Claim 1, H2b). All models include cluster-robust standard errors (cluster = cohort_id), subfamily fixed effects (5 levels, Dolichoderinae reference), and daily maximum temperature. a LRT χ 2 (1) vs. M0; b LRT χ 2 (1) vs. M0fr . Model

k

log-lik

AIC

Concordance

χ2 1

p

M0: logw_c + maxt + C(SF)

6

−12264.88

24541.8

0.637

—

—

0.637

0.42a

0.52

0.636

6.08b

0.014

M0fr : M0 + fr_c M2: M0fr + logw_c×fr_c

7 8

−12264.67 −12261.63

24543.3 24539.3

M0 is the primary model for Claims 1 and 3 (H1, H3, H7; see Methods). M2 (here denoted as in Methods §2.2) adds the body-size × foraging-rate interaction for H2b. Adding fr_c alone (M0 → M0fr ) yields no AIC improvement (∆AIC = +1.6, LRT p = 0.52); the interaction term in M2 is significant vs. M0fr (∆AIC = −4.1, p = 0.014) and vs. M0 (∆AIC = −2.5).

3.2

|

Claim 2 — Mortality trajectory is associated with circadian niche, not body size

Senescence is the rule rather than the exception: 14 of 18 species display senescent laboratory mortality trajectories (ρ > 1; mean ρ = 1.37, median ρ = 1.27; Table 6, Fig. 6). The explanation is not body size: across species, ρ is

Moscardi et al.

15

uncorrelated with body mass (Spearman ρ = −0.25, p = 0.32), a key dissociation — the trait that predicts duration (Claim 1) does not predict the shape of ageing. Trajectory is instead associated with circadian activity regime. ρ differs significantly among circadian niches (Kruskal–Wallis H = 9.35, p = 0.009), with matinal species senescing most steeply and crepuscular species most shallowly; the matinal vs. crepuscular contrast is significant (p = 0.002), whereas matinal vs. diurnal is not (p = 0.09). The latter is blunted by a single diurnal outlier, Iridomyrmex notialis, which has the steepest trajectory in the dataset (ρ = 2.51). The circadian signal is therefore best described as a graded matinal-to-crepuscular continuum rather than a clean three-way separation. A formal variance partitioning of ρlab (three OLS models: circadian niche only, subfamily only, and combined; N = 18 species) quantified the collinearity directly. Adjusted R 2 for circadian niche alone was 0.33 and for subfamily 2 = 0.19 (F alone 0.31, yet the full model achieved only R adj 6,11 = 1.67, p = 0.22), with a shared overlap fraction of −0.45.

The strongly negative overlap term indicates that circadian niche and subfamily are largely redundant predictors at N = 18 species, consistent with Cramér’s V = 0.85 (Supplementary Table S5). Finally, the field accelerates senescence uniformly: mortality trajectories are steeper in the field than in the laboratory in 17 of 18 species (Wilcoxon signed-rank p < 0.001; Table 7). In a sensitivity analysis restricted to the 15 species with independently estimated field ρ (excluding three species whose field estimate is based on a pooled-assemblage fallback), 14 of 15 show ∆ρ > 0 (Wilcoxon p < 0.001), indicating that the result is not driven by the fallback procedure. This amplification is near-universal rather than size-dependent: it is consistent with the shape of the trajectory being an intrinsic, species-typical property whose steepness is nonetheless amplified under field conditions. Rhytidoponera (Ectatomminae) carry among the highest laboratory trajectories (ρlab = 1.38–1.99; Fig. 6), i.e. steep senescence is already present before any thermal effect is considered.

3.3 | Claim 3 — Thermal risk is nonlinear and taxonomically patterned; body size offers no protection Temperature raises daily hazard by 3.6 % per °C in the main-effects model (HR = 1.036, 95 % CI [1.016–1.057], p < 0.001; Table 3) — equivalent to a ∼19 % increase across 5 °C of warming. But this linear summary conceals a strongly nonlinear reality. A piecewise spline with a breakpoint at T ∗ = 20 °C minimises AIC across the full search grid (T ∗ ∈ {15–34} °C; ∆AIC = −38 vs. the linear model; Davies permutation p < 0.001, B = 1,000; Table 8, Fig. 5A). Below 20 °C the marginal thermal hazard is steep (HR = 1.216 per °C, 95 % CI [1.117, 1.325], p < 0.001); above 20 °C it collapses to a near-plateau (HR = 1.016 per °C, 95 % CI [1.004, 1.027], p = 0.006). Because 65.7 % of cohort-day observations fall above this threshold (mean maxt = 22.8 °C), most of the realised thermal variation lies within the attenuated regime. The pattern is consistent with behavioural buffering — foragers retreating to the nest once surface temperatures exceed the activity optimum, which would cap realised thermal exposure (Andrew et al., 2013; Youngsteadt et al., 2023) — but this remains a hypothesis inferred from the mortality pattern rather than a direct behavioural measurement (we develop the interpretation in §3.3). External evidence for nest retreat under heat is well documented in genera present in this assemblage. Iridomyrmex purpureus (closely related to our three Iridomyrmex species) maintains body temperatures below trail temperatures during peak heat, with air temperatures at weatherstation height at least 10 °C below simultaneous ground-surface temperatures (Andrew et al., 2013). Under this calibration, our breakpoint T ∗ = 20 °C (air) corresponds to a ground-surface temperature of ≈30–35 °C — the range at which Rhytidoponera convexa (our genus Rhytidoponera) reaches its published upper foraging limit (Jayatilaka et al., 2011), and at which Myrmecia species transition from unimodal to bimodal foraging to avoid peak heat (Jayatilaka et al., 2011). These correspondences are consistent with the hypothesis that the plateau marks the onset of widespread

16

Moscardi et al.

