Fourth-corner latent variable models overstate confidence in trait–environment relationships and what to use instead

Traits differentially adapt plant species to particular conditions generating compositional shifts along environmental gradients. As a result, community‐scale trait values show concomitant shifts, termed trait‒environment relationships. Trait‒environment relationships are often assessed by evaluating community‐weighted mean (CWM) traits observed along environmental gradients. Regression‐based approaches (CWMr) assume that local communities exhibit traits centred at a single optimum value and that traits do not covary meaningfully. Evidence suggests that the shape of trait‒abundance relationships can vary widely along environmental gradients—reflecting complex interactions—and traits are usually interrelated. We used a model that accounts for these factors to explore trait‒environment relationships in herbaceous forest plant communities in Wisconsin (USA). We built a generalized linear mixed model (GLMM) to analyse how abundances of 185 species distributed among 189 forested sites vary in response to four functional traits (vegetative height—VH, leaf size—LS, leaf mass per area—LMA and leaf carbon content), six environmental variables describing overstorey, soil and climate conditions, and their interactions. The GLMM allowed us to assess the nature and relative strength of the resulting 24 trait‒environment relationships. We also compared results between GLMM and CWMr to explore how conclusions differ between approaches. The GLMM identified five significant trait‒environment relationships that together explain ~40% of variation in species abundances across sites. Temperature appeared as a key environmental driver, with warmer and more seasonal sites favouring taller plants. Soil texture and temperature seasonality affected LS and LMA; seasonality effects on LS and LMA were nonlinear, declining at more seasonal sites. Although often assumed for CWMr, only some traits under certain conditions had centred optimum trait‒abundance relationships. CWMr more liberally identified (13) trait‒environment relationships as significant but failed to detect the temperature seasonality‒LMA relationship identified by the GLMM. Synthesis . Although GLMM represents a more methodologically complex approach than CWMr, it identified a reduced set of trait‒environment relationships still capable of accounting for the responses of forest understorey herbs to environmental gradients. It also identified separate effects of mean and seasonal temperature on LMA that appear important in these forests, generating useful insights and supporting broader application of GLMM approach to understand trait‒environment relationships.

Compared to taxonomic approaches, traits‐based approaches are often implemented to gain a more mechanistic understanding of how biotic communities respond to changing environmental conditions, including landscape stressors. Studies evaluating trait–environment relationships have increased in prevalence over the last 2 decades, but such analyses have rarely been conducted across continental extents in which component regions exhibit substantial spatial variation in stressors and natural environmental conditions. Consequently, the role of biogeographic context in analyses of trait–environment relationships remains under studied. Here, we evaluated the generalisability of trait–environment relationships across the conterminous U.S.A. to advance our understanding of the functional biogeography of fluvial fishes. We sought to: (1) identify trait–environment relationships that were consistent regardless of regional differences in natural environmental conditions; and (2) address whether the multivariate structure of trait–environment relationships was comparable across regions. We used RLQ and fourth‐corner analyses to identify significant trait–environment relationships across nine large ecoregions comprising the conterminous U.S.A. These statistical approaches provided a framework to analyse co‐variation among 16 functional traits (e.g., trophic ecology, habitat preference) and 17 environmental variables representing 597 stream fish species whose abundances had been recorded in 45,183 stream reaches between 1990 and 2019. Trait–environment relationships varied among ecoregions such that few trait–environment relationships were consistent. Although some trait–environment relationships were significant in multiple ecoregions, the strength and multivariate structure of such relationships varied. These findings suggest that the functional traits most responsive to changing environmental conditions are determined by distinct biogeographic processes. Our findings add to a growing consensus that functional traits‐based approaches are context dependent. Understanding the complex relationships between functional traits and environmental conditions across large geographic regions can inform ecological theory and improve conservation efforts at broad spatial extents.

