We compiled and analysed a dataset of 17 functional traits with a sufficient number of records in the TRY database to characterize the main ecoregions of the world , that is, environmentally homogeneous areas with distinct biota . The dataset is based on 225,206 georeferenced observations comprising records of 20,655 species. The trait data were complemented with 21 climate variables and 107 soil variables . Trait–environment relationships were analysed for species medians aggregated to ecoregions using ridge regression , a robust method suitable to deal with high-dimensional, unbalanced and collinear predictors in combination with hierarchical partitioning .The rationale is that species presence indicates how the trait space can be realized in a given environment. Spatial aggregation is a suitable means to increase the detectability of global trait patterns , as described in earlier studies, where traits have been binned by temperature classes or for different altitudinal ranges . Extreme outliers, for instance towering trees such as the Californian Sequoia , may still exist far away from the equator, where precipitation is sufficiently high but their influence is outweighed in our approach by an increasing fraction of small-statured herbaceous species from tropical to temperate and boreal regions.To understand whether the axes of variation identified for the grouping of six traits also hold for the extended set of 17 traits, we cluster their trait–trait correlations and further represent these relations on the basis of their principal components . This analysis supports the clear distinction of size versus economics traits identified by Díaz and colleagues . The group of size traits contains two subclusters. The first includes height and seed size traits: plant height , seed mass, 10 plastic plant pots seed length and dispersal unit length . The second subset contains traits that are linked through plant hydraulic scaling relationships and contrasts high conduit density with high leaf area and leaf fresh mass .
Economics traits represent dry mass and nutrient investments in plant tissues, and the rate and duration of returns on those investments . They are represented by leaf nitrogen content per leaf area , leaf nitrogen , phosphorus and carbon content per dry mass, leaf N to P ratio and SLA. This study shows that the proposed global spectrum of plant form and function fits well to a substantially extended trait space compared to the original study , with seven traits that capture the whole-plant size spectrum and seven traits that capture the leaf economic spectrum and only three traits that do not fall along these dimensions . One explanation could be that the varying fraction of woody and non-woody species would drive these patterns. However, we showed that these two main trait groups remain clearly identifiable when the analysis is conducted separately, yet with fewer samples, for woody and non-woody species . However, we cannot discard the possibility that additional traits may add relevant axes of trait variation. For example, our study does not include carbon fixation rates or fire adaptation traits , nor does it include any root traits—representing an essential gap to be filled at the global scale . If such data were available they would have the potential to fundamentally change our perception of global plant form and function, and their relation to ecosystem functioning. Variation in size traits, represented by PC1 in Fig. 1b, shows a clear latitudinal gradient . In contrast, variation in economics traits does not show a latitudinal trend. Only a dip is apparent at around 35° , in addition to a decrease at high latitudes above 60° where available data become increasingly limited. However, comparison to a recent arctic dataset indicates that this decrease in variation at high latitudes reflects available observations . These patterns might represent a response to nutrient limitation and drought in water-scarce and nutrient-scarce deserts and Mediterranean regions or boreal and arctic areas characterized by short growing periods slowing down mineralization.
The dip at ~35° indeed can be related to low water availability . At high latitudes, cold winters and short growing seasons constrain plant height and require on average more conservative nutrient-use strategies and protection against frost damage than the global mean, despite the high functional diversity in economics traits observed at these latitudes . Additional datasets may shed more light on specific conditions, for example see Bjorkman et al. . Future studies should quantify how individual stressors, for example radiative stress or water stress, relate to global patterns of trait variation. The climate and soil factors used in this analysis explain up to 77% of observed trait variation—a high fraction given that trait variation is widely known to be determined also by other factors such as biotic interactions and anthropogenic effects or disturbances and local effects such as those of microclimate . Recent findings on how different trait groups vary with the environment indicate that size and economics traits vary differently and in particular respond differently to climate and soil . Our analyses reveal a dominant joint effect of climate and soil drivers on trait variation—as already suggested by a number of earlier studies but not yet quantified globally. The orthogonality of the two main dimensions of plant trait variation suggests that different aspects of climate and soil variables are relevant to explain plant trait patterns at the global scale . While latitude-related variables explain size traits, variables that share less explanatory power with latitude explain economics traits . The RDA presented in Fig. 4 provides some insight on the nature of these climate–soil interactions. The first RDA axis, which describes variation in size traits, resembles a latitudinal gradient. On one extreme end, ample water supply from high and frequent precipitation, abundant water vapour and constant rates of high solar radiation meet the fundamental requirements of plant physiology—water, sunlight and warm temperatures.
