Search

Rivers offer cultural ecosystem services (CES) that improve people's quality of life. Advancements in computing and data storage have primarily focused on terrestrial CES, neglecting riverine areas. This study aims to develop a methodology to assess CES in riverine landscapes from social media and citizen science images related to environmental information. We collected georeferenced pictures from Flickr and iNaturalist for three main test rivers in northwest Portugal (Minho, Lima and Cávado) and classified them based on content such as 'biodiversity', 'recreation/river beaches', 'historical heritage' and 'landscape', as well as environmental spatial variables. A multimodel inference approach was applied to predict the spatial distribution of the pictures and environmental variables to support CES mapping. The methodology was applied during two time periods, before and during the most restrictive period of the COVID‐19 pandemic. Results showed that estuaries were identified as 'hotspots' for CES related to rivers provision. There was distinct prevalence of pictures depending on the targeted river: pictures exhibiting 'recreation/river beaches' prevailed in Cávado (62%), 'biodiversity' in Lima (70%) and 'historical heritage' in Minho (39%). Only the values and patterns from the category 'biodiversity' were maintained on the two analysed periods, with the other categories not having posts in social media during COVID‐19 most restrictive period. The methodology for CES assessment in rivers can be replicated using different time periods and regions due to its simple stepwise framework. The study provides valuable insights for sociocultural approaches, aiding in decision‐making on freshwater environment management, despite potential limitations in image distribution. [ABSTRACT FROM AUTHOR]

Specimens of the Pacific white shrimp Litopenaeus vannamei raised in a polyculture pond of grass carp Ctenopharyngodon idellus were collected biweekly for five months. Six common candidate growth models including the specialized von Bertalanffy, the generalized von Bertalanffy, Brody, Gompertz, Logistic and Richards were used to fit the growth process of L. vannamei. The growth parameters of these models were estimated using maximum likelihood methods, and the model fitting was determined by the Akaike information criterion (AIC) and Bayesian Information Criterion (BIC). The results showed that body weight vs body length relationship of L. vannamei was W=0.0106L3.066 (R²=0.987), which presented the positive allometric growth (b>3) by t-test. According to AIC and BIC tests, the models of the specialized von Bertalanffy, the generalized von Bertalanffy and Brody essentially were not used because of no proted data, while the models of Richards, Logistic and Gompertz received strong support from the data with AIC and BIC tests. However, no one model of akaike weight exceeds 95%, and can be used separately to describe the growth relationship between the body length and culture day of L. vannamei. Therefore the model - averaged asymptotic length L of L. vannamei was estimated as a weighted average using three models. The growth parameter estimation was L∞ 13.741 cm (w∞ 32.694 g) of L. vannamei by multi-model inference approach. [ABSTRACT FROM AUTHOR]

Abstract Bayesian approaches to the modelling of ecological systems are increasingly popular, but there are competing methods for formal model comparisons. Here, we focus on the task of performing multimodel inference through estimating posterior model weights, which encompasses uncertainties in the choice of competing model structure into the inference outputs. Model‐based approaches such as reversible‐jump Markov chain Monte Carlo (RJ‐MCMC) are flexible and allow multimodel inference, but can be complex to implement and optimise, and so we translate a model‐based approach for ecological applications using Importance Sampling to estimate the marginal likelihood of the data given a particular model. This approach allows for model comparison through the estimation of Bayes' Factors or interpretable posterior model probabilities, yielding model weights that facilitate multimodel inference through Bayesian model averaging. We demonstrate Importance Sampling with two case study investigations in animal demography: censused analysis of banded mongoose (Mungos mungo) survival where missing data are uncommon, and capture–mark–recapture analysis of European badger (Meles meles) survival where data are commonly missing. We compare outcomes of the model comparison using the Importance Sampling approach to those obtained through single‐model inference approaches using Deviance information criteria and the Watanabe–Akaike information criteria. The results of the Importance Sampling method aligns with RJ‐MCMC model comparisons while often being more straightforward to fit and optimise, particularly if the competing models are non‐nested.

Bayesian approaches to the modelling of ecological systems are increasingly popular, but there are competing methods for formal model comparisons. Here, we focus on the task of performing multimodel inference through estimating posterior model weights, which encompasses uncertainties in the choice of competing model structure into the inference outputs.Model‐based approaches such as reversible‐jump Markov chain Monte Carlo (RJ‐MCMC) are flexible and allow multimodel inference, but can be complex to implement and optimise, and so we translate a model‐based approach for ecological applications using Importance Sampling to estimate the marginal likelihood of the data given a particular model. This approach allows for model comparison through the estimation of Bayes' Factors or interpretable posterior model probabilities, yielding model weights that facilitate multimodel inference through Bayesian model averaging.We demonstrate Importance Sampling with two case study investigations in animal demography: censused analysis of banded mongoose (Mungos mungo) survival where missing data are uncommon, and capture–mark–recapture analysis of European badger (Meles meles) survival where data are commonly missing.We compare outcomes of the model comparison using the Importance Sampling approach to those obtained through single‐model inference approaches using Deviance information criteria and the Watanabe–Akaike information criteria. The results of the Importance Sampling method aligns with RJ‐MCMC model comparisons while often being more straightforward to fit and optimise, particularly if the competing models are non‐nested. [ABSTRACT FROM AUTHOR]

An update of the age and growth for juveniles of the short fin mako shark (I. oxyrinchus) from the Mexican Pacific is presented, based on the analysis of growth band counts from dorsal vertebrae of 198 individuals [110 females, 74–231 cm of total length (TL) and 88 males, 72–231 cm TL) caught during 2008–2018. New available information on vertebral growth band periodicity (biannual deposition in juveniles) and the convenience of using vertebrae form the dorsal region over the cervical region to count growth bands, as well as a multimodel approach, were used. The von Bertalanffy (VB) growth model, Gompertz, logistic and two parameters of VB (2‐VB) were fitted to the length‐at‐age. Only ages ≤6 years were used for the fitting of the models and their performance was compared with the small‐sample bias‐corrected form of the Akaike information criterion (AICc), their differences (∆i) and weights (wi). Following a multimodel inference approach, the model averaged asymptotic length (L¯inf), length‐at‐age 0 (L¯0) and their unconditional standard error (SE¯), were estimated for each sex scenario using the three‐parameter version of each model. The precision of growth band counts was acceptable for the different methods used and by two different readers. The centrum edge analysis (CEA) and marginal increment analysis (MIA) did not support the hypothesis of biannual band pair formation for juveniles, likewise for adults the periodicity could not be verified due to the small sample of large animals. Age was estimated assuming the formation of two pairs of growth bands per year during the first 5 years and one pair of bands per year afterwards considering direct validation information. The estimated ages in years ranged from 0–14 for females and 0–6 for males. The Kimura likelihood ratio test showed no differences in the growth curves of juveniles by sex (P > 0.05). According to the AICc, the 2‐VB model better fitted the length‐at‐age data for combined sexes (Linf = 386.4 cm, k = 0.12 years−1, L0 = 70 cm). The model averaged L¯inf and L¯0 were 378.3 cm (SE¯=64.5) and 69.5 cm (SE¯=6.3), respectively. The growth parameters determined for juveniles of I. oxyrinchus are similar to those estimated in other regions, showing relatively fast growth rate as previously reported, medium longevity in comparison to other shark species and natural mortality close to that reported in the last stock assessment for the North Pacific Ocean. These life‐history parameters should be considered to evaluate the population in the region and to develop better fishery management and conservation measures. [ABSTRACT FROM AUTHOR]

