The Bureau of Meteorology (BoM) is currently developing a seasonal forecasting system of hydrological variables for Australia, using the AWRA-L land surface water balance model, forced with seasonal climate forecasts of precipitation, temperature, solar radiation and wind from the ACCESS-S model. The system generates monthly runoff forecasts for the Australian continent at 5km grid scale.

Due to the continental nature of the model, forecasts can exhibit biases when compared with observed runoff data. The objective of this project is to develop and evaluate a post-processing method to correct gridded runoff forecasts. Improving the accuracy of gridded runoff forecasts helps to improve the skill of predictions of extreme events, such as the risk of flooding or drought.

The project is divided in three phases:

  1. Analysis of geostatistical properties of forecast errors. The study will focus on one month ahead seasonal runoff forecasts in South-East Australia. This step will clarify the strength of spatial correlation in forecast errors and guide the development of the post-processing model.
  2. Development of a geostatistical post-processing model based on kriging interpolation of forecast errors. This step will explore the building of a statistical model of varying complexity to interpolate forecast errors spatially.
  3. Evaluation of corrected forecasts against observation in a leave-one-out validation process. This step will provide an estimate of expected performance gains provided by post-processed forecasts compared to raw forecasts.

The student will be based at the Bureau of Meteorology in Canberra or Melbourne (other Bureau offices might be possible).

The project requires experience in statistical or geostatistical modelling and familiarity with scientific programming environments (e.g. Python, R, Matlab). Experience in working with large datasets (e.g. on the NCI) would be preferable. The timing of the project can be arranged flexibly with the student.