Additionally, these conditions promote weathering of soil minerals but also microbial activity, contributing to fast turnover rates of organic matter supporting nutrient provisioning ; in brief, they represent conditions that allow plants to grow fast and tall in the race for light. Large vessels supporting large leaves promote high rates of water transport and thus growth, which is only possible because of the small risk of embolism under these benign water conditions . The high carbon gains can be invested in large fruits and seeds . Further along this gradient, the above-mentioned plant requirements become limited: water supply and temperatures are reduced and slow metabolic rates above ground and below ground. In ecoregions of the boreal and desert biomes, conduit diameter is constrained by the risk of cavitation during freeze–thaw cycles and water scarcity, amplified by little water holding capacity of gravel-rich soils. Our analysis thus indicates that size traits appear to be related to a latitudinal gradient of climatic favorability for plant growth determined by water and light availability. Important correlates of water and nutrient availability are associated with the second RDA axis, describing variation in economics traits. Traits associated with an acquisitive strategy are related to indicators of soil fertility, most importantly silt and organic matter concentration as well as pH . Soil pH is intermediate between the two axes, as might be expected given that pH reflects both broad-scale climate variation and a variety of processes related to nutrient availability and soil microbial communities . Silt forms the substrate of our most fertile soils as its structure is able to retain water against gravitation but renders it accessible to plants under drought conditions . The high fertility is associated with a high concentration of organic matter, which has a high cation exchange capacity especially under high pH . On the opposite end of the gradient, sandy soils require adaptations to both water and nutrient limitation. The trait configuration at the conservative end of the economics traits represents an adaptation to both . Various processes exist that lead to variation in the soil characteristics underlying the second RDA axis independent of latitude —for example, sandstone as a geological substrate giving rise to sandy soils exists from the tropics to the arctic . However, different climate variables related to solar radiation, temperature and precipitation, plastic pots large which influence long and short-term soil development processes directly and indirectly via soil biology , are related to this axis. Variation in economic traits is most probably the evolutionary response to exploiting this partly climate-independent edaphic niche axis. Size traits are on average explained better than economics traits by the environmental variables considered in this study. The lower fraction of explained variance for economics traits could have several causes. Firstly, data on soil factors that are likely to be very important, such as soil nitrogen and phosphorus availability , are not yet available at a global scale. Secondly, economics traits show relatively more within-site variation than across-site variation in comparison to size traits , probably because economics traits vary more than size traits within one plant; for example, leaf N per area and SLA vary with age and light availability . Thirdly, soil heterogeneity within ecoregions—both abiotic and biotic—may weaken the relationship between economics traits and environmental variables .
Reasons for small-scale soil variation are, for example, topography, soil age and thus fertility but also abundance of microbial communities and mycorrhiza that interact with climate, pH, soil properties and also plant traits . Trait–environment relationships due to smaller scale variation require well-resolved soil data. However, we note that soil physics and chemistry explain a large portion of variance along the trait PC axis three ; Supplementary Figs. 5, 6 and 38). We expect that with improved soil datasets and a higher resolution, the joint control of climate and soil on trait variation will probably appear even stronger and more evenly distributed between the two groups of driver variables. Our analysis can serve as reference for model developments that increasingly consider plant functional traits as part of vegetation dynamics under climate change . Individual plants and their trait syndromes are considered to be viable only within specific environmental conditions2 . Therefore trait–environment relationships should be scale-independent. However, different plant strategies can be successful under given environmental conditions, which in addition are often confounded by small-scale variation. In analyses to date, trait–environment relationships become more apparent for aggregations higher than the community scale , where most of the small-scale variation is averaged out. In addition the difference between potential and actual vegetation is suggested to explain some of this gap . Dynamic global vegetation models predict individual plant processes well but fail to produce reliable forecasts with a changing environment . Deciphering at which spatial and temporal scale, or conditions, actual vegetation is representative of potential vegetation may advance our understanding of community assembly and necessary model complexity. Yet, the ubiquitous importance of climate variables for explaining current differences in trait expression at ecoregion scale, suggests that trait shifts will occur with climate change. Trait shifts are constrained by available trait combinations in addition to other constraints such as species dispersal. For example, our results indicate that plant size increases with temperature so long as sufficient water is available , in line with the finding that species become larger and large species are more prevalent at warmer and wetter sites in the tundra . Global change is also reflected by soil degradation. Changes in soil parameters can be considered to also correspond with trait shifts, especially for economics traits. Human-induced soil degradation has many facets: often fertile topsoil is lost or toxic substances accumulate; rooting is impeded and alThered by artificial fertilizers; while soil formation takes millenia . The trait shifts may thus be similarly complex and depend on the extent and type of soil degradation. For example, in areas of wind and water erosion, species that tolerate lower nutrient availability may be more successful and this may be reflected in lower leaf nutrient contents . The fertilization of nutrient-poor grasslands, for example resulting from agricultural run-off, may shift these areas from more conservative to more competitive species with higher leaf nutrient contents. Plants as a whole need to balance both size and economics traits. To sustain human livelihoods, it may be important to understand the local expression of trait shifts and their global consequences for biodiversity when viable trait combinations change. In conclusion, the insights extracted here advance our understanding of broad-scale plant functional patterns.