To describe the growth pattern of the bonnethead shark (Sphyrna tiburo) in the Gulf of Mexico, a von Bertalanffy (VB) model has been automatically fit, which indicated a single−phase continuous growth without oscillations, though this would generate biases if this hypothesis is not confirmed. The objective of this research was to describe the growth pattern of S. tiburo under a multimodel approach based on information theory and contrasting single−phase models (VB, Gompertz, logistic models, and variants) and biphasic models (Soriano model and variants). The VB model was not supported. The Soriano model, with the variant in growth rate (k) and including length at birth (L0), was selected with 100% supporting evidence. The hypothesis of the two−phase growth of S. tiburo with an increase in k, more than L∞, fitted to L0, is confirmed, and a correspondence was identified between growth−phase change sizes and the sizes reported in the literature for change in the juvenile–adult stages in females and for onset of reproductive maturity in males and both sexes. [ABSTRACT FROM AUTHOR]

Abstract The natural and seminatural components of agricultural landscapes play a key role in maintaining a high level of biodiversity. Being the Po Valley one of the most human‐dominated and intensively cultivated landscapes in Europe, we investigated the effect of no‐crop habitats on carabid richness and composition and evaluated the role of tree row as corridor for forest carabid dispersion. Carabids were sampled with 70 pitfall traps arranged in 35 sampling plots along three parallel transects (80, 100, and 140 m long) and encompassing five different habitats: tree row, tree row edge, grassland, forest edge, and forest. We found 5,615 individuals belonging to 55 species. Despite the similarity in species richness, all the habitats investigated showed a peculiar and distinct species assemblage. The main distinction was between the "open habitat" cluster composed of grassland and tree row edge and the “forest" cluster composed of forest, tree row, and forest edge. We found that forest species are able to penetrate the grassland matrix up to 30 m from the forest edge and that a distance of no more than 60 m between tree row and forest can allow the passage of up to 50% of the forest species. Beyond this distance, the grassland matrix becomes a barrier, preventing them from reaching other suitable habitats. Our findings confirm the importance of maintaining different types of natural habitats to significantly increase biodiversity in an intensively cultivated agroecosystem and demonstrated the role of linear elements as a corridor and “stepping stones” for many forest species.

The multimodel inference makes statistical inferences from a set of plausible models rather than from a single model. In this paper, we focus on the multimodel inference based on smoothed information criteria proposed by seminal monographs (see Buckland et al. (1997) and Burnham and Anderson (2003)), which are termed as smoothed Akaike information criterion (SAIC) and smoothed Bayesian information criterion (SBIC) methods. Due to their simplicity and applicability, these methods are very widely used in many fields. By using an illustrative example and deriving limiting properties for the weights in the linear regression, we find that the existing variance estimation for SAIC is not applicable because of a restrictive condition, but for SBIC it is applicable. Especially, we propose a simulation-based inference for SAIC based on the limiting properties. Both the simulation study and real data example show the promising performance of the proposed simulation-based inference. [ABSTRACT FROM AUTHOR]

A fundamental goal of scientific research is to identify the underlying variables that govern crucial processes of a system. This is especially difficult in ecology, which is intrinsically rich in candidate predictors. An efficient statistical procedure to evaluate the relative importance of predictors in regression models is highly desirable. However, previous studies criticised the most universally applicable method, by pointing out the low discriminating power of the importance index in simulated datasets. Here we proposed a new index, WiBB, which integrates the merits of several existing methods. WiBB combines a model‐weighting method from information theory (Wi), a standardised regression coefficient method measured by β* (B), and bootstrap resampling technique (B). We applied the WiBB in simulated datasets with known correlation structures, for both linear models (LM) and generalized linear models (GLM), to evaluate its performance. We also applied it to an empirical dataset of a plant genus Mimulus to select bioclimatic predictors of species' presence across the landscape. Results in the simulated datasets showed that the bootstrap resampling technique significantly improved the discriminant ability by correctly sorting the orders of relative importance of predictors. The WiBB method outperformed the β* and the relative sum of weights (SWi, a standardised version of sum of weights) methods in scenarios with small and large sample sizes, respectively. When testing WiBB in the empirical dataset with GLM, it sensibly identified four important predictors with high credibility out of six candidates in modelling geographical distributions of 71 Mimulus species. This integrated index has great advantages in evaluating predictor importance and hence reducing the dimensionality of data, without losing interpretive power. The simplicity of calculation of the new metric over more sophisticated statistical procedures makes it a handy method in the statistical toolbox. [ABSTRACT FROM AUTHOR]

Obtaining the best possible estimates of individual growth parameters is essential in studies of physiology, fisheries management, and conservation of natural resources since growth is a key component of population dynamics. In the present work, we use data of an endangered fish species to demonstrate the importance of selecting the right data error structure when fitting growth models in multimodel inference. The totoaba (Totoaba macdonaldi) is a fish species endemic to the Gulf of California increasingly studied in recent times due to a perceived threat of extinction. Previous works estimated individual growth using the von Bertalanffy model assuming a constant variance of length-at-age. Here, we reanalyze the same data under five different variance assumptions to fit the von Bertalanffy and Gompertz models. We found consistent significant differences between the constant and nonconstant error structure scenarios and provide an example of the consequences using the growth performance index -0 to show how using the wrong error structure can produce growth parameter values that can lead to biased conclusions. Based on these results, for totoaba and other related species, we recommend using the observed error structure to obtain the individual growth parameters. [ABSTRACT FROM AUTHOR]

