Ensuring We Know When the Glass Is Half Full: Some Statistical Challenges in Water Resources in Australia
Brent Henderson
CSIRO Mathematics, Informatics and Statistics, CSIRO, Canberra, ACT, Australia

Water scarcity is a major national challenge in Australia. Statistical science is playing a key role in improving the information that is available and helping address this challenge. Accurate and timely water resource information underpins the Australian Bureau of Meteorology's annual National Water Accounts and regular water resource assessments, and is vital to water management practitioners, policy makers and researchers. This talk will illustrate three areas in which statistical science in CSIRO is contributing strongly to this. We will discuss an improved continental precipitation product created by the blending of observed sparse rain gauge data with less accurate but spatially exhaustive satellite-derived rainfall estimation. We will describe how model-data fusion methods are being developed to reconcile soil moisture estimated from a deterministic dynamic landscape hydrologic model with observed soil moisture information from satellite platforms or in situ measurements, and provide enhanced estimates with quantified uncertainty. Finally, we will consider water accounts that summarise changes in various inflows, outflows, losses (e.g. evaporation, irrigation) and storages for a spatial region over a specific period of time. These accounts rarely balance, with individual items in an account often subject to substantial uncertainty or estimated in different ways. We will discuss a Bayesian methodology developed to uniquely reconcile unbalanced water accounts, that respects the mass balance constraint, the uncertainty in individual components of the account, the potentially the correlations between them.

Keywords: Water resources; Data assimilation; Spatial blending; Uncertainty analysis

Biography: Brent Henderson is a statistical scientist based in Canberra with CSIRO's Mathematics, Informatics and Statistics Division. He has been working as an environmental statistician within CSIRO for 10 years. His primary focus has been on issues related to water quality and ecosystem health, and more recently water resources. He currently leads a Model-Data Fusion Project for assessing Australia's water resources that is part of an alliance between CSIRO and the Australian Bureau of Meteorology. Brent is also currently Group Leader of the Environmetrics Research Group within CSIRO's Mathematics, Informatics and Statistics Division.