Towards a theory of carbon allocation based on eco-evolutionary optimality principles

Understanding the distribution of assimilated carbon (C) among different plant parts is essential in explaining the responses of multiple functional traits to climate change. C allocation is not adequately represented by current ecosystem models, and the general explanatory framework of C allocation with environmental conditions is fragmented. Machine learning approaches applied to large data sets have failed to reveal general principles underlying C allocation. Here, we analyse a large global set of data derived from several previous compilations to test eco-evolutionary optimality hypotheses that potentially account for the environmental controls on root:shoot biomass ratios (R:S) in both woody and herbaceous plants. These controls are expressed in terms of statistical predictors describing aspects of the environment relevant to plant stimuli. Thus, for example, we consider growing-season temperatures rather than annual means; and we include modelled gross primary production (GPP) and root-zone water capacity (RZWC) among the candidate predictors. We hypothesize that increasing gross primary productivity (GPP) permits increased C allocation to stems, automatically reducing R:S. Demand for C allocation to roots is less in warmer climates because of faster nutrient turnover in warmer soils. On acid soils, the need for roots to take up nutrients is reduced due to more open stomata and thus lower optimal photosynthetic capacity. More C is allocated to roots in climates with seasonal mismatches between water supply and demand, where increased RZWC is required to maintain water availability during the dry season. On sandy soils, low water-holding capacity implies a need for further investment in roots for water uptake. Our analysis broadly supports these hypotheses, and an ordinary least-squares multiple linear regression model explains nearly three-quarters of the observed global variation in R:S. However, the allocation strategies of woody and herbaceous plants differ. The expected negative relationship of R:S to growth temperature, and the positive relationship of R:S to sand content, are shown only in woody plants; while the expected positive relationship of R:S to soil pH is shown only in herbaceous plants. These findings constitute a first step towards a theory of C allocation response to resource availability, and a parsimonious model for inclusion in next-generation C cycle models.