The natural and seminatural components of agricultural landscapes play a key role in maintaining a high level of biodiversity. Being the Po Valley one of the most human‐dominated and intensively cultivated landscapes in Europe, we investigated the effect of no‐crop habitats on carabid richness and composition and evaluated the role of tree row as corridor for forest carabid dispersion. Carabids were sampled with 70 pitfall traps arranged in 35 sampling plots along three parallel transects (80, 100, and 140 m long) and encompassing five different habitats: tree row, tree row edge, grassland, forest edge, and forest. We found 5,615 individuals belonging to 55 species. Despite the similarity in species richness, all the habitats investigated showed a peculiar and distinct species assemblage. The main distinction was between the "open habitat" cluster composed of grassland and tree row edge and the "forest" cluster composed of forest, tree row, and forest edge. We found that forest species are able to penetrate the grassland matrix up to 30 m from the forest edge and that a distance of no more than 60 m between tree row and forest can allow the passage of up to 50% of the forest species. Beyond this distance, the grassland matrix becomes a barrier, preventing them from reaching other suitable habitats. Our findings confirm the importance of maintaining different types of natural habitats to significantly increase biodiversity in an intensively cultivated agroecosystem and demonstrated the role of linear elements as a corridor and "stepping stones" for many forest species. [ABSTRACT FROM AUTHOR]

Abstract Background Invasive plants commonly occupy diverse habitats and thus must adapt to changing environmental pressures through altering their traits and economics spectra, and addressing these patterns and their drivers has an importantly ecological and/or evolutionary significance. However, few studies have considered the role of multiple biotic and abiotic factors in shaping trait variation and spectra. In this study, we determined seven leaf traits of 66 Solidago canadensis populations, and quantified the relative contributions of climate, soil properties, native plant diversity, and S. canadensis–community interactions (in total 16 factors) to leaf trait variation and spectrum with multimodel inference. Results Overall, the seven leaf traits had high phenotypic variation, and this variation was highest for leaf dry matter content and lowest for leaf carbon concentration. The per capita contribution of climate to the mean leaf trait variation was highest (7.5%), followed by soil properties (6.2%), S. canadensis–community interactions (6.1%), and native plant diversity (5.4%); the dominant factors underlying trait variation varied with leaf traits. Leaf production potential was negatively associated with leaf stress-tolerance potential, and the relative contributions to this trade-off followed in order: native plant diversity (7.7%), climate (6.9%), S. canadensis–community interactions (6.2%), and soil properties (5.6%). Climate, diversity, soil, and interactions had positive, neutral or negative effects. Conclusions Climate, soil, diversity, and interactions contribute differentially to the leaf trait variation and economics spectrum of S. canadensis, and their relative importance and directions depend on plant functional traits.

Meta‐regression can be used to examine the association between effect size estimates and the characteristics of the studies included in a meta‐analysis using regression‐type methods. By searching for those characteristics (i.e., moderators) that are related to the effect sizes, we seek to identify a model that represents the best approximation to the underlying data generating mechanism. Model selection via testing, either through a series of univariate models or a model including all moderators, is the most commonly used approach for this purpose. Here, we describe alternative model selection methods based on information criteria, multimodel inference, and relative variable importance. We demonstrate their application using an illustrative example and present results from a simulation study to compare the performance of the various model selection methods for identifying the true model across a wide variety of conditions. Whether information‐theoretic approaches can also be used not only in combination with maximum likelihood (ML) but also restricted maximum likelihood (REML) estimation was also examined. The results indicate that the conventional methods for model selection may be outperformed by information‐theoretic approaches. The latter are more often among the set of best methods across all of the conditions simulated and can have higher probabilities for identifying the true model under particular scenarios. Moreover, their performance based on REML estimation was either very similar to that from ML estimation or at times even better depending on how exactly the REML likelihood was computed. These results suggest that alternative model selection methods should be more widely applied in meta‐regression. [ABSTRACT FROM AUTHOR]

To describe the growth pattern of the bonnethead shark (Sphyrna tiburo) in the Gulf of Mexico, a von Bertalanffy (VB) model has been automatically fit, which indicated a single−phase continuous growth without oscillations, though this would generate biases if this hypothesis is not confirmed. The objective of this research was to describe the growth pattern of S. tiburo under a multimodel approach based on information theory and contrasting single−phase models (VB, Gompertz, logistic models, and variants) and biphasic models (Soriano model and variants). The VB model was not supported. The Soriano model, with the variant in growth rate (k) and including length at birth (L0), was selected with 100% supporting evidence. The hypothesis of the two−phase growth of S. tiburo with an increase in k, more than L∞, fitted to L0, is confirmed, and a correspondence was identified between growth−phase change sizes and the sizes reported in the literature for change in the juvenile–adult stages in females and for onset of reproductive maturity in males and both sexes.

Tree plantations have expanded rapidly during the last decades, specially in Asia and South America, and Chile has shown one of the largest increases in tree plantations in the world. The aim of this study was to analyze the dynamic and factors statistically associated with the expansion of tree plantations in two coastal basins in south-central Chile. We used logistic regression and a multimodel inference approach to assess the association of 13 location factors with tree plantation expansion in two periods (1987–2001; 2001–2015). The area of tree plantations increased by 292% and 196% in the Budi and Lingue basins, respectively. The presence of agriculture fields before conversion to tree plantations showed the greatest effect in both basins, followed by the suitability of the soil for forestry. Likewise, tree plantations were highly associated with the replacement of native vegetation, contributing to ongoing deforestation, and changes in forest policy increased the expansion in indigenous land. [ABSTRACT FROM AUTHOR]

Abstract In the “digital conservation” age, big data from Earth observations and from social media have been increasingly used to tackle conservation challenges. Here, we combined information from those two digital sources in a multimodel inference framework to identify, map, and predict the potential for nature's cultural contributions to people in two contrasting UNESCO biosphere reserves: Doñana and Sierra Nevada (Spain). The content analysis of Flickr pictures revealed different cultural contributions, according to the natural and cultural values of the two reserves. Those contributions relied upon landscape variables computed from Earth observation data: the variety of colors and vegetation functioning that characterize Doñana landscapes, and the leisure facilities, accessibility features, and heterogeneous landscapes that shape Sierra Nevada. Our findings suggest that social media and Earth observations can aid in the cost‐efficient monitoring of nature's contributions to people, which underlie many Sustainable Development Goals and conservation targets in protected areas worldwide.

