Decision tree based uncertainty framework for water accounting

Water accounting is becoming a common term in water resources management, especially in Australia after the Council of Australian Governments (COAG) reached an intergovernmental agreement on the National Water Initiative (NWI) in 2004. Water accounting aims to provide a complete picture of water resources for a region over a reporting period. It provides quantitative information about all water resource components such as inflows, outflows, rainfall, evaporation, water use, surface water availability and change, groundwater availability and change etc on a consistent basis. It has already been occasionally used in water resources management and planning processes. For example, the Commonwealth Water Act 2007 assigns the Bureau of Meteorology (the Bureau hereinafter) the role of "compiling and maintaining water accounts for Australia, including a set of water accounts to be known as the National Water Account." Among other obligations under the Act, the Bureau is required to publish this National Water Account (NWA) annually in a form readily accessible by the public. The NWA will provide reliable information such as water rights, water availability and water use that has previously been difficult to access or unavailable to general users in a standardised form. The Bureau, together with water agencies and other government agencies, has successfully prepared and published Pilot NWA for 2007-2008, and will publish the first NWA in 2011. Uncertainty of a measurand is "a parameter associated with the result of a measurement or estimation that characterises the dispersion of the values that could reasonably be attributed to the measurand." Most, if not all, line items in water accounting are subject to uncertainties. The sources of uncertainty vary, depending on the origin of data and processing procedures. Possible uncertainty sources include direct measurement uncertainty, indirect measurement uncertainty (processing errors from recording, estimation, interpolation, aggregation, sampling, etc), and hydrological model uncertainty (model structure and model parameters). Quantitative uncertainty information in water accounting can encourage users to query the precision of the underlying data and information, thereby improving their understanding of the nature of water stocks and flows. It can help identify the main contributors to the uncertainty thus enabling resources to be allocated appropriately in order to improve the accounting in future. It can also improve risk control in informed water management and planning. Thus, it is crucial to include uncertainty in water accounting, such as the NWA. As an activity under the Water Resources Information Research and Development Alliance (WIRADA) between the Bureau and CSIRO, we develop a consistent uncertainty framework for assessing, quantifying and reporting of quantitative uncertainty for water accounting. The proposed hierarchical decision-tree-based uncertainty framework is intended to help systematically choose and apply consistent and reliable approaches for quantifying and reporting uncertainty information about water accounting items. For each line item, the decision tree considers its nature, source and logical foundation. It will provide a pathway to some end point through a couple of simple decisions. For each end point (leaf node) in the decision-tree, we recommend a practical uncertainty analysis technique, ranging from classical uncertainty analysis techniques, uncertainty propagation law, Monte Carlo, literature to expert elicitation. Hierarchical structures are ubiquitous in the water accounting. They may originate from geographical hierarchy of reporting entities like subcatchment, catchment, and a nation, or from reporting detail hierarchy like detailed, intermediate, and simple water accounts, or from temporal hierarchy like monthly and yearly water accounts. We also suggest techniques to aggregate uncertainty from a lower level of accounting items. We simply list software implementation possibilities, including add-ins for spreadsheet like Microsoft Excel, standalone software, and general statistical software. For the work in progress, we also discuss a couple of issues in this preliminary uncertainty framework for water accounting.