Fig. 4 Weibull shape parameter by species (laboratory data) (95% CI from Weibull MLE; n = individuals per species) Iridomyrmex notialis Rhytidoponera tasmaniensis Meranoplus fenestratus Crematogaster laeviceps Rhytidoponera victoriae Anonychomyrma nitidiceps Meranoplus d Chelaner cinctum Rhytidoponera metallica Camponotus claripesA Iridomyrmex splendens Camponotus consobrinusus Iridomyrmex septentrionalis Papyrius A Melophorus turneri Camponotus nigroaeneus Myrmecia pyriformis Myrmecia piliventris

n = 30 n = 90 n = 90 n = 30 n = 60 n = 60 n = 60 n = 90 n = 70 n = 60 n = 60 n = 90 n = 90 n = 90 n = 60 n = 30

Matinal (colour + shape) Diurnal (colour + shape) Crepuscular (colour + shape) = 1 constant hazard

n = 58 n = 30

0.5

1.0

1.5

2.0

2.5

Weibull shape parameter (95% CI from MLE)

3.0

F I G U R E 6 Weibull shape parameters ρ by species, estimated from laboratory survival data (Claim 2, H4, H5). Species are sorted in ascending order of ρ̂lab . Horizontal bars are 95 % confidence intervals from maximum-likelihood Weibull fits to individual-level laboratory survival; sample sizes (n individuals per species) are annotated to the right of each estimate. The dashed vertical line at ρ = 1 separates frailty selection (ρ < 1) from senescence (ρ > 1). Points are coloured and shaped by circadian activity niche: matinal (blue triangle ▲), diurnal (orange square ■), crepuscular (purple circle •) — dual encoding makes the grouping legible in greyscale and for colour-vision deficiencies. Fourteen of 18 species have ρ > 1 (actuarial senescence). The distribution of ρ across niches illustrates the circadian structuring of trajectory: matinal species concentrate at higher ρ values (Kruskal–Wallis H = 9.35, p = 0.009; matinal vs. crepuscular p = 0.002).

Moscardi et al.

17

TA B L E 6 Weibull shape parameter ρlab estimated from laboratory mortality data, with species ordered by descending ρlab (Claims 2, H4, H5). Circadian niche is assigned from peak foraging hour (M = matinal, D = diurnal, C = crepuscular). The shape parameter classifies the trajectory: ρ > 1, senescence (rising hazard); ρ ≈ 1, constant hazard; ρ < 1, frailty selection (declining hazard). λ lab is the Weibull scale parameter (characteristic lifetime, in days); Median (d) is the model-implied median lifespan. 95 % CIs are from the maximum-likelihood Weibull fit (lifelines WeibullFitter); the parenthetical value in the Pattern column is n, the number of individuals assayed per species. The two bottom rows summarise the assemblage: mean / median ρlab , and the count of senescent species (ρlab > 1) out of 18. Species

Subfamily

Niche

ρlab

95 % CI

λlab

Median (d)

Pattern (n)

Iridomyrmex notialis

Dolichoderinae

D

2.51

[1.72, 3.30]

22.5

19.4

Senescence (30)

Rhytidoponera tasmaniensis

Ectatomminae

M

1.99

[1.65, 2.33]

58.5

48.7

Senescence (90)

Meranoplus fenestratus

Myrmicinae

M

1.85

[1.52, 2.17]

83.6

68.6

Senescence (90)

Crematogaster laeviceps

Myrmicinae

M

1.82

[1.30, 2.34]

35.5

29.0

Senescence (30)

Rhytidoponera victoriae

Ectatomminae

M

1.70

[1.33, 2.07]

55.6

44.8

Senescence (60)

Anonychomyrma nitidiceps

Dolichoderinae

M

1.56

[1.25, 1.86]

8.6

6.8

Senescence (60)

Meranoplus diversus

Myrmicinae

M

1.47

[1.17, 1.78]

89.3

69.6

Senescence (60)

Chelaner cinctum

Myrmicinae

M

1.46

[1.22, 1.71]

69.4

54.0

Senescence (90)

Rhytidoponera metallica

Ectatomminae

M

1.38

[1.11, 1.64]

68.2

52.3

Senescence (70)

Camponotus claripes A

Formicinae

C

1.15

[0.91, 1.40]

44.3

32.2

Senescence (60)

Iridomyrmex splendens

Dolichoderinae

D

1.12

[0.90, 1.33]

19.8

14.3

Senescence (60)

Camponotus consobrinus

Formicinae

C

1.08

[0.89, 1.27]

98.7

70.3

≈Constant (90)

Iridomyrmex septentrionalis

Dolichoderinae

D

1.07

[0.92, 1.23]

21.8

15.5

≈Constant (90)

Papyrius A

Dolichoderinae

D

1.01

[0.85, 1.17]

22.6

15.7

≈Constant (90)

Melophorus turneri

Formicinae

D

0.96

[0.78, 1.15]

28.4

19.5

≈Constant (60)

Camponotus nigroaeneus

Formicinae

C

0.89

[0.60, 1.17]

82.5

54.6

Frailty (30)

Myrmecia pyriformis

Myrmeciinae

C

0.83

[0.67, 1.00]

18.4

11.8

Frailty (58)

Myrmecia piliventris

Myrmeciinae

C

0.81

[0.57, 1.06]

34.3

21.8

Frailty (30)

Mean / median

1.37 / 1.27

Senescent (ρ > 1)

14/18

Circadian niche: M = matinal, D = diurnal, C = crepuscular. 95 % CI: confidence interval from the Weibull MLE summary; n (in Pattern column): number of individuals assayed. Kruskal–Wallis across niche groups: H = 9.35, p = 0.009. Matinal vs. crepuscular: p = 0.002; matinal vs. diurnal: p = 0.09. Spearman correlation of ρ vs. body mass: ρ = −0.25, p = 0.32.

18

Moscardi et al.

TA B L E 7 Field amplification of mortality trajectory (H6): ∆ρ = ρfield −ρlab per species. Positive ∆ρ indicates steeper senescence in the field. Species are grouped by circadian niche. † = field ρ from pooled AFT fallback. Species

