Seeing the grass through the trees: an investigation of tree-grass productivity and phenology in an Australian tropical savanna

Tree-grass savanna ecosystems are a widespread terrestrial biome and are highly valued for their ecosystem services. They support one fifth of the global human population through food and timber production and are a key biome for biodiversity, the water cycle and carbon sequestration. At the global scale, savannas account for 25 % of terrestrial carbon uptake, which is a product of the interplay between trees and grasses and is maintained through interactions with climate and disturbance (fire, herbivory, land use change). As the climate changes into the 21st century, the future of savannas in their current form is uncertain. Modifications to the timing and amount of rainfall delivered to savanna ecosystems will have implications for species phenology and current structural (i.e. tree-grass) dynamics. Therefore, it is important to separate the savanna tree and grass components to understand what governs these dynamics over time to determine if there is a differential response of trees and grasses to climate change.
   
   This thesis explores tree and grass dynamics in detail for an Australian tropical savanna. The eddy covariance technique was employed to monitor fluxes of carbon, water and energy between the savanna ecosystem and the atmosphere at the Howard Springs OzFlux site. An understory tower was utilised to separate the overstory (i.e. tree) and understory (i.e. grass) contributions into ecosystem fluxes, with particular attention paid to the gross primary productivity (GPP) component. The partitioned fluxes showed that the understory was more seasonally dynamic than the overstory, contributing 40 % to ecosystem GPP in the wet season and only 18 % in the dry season. Understory GPP did not completely cease in the dry season due to contributions from woody species that occupy 20 % of the understory biomass. To capture the different phenological signals displayed by the overstory and understory, a suite of time-lapse cameras (i.e. phenocams) were also installed, where colour indices were calculated from each camera image to provide a time series of in situ variability in vegetation greenness. The greenness data closely tracked GPP over time, and when used in a light use efficiency (LUE) model, tree and grass GPP estimates from the model were improved. This result reinforces the importance of vegetation phenology for determining variability in savanna productivity.
   
   The Howard Springs flux site has been in continuous operation since 2001 and this 15-year flux record was partitioned into overstory and understory contributions. The partitioned dataset was used to explore whether the tree-grass ratio had changed over the 15-year period and what the main meteorological drivers were over time. For the wet season, productivity at Howard Springs was light limited due to increased cloud cover during the summer monsoon. In contrast, productivity in the dry season was water limited due to depletion of soil water stores. Inter-annually, productivity was determined by soil moisture availability linked with annual rainfall and rainy season length, and the tree-grass ratio varied in line with changes in the Southern Oscillation Index (SOI). This research has provided a missing link in our understanding of tree-grass dynamics in Australian savannas, which is vital if savanna ecosystems are to be successfully managed in the coming century.