There is growing evidence that species and communities are responding to, and will continue to be affected by, climate change. For species at risk, vulnerability can be reduced by ensuring that their habitat is extensive, connected and provides opportunities for dispersal and/or gene flow, facilitating a biological response through migration or adaptation. For woodland epiphytes, vulnerability might also be reduced by ensuring sufficient habitat heterogeneity, so that microhabitats provide suitable local microclimates, even as the larger scale climate continues to change (i.e. microrefugia). This study used fuzzy set ordination to compare bryophyte and lichen epiphyte community composition to a large-scale gradient from an oceanic to a relatively more continental macroclimate. The residuals from this relationship identified microhabitats in which species composition reflected a climate that was more oceanic or more continental than would be expected given the prevailing macroclimate. Comparing these residuals to features that operate at different scales to create the microclimate (landscape, stand and tree-scale), it was possible to identify how one might engineer microrefugia into existing or new woodland, in order to reduce epiphyte vulnerability to climate change. Multimodel inference was used to identify the most important features for consideration, which included local effects such as height on the bole, angle of bole lean and bark water holding capacity, as well as tree species and tree age, and within the landscape, topographic wetness and physical exposure. [ABSTRACT FROM AUTHOR]

Objective: To explore and apply multimodel inference to test the relative contributions of latent genetic, environmental and direct causal factors to the covariation between two variables with data from the classical twin design by estimating model-averaged parameters. Methods: Behavior genetics is concerned with understanding the causes of variation in phenotypes and the causes of covariation between two or more phenotypes. Two variables may correlate as a result of genetic, shared environmental or unique environmental factors contributing to variation in both variables. Two variables may also correlate because one or both directly cause variation in the other. Furthermore, covariation may result from any combination of these sources, leading to 25 different identified structural equation models. OpenMx was used to fit all these models to account for covariation between two variables collected in twins. Multimodel inference and model averaging were used to summarize the key sources of covariation, and estimate the magnitude of these causes of covariance. Extensions of these models to test heterogeneity by sex are discussed. Results: We illustrate the application of multimodel inference by fitting a comprehensive set of bivariate models to twin data from the Virginia Twin Study of Psychiatric and Substance Use Disorders. Analyses of body mass index and tobacco consumption data show sufficient power to reject distinct models, and to estimate the contribution of each of the five potential sources of covariation, irrespective of selecting the best fitting model. Discrimination between models on sample size, type of variable (continuous versus binary or ordinal measures) and the effect size of sources of variance and covariance. Conclusions: We introduce multimodel inference and model averaging approaches to the behavior genetics community, in the context of testing models for the causes of covariation between traits in term of genetic, environmental and causal explanations. [ABSTRACT FROM AUTHOR]

In the "digital conservation" age, big data from Earth observations and from social media have been increasingly used to tackle conservation challenges. Here, we combined information from those two digital sources in a multimodel inference framework to identify, map, and predict the potential for nature's cultural contributions to people in two contrasting UNESCO biosphere reserves: Doñana and Sierra Nevada (Spain). The content analysis of Flickr pictures revealed different cultural contributions, according to the natural and cultural values of the two reserves. Those contributions relied upon landscape variables computed from Earth observation data: the variety of colors and vegetation functioning that characterize Doñana landscapes, and the leisure facilities, accessibility features, and heterogeneous landscapes that shape Sierra Nevada. Our findings suggest that social media and Earth observations can aid in the cost‐efficient monitoring of nature's contributions to people, which underlie many Sustainable Development Goals and conservation targets in protected areas worldwide. [ABSTRACT FROM AUTHOR]

Urban biodiversity plays a crucial role in the functioning of urban ecosystems and significantly impacts the well-being and quality of life for city residents. By focusing on the city of Porto as a case study, the influence of local-scale land cover evolution on urban biodiversity is investigated, using species richness of birds, reptiles, and amphibians as indicators, within a multimodel inference framework. The results underscore the importance of past land cover and land cover legacy in shaping urban biodiversity. Birds, reptiles, and amphibians respond positively to older vegetation patches but negatively to long-established urban areas. Birds exhibit adaptability by positively responding to recent vegetation, while amphibians tend to avoid newly vegetated zones. The fragmented distribution of amphibians and reptiles in Porto suggests limited mobility and a potential delayed response to habitat loss and isolation. Reptiles benefit from both wooded and herbaceous habitats, emphasizing the importance of local spatial diversity, while water elements are critical for amphibians, although many aquatic habitats in the urban context may not be suitable. Modeling results guide the identification of priority areas for urban biodiversity conservation in Porto, informing decision-makers, urban planners, and conservationists. This spatially explicit research aids efforts to create more ecologically resilient and biodiverse urban environments. It highlights the role of historical land cover and the unique responses of different vertebrate groups to urbanization, contributing to our understanding of urban biodiversity dynamics and sustainable urban development. • Vertebrate groups respond positively to vegetation, negatively to artificial cover. • Birds adapt to both older and recent vegetation, due to higher mobility. • Herpetofauna relies on small-scale habitat diversity, such as water features. • Herpetofauna's fragmented distribution suggests delayed response to habitat loss. • Spatial prioritization is particularly relevant for urban planning and management. [ABSTRACT FROM AUTHOR]

The age and growth rate of the giant electric ray, Narcine entemedor, was estimated using growth bands deposited in the vertebral centra of 245 specimens. Differences in size and age distribution were found between the sexes, a pattern that suggests the annual deposition of band pairs, possibly occurring in April. Multimodel inference and back-calculation were performed to three age data sets of females considering their reproductive cycle and time of capture, among which the von Bertalanffy growth function was found to be the most appropriate (L∞ = 81.87 cm TL, k = 0.17 year−1). Our research supports the idea that age can be determined via biological features such as birth date and growth band periodicity. We concluded that N. entemedor is of a moderate body size, moderate longevity and is a fast-growing elasmobranch species.

Obtaining the best possible estimates of individual growth parameters is essential in studies of physiology, fisheries management, and conservation of natural resources since growth is a key component of population dynamics. In the present work, we use data of an endangered fish species to demonstrate the importance of selecting the right data error structure when fitting growth models in multimodel inference. The totoaba (Totoaba macdonaldi) is a fish species endemic to the Gulf of California increasingly studied in recent times due to a perceived threat of extinction. Previous works estimated individual growth using the von Bertalanffy model assuming a constant variance of length-at-age. Here, we reanalyze the same data under five different variance assumptions to fit the von Bertalanffy and Gompertz models. We found consistent significant differences between the constant and nonconstant error structure scenarios and provide an example of the consequences using the growth performance index ϕ′ to show how using the wrong error structure can produce growth parameter values that can lead to biased conclusions. Based on these results, for totoaba and other related species, we recommend using the observed error structure to obtain the individual growth parameters.

