Skip to main content
  • Home
  • Thesis Topics 
  • Downscaling Drought Risk with Deep Learning for Sustainable Agriculture and Farm Applications

Research Thesis Topic

Downscaling Drought Risk with Deep Learning for Sustainable Agriculture and Farm Applications

Topic ID

Thesis Topic/Title
Downscaling Drought Risk with Deep Learning for Sustainable Agriculture and Farm Applications


Agricultural commodity, ecosystems, health and well-being are impacted by drought. Improving confidence in spatial modelling is necessary. Local drought information using regional climate prediction is difficult to achieve due to coarse resolution of models, unique geographical location and lack of real-time observations over large regions. Local-scale predictive models using machine learning (computational programs) can achieve accurate outcomes. Efficient tools for local assessment as alert or predictive systems are needed to operate locally using point-based input data and modelling that can be used by farmers and local government to be forewarned of drought in their localised regions.

To address deficits in methods, students will forecast Rainfall-Decile Drought Index and Standardized Precipitation and Evapotranspiration Index. This information acts as crucial knowledge for farmers and Government Ministries. The project develops hybrid models using meteorological data, sea surface temperature and climate indices with spatial mapping of accuracy, limitations and model performances.

Objectives are: (1) To test a series of models including Artificial Neutral Network, Support Vector Regression and Extreme Learning Machine, (2) Address issues related to ‘‘noise’’, non-stationary or contaminated inputs that deteriorate model performance by novel wavelet technique for “cleaning” the input variables, (3) Validate input/output data and model uncertainties arise from quality and representativeness of data, model structure (ability of model to describe the input variable’s response) (4) Combine several models into a hybrid framework that presents the best prediction of climate-risk.

The use of hybrid models new addition to drought modelling for climate-risk management.

Principal Supervisor

Associate Supervisors

Research Affiliations
  • Institute for Agriculture and the Environment
  • International Centre for Applied Climate Sciences
  • School of Agricultural, Computational and Environmental Sciences

Field of Research
  • Atmospheric Sciences
  • Civil Engineering
  • Physical Geography and Environmental Geoscience

Application Open Date

Application Close Date

USQ Scholarship Applications

Other Scholarship Funding Details
Australia Awards (Country-Specific Closing Dates)

Pre-approved for Ethics
Not Applicable

Admission Requirements

Please review the admission requirements for the academic program associated with this Thesis Topic

Suits students interested with background in physics, atmospheric science, environmental physics, climate science, computing and agriculture.

Back to List