Niche

ρlab

ρfield

∆ρ

Direction

Meranoplus fenestratus

Matinal

1.85

3.28

+1.43

↑

Rhytidoponera metallica

Matinal

1.38

0.54

−0.84

↓

Rhytidoponera tasmaniensis

Matinal

1.99

2.94

+0.95

↑

Rhytidoponera victoriae

Matinal

1.70

3.59

+1.88

↑

Anonychomyrma nitidiceps

Matinal

1.56

2.12

+0.57

↑

Chelaner cinctum

Matinal

1.46

1.34

+0.13

↑

Crematogaster laeviceps†

Matinal

1.82

5.65

+3.83

↑

Meranoplus diversus

Matinal

1.47

2.60

+1.13

↑

Iridomyrmex notialis†

Diurnal

2.51

5.65

+3.14

↑

Iridomyrmex septentrionalis

Diurnal

1.07

4.69

+3.62

↑

Iridomyrmex splendens

Diurnal

1.12

3.99

+2.87

↑

Melophorus turneri

Diurnal

0.96

5.15

+4.19

↑

Papyrius A

Diurnal

1.01

3.79

+2.78

↑

Camponotus claripes A

Crepuscular

1.15

4.89

+3.74

↑

Camponotus consobrinus

2.85

+1.77

↑

Crepuscular

1.08

Camponotus nigroaeneus†

Crepuscular

0.89

5.65

+4.76

↑

Myrmecia piliventris

Crepuscular

0.81

4.47

+3.66

↑

Myrmecia pyriformis

Crepuscular

0.83

1.25

+0.42

↑

Species with ∆ρ > 0 Wilcoxon signed-rank test vs. 0 †

Field ρ estimated from pooled assemblage AFT fallback.

17/18 p < 0.001

Moscardi et al.

19

behavioural retreat rather than a hard physiological ceiling (Parr and Bishop, 2022; Andrew et al., 2013; Jayatilaka et al., 2011), the lineages with ready access to nest retreats being those most buffered against gradual warming; we did not, however, observe retreat behaviour directly, and confirming the mechanism would require behavioural assays under progressive warming. TA B L E 8 Piecewise spline model of the temperature–hazard function (H8, Claim 3), in two panels. Upper panel: model comparison; the best breakpoint T ∗ = 20 °C was selected by minimising AIC across T ∗ ∈ {15, 16, . . . , 34} °C. Lower panel: marginal thermal hazard within each spline segment. ∆AIC, change in Akaike information criterion relative to the linear baseline (negative values favour the piecewise model); HR/°C, hazard ratio per 1 °C increase in daily maximum temperature; T ∗ , estimated breakpoint; LRT, likelihood-ratio test; a dash (—) denotes a quantity that is not applicable (the linear model is the reference). AIC

∆AIC

LRT p

Linear (M0, reference)

24,542

0

—

Piecewise spline, T ∗ = 20 °C

24,504

−38

< 0.001

Model

Note Baseline Best breakpoint

Spline segment

HR/°C

95 % CI

p

Below T ∗ = 20 °C

1.216

[1.117, 1.325]

< 0.001

Steep thermal hazard

Above T ∗ = 20 °C

1.016

[1.004, 1.027]

0.006

Near-plateau

Interpretation

∆AIC = −37.7 vs. linear model (reported as −38 in text); Davies permutation test (permute maxt within cohorts; B = 1,000): p < 0.001 (0 of 1,000 permuted datasets reached ∆AIC ≤ −37.7). The below-T ∗ HR is the maxt coefficient; the above-T ∗ HR and its 95 % CI are the linear combination (β maxt + β ( maxt−T ∗ ) + ) with cluster-robust variance propagation. 65.7 % of cohort-day observations fall above T ∗ = 20 °C (mean maxt = 22.8 °C).

This buffer is not shared equally. For between-species comparisons of thermal sensitivity, the phylogenetically appropriate primary framework is PGLS. Pagel’s λ = 0 (LRT p = 1.00) indicates no detectable phylogenetic signal in species-level thermal sensitivities, so PGLS reduces to OLS; nonetheless, using species as the unit of analysis is more conservative than pooling all worker-days. In this species-level analysis, the Rhytidoponera lineage (Ectatomminae) shows directionally elevated thermal sensitivity: PGLS βˆEctatomm = +0.048/°C (p = 0.081, N = 14 species; Table S3). The result does not reach α = 0.05 — reflecting limited statistical power at N = 14 species — but the direction is consistent across all specifications. Corroborating this genus-level signal, the individual-level Cox model draws on the full worker-day dataset (N = 2,363 observations; 1,826 deaths) and thus has substantially greater power: the joint subfamily × temperature interaction is significant (LRT χ 2 (4) = 12.0, p = 0.017; Table 9, Fig. 5), and only the Ectatomminae contrast — here represented entirely by Rhytidoponera — shows a significant positive term: elevated thermal sensitivity of 5 % per °C (HR = 1.05, 95 % CI [1.01–1.09], p = 0.015) above the assemblage response. During extreme heat events (maxt > 30 °C), Rhytidoponera experience a 66 % excess hazard (HR = 1.66, 95 % CI [1.11–2.48], p = 0.014). Together, the PGLS direction and individual-level significance point to elevated thermal sensitivity observed in the Rhytidoponera lineage (Ectatomminae) — not a phylogenetically conserved clade-wide pattern, but a consistent genus-level signal visible across both analytical frameworks. Body size provides no protection at any temperature (Claim 1, H3). A parallel PGLS for the body-size log-hazard signal (βˆlogw per species, N = 14) yielded λ̂ = 0.06 (LRT p = 0.90), indicating that the size–longevity pattern likewise lacks detectable phylogenetic structure (Supplementary Table S4).

20

Moscardi et al.

TA B L E 9 Subfamily × temperature interaction model coefficients (H9, Claim 3). Reference: Dolichoderinae × temperature. Panel A: continuous maxt_c interaction; Panel B: binary heat-event indicator (heat30: maxt > 30 °C). Both models include body size and subfamily fixed effects. Panel A: Continuous temperature interaction

Subfamily interaction

p

β

HR

95 % CI

maxt_c (Dolichoderinae reference slope)

+0.014

1.014

[0.986, 1.044]

maxt_c × Ectatomminae

+0.048

1.050

[1.009, 1.091]

maxt_c × Formicinae

+0.023

1.023

[0.974, 1.075]

0.367

maxt_c × Myrmeciinae

+0.013

1.013

[0.966, 1.064]

0.589

maxt_c × Myrmicinae

+0.019

1.019

[0.962, 1.078]

0.524

HR

95 % CI

p

heat30 (Dolichoderinae reference)

1.207

[0.918, 1.587]

heat30 × Ectatomminae

1.656

[1.106, 2.479]

heat30 × Formicinae

1.131

[0.700, 1.828]

0.616

heat30 × Myrmeciinae

1.130

[0.651, 1.963]

0.663

heat30 × Myrmicinae

0.899

[0.433, 1.863]

