Vegetation–climate interactions in arid and semi–arid regions

Abstract

Vegetation–climate interactions play an important role in earth system dynamics. It includes the effects of climate on vegetation and feedbacks from vegetation to climate. Such complex and nonlinear processes can enhance climate variation and lead to alternative stable states under given climate regimes. In this thesis, both model and statistical approaches are applied to demonstrate specific mechanisms of vegetation–climate interactions and related consequences. Research area includes arid and semi–arid regions in West Africa, where vegetation and climate are found to be strongly coupled. The thesis starts from a Balanced Optimality Structure Vegetation Model (BOSVM) considering water, energy and carbon balances. The BOSVM is used to demonstrate: 1) How vegetation adjusts to local climate by optimizing its spatial structure to maximize equilibrium biomass? 2) How the optimal structure shifts with climate regimes? Results show that vegetation with a low shoot–total biomass ratio and a vertical canopy can reach the maximal biomass under water–limited conditions. However the optimal structure shifts to high shoot–total biomass ratio and horizontal canopy with an increase of mean annual precipitation. A positive and a negative feedback are found in the water competition between vegetation and bare soil, which makes vegetation grow into patches to maximize water use efficiency, or to extend vegetation cover to stop water loss from bare soil. One important consequence of vegetation–climate interactions is the formation of alternative stable states under a given climate regime. It implies that vegetation can abruptly shift from one stable state to another with a change of climate. The point where the critical transition occurs, is called the tipping point. A complex network approach is applied to monitor the stability of vegetation state and provide early warning signals of upcoming tipping point in a coupled land–atmosphere model. Comparing with two classical indicators, network indicators show higher sensitivity to potential critical transitions and yielded early warning signals can be easier distinguished from local variability. One evidence of alternative stable states is the observed bimodal distribution of woody cover under the same rainfall band in tropical regions. Simulated biomass dynamics affected by fire is compared with observations to understand the observed bimodality. Results suggest that growth rate of woody cover varies with the age of woody plants, which also can lead to the observed bimodality of woody cover. The shift of vegetation structure is the necessary component for the formation of bimodality. Finally, the spatial distribution of land cover types in West Africa is illustrated. Six climatic indicators are analyzed to demonstrate their ability to distinguish land cover types. The mean annual precipitation has large uncertainty to predict specific land cover type. Simultaneously, prediction accuracies of other climatic indicators vary significantly with the change of land cover types. Several indicators are chosen and combined to improve land cover prediction. Patterns of potential land cover change in West Africa are illustrated, where forest is under stress while savanna and grassland show a tendency to extend to the north.