Abstract Ecologists routinely fit complex models with multiple parameters of interest, where hundreds or more competing models are plausible. To limit the number of fitted models, ecologists often define a model selection strategy composed of a series of stages in which certain features of a model are compared while other features are held constant. Defining these multi‐stage strategies requires making a series of decisions, which may potentially impact inferences, but have not been critically evaluated. We begin by identifying key features of strategies, introducing descriptive terms when they did not already exist in the literature. Strategies differ in how they define and order model building stages. Sequential‐by‐sub‐model strategies focus on one sub‐model (parameter) at a time with modeling of subsequent sub‐models dependent on the selected sub‐model structures from the previous stages. Secondary candidate set strategies model sub‐models independently and combine the top set of models from each sub‐model for selection in a final stage. Build‐up approaches define stages across sub‐models and increase in complexity at each stage. Strategies also differ in how the top set of models is selected in each stage and whether they use null or more complex sub‐model structures for non‐target sub‐models. We tested the performance of different model selection strategies using four data sets and three model types. For each data set, we determined the "true" distribution of AIC weights by fitting all plausible models. Then, we calculated the number of models that would have been fitted and the portion of "true" AIC weight we recovered under different model selection strategies. Sequential‐by‐sub‐model strategies often performed poorly. Based on our results, we recommend using a build‐up or secondary candidate sets, which were more reliable and carrying all models within 5–10 AIC of the top model forward to subsequent stages. The structure of non‐target sub‐models was less important. Multi‐stage approaches cannot compensate for a lack of critical thought in selecting covariates and building models to represent competing a priori hypotheses. However, even when competing hypotheses for different sub‐models are limited, thousands or more models may be possible so strategies to explore candidate model space reliably and efficiently will be necessary.

Ecologists routinely fit complex models with multiple parameters of interest, where hundreds or more competing models are plausible. To limit the number of fitted models, ecologists often define a model selection strategy composed of a series of stages in which certain features of a model are compared while other features are held constant. Defining these multi‐stage strategies requires making a series of decisions, which may potentially impact inferences, but have not been critically evaluated. We begin by identifying key features of strategies, introducing descriptive terms when they did not already exist in the literature. Strategies differ in how they define and order model building stages. Sequential‐by‐sub‐model strategies focus on one sub‐model (parameter) at a time with modeling of subsequent sub‐models dependent on the selected sub‐model structures from the previous stages. Secondary candidate set strategies model sub‐models independently and combine the top set of models from each sub‐model for selection in a final stage. Build‐up approaches define stages across sub‐models and increase in complexity at each stage. Strategies also differ in how the top set of models is selected in each stage and whether they use null or more complex sub‐model structures for non‐target sub‐models. We tested the performance of different model selection strategies using four data sets and three model types. For each data set, we determined the "true" distribution of AIC weights by fitting all plausible models. Then, we calculated the number of models that would have been fitted and the portion of "true" AIC weight we recovered under different model selection strategies. Sequential‐by‐sub‐model strategies often performed poorly. Based on our results, we recommend using a build‐up or secondary candidate sets, which were more reliable and carrying all models within 5–10 AIC of the top model forward to subsequent stages. The structure of non‐target sub‐models was less important. Multi‐stage approaches cannot compensate for a lack of critical thought in selecting covariates and building models to represent competing a priori hypotheses. However, even when competing hypotheses for different sub‐models are limited, thousands or more models may be possible so strategies to explore candidate model space reliably and efficiently will be necessary. [ABSTRACT FROM AUTHOR]

Young and Holsteen (YH) introduce a number of tools for evaluating model uncertainty. In so doing, they are careful to differentiate their method from existing forms of model averaging. The fundamental difference lies in the way in which the underlying estimates are weighted. Whereas standard approaches to model averaging assign higher weight to better fitting models, the YH method weights all models equally. As I show, this is a nontrivial distinction, in that the two sets of procedures tend to produce radically different results. Drawing on both simulation and real-world examples, I demonstrate that in failing to distinguish between numerical variation and statistical uncertainty, the procedure proposed by YH will tend to overstate the amount of uncertainty resulting from variation across models. In standard circumstances, the quality of estimates produced using this method will tend to be objectively worse than that of conventional alternatives. [ABSTRACT FROM AUTHOR]

Abstract The aim of the presented studies was to investigate the relationships between mindfulness, trait anxiety, attentional control and cognitive failures. In two studies the structural equation modelling and multimodel inference were applied. The results of the first study (N = 203) led us to conclude that trait anxiety is a strong predictor of both receptive attention and decentration; these two facets of mindfulness are not correlated with each other when anxiety is controlled; attentional control partially mediates the relationship between anxiety and decentration but attentional control does not relate to receptive attention. In study two (N = 216) the findings of the first study were replicated, and models with cognitive failures as the dependent variable were tested. The results strongly suggested that both facets of mindfulness are independent mediators of the relationship between trait anxiety and cognitive failures. Moreover, decentration, but not receptive attention, was also a full mediator between attentional control and cognitive failures. Both studies revealed that anxiety is a more important predictor of mindfulness than attentional control. They also showed that mindfulness explains some effects of anxiety independently of attentional control. These results are discussed in the context of previous findings and the metacognitive model of mindfulness. [ABSTRACT FROM AUTHOR]

Investment in immune function can be costly, and life-history theory predicts trade-offs between immune function and other physiological demands. Environmental heterogeneity may constrain or change the optimal strategy and thereby alter baseline immune function (possibly mediated by stress responses). We tested several hypotheses relating variation in climatic, ecological, and social environments to chronic stress and levels of baseline innate immunity in awild, cooperatively breeding bird, the purple-crowned fairy-wren (Malurus coronatus coronatus). From samples collected biannually over 5 yr, we quantified three indexes of constitutive innate immune function (haptoglobin/PIT54, natural antibodies, complement activity) and one index of chronic stress (heterophil-lymphocyte ratio; n = 513–647). Using an information-theoretic and multimodel inference statistical approach, we found that habitat quality and social group size did not affect any immune index, despite hypothesized links to resource abundance and parasite pressure. Rather, short-term variation in temperature and rainfall was related to immune function, while overall differences between seasons were small or absent, despite substantial seasonal variation in climate. Contrary to our expectation, we found no evidence that physiological stress mediated any effects of short-term climatic variables on immune indexes, and alternative mechanisms may be involved. Our results may be interpreted from the perspective of reactive scope models, whereby predictive homeostasis maintains standing immune function relative to long-term demands, while short-term environmental change, being less predictable, has a greater influence on baseline immune function. [ABSTRACT FROM AUTHOR]

