The Future of Terrestrial Carbon in Australia

Download files
Access & Terms of Use
open access
Copyright: Teckentrup, Lina
Terrestrial ecosystems sequester about one third of anthropogenic greenhouse gas emissions every year, and strongly influence the interannual variability in the growth rate of atmospheric CO2. Ecosystems in semi-arid regions of the Southern Hemisphere have a disproportionately large impact on the year-to-year variability and trend in the net global carbon sink. In these regions, the carbon balance is linked to circulation-driven variations in both precipitation and temperature that in turn are influenced by climate modes of variability, such as the El Nino-Southern Oscillation. Typically, future carbon cycle predictions depend on terrestrial biosphere models (TBMs), and on climate predictions based on simulations by Global Circulation Models (GCMs). However, GCM simulations are associated with large biases in both the representation of climate modes of variability, and in the averages of climate variables, such as temperature and precipitation. Studies have also shown significant uncertainties in the representation of the terrestrial carbon cycle across different TBMs. This thesis explores the degree to which uncertainty in i) climate modes of variability, ii) climate simulations based on GCMs, and iii) terrestrial biosphere models represent a source of uncertainty in simulations of the terrestrial carbon cycle. The overarching goal is to achieve a constrained estimate of the future carbon cycle over Australia. This thesis first investigates whether the expression (or flavour) of El Nino (as distinct from the El Nino-La Nina cycle) affects the interannual carbon cycle variability. Using the dynamic global vegetation model LPJ-GUESS within a synthetic experimental framework, the results show that different expressions of El Nino affect interannual variability in the terrestrial carbon cycle, but the effect on longer timescales is small. This suggests that capturing the characteristics specific to the expression of El Nino may not be critical for robust simulations of the terrestrial carbon cycle on multidecadal timescales. Known as a hotspot for terrestrial carbon cycle variability, and strongly influenced by climate modes of variability, the remainder of this thesis then focuses on Australia as a testbed to study areas of uncertainty in regional carbon cycle projections. At regional scales, climate projections display large biases, which hamper predictive capacity in impact studies. Many methods exist to either remove biases in the climate forcing, or to achieve informed ensemble averages, but it is not obvious whether some methods are preferable to others. Simulations using LPJ-GUESS and climate output from the Coupled Model Intercomparison Project Phase 6 (CMIP6) show that all bias correction methods reduce the bias in simulated carbon cycle to similar degrees but can lead to different vegetation distributions in the individual simulations. Bias corrections do not influence the ensemble average, but do reduce the ensemble uncertainty significantly. Choosing an informed ensemble averaging method, such as a weighted or random forest approach, is preferential to calculating a simple arithmetic ensemble average. However, suitable target datasets for carbon cycle variables covering both the spatial and temporal scales necessary are sparse, limiting the applicability of these methods for future studies. In addition, the representation of the Australian carbon cycle in TBMs, namely those part of the TRENDY v8 ensemble, was analysed. Land-use change is the main driver for discrepancies in the simulated long-term accumulated net carbon balance across TBMs. The TBMs also have different sensitivities to atmospheric carbon dioxide (CO2) concentration, but climate drives the year-to-year variability in the net carbon sink rather than the trend. Further, differences in the timing of simulated phenology and fire dynamics, as well as simulated vegetation carbon, and apparent carbon residence time are associated with differences in simulated or prescribed vegetation cover and process representation. These results highlight the need to evaluate parameter assumptions and the key processes that drive vegetation dynamics, such as phenology, mortality, and fire, in an Australian context to reduce uncertainty across models. Since none of the TBMs investigated clearly outperforms the others, LPJ-GUESS was then taken as the model with which to constrain the Australian carbon cycle. Observed plant traits were prescribed to achieve an improved representation of the vegetation cover in LPJ-GUESS. A comparison between the model and satellite-derived datasets showed reasonable agreement for gross primary productivity and leaf area index. LPJ-GUESS further captured the woody and non-woody cover over Australia. This allowed the model to be used to explore the future terrestrial carbon cycle over Australia. Based on the above findings, this thesis then explores the future Australian terrestrial carbon cycle using the CMIP6 ensemble together with the regionally parametrised LPJ-GUESS. The uncertainty in Australia’s future carbon cycle is strongly linked to biases in the meteorological forcing, and can be significantly reduced via bias correction. However, implementing bias correction methods still leads to an unresolved uncertainty in carbon storage in the vegetation at the end of the century. Variations in carbon residence time, and model sensitivities to CO2, temperature, and precipitation are the key drivers for the discrepancy in simulated carbon stored in vegetation. Reducing this uncertainty will require improved terrestrial biosphere models, but also major improvements in the simulation of regional precipitation by global circulation models. The thesis concludes with suggestions of future work that should help to resolve the large uncertainties in the future carbon stored in vegetation over Australia.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
Resource Type
Degree Type
PhD Doctorate
UNSW Faculty
download public version.pdf 30.19 MB Adobe Portable Document Format
Related dataset(s)