0.774

Panel B: Extreme heat events (maxt > 30 °C)

Coefficient

0.177 ⋆

0.014

⋆ Significant at α = 0.05. Reference subfamily: Dolichoderinae. HR, hazard ratio; CI, clusterrobust 95 % confidence interval; maxt_c, daily maximum temperature mean-centred at the assemblage average (22.8 °C); heat30, binary indicator equal to 1 when daily maximum temperature exceeds 30 °C and 0 otherwise. In Panel A the maxt_c row is the reference-group (Dolichoderinae) temperature slope from the interaction model; the assemblage-average temperature effect (HR = 1.036) is reported for the pooled model M0 in Table 3. Joint likelihoodratio test for the subfamily × temperature interaction (Panel A): χ 2 (4) = 12.0, p = 0.017 (vs. the no-interaction model). Only the Rhytidoponera/Ectatomminae contrast departs significantly in either panel; all other subfamily interactions are non-significant.

0.328 ⋆

0.015

Moscardi et al.

4

21

DISCUSSION

|

Our central result is a three-axis reorganisation: no single predictor accounts for all three components of worker mortality risk, and body size — the canonical predictor — accounts for only one of them. This reorganises a sizecentric framework into a three-axis one, with direct consequences for how we predict ant responses to environmental change. We emphasise that the conceptual contribution — evidence consistent with duration, senescence trajectory, and thermal vulnerability being decoupled and associated with different predictors — is the primary advance. The specific identity of the most vulnerable lineage (Rhytidoponera, Ectatomminae; Claim 3) is a secondary, provisional finding that motivates a testable hypothesis rather than a settled conclusion.

4.1

|

Duration is associated with intrinsic physiology (Claim 1)

The convergence of a robust main effect with two null moderators (colony size, temperature) and a weak but significant foraging-rate interaction (LRT p = 0.014) narrows the mechanistic window for the size–longevity rule. The pattern is more consistent with intrinsic physiology — mass-specific metabolic scaling and reactive-oxygen-species production (Brown et al., 2004; Kramer et al., 2021) — than with body size acting primarily as an extrinsic buffer (Lighton et al., 1994; Bujan et al., 2016), though foraging intensity provides a modest contextual qualification. Crucially, the CI on the size × temperature interaction [0.98, 1.03] is inconsistent with any thermal moderation of the size benefit larger than ≈ 3 %/°C. With 2,363 cohort-day observations this test has ≈97 % power to detect an interaction as small as HR = 1.05 (§4, Limitations): the null is therefore positive evidence against ecologically meaningful thermal buffering by body size, not a failure of detection. We cannot formally exclude a weaker buffering effect, or one operating at finer temporal or spatial resolution than captured by maxt. The directional field amplification (1.28× on the loghazard scale) is most parsimoniously read as extrinsic mortality filtering through an intrinsic size advantage, rather than body size buffering any specific extrinsic stressor. Frailty models confirm that this is a between-species, not within-species, pattern (Table S2): intraspecific body-size variation in this assemblage is too narrow to detect an independent individual-level effect, and the HR = 0.67 should be interpreted as a species-level comparative signal alongside other between-species analyses.

4.2

|

Trajectory is associated with circadian regime, not size (Claim 2)

The dissociation between body size and trajectory (Spearman ρ = −0.25, p = 0.32) is the conceptual hinge of the paper: duration and trajectory are different axes with different predictors. Trajectory shape is associated with circadian niche, with matinal species senescing fastest. Matinal foragers, active as surface temperatures rise steeply in the early morning, may experience a more variable and physiologically taxing thermal environment, accelerating senescence. This parallels ‘pace-of-life’ syndromes described in vertebrates (Ricklefs and Wikelski, 2002; Debecker and Stoks, 2018), but expressed here at the level of species-specific activity regimes. We are cautious about over-reading the three-way niche contrast: the matinal–diurnal difference is not individually significant and one diurnal species departs strongly from its group, so we describe the signal as a graded matinal-to-crepuscular continuum.

4.3

|

Thermal risk is nonlinear — and the buffer is taxon-specific (Claim 3)

The thermal plateau above T ∗ = 20 °C is the most novel result of this study. It departs from the monotonically rising limb of the thermal performance curve expected below the activity optimum (Martin and Huey, 2008; Deutsch

22

Moscardi et al.

et al., 2008; Arnoldi et al., 2025): rather than mortality continuing to climb with temperature, hazard saturates. The pattern is consistent with behavioural buffering (thermoregulation) — workers retreating below ground when surface temperatures exceed foraging optima would impose a ceiling on realised thermal exposure (Andrew et al., 2013; Jayatilaka et al., 2011). This hypothesis is supported by the correspondence between T ∗ = 20 °C (air) and published foraging-limit temperatures for this assemblage (see Results), but is not directly demonstrated here. The plateau onset corresponds to a daily maximum air temperature of 20 °C, at which point ground surface temperatures already substantially exceed air temperature — consistent with conditions that trigger nest retreat in Australian ants, which restrict field activity well below their physiological thermal maxima (Jayatilaka et al., 2011) — a pattern documented more broadly across ant assemblages globally (Nascimento et al., 2022). Importantly, the plateau does not imply invulnerability above 20 °C. First, the hazard ratio remains above 1 throughout — thermal mortality continues, merely attenuated. Second, elevated thermal sensitivity in the Rhytidoponera lineage (excess hazard 5 % per °C) is consistent with some lineages being unable to fully execute behavioural retreat, possibly because their matinal schedule exposes them to morning temperature ramps before retreat is possible — a hypothesis requiring direct behavioural testing. Third, extreme heat events (> 30 °C) produce large excess mortality (66 % for Rhytidoponera) that likely overwhelms behavioural buffering altogether.