Highlights • A framework to guide forest management to deal more effectively with plant invasions. • Models based on age/size classes represent a significant improvement. • Height classes are influenced by distinct factors operating at different scales. • Step forward for establishing management actions at appropriate spatial scales. • Way to enhance the efficiency of control actions on major forest invaders. Abstract Identifying the key factors driving invasion processes is crucial for designing and implementing appropriate management strategies. In fact, the importance of (model-based) prevention and early detection was highlighted in the recent European Union regulation on Invasive Alien Species. Models based on abundance estimates for different age/size classes would represent a significant improvement relative to the more usual models based only on species' occurrence data. Here, we evaluate the relative contribution of different environmental drivers to the spatial patterns of abundance of several height classes (or life-stages) of invasive tree populations at the regional scale, using a data-driven hierarchical modelling approach. A framework for modelling life-stages to obtain spatial projections of their potential occurrence or abundance has not been formalized before. We used Acacia dealbata (Silver-wattle) as a test species in northwest of Portugal, a heavily invaded region, and applied a multimodel inference to test the importance of various environmental drivers in explaining the abundance patterns of five plant height classes in local landscape mosaics. The ensemble of height classes is considered here as a proxy for population dynamics, life-stages and age of adult trees. In this test with A. dealbata , we used detailed field data on population height structure and calibrated an independent model for each height class. We found evidence to support our hypothesis that the distribution of height classes is mostly influenced by distinct factors operating at different scales. The spatial projections which resulted from several height class models provide an overview of population structure and invasion dynamics considering various life-stages, that is widely used in biodiversity and invasion research. The approach proposed here provides a framework to guide forest management to deal more effectively with plant invasions. It allows to test the effects of key invasion factors (depending on the focal species and on data availability) and supports the spatial identification of suitable areas for invasive species' occurrence while also accounting for the structural complexity of invasive species populations, thereby anticipating future invasion dynamics. The approach thus constitutes a step forward for establishing management actions at appropriate spatial scales and for focusing on earlier stages of invasion and their respective driving factors (regeneration niche), thereby enhancing the efficiency of control actions on major forest invaders. [ABSTRACT FROM AUTHOR]

Graphical abstract Highlights • Candidate kinetic models are obtained from a systematic approach. • All combinations of plausible elementary steps are considered. • Mathematical and physico-chemical criteria are used to discriminate among models. • Multimodel inferences based on Akaike Information Criterion are applied. • The best kinetic model stems from an Eley–Rideal mechanism. Abstract The kinetics of the liquid-phase addition of 1-propanol to isobutene to produce propyl tert -butyl ether (PTBE) has been studied using the ion-exchange resin Amberlyst™ 35 as the catalyst. Reaction rates free from mass transfer limitations have been experimentally determined in the temperature range 303–352 K, for different initial proportions of alcohol and isobutene, and using two different reactor types (i.e., a batch stirred tank reactor, to obtain most of the experimental data, and a tubular reactor, to validate those results). To find out the best kinetic model, a systematic approach has been adopted. The overall etherification reaction has been decomposed as the result of elementary steps based on Langmuir–Hinshelwood–Hougen–Watson or Eley–Rideal mechanisms. Candidate kinetic equations have been originated from all possible combinations of adsorbed and non-adsorbed compounds, and rate-determining step. The possible effect of the interaction between the reaction medium and the resin on reaction rates has been also examined. Since all experimental data have been used at once in the fit of the kinetic equations, all combinations of significant or non-significant temperature dependence of model parameters have been also considered. As a result, 1404 candidate kinetic equations have been fitted separately to experimental data. Discrimination among models is based on mathematical and physico-chemical criteria. The final choice of the best kinetic model involves multimodel inference. It corresponds to an Eley–Rideal mechanism where one 1-propanol molecule adsorbed on the catalyst reacts with one isobutene molecule from the liquid phase to form one adsorbed PTBE molecule, the surface reaction being the rate-determining step. [ABSTRACT FROM AUTHOR]

Fisheries management is dominated by the need to forecast catch and abundance of commercially and ecologically important species. The influence of spatial information and environmental factors on forecasting error is not often considered. I propose a forecasting method called spatiotemporally explicit model averaging (STEMA) to combine spatial and temporal information through model averaging. I examine the performance of STEMA against two popular forecasting models and a modern spatial prediction model: the autoregressive integrated moving averages with explanatory variables (ARIMAX) model, the Bayesian hierarchical model, and the varying coefficient model. I focus on applying the methods to four species of Alaskan groundfish for which catch data are available. My method reduces forecasting errors significantly for most of the tested models when compared to ARIMAX, Bayesian, and varying coefficient methods. I also consider the effect of sea surface temperature (SST) on the forecasting of catch, as multiple studies reveal a potential influence of water temperature on the survival and growth of juvenile groundfish. For most of the preferred models, inclusion of SST in the model improved forecasting of catch. It is advisable to consider both spatial information and relevant environmental factors in forecasting models to obtain more accurate projections of population abundance. The STEMA method is capable of accounting for spatial information in forecasting and can be applied to various types of data because of its flexible varying coefficient model structure. It is therefore a suitable forecasting method for application to many fields including ecology, epidemiology, and climatology. We propose a forecasting method called spatiotemporally explicit model averaging (STEMA) to combine spatial and temporal information through model averaging. We focus on applying the method to four species of Alaskan groundfish and consider the effect of sea surface temperature (SST) on the forecasting of catch. Our method reduces forecasting errors significantly for most of the tested models when compared to two popular forecasting models and a modern spatial prediction model. For most of the preferred models, inclusion of SST in the model improved forecasting of catch. [ABSTRACT FROM AUTHOR]

