Maternal, social and abiotic environmental effects on growth vary across life stages in a cooperative mammal

Resource availability plays a key role in driving variation in somatic growth and body condition, and the factors determining access to resources vary considerably across life stages. Parents and carers may exert important influences in early life, when individuals are nutritionally dependent, with abiotic environmental effects having stronger influences later in development as individuals forage independently. Most studies have measured specific factors influencing growth across development or have compared relative influences of different factors within specific life stages. Such studies may not capture whether early-life factors continue to have delayed effects at later stages, or whether social factors change when individuals become nutritionally independent and adults become competitors for, rather than providers of, food. Here, we examined variation in the influence of the abiotic, social and maternal environment on growth across life stages in a wild population of cooperatively breeding meerkats. Cooperatively breeding vertebrates are ideal for investigating environmental influences on growth. In addition to experiencing highly variable abiotic conditions, cooperative breeders are typified by heterogeneity both among breeders, with mothers varying in age and social status, and in the number of carers present. Recent rainfall had a consistently marked effect on growth across life stages, yet other seasonal terms only influenced growth during stages when individuals were growing fastest. Group size and maternal dominance status had positive effects on growth during the period of nutritional dependence on carers, but did not influence mass at emergence (at 1 month) or growth at independent stages (>4 months). Pups born to older mothers were lighter at 1 month of age and subsequently grew faster as subadults. Males grew faster than females during the juvenile and subadult stage only. Our findings demonstrate the complex ways in which the external environment influences development in a cooperative mammal. Individuals are most sensitive to social and maternal factors during the period of nutritional dependence on carers, whereas direct environmental effects are relatively more important later in development. Understanding the way in which environmental sensitivity varies across life stages is likely to be an important consideration in predicting trait responses to environmental change. © 2013 The Authors. Journal of Animal Ecology © 2013 British Ecological Society. This study was conducted using long-term data from a wild population of meerkats inhabiting private ranch land in the South African Kalahari Desert (26°58′S, 21°49′E). All individuals in the population were tagged with unique subcutaneous transponder chips and were identifiable in the field through dye marks on their fur. Groups were visited approximately three times per week and all life-history events, including births, deaths, immigration and emigration, were recorded. Further details on the study site and population are described elsewhere (Clutton-Brock et al. 1998; Russell et al. 2002). In this study, factors affecting body mass and growth were investigated between birth and 18 months of age in a total of 1378 individuals from 119 mothers in 26 social groups, born between January 1998 and December 2009. Five separate analyses were conducted to investigate factors influencing mass (first stage) or growth (all other stages) in the following life stages: (i) ‘emergence’, at 1 month of age (when pups are first weighed, shortly after emerging from the natal burrow); (ii) ‘pups’, between 1 and 3 months of age (when individuals are still nutritionally dependent on adults); (iii) ‘juveniles’, between 4 and 6 months of age (when individuals are foraging independently, yet contribute little to cooperative care); (iv) ‘subadults’ between 10 and 12 months of age (when individuals are sexually mature and have started helping); and (v) ‘adults’, between 16 and 18 months of age (beyond which age few individuals remain in their natal group as subordinates). Two-month, fixed windows for growth were selected to assess the effects of short-term fluctuations in abiotic and social environmental factors and to compare them across the different stages of development. While meerkat growth is nonlinear overall, best described by a modified monomolecular curve (English, Bateman & Clutton-Brock 2012), linear approximations of growth on two-month time windows allowed for straightforward assessment of relevant effects (see Fig. S1, Supporting information). Mass measurements were obtained without the need for capture, as most individuals (>95%) in the population were trained to step onto a top-pan electronic scale in return for a small reward (<1 g) of egg or water. In this study, pre-foraging mass measurements taken in the morning were used, to avoid any short-term fluctuations in mass due to variable foraging success. To avoid error due to missing data or variation in sampling effort, an interpolated monthly mass measure was calculated for individuals for each age in months (for a similar approach, see Ozgul et al. 2010). This monthly measure was calculated by first conducting linear mixed-effect models for all individuals including mass measurements for 1 month before and after each monthly age, with age and age2 as fixed-effect terms, and individual as a random term. A quadratic term of age was included to account for potential deceleration of growth across the period. These models were then used to estimate a best linear unbiased predictor for each individual's mass for its exact monthly age, conditional both on the fixed-effect terms and individual-level variation. Growth measures were calculated as the difference between monthly mass measures at the appropriate ages. All analyses on growth accounted for mass at the start of the period of interest. Previous work on meerkats has demonstrated that long-term growth is influenced by both season and rain (English, Bateman & Clutton-Brock 2012). A sine-plus-cosine function was included to account for intra-annual seasonal periodicity, by fitting two coefficients multiplied by sin(2·π·day/365·25) and cos(2·π·day/365·25), respectively, where ‘day’ represents the day-of-year when an individual turned the end-age of the life stage in question. Total rainfall in the two-month window prior to the mid-point of the focal period was also included. Rainfall data were obtained from the NASA GES DISC (Goddard Earth Sciences Data and Information Services Center) Giovanni online data system (described in Acker & Leptoukh 2007). The effects of both nutritionally dependent and independent group members on growth were considered by including the number of individuals younger than 3 months of age (number of pups) and the number of individuals over 6 months of age (number of adults, i.e. potential helpers), as well as a quadratic term on the latter to account for potential negative effects of resource competition in large groups. Mean values during the two-month window prior to the mid-point of the focal period were used in all analyses. Maternal age (in days) and dominance status at birth were both included in all analyses. A quadratic term of maternal age was also considered, to test for effects of senescence (Sharp & Clutton-Brock 2010). Maternal dominance status was assessed primarily through field observation, as one female (usually the dominant) tended to give birth at a time. In the rare cases where several females bred at the same time, maternity was inferred based on genetic data (details on molecular genetic analysis are described in Nielsen et al. 2012). The focal individual's sex was also included in order to assess whether sex differences, if any, emerge across development in this relatively size-monomorphic species. Linear mixed models, created in MCMCglmm (v. 2.16, Hadfield 2010) in R (v. 2.15, R Core Team 2012), were used to analyse the data. Continuous predictor variables were mean-centred and standardized for each data set for a particular growth period, for ease of comparison within and among models. All predictor variables were retained in each model, as our aim was not to derive the best predictive model of growth at each stage, but to compare the relative influence of predictor variables across different stages. MCMCglmm was therefore used calculate 95% credible intervals for each fixed parameter. MCMCglmm iterations were run with default inverse Wishart priors set at V = 1 and nu = 0·002 for all random effects (Gelman & Hill 2007). For each model, three separate chains were run and convergence of model parameters assessed by calculating the Gelman–Rubin statistic (Gelman 1996). For each chain, 2 000 000 iterations were run, with samples taken every 500 iterations and the first 1 500 000 removed as burn-in. This resulted in 1000 samples, which were used to calculate posterior modes and 95% credible intervals for the parameters. When credible intervals did not span zero, the parameter's effect was deemed to be statistically significant. Collinearity among predictor variables was assessed prior to analysis by calculating variance inflation factors (Zuur et al. 2009). As these were all less than 1·8, collinearity was deemed unlikely to affect the results. Random intercept terms for litter identity, mother identity and group identity were included in all models. The former two terms accounted for unexplained variation based on common genetic and environmental factors shared by littermates and individuals born to the same mother. Group identity accounted for unexplained variation affecting members of the same group. Repeatability estimates and 95% credibility intervals for each random-effect term were calculated following Nakagawa & Schielzeth (2010).