4.4

|

Synthesis: circadian regime and lineage identity — collinear climate-relevant axes

The three claims converge on a single reframing (Fig. 1). Body size, the trait most often used as a proxy for risk, predicts only duration; it is silent on trajectory and on thermal vulnerability. Those two climate-relevant axes are associated with circadian activity regime and lineage identity, respectively — axes that are strongly collinear in this assemblage (Cramér’s V = 0.85) and intersect in one lineage. Importantly, circadian regime and lineage identity cannot be fully disentangled in this assemblage. The three signals are not equally robust. The body-size effect (HR = 0.67, p = 0.002) and the null size × temperature interaction (CI [0.98, 1.03]) are the most firmly established results; the Kruskal–Wallis circadian-niche effect on ρ (p = 0.009) and the Rhytidoponera/Ectatomminae Cox interaction (p = 0.015) are well supported at the individual level but rest on N = 18 species or a single genus; and the PGLS signal for this lineage (p = 0.081) is suggestive only. With that ranking in mind, three signals converge on Rhytidoponera (Ectatomminae): steep senescent trajectories (Claim 2), a directionally elevated but non-significant PGLS thermal coefficient (Claim 3, βˆ = +0.048/°C, p = 0.081, N = 14 species; power limited to detect genus-level signals at this sample size), and a matinal activity regime — three signals that converge without individually constituting strong evidence. The individuallevel Cox model corroborates this with greater statistical power (p = 0.015, N = 2,363 cohort-days), but the PGLS — the appropriate framework for between-species comparisons — remains the primary test; together they indicate elevated thermal sensitivity observed in the Rhytidoponera lineage (Ectatomminae) rather than a phylogenetically conserved clade-wide pattern. The genus Rhytidoponera is widespread across eastern and south-eastern Australia and ecologically important as predators and seed dispersers (Lubertazzi et al., 2010; Parr et al., 2007). Because the mean daily maximum already sits at the plateau onset, the most consequential climate pathway is not gradual warming of the mean but the rising frequency of extreme-heat days (> 30 °C), where behavioural buffering fails and where Rhytidoponera are most exposed (IPCC, 2021; Herold et al., 2018; Perkins-Kirkpatrick and Lewis, 2020).

4.5

|

Implications: what changes if risk is three-dimensional

Treating mortality risk as a single, size-indexed axis is incomplete. Our results show that purely size-based vulnerability frameworks are insufficient: they capture only the duration axis and can be improved by incorporating the two addi-

Moscardi et al.

23

tional axes identified here (senescence trajectory and thermal vulnerability structured by circadian regime and lineage identity). Trait-based vulnerability frameworks routinely assess thermal tolerance (Deutsch et al., 2008; Diamond and Chick, 2017); our results show that such indices are, by construction, silent on the two axes that structure climate exposure here (size × temperature p = 0.72). A framework that substituted circadian activity regime and lineage identity for body size would have flagged matinal Rhytidoponera (Ectatomminae) as most exposed; a size-based index would not, and would in fact rank them as intermediate. Two consequences follow for forecasting. First, because the temperature–hazard function plateaus near the assemblage mean (T ∗ = 20 °C; mean maxt = 22.8 °C; a result firmly supported by ∆AIC = −38 and Davies p < 0.001), gradual warming of the mean should have limited effect on realised mortality, whereas the rising frequency of extremeheat days (> 30 °C) — the regime in which behavioural buffering fails and Rhytidoponera carry a 66 % excess hazard (Cox HR = 1.66 [1.11, 2.48], individual-level only; to be treated as preliminary pending broader sampling) — is the operative climate variable (Perkins-Kirkpatrick and Lewis, 2020; IPCC, 2021). Vulnerability projections built on meantemperature anomalies will therefore understate risk for heat-sensitive lineages. Second, because the most exposed lineage (Rhytidoponera) is a dominant predator and seed disperser in south-eastern Australia (Lubertazzi et al., 2010; Parr et al., 2007), lineage-specific heat mortality has a plausible pathway to ecosystem function — myrmecochory and predation — that an assemblage-average analysis would obscure. We frame both as actionable hypotheses rather than firm projections: at N = 18 species, with circadian niche and lineage identity strongly collinear (Cramér’s V = 0.85), the priority is to redirect monitoring and trait-collection effort, not to issue lineage-specific risk verdicts.

4.6 1.

|

Limitations

Sample size and genus-driven signals. Eighteen species limit power for subfamily-level inference. The thermal signal attributed to Ectatomminae in the models is driven entirely by three Rhytidoponera species, and the corresponding PGLS coefficient (p = 0.081) does not cross the conventional significance threshold at this N . Claims 2 and 3 should therefore be read as provisional patterns rather than settled conclusions: replication across other matinal, ground-foraging lineages in different biogeographic regions is needed before general statements about subfamily-level thermal vulnerability can be made.

2.

Temperature resolution and microhabitat mismatch. maxt is a daily maximum recorded at a weather station ∼1 km from the study site, which can diverge from microhabitat temperatures by 10–20 °C depending on nest depth and substrate (Andrew et al., 2013). Surface-foraging species (many Ectatomminae) are likely more exposed to stationlevel maxt than subterranean or nocturnal foragers, so our thermal-hazard estimates may be partly confounded with foraging guild, and the breakpoint T ∗ = 20 °C may reflect guild-level heterogeneity as well as a physiological threshold.

3.

Field vs. laboratory ρ. Laboratory conditions (controlled temperature, ad libitum food) differ markedly from the field, and several field ρ estimates rest on few cohorts; ∆ρ therefore reflects both genuine extrinsic acceleration and environmental confounding, and should be read as a qualitative direction, not a precise magnitude. For three species with fewer than three cohorts (Camponotus nigroaeneus, Crematogaster laeviceps, Iridomyrmex notialis), ρfield was estimated from a pooled fallback model (flagged †); a sensitivity analysis excluding these three species confirmed that 14 of 15 remaining species show ∆ρ > 0 (Wilcoxon p < 0.001), indicating that the conclusion of field amplification of senescence is not driven by the fallback estimates.

4.

Collinearity (foraging rate). The foraging-rate × body-size interaction carries moderate collinearity (VIF ≈ 4), so the corresponding interaction (LRT p = 0.014) should be read cautiously.

5.