Copyright of Avian Conservation & Ecology is the property of Resilience Alliance and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Anthropogenic global warming, nitrogen addition, and overgrazing alter plant communities and threaten plant biodiversity, potentially impacting community productivity, especially in sensitive mountain grassland ecosystems. However, it still remains unknown whether the relationship between plant biodiversity and community productivity varies across different anthropogenic influences, and especially how changes in multiple biodiversity facets drive these impacts on productivity. Here, we measured different facets of biodiversity including functional and phylogenetic richness and evenness in mountain grasslands along an environmental gradient of elevation in Yulong Mountain, Yunnan, China. We combined biodiversity metrics in a series of linear mixed‐effect models to determine the most parsimonious predictors for productivity, which was estimated by aboveground biomass in community. We examined how biodiversity–productivity relationships were affected by experimental warming, nitrogen addition, and livestock‐grazing. Species richness, phylogenetic diversity, and single functional traits (leaf nitrogen content, mg/g) represented the most parsimonious combination in these scenarios, supporting a consensus that single‐biodiversity metrics alone cannot fully explain ecosystem function. The biodiversity–productivity relationships were positive and strong, but the effects of treatment on biodiversity–productivity relationship were negligible. Our findings indicate that the strong biodiversity–productivity relationships are consistent in various anthropogenic drivers of environmental change. Our study emphasized the strong and consistent effects of biodiversity on productivity under experimental warming, nitrogen addition, and grazing in mountain grasslands. Here, species richness, phylogenetic diversity, and single functional traits represented the most parsimonious combinations in various anthropogenic drivers of environmental change. [ABSTRACT FROM AUTHOR]

Abstract The assessment of fetal size and the accurate estimation of gestational age are of crucial importance for proper pregnancy management. The information is almost exclusively based on ultrasound measurements of fetal biometric parameters and the means for evaluating these measurements are age-related reference charts (centile charts) allowing interpretation of obtained fetal measurement in comparison with the expected average measurement in the reference population. The construction of such reference charts requires an appropriate statistical methodology. The most frequent method for the construction of fetal reference charts from cross-sectional data is the parametric approach with fractional polynomials regression functions for the mean and standard deviation of each fetal measurement. This article suggests how this method can be extended to longitudinal data using fractional polynomials in linear mixed effect regression. The presented approach includes maximum likelihood estimation for fitting first- and second-order fractional polynomial models, and multimodel inference using Akaike's information criterion and related tools as a suitable strategy for model selection. Finally, an example of the suggested approach is presented. [ABSTRACT FROM AUTHOR]

Abstract: A fundamental goal of disease ecology is to determine the landscape and environmental processes that drive disease dynamics at different biological levels to guide management and conservation. Although ranaviruses (family Iridoviridae) are emerging amphibian pathogens, few studies have conducted comprehensive field surveys to assess potential drivers of ranavirus disease dynamics. We examined the factors underlying patterns in site‐level ranavirus presence and individual‐level ranavirus infection in 76 ponds and 1,088 individuals representing five amphibian species within the East Bay region of California. Based on a competing‐model approach followed by variance partitioning, landscape and biotic variables explained the most variation in site‐level presence. However, biotic and individual‐level variables explained the most variation in individual‐level infection. Distance to nearest ranavirus‐infected pond (the landscape factor) was more important than biotic factors at the site level; however, biotic factors were most influential at the individual level. At the site level, the probability of ranavirus presence correlated negatively with distance to nearest ranavirus‐positive pond, suggesting that the movement of water or mobile taxa (e.g., adult amphibians, birds, reptiles) may facilitate the movement of ranavirus between ponds and across the landscape. Taxonomic richness associated positively with ranavirus presence at the site level, but vertebrate richness associated negatively with infection prevalence in the host population. This might reflect the contrasting influences of diversity on pathogen colonisation versus transmission among hosts. Amphibian host species differed in their likelihood of ranavirus infection: American bullfrogs (Rana catesbeiana) had the weakest association with infection while rough‐skinned newts (Taricha granulosa) had the strongest. After accounting for host species effects, hosts with greater snout–vent length had a lower probability of infection. Our study demonstrates the array of landscape, environmental, and individual‐level factors associated with ranavirus epidemiology. Moreover, our study helps illustrate that the importance of these factors varies with biological level. [ABSTRACT FROM AUTHOR]

Although spatial variability in the distribution of soil organisms is often regarded as random, recent attempts have found significant spatial structure at several scales. Understanding the drivers of this variability at large scale could help to evaluate the ecological and socio-economic impacts of soil organisms in the ecosystems. In the present study we aim to (i) understand the relative role of environmental factors and human influence on the distribution of one of the World’s most destructive and invasive plant pathogens, Phytophthora cinnamomi Rands, and (ii) to estimate its potential geographic distribution in the Iberian Peninsula as a proxy of its potential impact. For this purpose, we used a total of 277 records of P. cinnamomi (presence: n  = 157; absences: n  = 120) in the Iberian Peninsula and three sets of explanatory variables reflecting abiotic conditions (climate and soil), biotic conditions (main susceptible host tress distribution and vegetation cover based on NDVI) and human influence. The current distribution of P. cinnamomi in the Iberian Peninsula seems to be influenced principally by fine texture soil and climate, following the land use and lastly the presence of its main host forest species. Its potential distribution across the Iberian Peninsula suggests potential for further expansion along the northeast and southeast of Spain and central Portugal. Given the significant impact of this pathogen on forest ecosystems, the modelling of its distribution in the Iberian Peninsula offers an important decision tool for the monitoring and restoration of declining Mediterranean oak forests. [ABSTRACT FROM AUTHOR]

Abstract: Full‐subsets information theoretic approaches are becoming an increasingly popular tool for exploring predictive power and variable importance where a wide range of candidate predictors are being considered. Here, we describe a simple function in the statistical programming language R that can be used to construct, fit, and compare a complete model set of possible ecological or environmental predictors, given a response variable of interest and a starting generalized additive (mixed) model fit. Main advantages include not requiring a complete model to be fit as the starting point for candidate model set construction (meaning that a greater number of predictors can potentially be explored than might be available through functions such as dredge); model sets that include interactions between factors and continuous nonlinear predictors; and automatic removal of models with correlated predictors (based on a user defined criterion for exclusion). The function takes continuous predictors, which are fitted using smoothers via either gam, gamm (mgcv) or gamm4, as well as factor variables which are included on their own or as two‐level interaction terms within the gam smooth (via use of the “by” argument), or with themselves. The function allows any model to be constructed and used as a null model, and takes a range of arguments that allow control over the model set being constructed, including specifying cyclic and linear continuous predictors, specification of the smoothing algorithm used, and the maximum complexity allowed for smooth terms. The use of the function is demonstrated via case studies that highlight how appropriate model sets can be easily constructed and the broader utility of the approach for exploratory ecology. [ABSTRACT FROM AUTHOR]

