Towards an integrated ecology through mechanistic modelling of ecosystem structure and functioning

Abstract: Plant physiology, population dynamics and the structure and functioning of ecosystems are closely interrelated. However, these processes are often investigated independently within different subdisciplines of ecology. Mechanistic models of ecosystem structure and functioning combine representations of processes operating at a range of scales and levels of organisation, providing a tool for the description and study of ecosystems as integrated entities. This thesis is based on the development and application of LPJ-GUESS, a modelling framework which integrates processes such as leaf-level photosynthesis, the dynamics of populations competing for resources, and the fluxes of carbon and water between soil layers, vegetation and the atmosphere. LPJ-GUESS currently includes as alternative configurations the Lund-Potsdam-Jena dynamic global vegetation model (LPJ-DGVM) and the General Ecosystem Simulator (GUESS). The results from a study of potential future ecosystem dynamics on the continental scale projected by two DGVMs (LPJ and MC1) suggest that changes in climate and atmospheric CO2 concentrations may severely impact North American ecosystems within the next century (e.g. through vegetation dieback caused by drought), but simulation results depended heavily on the climate scenario and alternative assumptions as to the effects of increasing atmospheric CO2 on carbon assimilation and ecosystem water balance. LPJ-DGVM simulated similar effects of elevated CO2 on forest productivity as observed during four years of CO2 fumigation at a free air CO2 enrichment (FACE) experiment, but it remains uncertain if a CO2 effect of the magnitude commonly simulated by ecosystem models is realistic in the longer term. A comparison of the structure and composition of pristine forests simulated by a number of forest gap models revealed that most of the models have restricted generality because the empirical relationships used to calculate growth are usually not applicable beyond the climatic region for which each particular model was developed. More generally applicable gap models should incorporate process representations derived from generalised plant-physiological mechanisms. The physiology-based gap-type model GUESS was demonstrated to reproduce observed patterns of vegetation dynamics at the plant-type as well as the species level. Finally, a comprehensive hypothesis on the effects of plant hydraulic architecture on water uptake in different types of plants was implemented within the LPJ-DGVM. The hypothesis was validated through a comparison of simulated ecosystem structural and functional features with available data. Comparing the predictions from a number of models that have been applied under standardised conditions has been useful for identifying major drivers and uncertainties in mechanistic ecosystem models, but the future development of the field will crucially depend on evaluation of individual process formulations by field and experimental scientists in close collaboration with modellers. Such approaches could benefit ecology through integration of knowledge from different subdisciplines and by tightening the coupling between empirical studies and models.