Collinearity (circadian niche and subfamily). Circadian niche and subfamily identity are strongly collinear in this

24

Moscardi et al.

assemblage: all three Rhytidoponera (Ectatomminae) are matinal and both Myrmecia (Myrmeciinae) are crepuscular (Cramér’s V = 0.85, χ 2 (8) = 25.9, p = 0.001). Claims 2 and 3 cannot be fully separated with N = 18 species, so the attribution to circadian niche (Claim 2) versus subfamily identity (Claim 3) is operationally indistinguishable at this scale. Two of our conclusions are nonetheless robust to this collinearity: the body-size HR (an interspecific effect, independent of niche classification) and the null size × temperature interaction. Two are tentative: the specific attribution of trajectory structuring to circadian niche rather than lineage identity, and the attribution of thermal vulnerability to Ectatomminae as a subfamily rather than to matinal Rhytidoponera specifically. Disentangling these will require cross-phylogenetic sampling — multiple matinal lineages drawn from different subfamilies. Reviewers may ask: “How can you separate circadian niche from lineage identity?” The honest answer is: we cannot, with N = 18 from a single assemblage. We present the attribution to circadian niche (Claim 2) and to lineage identity (Claim 3) as two complementary framings of the same collinear signal, and propose the separation as the key empirical question for future cross-phylogenetic work. 6.

Hypothesis-generating scope. This study is confirmatory for Claim 1 (body-size duration effect: HR = 0.67, p = 0.002, robust to multiple sensitivity analyses) and for the null size × temperature interaction (CI [0.98, 1.03]). It is hypothesis-generating for Claims 2 and 3: the circadian-niche structuring of trajectory and the thermal excess in Rhytidoponera are consistent, directional signals that motivate targeted replication, but cannot be treated as confirmed at the subfamily or clade level from a single 18-species assemblage. Reviewers should assess Claims 2 and 3 by the standard of pattern discovery, not by the standard of mechanistic proof.

7.

Statistical power at N = 18 (simulation-based). We estimated retrospective power via Monte Carlo simulation (3,000 replicates per design point; block0_power/power_analysis.py; fixed seed = 42). (i)

Spearman ρlab vs. body mass (N = 18, two-tailed α = 0.05). Estimated power at the observed |ρ | = 0.25 is 15 %; the minimum detectable correlation at 80 % power is |ρ | = 0.65. The null result (ρ = −0.25, p = 0.32) therefore cannot exclude true associations as large as |ρ | ≈ 0.60.

(ii)

PGLS Ectatomminae thermal coefficient (N = 14: three Ectatomminae vs. eleven others; λ̂ ≈ 0, so a t -test is a valid proxy). Power at Cohen’s d = 1.0 is 41 % (one-sided); the minimum detectable effect at 80 % power is d ≈ 1.8. The observed p = 0.081 is fully consistent with a genuine but underpowered elevation.

(iii)

Cox size × temperature interaction (N = 2,363 cohort-days; analytical Wald test). The interaction SE = 0.013 affords 80 % power to detect HR ≥ 1.04; HR = 1.05 is detectable at 97 % power. The null result (HR = 1.00 [0.98, 1.03]) therefore effectively rules out biologically meaningful size × temperature interactions.

(iv)

Cox body-size main effect (N = 2,363). Power at the observed HR = 0.67 is 87 %, confirming the robustness of Claim 1.

We include tests with < 50 % power for completeness and directionality; they should not be read as decisive evidence. 8.

Phylogenetic power. The PGLS robustness check used N = 14 of 18 species (four failed to converge due to near-zero intraspecific body-size variance); Pagel’s λ = 0 precludes phylogenetically informed weighting. The parallel PGLS for the body-size log-hazard signal also returned λ̂ = 0.06 (p = 0.90), confirming that neither size– longevity nor size–temperature sensitivity shows detectable phylogenetic structure. Elevated thermal sensitivity in Rhytidoponera (p = 0.015 in the individual-level Cox) should therefore be treated as a genus-level hypothesis rather than a confirmed clade-wide pattern.

Moscardi et al.

5

|

25

CONCLUSION

Worker mortality risk in ants is organised along three axes, each associated with a different predictor. Body size predicts longevity duration, consistent with intrinsic physiology, independently of colony size and temperature (with a weak foraging-rate qualification). The trajectory of mortality — the rate of senescence — is associated with circadian activity regime, not body size. Thermal risk is nonlinear, plateauing above 20 °C in a pattern consistent with behavioural buffering, with elevated thermal sensitivity observed in the Rhytidoponera lineage (Ectatomminae; Pagel’s λ = 0) that persists above the plateau and intensifies during extreme heat. By showing that the trait predicting how long a worker lives says nothing about how it ages or how it dies from heat, these results reframe the size–longevity paradigm. The practical consequence is concrete: predicting ant responses to warming requires tracking the frequency of extremeheat events rather than shifts in mean temperature, and weighting circadian regime and lineage identity — not body size — when ranking which taxa, and which ecosystem functions, are most at risk.

Acknowledgements [To be completed at submission; must include land acknowledgement and any technical assistance.]

Conflict of Interest The authors declare no competing interests.

Data and Code Availability All raw data are available via Dryad (Riskas et al., 2026, doi: 10.5061/dryad.j0zpc86s4). Analysis code is available at https://github.com/[FILL-BEFORE-SUBMISSION] as a fully reproducible Python pipeline (source_files/), with a frozen archive deposited on Zenodo (https://doi.org/[FILL-BEFORE-SUBMISSION]). The repository includes a README documenting execution order and expected outputs.

references Andrew, N. R., Hart, R. A., Jung, M.-P., Hemmings, Z. and Terblanche, J. S. (2013) Can temperate insects take the heat? A case study of the physiological and behavioural responses in a common ant, Iridomyrmex purpureus (Formicidae), with potential climate change. Journal of Insect Physiology, 59, 870–880. Arnoldi, J.-F., Jackson, A. L., Peralta-Maraver, I. and Payne, N. (2025) A universal thermal performance curve arises in biology and ecology. Proceedings of the National Academy of Sciences, 122. URL: https://doi.org/10.1073/pnas.2513099122. Baudier, K. M. and O’Donnell, S. (2018) Complex body size differences in thermal tolerance among army ant workers (Eciton burchellii parvispinum). Journal of Thermal Biology, 78, 277–280. Brown, J. H., Gillooly, J. F., Allen, A. P., Savage, V. M. and West, G. B. (2004) Toward a metabolic theory of ecology. Ecology, 85, 1771–1789. URL: https://doi.org/10.1890/03- 9000. Bujan, J., Yanoviak, S. P. and Kaspari, M. (2016) Desiccation resistance in tropical insects: causes and mechanisms underlying variability in a Panama ant community. Ecology and Evolution, 6, 6282–6291. URL: https://doi.org/10.1002/ece3.2355. Bureau of Meteorology (2024) Climate statistics for Australian locations: Bundoora (La Trobe University), station 086068. URL: http://www.bom.gov.au/climate/averages/tables/cw_086068.shtml. 1991–2020 climate normals; accessed 2024.