This paper outlines a methodology for Bayesian multimodel uncertainty quantification (UQ) and propagation and presents an investigation into the effect of prior probabilities on the resulting uncertainties. The UQ methodology is adapted from the information-theoretic method previously presented by the authors (Zhang and Shields, 2018) to a fully Bayesian construction that enables greater flexibility in quantifying uncertainty in probability model form. Being Bayesian in nature and rooted in UQ from small datasets, prior probabilities in both probability model form and model parameters are shown to have a significant impact on quantified uncertainties and, consequently, on the uncertainties propagated through a physics-based model. These effects are specifically investigated for a simplified plate buckling problem with uncertainties in material properties derived from a small number of experiments using noninformative priors and priors derived from past studies of varying appropriateness. It is illustrated that prior probabilities can have a significant impact on multimodel UQ for small datasets and inappropriate (but seemingly reasonable) priors may even have lingering effects that bias probabilities even for large datasets. When applied to uncertainty propagation, this may result in probability bounds on response quantities that do not include the true probabilities. [ABSTRACT FROM AUTHOR]

The small-island effect (SIE) has become more and more part of the theoretical framework of island biogeography and biodiversity research. However, previous methods for the detection of SIEs are often flawed in one way or another, including not accounting for model complexity, not comparing all relevant models, and not including islands with no species. Therefore, the existence and the prevalence of the SIE may be dubious. In this study, after controlling for all these methodological shortcomings, we tested for the existence of the SIE in amphibian assemblages on subtropical landbridge islands created by the inundation of the Thousand Island Lake, China. We used the line transect method to determine the distribution of amphibian assemblages on 23 study islands during 3 breeding seasons from 2009 to 2011. To evaluate whether an SIE exists in amphibian assemblages, we compared the fit of a simple linearized power model with two most widely used breakpoint regression models. The information-theoretic multimodel inference approach based on Akaike's information criterion identified the left-horizontal SIE model as the best single model. Thus, we found strong evidence for the existence of an SIE in our system. The upper limit of the SIE for amphibian assemblages was 39.95ha. Below this threshold area, amphibian richness varied independently of island size. The SIE in amphibian assemblages may be due to episodic disturbances, stochastic events, and nutrient subsidies from the lake. Our results indicate that all the islands >39.95 ha should be protected for the effective conservation of amphibian assemblages in our system. [ABSTRACT FROM AUTHOR]

The effect of multiple stressors on marine ecosystems is poorly understood. To partially bridge this knowledge gap we investigated the relative contribution of environmental variables to density and diversity dynamics of the zooplankton community in the Belgian part of the North Sea. We applied multimodel inference on generalized additive models to quantify the relative contribution of chlorophyll a , temperature, nutrients, salinity and anthropogenic chemicals (i.e. polychlorinated biphenyls and polycyclic aromatic hydrocarbons) to the dynamics of calanoid copepod species in the Belgian part of the North Sea. Temperature was the only predictor consistently showing a high importance in all models predicting the abundances of the selected copepod species. The relative contribution of other predictors was species-dependent. Anthropogenic chemicals were important predictors for three out of six species indicating that chemical mixtures at low concentrations should not be left unattended when performing risk assessments in a natural environment. [ABSTRACT FROM AUTHOR]

The destruction of forest for agricultural expansion has created a vast estate of human-modified land in tropical regions. One group of organisms that are particularly vulnerable to the loss of forest habitat are insectivorous birds. Despite this, few conservation strategies have been identified for this group in human-modified landscapes. We survey the use of 104 isolated trees by insectivorous birds in rural Assam, India. We used an information theoretic model comparison approach to determine the important variables driving insectivorous bird diversity within these isolated trees. Our work demonstrates that the conservation of large trees in human-modified landscapes may play an important role in maintaining bird diversity and ecological function beyond the forest edge. More specifically, we found that isolated Ficus trees hold assemblages with particularly high insectivore abundance, richness and functional diversity when compared to other isolated fruit and large trees. We argue that, where present, Ficus trees should be actively conserved in human-modified landscapes to maintain the composition of insectivore communities in a ' Ficus first' strategy. [ABSTRACT FROM AUTHOR]

The present paper presents a multimodel inference approach to linguistic variation, expanding on prior work by Kuperman and Bresnan (2012). We argue that corpus data often present the analyst with high model selection uncertainty. This uncertainty is inevitable given that language is highly redundant: every feature is predictable from multiple other features. However, uncertainty involved in model selection is ignored by the standard method of selecting the single best model and inferring the effects of the predictors under the assumption that the best model is true. Multimodel inference avoids committing to a single model. Rather, we make predictions based on the entire set of plausible models, with contributions of models weighted by the models' predictive value. We argue that multimodel inference is superior to model selection for both the I-Language goal of inferring the mental grammars that generated the corpus, and the E-Language goal of predicting characteristics of future speech samples from the community represented by the corpus. Applying multimodel inference to the classic problem of English auxiliary contraction, we show that the choice between multimodel inference and model selection matters in practice: the best model may contain predictors that are not significant when the full set of plausible models is considered, and may omit predictors that are significant considering the full set of models. We also contribute to the study of English auxiliary contraction. We document the effects of priming, contextual predictability, and specific syntactic constructions and provide evidence against effects of phonological context. [ABSTRACT FROM AUTHOR]

The main component of mammalian maternal care is milk production. The composition of milk has been hypothesized to determine offspring survival, and to vary with offspring sex. Few studies, however, have examined variation in milk composition of wild mammals in seasonal environments, where environmental conditions can impact the ability of mothers to provide care. We investigated individual differences in milk composition and offspring survival in wild eastern grey kangaroos ( Macropus giganteus). We analyzed total protein and lipid concentrations in 103 milk samples from 91 females over 2 years. In a year of low forage production, few females that lactated were in poorer condition and produced milk of lower energy content compared to females lactating in a year of high forage production. Females nursing in late winter produced milk with more lipids compared to females at the same stage of lactation in late spring, whose milk had a higher proportion of protein. The milk of larger females contained more protein than lipid, and females in better condition allocated higher proportions of protein to sons than to daughters. Increased protein concentration in milk was correlated with offspring longevity, and this effect was stronger for sons than daughters. A seasonally modulated lactation strategy enables the adjustment of milk composition to available resources. Condition-specific protein allocation in favor of sons for mothers of higher caring ability suggests adaptive sex-biased maternal care in this highly sexually dimorphic mammal. [ABSTRACT FROM AUTHOR]