One of J.P. Grime's greatest achievements was demonstrating the importance of the relationship between the environment and plant functional traits for understanding community assembly processes and the effects of biodiversity on ecosystem functioning. A popular approach assessing trait–environment relationships is the community weighted means (CWMs) method, which evaluates changes in communities' average trait values along gradients, with Grime being among its first practitioners. Today the CWM method is well‐established but some scholars have criticized it for inflated Type I errors. That is, in some scenarios of compositional turnover along a gradient, CWM tests can provide significant results even for randomly generated traits. Null models have been proposed to correct for such effects by randomizing trait values across species (CWM‐sp). We review different approaches relating traits to the environment within the framework of the accepted dichotomy between species‐level (observations are species) versus community‐level (observations are community parameters) analyses. Between these families of analyses and their combinations, a great variety of methods exist that test different trait–environment relationships, each with different null hypotheses and ecological questions. In classic CWM tests, the null hypothesis focuses on characteristics of trait distributions at the community level along gradients. The Type I error rate should not be a priori considered inflated when this test is used to identify changes in community trait structure affecting the functioning of communities. Trait changes observed with CWM tests may be accurate, but the interpretation that a specific trait drives turnover may be fallacious. Approaches like CWM‐sp may be more appropriate for testing other ecological hypotheses, such as whether trait–environment relationships are widespread across species. In effect, this moves the ecological focus towards species‐level analyses, that is on the adaptive value of traits and their relation to species niches. Synthesis. There is no single trait–environment relationship. Species‐level and community‐level analyses, including variants within them, test different relationships with different null hypotheses, such that the potential for inflated error rates can be misleading. Using a spectrum of methods provides a comprehensive picture of the diversity of trait–environment relationships.

Understanding the trait–environment relationships has been a core ecological research topic in the face of global climate change. However, the strength of trait–environment relationships at the local and regional scales in temperate forests remains poorly known. In this study, we investigated the local and regional scale forest plots of the natural broad-leaved temperate forest in northeastern China, to assess what extent community-level trait composition depends on environmental drivers across spatial scales. We measured five key functional traits (leaf area, specific leaf area, leaf carbon content, leaf nitrogen content, and wood density) of woody plant, and quantified functional compositions of communities by calculating the “specific” community-weighted mean (CWM) traits. The sum of squares decomposition method was used to quantify the relative contribution of intraspecific trait variation to total trait variation among communities. Multiple linear regression model was then used to explore the community-level trait–environment relationships. We found that (i) intraspecific trait variation contributed considerably to total trait variation and decreased with the spatial scale from local to regional; (ii) functional composition was mainly affected by soil and topography factors at the local scale and climate factor at the regional scale, while explaining that variance of environment factors were decreased with increasing spatial scale; and (iii) the main environment driver of functional composition was varied depending on the traits and spatial scale. This work is one of the few multi-scale analyses to investigate the environmental drivers of community functional compositions. The extent of intraspecific trait variation and the strength of trait–environment relationship showed consistent trends with increasing spatial scale. Our findings demonstrate the influence of environmental filtering on both local- and regional-scale temperate forest communities, and contribute to a comprehensive understanding of trait–environment relationships across spatial scales.

AimThe characterization of trait–environment relationships over broad‐scale gradients is a critical goal for ecology and biogeography. This implies the merging of plot and trait databases to assess community‐level trait‐based statistics. Potential shortcomings and limitations of this approach are that: (i) species traits are not measured where the community is sampled and (ii) the availability of trait data varies considerably across species and plots. Here we address the effect of trait data representativeness [the sampling effort per species and per plot] on the accuracy of (i) species‐level and (ii) community‐level trait estimates and (iii) the consequences for the shape and strength of trait–environment relationships across communities.InnovationWe combined information existing in databases of vegetation plots and plant traits to estimate community‐weighted means [CWMs] of four key traits [specific leaf area, plant height, seed mass and leaf nitrogen content per dry mass] in permanent grasslands at a country‐wide scale. We propose a generic approach for systematic sensitivity analyses based on random subsampling and data reduction to address the representativeness of incomplete and heterogeneous trait information when exploring trait–environment relationships across communities.Main conclusionsThe accuracy of the CWMs was little affected by the number of individual trait values per species [NIV] but strongly affected by the cover proportion of species with available trait values [PCover]. A PCover above 80% was required for all four traits studied to obtain an estimation bias below 5%. Our approach therefore provides more conservative criteria than previously proposed. Restrictive criteria on both NIV and PCover primarily excluded communities in harsh environments, and such reduction of the sampled gradient weakened trait–environment relationships. These findings advocate systematic measurement campaigns in natural environments to increase species coverage in global trait databases, with special emphasis on species occurring in under‐sampled and harsh environmental conditions.

