Century-long cod otolith biochronology reveals individual growth plasticity in response to temperature

Otolith biochronologies combine growth records from individual fish to produce long-term growth sequences, which can help to disentangle individual from population-level responses to environmental variability. This study assessed individual thermal plasticity of Atlantic cod (Gadus morhua) growth in Icelandic waters based on measurements of otolith increments. We applied linear mixed-effects models and developed a century-long growth biochronology (1908–2014). We demonstrated interannual and cohort-specific changes in the growth of Icelandic cod over the last century which were mainly driven by temperature variation. Temperature had contrasting relationships with growth—positive for the fish during the youngest ages and negative during the oldest ages. We decomposed the effects of temperature on growth observed at the population level into within-individual effects and among‐individual effects and detected significant individual variation in the thermal plasticity of growth. Variance in the individual plasticity differed across cohorts and may be related to the mean environmental conditions experienced by the group. Our results underscore the complexity of the relationships between climatic conditions and the growth of fish at both the population and individual level, and highlight the need to distinguish between average population responses and growth plasticity of the individuals for accurate growth predictions.


Supplementary results
Table S1. Selection of the optimal random effects of the Atlantic cod growth model. Series of models were fitted to the data with the full intrinsic fixed-effects structure (age in the interaction with sex). Based on Akaike Information Criterion corrected for the small sample sizes (AICc) the best model was selected (in bold). The random Age slopes for each FishID, Year or Cohort are denoted by "|", and "+" indicates that random term was included in the given model.

Random intercept
Random slope  Table S2. Selection of the optimal fixed intrinsic effects of the Atlantic cod growth model.
Series of models were fitted to the data with the optimal random effects structure (Table S1).
Based on Akaike Information Criterion corrected for the small sample sizes (AICc) the best model was selected (in bold). Parameter estimates of continuous variables are given in the selection table, "Age:Sex" indicates interaction, and "+" indicates that a term was included in the given model.  Table S4. Selection of the optimal fixed extrinsic effects of the Atlantic cod growth model.
Series of models were fitted to the data with the optimal fixed intrinsic and random effects structure (Tables S1, S2) and optimal time window of sea surface temperature (Table S3).
Based on Akaike Information Criterion corrected for the small sample sizes (AICc) the best model was selected (in bold). Parameter estimates are given in the selection  Table S5. Estimates of fixed effects provided by optimal extrinsic model (Tables S1, S2, S3, S4) with the scaled and centered response and explanatory variables. "CI" -confidence intervals for the fixed effect estimates, ":" indicates interaction.  Table S7. Comparison of combined, within and between-individual thermal effects on Atlantic cod growth. Series of models were fitted to the data with the optimal fixed and random effects structure and optimal time window of sea surface temperature (Tables S1, S2 Table S8. Selection of the optimal random effect structure to test for individual differences in growth plasticity. Series of models were fitted to the data with the previously identified optimal fixed and random effects structure (Tables S1, S2,    14 Fig. S7. Individual growth trajectories. The first and last increment measurements were excluded from the analysis because they showed incomplete growth.

Note 1: Stock dynamics and environmental data
The reconstruction of the stock dynamics of the Icelandic cod stock was based on combining the catch at age (age 3-14) matrix for years 1928-1954(Schopka 1994) and 1955-2017(ICES 2019. Tuning indices were based on age groups 1 to 10 from the Icelandic spring groundfish survey and Icelandic autumn groundfish survey (NWWG 2019).
16 Fig. S8. Environmental, fishing and stock dynamics data used in the study. Mean sea surface temperature in the spawning area (a); mean sea surface temperature in the Icelandic shelf (b); harvest rate for the stock (c); number of individuals at age groups (d); number of individuals scaled within age groups (e).