Supervisors: Dr Elizabeth Vogel (elizabeth.vogel@bom.gov.au), Dr Gab Abramowitz (gab@unsw.edu.au), Prof Craig Bishop (craig.bishop@unimelb.edu.au)

Climate change affects the frequency and severity of hydrological extremes, including the risk of flooding and drought. These changes in hydrological extremes are a concern for many sectors that are highly dependent on hydrological conditions or affected by drought and flooding, such as water resources management, energy production, the transport sector, infrastructure planning or agriculture. In order to prepare for these changes, it is crucial to gain a better understanding of the spatial and temporal pattern of impacts of hydrological extremes on these sectors.

Machine learning methods (such as Neural Networks, Random Forests) can support climate and hydrological impacts assessment in multiple ways: 1) Machine learning can be used to better understand downstream impacts of hydro-climatic extremes. For example, machine learning has been used in previous studies to understand the impact of climate extremes on agricultural yields, to predict flood risks or develop flood inundation maps. 2) Machine learning has also been used in previous studies for statistical downscaling of climate model outputs, which may help in identifying impacts of large scale climate change on local climate. 3) Finally, machine learning approaches can be used to approximate bio-physical processes in climate models or in downstream impact models (such as ecosystem models or hydrological models).

This project will explore the use of supervised and unsupervised statistical learning methods (such as neural networks, random forest, clustering) to understand the impact of climate change on hydrological extremes and/or to simulate downstream impacts on affected sectors, such as agriculture, energy, transport, water resources management.

This project will use outputs of the AWRA-L hydrological model, which underpins the BoM’s Australian Landscape Water Balance website. The model simulates the land surface water balance and outputs hydrological stores and fluxes, including run-off, evapotranspiration and soil moisture in three soil layers (0m–0.1m, 0.1m–1.0m, 1.0m–6.0m). As part of the Bureau’s Hydrological Projections project, AWRA-L was forced with an ensemble of climate data based on: a) two scenarios for future greenhouse gas concentrations, b) four general circulation models (GCM) that have been assessed to be skilful for the Australian domain, and c) a range of statistical and dynamical bias-correction and downscaling methods. Depending on the interest of the student, possible analyses using machine learning could include:

  • Assessing the risk of agricultural production losses related to hydro-climatic extremes under climate change
  • Flood risk / flood inundation modelling and analysis of impacts on the transport network
  • Understanding changes in rainfall patterns leading to hydrological extremes under climate change

The student will ideally be based at the Bureau of Meteorology in Melbourne (however, remote work and/or placement in other Bureau offices is possible). The project would ideally suit a student with some experience in programming and data visualisation (e.g. using Python, R, Matlab). Experience in working with large datasets (e.g. on the NCI) and application of machine learning methods would be preferable. The timing of the project can be arranged flexibly with the student.

To apply: the Undergraduate Scholarship application form can be found here