Ecological flows across ecosystem boundaries are typically studied at spatial scales that limit our understanding of broad geographical patterns in ecosystem linkages. Aquatic insects that metamorphose into terrestrial adults are important resource subsidies for terrestrial ecosystems. Traits related to their development and dispersal should determine their availability to terrestrial consumers. Here, we synthesize geospatial, aquatic biomonitoring and biological traits data to quantify the relative importance of several environmental gradients on the potential spatial and temporal characteristics of aquatic insect subsidies across the contiguous United States. We found the trait composition of benthic macroinvertebrate communities varies among hydrologic regions and could affect how aquatic insects transport subsidies as adults. Further, several trait–environment relationships were underpinned by hydrology. Large bodied taxa that could disperse further from the stream were associated with hydrologically stable conditions. Alternatively, hydrologically variable conditions were associated with multivoltine taxa that could extend the duration of subsidies with periodic emergence events throughout the year. We also found that anthropogenic impacts decrease the frequency of individuals with adult flight but potentially extend the distance subsidies travel into the terrestrial ecosystem. Collectively, these results suggest that natural and anthropogenic gradients could affect aquatic insect subsidies by changing the trait composition of benthic macroinvertebrate communities. The conceptual framework and trait–environment relationships we present shows promise for understanding broad geographical patterns in linkages between ecosystems.

QuestionsResearch efforts have sought to understand trait–trait relationships among species and trait–environment relationships. However, connections between these two approaches are rare, despite the fact that species‐level trait–trait correlations constrain the possible trait–environment correlations. We ask how functional traits of grasses are related to each other and to environmental variation.LocationGlobal, with particular focus on the continental United States.MethodsWe compiled distribution data for grasses with three spatial grains – TDWG Level 3 ‘botanical countries’, US counties and vegetation plots within the US. We combined these data with trait data compiled from published sources for 14 traits describing physical and chemical features of the leaves, seeds, roots and entire plant. Trait–trait relationships were explored using correlations and PCA, and trait–environment relationships using regression. Finally, we implemented a null model to predict trait–trait correlations at the assemblage level from those at the species level.ResultsThe functional trait composition of grass species varied strongly along environmental gradients. At the species level, there were two main clusters of related traits – one describing general plant size (including height, seed mass, leaf size and rooting depth), and one describing the leaf economics spectrum (including specific leaf area, Nmass and Pmass). Most trait–trait correlations at the assemblage level did not differ significantly from that predicted from the species level, suggesting that the former are strongly constrained by the latter. Trait–trait and trait–environment relationships in grasses were broadly similar to those observed for other groups, with some exceptions related to the particular growth form, physiology and ecology of grass species.ConclusionsThe unique evolutionary history and ecological role of grasses has led to some unusual trait–climate relationships in the group. Co‐variation among traits at the species level is an important template upon which environmental filters act to determine assemblage trait composition.

AimThe aim was to test trait–environment relationships in hydroids across large spatial and environmental gradients and to evaluate associations between traits, environmental variables, space and phylogeny.LocationAtlantic Ocean and adjacent polar seas.Time periodPresent day.Major taxa studiedHydrozoa.MethodsTrait–environment relationships and their spatial and phylogenetic contexts were assessed in hydroids by using a combination of a fourth‐corner test and extended‐RLQ approach, in a multivariate ordination method that uses five matrices: (1) species across sites, (2) environmental data across sites, (3) traits across species, (4) latitude and longitude for each site, and (5) phylogenetic distance among species. We based our analyses on 3680 records of 431 species distributed at 1440 sites, 16 species traits, nine environmental variables and a phylogenetic tree of the studied species.ResultsHydroid traits are significantly correlated with environmental variables and geographical space, and hydroid phylogeny is correlated with the environment and geographical space. More evidently, we observed an increased presence of larger species (taller, more branched, with greater base diameter and polysiphonic) at sites richer in nitrate and silicate, with higher dissolved oxygen, lower temperatures, lower salinities and lower current velocities. The observation of phylogenetically related species with similar traits and distributions corroborates that historical processes interact with the environment in determining community assembly and explaining patterns of distribution of species traits.Main conclusionsSeveral environmental variables combined affect the distribution of hydroid traits, which are also influenced by spatial and historical factors. In the first analysis of its kind for hydrozoans, we have provided an overview of how hydroids are distributed across environments according to their traits and revealed the spatial and phylogenetic components of these trait–environment relationships.

