Linking Population Temporal Dynamics to Extinction Risk: A Cross-Realm Analysis

A recent Nature Communications investigation examined how temporal patterns of population abundance relate to extinction risk across marine and terrestrial species. The researchers modeled extinction risk as a function of defined temporal categories, testing whether the relationship differs between realms.

Nature

The analysis used a compiled dataset of species assemblages, each containing multiple population time series. Extinction risk was derived from the International Union for Conservation of Nature (IUCN) Red List categories. To reduce dominance of Least Concern species, the authors simplified the IUCN categories into a binary threat classification: non-threatened (Least Concern + Near Threatened) versus threatened (higher-risk categories). Temporal dynamics were assigned to each population based on prevalence within the assemblage, producing four temporal categories (Fig. 1D).

Nature

• Binary threat classification combined IUCN categories 1-2 (Least Concern, Near Threatened) as non-threatened and categories 3-4 as threatened.

Nature

• Temporal categories captured distinct prevalence patterns of populations over time.

Nature

• The number of populations in each temporal-risk combination was summarized in Table S2.

Nature

Linear mixed-effects models were fitted with the lme4 package’s `lmer()` function. The model specification was:

Nature

\[

Nature

y_{i,j} \sim N(\mu_{i,j},\sigma^2),\quad

Nature

\mu_{i,j}= \beta_0 + \beta_1\text{TemporalCategory} + \beta_2(\text{TemporalCategory}\times\text{Realm}) + \alpha_{i,j}

Nature

\]

Nature

where \(y_{i,j}\) is the continuous extinction-risk score for population \(i\) in assemblage \(j\); \(\beta_0\) is the global intercept; \(\beta_1\) represents effects of temporal categories; \(\beta_2\) captures deviations for the terrestrial realm relative to the marine realm; and \(\alpha_{i,j}\) denotes assemblage-level random effects. Two model versions were run—one with four extinction-risk categories and one with the binary classification. Effect sizes were expressed as Cohen’s d, calculated from model t-values and degrees of freedom. Analyses were performed on the full dataset and separately on temperate-region data (Fig. S1), yielding consistent outcomes.

Nature

The interaction term indicated that the influence of temporal dynamics on extinction risk varies between marine and terrestrial species. Both the four-category and binary-category models produced similar effect-size estimates, suggesting robustness of the temporal-risk relationship. Results for temperate regions mirrored those for the entire dataset, despite the latter containing the majority of observations.

Nature

The study’s primary data are concentrated in temperate latitudes (defined as locations wholly between ± 23.5° and ± 65°). Non-temperate studies were classified as polar/temperate or tropical/temperate, indicating limited representation of polar and tropical assemblages. Consequently, the generalizability of the temporal-risk relationship to those biogeographic zones remains uncertain.

Nature

• “1= Least Concern and Lower Risk/Least Concern; 2= Near Threatened and Lower Risk/Near Threatened; To ensure that our results were not overly influenced by the predominance of Least Concern species in our data, we further simplified the extinction risk categories into a binary threat classification of non-threatened vs threatened species, combining 1-2 and 3-4 in the above classification, respectively.” — Authors, Nature Communications

Nature

• “We also included an interaction term between temporal category and realm to test if the effects differ between marine and terrestrial species, and a random effect for assemblage to account for multiple species being sampled within each time series.” — Authors, Nature Communications

Nature

• “We calculated effect sizes using Cohen’s d, which measures the extent of difference between two group means relative to their standard deviation85: $$Cohen^\prime sd=\sqrt\fract^2t^2+df$$ (5) where t is t-value and df degrees of freedom from the model.” — Authors, Nature Communications

Nature

• “We used R packages ggeffects86 for plotting model results, tidyverse87 for data wrangling and rphylopic88 for the species icons used in Fig.” — Authors, Nature Communications

Nature

The article’s methodological transparency, including model equations, software tools, and data categorization, is documented in the Nature Portfolio Reporting Summary linked to the publication.

Nature