26

Moscardi et al.

— (2025) Annual climate statement 2024: Analysis of Australia’s oceans, atmosphere, temperature, rainfall, water, and significant weather during 2024. URL: https://www.bom.gov.au/climate/current/annual/aus/2024/. Published 6 February 2025. Dahlgren, J. P., Colchero, F., Jones, O. R., Øien, D.-I., Moen, A. and Sletvold, N. (2016) Actuarial senescence in a long-lived orchid challenges our current understanding of ageing. Proceedings of the Royal Society B, 283, 20161217. URL: https: //doi.org/10.1098/rspb.2016.1217. Debecker, S. and Stoks, R. (2018) Pace of life syndrome under warming and pollution: integrating life history, behavior, and physiology across latitudes. Ecological Monographs, 88, 532–553. URL: https://doi.org/10.1002/ecm.1332. Deutsch, C. A., Tewksbury, J. J., Huey, R. B., Sheldon, K. S., Ghalambor, C. K., Haak, D. C. and Martin, P. R. (2008) Impacts of climate warming on terrestrial ectotherms across latitude. Proceedings of the National Academy of Sciences, 105, 6668– 6672. URL: https://doi.org/10.1073/pnas.0709472105. Diamond, S. E. and Chick, L. D. (2017) The Janus of macrophysiology: stronger effects of evolutionary history on upper thermal limits are reversed for lower thermal limits in ants. Current Zoology, 64, 223–230. URL: https://doi.org/10.1093/cz/ zox072. Herold, N., Downes, S., Gross, M. H., Ji, F., Nishant, N., Macadam, I., Ridder, N. and Beyer, K. (2021) Projected changes in the frequency of climate extremes over southeast Australia. Environmental Research Communications, 3. URL: https: //doi.org/10.1088/2515- 7620/abe6b1. Herold, N., Ekström, M., Kala, J., Goldie, J. and Evans, J. P. (2018) Australian climate extremes in the 21st century according to a regional climate model ensemble: implications for health and agriculture. Weather and Climate Extremes, 20, 54–68. URL: https://doi.org/10.1016/j.wace.2018.01.001. IPCC (2021) Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press. URL: https://doi.org/10.1017/9781009157896. Jayatilaka, P., Narendra, A., Reid, S. F., Cooper, P. and Zeil, J. (2011) Different effects of temperature on foraging activity schedules in sympatric Myrmecia ants. Journal of Experimental Biology, 214, 2730–2738. URL: https://doi.org/10.1242/ jeb.053710. Kaspari, M., Clay, N. A., Lucas, J., Yanoviak, S. P. and Kay, A. (2015) Thermal adaptation generates a diversity of thermal limits in a rainforest ant community. Global Change Biology, 21, 1001–1007. URL: https://doi.org/10.1111/gcb.12750. Kramer, B. H., Nehring, V., Buttstedt, A., Heinze, J., Korb, J., Libbrecht, R., Meusemann, K., Paxton, R., Séguret, A., Schaub, F. and Bernadou, A. (2021) Oxidative stress and senescence in social insects: a significant but inconsistent link? Philosophical Transactions of the Royal Society B, 376. URL: https://doi.org/10.1098/rstb.2019.0732. Legendre, P. and Legendre, L. (2012) Numerical Ecology. Amsterdam: Elsevier, 3rd edn. Lemaitre, J.-F., Moorad, J., Gaillard, J.-M., Maklakov, A. A. and Nussey, D. (2024) A unified framework for evolutionary genetic and physiological theories of aging. PLOS Biology, 22. URL: https://doi.org/10.1371/journal.pbio.3002513. Lighton, J. R. B., Quinlan, M. C. and Feener, D. H. (1994) Is bigger better? Water balance in the polymorphic desert harvester ant Messor pergandei. Physiological Entomology, 19, 325–331. URL: https://doi.org/10.1111/j.1365- 3032.1994. tb01059.x. Lubertazzi, D., Lubertazzi, M. A. A., McCoy, N., Gove, A., Majer, J. and Dunn, R. R. (2010) The ecology of a keystone seed disperser, the ant Rhytidoponera violacea. Journal of Insect Science, 10. URL: https://doi.org/10.1673/031.010.14118. Martin, T. L. and Huey, R. B. (2008) Why “suboptimal” is optimal. American Naturalist, 171, E102–E118. URL: https://doi. org/10.1086/527502.

Moscardi et al.

27

Nascimento, G., Câmara, T. and Arnan, X. (2022) Critical thermal limits in ants and their implications under climate change. Biological Reviews, 97, 1929–1952. URL: https://doi.org/10.1111/brv.12843. Parr, C. L., Andersen, A. N., Chastagnol, C. and Duffaud, C. (2007) Savanna fires increase rates and distances of seed dispersal by ants. Oecologia, 151, 33–41. URL: https://doi.org/10.1007/s00442- 006- 0570- 5. Parr, C. L. and Bishop, T. R. (2022) The response of ants to climate change. Global Change Biology, 28, 3188–3205. Perkins-Kirkpatrick, S. E. and Lewis, S. C. (2020) Increasing trends in regional heatwaves. Nature Communications, 11, 3357. URL: https://doi.org/10.1038/s41467- 020- 16970- 7. Ricklefs, R. E. (2010) Life-history connections to rates of aging in terrestrial vertebrates. Proceedings of the National Academy of Sciences, 107, 10314–10319. URL: https://doi.org/10.1073/pnas.1005862107. Ricklefs, R. E. and Wikelski, M. (2002) The physiology/life-history nexus. Trends in Ecology & Evolution, 17, 462–468. URL: https://doi.org/10.1016/S0169- 5347(02)02578- 8. Riskas, K. A. et al. (2026) Body size and worker survival in Australian ants. rspb.2025.1907. URL: https://doi.org/10.1098/rspb.2025.1907.