Summary The extent of intraspecific variation in trait–environment relationships is an open question with limited empirical support in crops. In organic agriculture, with high environmental heterogeneity, this knowledge could guide breeding programs to optimize crop attributes. We propose a three‐dimensional framework involving crop performance, crop traits, and environmental axes to uncover the multidimensionality of trait–environment relationships within a crop.We modeled instantaneous photosynthesis (A sat) and water‐use efficiency (WUE) as functions of four phenotypic traits, three soil variables, five carrot (Daucus carota) varieties, and their interactions in a national participatory plant breeding program involving a suite of farms across Canada. We used these interactions to describe the resulting 12 trait–environment relationships across varieties.We found one significant trait–environment relationship for A sat (taproot tissue density–soil phosphorus), which was consistent across varieties. For WUE, we found that three relationships (petiole diameter–soil nitrogen, petiole diameter–soil phosphorus, and leaf area–soil phosphorus) varied significantly across varieties. As a result, WUE was maximized by different combinations of trait values and soil conditions depending on the variety.Our three‐dimensional framework supports the identification of functional traits behind the differential responses of crop varieties to environmental variation and thus guides breeding programs to optimize crop attributes from an eco‐evolutionary perspective.

Future changes in key plant traits across Central Europe vary with biogeographical status, woodiness, and habitat type

Assessing trait–environment relationships is crucial for predicting effects of natural and human‐induced environmental change on biota. We compiled a global database of fish assemblages in estuaries, functional traits of fishes and ecosystem features of estuaries. And we quantified the relative importance of ecosystem features as drivers of patterns of fish functional traits among estuaries worldwide (i.e. drivers of the proportions of fish traits). In addition to biogeographical context, two main environmental gradients regulate traits patterns: firstly temperature, and secondly estuary size and hydrological connectivity of the estuary with the marine ecosystem. Overall, estuaries in colder regions, with larger areas and with higher hydrological connectivity with the marine ecosystem, have higher proportions of marine fish (versus freshwater), macrocarnivores and planktivores (versus omnivores, herbivores and detritivores) and larger fish, with greater maximum depth of distribution and longer lifespan. The observed trait patterns and trait–environment relationships are likely generated by multiple causal processes linked to physiological constraints due to temperature and salinity, size‐dependent biotic interactions, as well as habitat availability and connectivity. Biogeographical context and environmental conditions drive species richness and composition, and present results show that they also drive assemblage traits. The observed trait patterns and trait–environment relationships suggest that assemblage composition is determined by the functional role of species within ecosystems. Conservation strategies should be coordinated globally and ensure protection of an array of estuaries that differ in ecosystem features, even if some of those estuaries do not support high species richness.

It is assumed that widespread, generalist species have high phenotypic variation, but we know little about how intraspecific trait variation (ITV) relates to species abundance and niche breadth. In the temperate rainforest of southern Chile, we hypothesized that species with wide niche breadth would exhibit 1) high among‐plot ITV, 2) a strong relationship between trait values and the environment, and 3) a close fit between traits and local environment trait optima. We measured leaf functional traits (leaf area, LMA, leaf N and P concentrations) of saplings in woody species, and compared the relative abundance of each species with its niche breadth, measured as the range of light, soil N and P availability. We used the slope of the linear regression of species’ trait–environment relationships to assess the strength and direction of these relationships, and measured the degree to which species’ trait values track the environmental optimum across plots. In some cases, species having wide niche breadth had high ITV in leaf N and also matched traits (LMA and leaf P) to local optima along the light gradient; they also had high ITV in general and matched leaf P to local optima along the soil P gradient. The relationship between species with wide niche breadth and the strength of intraspecific trait–environment relationships was generally weak and varied depending on the niche dimension and trait in question. Species varied considerably in the strength of trait–environment relationships and total magnitude of ITV, and this variation was not generally strongly related to species abundances or niche breadth patterns. In conclusion, trait variation at the community level is not driven by a few abundant, widely distributed species, but depends on the aggregate trait responses of both abundant and rare species. This makes it difficult to scale individual species trait responses up to the community level.

Summary Two primary lines of indirect evidence for contemporary evolution in alien species are based on differences between native and introduced ranges in one or more functional traits or a shift in environmental niche. Although the integration of trait and environmental niche perspectives is increasingly recognised as a key to understanding the role of life‐history evolution in range‐shifting populations, there has been no attempt to bring together these perspectives on the contemporary evolution of alien plant species. We develop a set of scenarios that contrast trait–environment relationships observed in the field for alien species in the native and introduced range, and explore how they might be shaped by contemporary evolution. In each case, the limitations of uniquely trait or environmental niche perspectives are highlighted. The scenarios are examined in relation to long‐term trends in covariation between temperature and first flowering date of European plant species introduced into the USA. Support for four of the scenarios is found. Field studies examining how species traits respond to environmental variation along natural gradients cannot by themselves distinguish relationships arising from genetic variation, phenotypic plasticity or genetic variation for such plasticity. This is best assessed by reciprocal transplant experiments. However, trait–environment relationships provide a basis for better targeted common garden studies that are more hypothesis driven and that pinpoint the traits of interest, ascertain the appropriate selection gradients and the range over which they need to be observed, as well as identify candidate species for further study. Synthesis. The need to improve species distribution models through a better understanding of underlying ecological and evolutionary processes makes assessments of trait–environment relationships, in both the native and introduced ranges, significant. They are of paramount importance when explaining the differential success of alien plants in novel environments as well as when predicting the potential for future range shifts following introduction. The current paucity of such comparisons represents a significant gap in our understanding of biological invasions.

