Aspects of Land Surface Modelling: Role of Biodiversity in Ecosystem Resilience to Environmental Change and a Robust Ecosystem Demography Model

Abstract

Earth's species are disappearing at a rate unprecedented in human history, yet whether this loss will make the ecosystem "services" that support our civilisation more vulnerable to environmental change is poorly understood. This thesis investigates two different aspects of land surface modelling. It firstly models the role of biodiversity in ecosystem resilience using the Lotka-Volterra and single resource models to model diversity using competition coeffcients, stochastic noise and evolution inspired trait diffusion and then examines if higher diversity makes these simple models more resistant to temperature increases. It secondly develops a theoretical plant demography model, based on the continuity equation, to robustly represent forest size diversity. This avoids both the complexity and maintainability issues seen in Forest Gap models and improves the representation of land use and land cover change and of regrowth time-scales after disturbance, which can be unrealistic in some of the previous generation of Dynamic Global Vegetation Models (DGVMs), such as TRIFFID (Cox et al., 2001). While the Lotka-Volterra with competition coeffcients and the single resource with stochastic noise approaches are found to be impractical, the single resource model with trait diffusion successfully shows that higher diversity requires a faster critical rate of temperature change before system net primary productivity (NPP) collapses. The continuity equation model of vegetation demography is solved analytically with the size dependence of the growth rate approximated first by a power law and then with a quadratic. The power law solution can be reduced to a "self-thinning" trajectory, and the quadratic solution gives either a rotated sigmoid or 'U-shape' distribution of plant sizes, depending on the ratio of mortality to maximum growth gradient. The model is then extended to produce the basis of a new Dynamic Global Vegetation Model (DGVM) called "Robust Ecosystem Demography" (RED), adapting the plant physiology from TRIFFID DGVM to generate a size-dependent growth function. A proportion of the NPP from this growth is used for reproduction and the shading is modelled simply by random overlap. The model is found to better represent regrowth time-scales compared to TRIFFID and is also found to demonstrate an optimum proportion of NPP to reproduction which decreases with plant lifetime.