Proceedings of the Royal Society B, 293,

West, G. B., Woodruff, W. H. and Brown, J. H. (2001) Allometric scaling of metabolic rate from molecules and mitochondria to cells and mammals. Proceedings of the National Academy of Sciences, 99, 2473–2478. URL: https://doi.org/10.1073/ pnas.012579799. Youngsteadt, E., Prado, S. G., Keleher, K. J. and Kirchner, M. (2023) Can behaviour and physiology mitigate effects of warming on ectotherms? A test in urban ants. Journal of Animal Ecology, 92, 568–579.

28

Moscardi et al.

Supplementary Tables TA B L E S 1 Proportional hazards assumption test via Schoenfeld residuals for the main Cox model M0 (logw_c + maxt + C(SF), N = 2,363 cohort-day observations, 1,826 events, cluster = cohort_id). Test statistic and two-sided pvalue from lifelines.statistics.proportional_hazard_test using the rank-transformed time axis (Grambsch– Therneau test, rank variant); the rank transform reduces sensitivity to the baseline hazard shape and improves power against smooth time-varying alternatives. logw_c: log10 body mass, mean-centred; maxt: daily maximum temperature (°C). No covariate shows evidence of a time-varying hazard ratio at α = 0.05, supporting the PH assumption for M0. Covariate

p

Test statistic

Body size (logw_c)

1.854

0.173

Max. temperature (maxt)

1.410

0.235

Subfamily: Ectatomminae

0.292

0.589

Subfamily: Formicinae

0.323

0.570

Subfamily: Myrmeciinae

0.569

0.451

Subfamily: Myrmicinae

1.529

0.216

All p ≥ 0.17; the PH assumption is not violated for any primary predictor.

TA B L E S 2 Species-frailty robustness check for H1. Three nested Cox models are compared to assess whether the body-size hazard ratio (HR = 0.67 in M0) is robust when species identity is controlled. M0_spp adds species fixed effects (18 dummies absorbing all between-species variation); M0_strat uses species as strata, allowing speciesspecific baseline hazards. Both variants yield a directionally consistent but statistically non-significant HR, indicating that the M0 coefficient reflects a predominantly interspecific size–longevity relationship. Model

Species control

HRlogw

p

M0

Subfamily fixed effects (ref)

0.672

0.002

M0_spp

Species fixed effects (18 dummies)

0.770

0.471

M0_strat

Stratified Cox by species

0.710

0.354

HR direction is consistent across all models (< 1, indicating larger workers live longer). Non-significance in M0_spp and M0_strat reflects narrow intraspecific variation in logw_mass rather than absence of a size effect. All models cluster standard errors by cohort.

Moscardi et al.

29

TA B L E S 3 Phylogenetic generalised least-squares (PGLS) analysis of subfamily differences in thermal sensitivity (primary analysis for between-species comparisons, Claim 3). Response variable: per-species βˆmaxt from individualspecies Cox models (N = 14 of 18 species converged). Predictor: subfamily (reference = Dolichoderinae). Phylogenetic covariance matrix derived from the ultrametric mr_phylo.tre tree (Riskas et al., 2026), scaled by Pagel’s λ̂ = 0 (LRT p = 1.00; no detectable phylogenetic signal), so PGLS reduces to OLS. Term

Coef.

SE

t

p

Intercept (Dolichoderinae)

0.018

0.016

1.16

0.277

Ectatomminae

0.048

0.024

1.97

0.081

Formicinae

0.024

0.024

0.98

0.355

Myrmeciinae

0.034

0.036

0.95

0.367

Myrmicinae

0.035

0.024

1.44

0.185

Joint F (4, 9) = 1.09

0.416

Pagel’s λ̂ = 0.00 (LRT χ 2 (1) = 0.00, p = 1.00). Four species excluded due to convergence failure in species-specific Cox models (C. nigroaeneus, C. laeviceps, I. notialis, M. piliventris). The Rhytidoponera/Ectatomminae PGLS coefficient is directionally consistent with the individual-level Cox result (HR = 1.05, p = 0.015) but does not reach α = 0.05 at N = 14 species.

TA B L E S 4 Phylogenetic generalised least-squares (PGLS) for the body-size log-hazard signal (A1 robustness check). Response variable: per-species βˆlogw from individual-species Cox models (logw_c + maxt; N = 14 converged). Model: intercept-only PGLS to obtain the assemblage-level phylogenetically corrected mean. Pagel’s λ̂ = 0.06 (LRT χ 2 (1) = 0.02, p = 0.90): no detectable phylogenetic signal, supporting retention of the M0 Cox analysis as the primary model. λ̂

LRT p

Conclusion

0.06

0.90

No phylogenetic structure; M0 retained

Term Intercept (assemblage β logw )

Four species excluded due to convergence failure (C. nigroaeneus, C. laeviceps, I. notialis, M. piliventris). λ̂ < 0.3 indicates that the body-size mortality signal does not co-vary appreciably with phylogenetic distance, supporting the non-phylogenetic Cox approach as the primary analysis.

TA B L E S 5 Variance partitioning of ρlab between circadian niche and subfamily (A3 robustness check; N = 18 species). Adjusted R 2 for three OLS models: niche-only, subfamily-only, and combined. Negative shared fraction confirms high collinearity between predictors (Cramér’s V = 0.85), consistent with the non-significant full model (F6,11 = 1.67, p = 0.22). Component

2 R adj

Interpretation

Circadian niche only

0.33

Unique niche contribution

Subfamily only

0.31

Unique subfamily contribution

Full model (niche + SF)

0.19

Combined (not additive due to collinearity)

Shared (overlap)

−0.45

Redundancy; collinearity confirmed

Variance partitioning follows Legendre and Legendre (2012). The strongly negative shared fraction (−0.45) arises when predictors are highly collinear: adding a redundant predictor to an already-saturated model reduces adjusted R 2 . This is not a modelling artefact but a quantitative expression of the Cramér’s V = 0.85 collinearity reported in the Limitations.

Related documents

• How long versus when: worker mortality synchrony is widespread but decoupled from body size in Australian ants#1062177
• Data from: Dress for success: Climate pressures predict fur insulation and body size in natural and reintroduced populations of a threatened marsupial#124804
• Looking Backward and Looking Forward: Anthropometric Research and the Development of Social Science History#646086
• Sociality influences thermoregulation and roost switching in a forest bat using ephemeral roosts#643947
• Development of size constancy in children: A review and meta-analysis of size-matching studies.#125288