ABSTRACTAimTo investigate how trait correlations between life stages associated with complex life cycles (aquatic nymph and terrestrial adult) shape the functional diversity and trait–environment relationships of European dragonflies (Odonata: Anisoptera).LocationEuropean mainland.Time PeriodPre‐1990 and post‐1990.Major Taxa StudiedDragonflies (Odonata: Anisoptera).MethodsBased on functional traits linked to dispersal and microhabitat preference, we use trait hypervolumes and structural equation modelling to estimate spatial and temporal trait correlations between terrestrial (adult) and aquatic (nymphal) life stages, and potential complex trait–environment relationships across life stages.ResultsAdult and nymphal functional diversity were positively correlated and trait variation between life stages did show reciprocal causality. Cross‐lagged correlations showed that historical nymphal traits most strongly impacted present nymphal and adult diversity, suggesting that functional diversity patterns are influenced by carryover effects and differential selection pressures on nymphs relative to adults. Between the two life stages, we find both parallel and contrasting patterns between direct and indirect trait–environment relationships. The effect of mean annual temperature on adult trait diversity is largely driven by its positive correlation with nymphal traits. Positive nymphal trait correlations with habitat availability and topography are reducing the direct negative effects these variables have on adult trait diversity.Main ConclusionsWe show that constraints inherent to complex life cycles significantly influence functional diversity patterns in European dragonflies, creating indirect trait–environment relationships across life stages. Spatial patterns in functional diversity were determined by both life stages, not just adults or nymphs, via a combination of independent and interactive trait–environment relationships.These findings challenge conventional functional biogeography models focused solely on direct environmental filtering. Consequently, integrating reciprocal trait relationships enhances causal claims when predicting functional biodiversity responses to environmental changes.

QuestionHow do multivariate methods perform in relating species‐ and community‐level trait responses to the environment?Location(1) Field data from grazed semi‐natural grasslands, NE Germany; (2) artificial data.MethodsResearch questions associated with trait–environment relationships were briefly reviewed and seven available methods evaluated. The main distinction between research questions is whether trait–environment relationships should be addressed at community or species level. A redundancy analysis (RDA) of mean trait values of species in a plot weighted by their abundances (CWM‐RDA) is exclusively suitable for the community level. The other six methods address the species level. A double inertia analysis of two arrays (RLQ) and double canonical correspondence analysis (doubleCCA) use combinations of ordinations to simultaneously analyse species and trait responses to the environment. A combination of the outlying mean index with generalized additive models (OMI‐GAM) predicts the response of species to environmental variables on trait gradients.RDA‐RegTree first analyses species responses to the environment withRDAand then uses a regression tree to classify trait expressions according to scores of species responses on the ordination axes. Cluster regression uses cluster analyses and logistic regression to search for trait combinations with the best response to the environmental variables. This method models the distribution of functional groups on environmental gradients. All methods and data are available as R scripts.ResultsAll methods consistently revealed the main trait responses to environment in the field data set, namely that life history was associated with available phosphorus while grazing intensity was related to leafC:Nratio and canopy height. At community level,CWM‐RDAgave a good overview of trait–environment relationships, as also provided by the species‐based methodsRLQand doubleCCA.OMI‐GAMrevealed non‐linear relationships in the field data set. Field and artificial data gave that the number and stability of functional groups produced byCluster regression andRDA‐RegTree varied more strongly thanRLQ, doubleCCAandOMI‐GAM.ConclusionsEach method addresses particular ecological concepts and research questions. If a user asks for the response of average trait expressions of communities to environmental gradients,CWM‐RDAmay be the first choice. However, species‐based methods should be applied to address questions regarding co‐existence of different life histories or to assess how groups of species respond to environmental changes. The artificial data set revealed that the methods differed in sensitivity to gradient lengths and random data.