Trading water for carbon in a changing climate: can optimality theory improve the predictability of land surface models?

Terrestrial ecosystems are facing unprecedented threats from climate change. Projected increases in the intensity, frequency, and duration droughts and heatwaves could drive future mass forest die-back. However, assessing future risk remains a challenge because climate models demonstrate systematic errors in predicting plant function when resources become limiting (e.g., water). To improve predictions of ecosystem resilience, major model weaknesses in the formulation of plant responses to climatic stress must be resolved. The aim of this thesis is to examine whether optimality principles can improve land surface model (LSM) predictability, through improved process-representation of vegetation function, and by reducing existing reliance on poorly supported empirical functions. Major contributions from the thesis include: 1. The development and testing of a novel canopy gas-exchange scheme, that optimises stomatal conductance with respect to photosynthetic and hydraulic functions, for use in LSMs. The new scheme improves the simulation of forest evaporative fluxes over a mesic to xeric ecosystem gradient in Europe, reducing errors by over 60% compared to a reference empirical scheme. 2. A multi-model comparison study of potential alternative stomatal coupling schemes, including empirical (n=3) and optimal (n=9) formulations, carried out in a single LSM framework. The findings provide important insight into: (i) model identifiability and parameter redundancy; (ii) how and why models diverge from each other and observations; and (iii) provide guidance for future model development. 3. The combination and evaluation of competing optimisation approaches, with contrasting fitness targets, in a LSM framework – competing approaches have been proposed and evaluated in isolation, but seldom implemented together in LSMs. Here, the canopy gas-exchange scheme tested in 1) is coupled to a scheme that optimises leaf nitrogen investment, and both approaches are extended to account for system legacies from drought, before being evaluated for their ability to coherently interact with one another. This thesis demonstrates the wide-ranging scope for optimality theory to be incorporated into LSMs. Plant optimality theory is grounded in physiological principles, and it adjusts to changing environmental constraints, providing a pathway to improve both the theoretical underpinning and the predictability of LSMs