River water balance accounts for the murray-darling basin to support water assessment modelling

Abstract

Monthly water balance accounts (1990-2006) were developed for 145 river reaches within the Murray-Darling Basin (MDB) as part of the Murray-Darling Basin Sustainable Yields project. They were used to assess how well river hydrology is measured and understood, and to identify the main uncertainties in river modelling. The purpose of accounting was different compared to water accounting systems set up for natural resources management, such as those that will be produced by the Bureau of Meteorology. Many aspects are still relevant however. A brief overview of the methods and data sources is provided, with emphasis on the methods used to combine different data sources in water balance accounts. Some benefits of the approach taken were identified as well as deficiencies requiring improvement. The aggregate water balance accounts are presented for all reaches that could be assessed, which covered most but not all of the MDB's river sections. The components of the water balance for which on-ground hydrometric data are available were compared to those which could be attributed with less direct observations, and apparent gains and losses that could not be attributed, respectively (noting that these will have included both real water and estimation errors in all terms). Approximately 48% of the overall water balance appeared to be gauged. Another 21% could be attributed, whilst the remaining 31% was unattributed. Comparison of the accounted and simulated water balance was not straightforward but provided useful insights into the assumptions made in modelling, and the uncertainty in these assumptions. Some implications for 'formal' water accounting are discussed. Unlike perhaps financial accounting, all numbers in a water account will be indirectly derived using estimation methods, because (1) every hydrological observation is indirect and involves a varying degree of estimation; and (2) not including even more indirect estimation methods (e.g. hydrological models) will produce water accounts that are incomplete and probably not useful. The suggested alternative is to use the full range of observations in a modelling framework that considers the uncertainty in the model as well as in all observations, to maximise both accuracy and precision in the accounts. Water accounts were most uncertain towards the end of inland river systems, particularly where anabranching and wetlands occur alongside irrigated areas: through unregulated or distributed diversions and extractions, losses to floodplains and wetlands, and groundwater recharge. These can all occur within the same reach and at the same time. In addition accurate streamflow gauging can also be challenging in this environment. Satellite observations of land use, evapotranspiration and inundation offer the best opportunity for further reducing uncertainty in these areas. Uncertainty is least in the unregulated, wetter headwaters of most of the regions, where gauging is better and processes better known. A degree of model 'over fitting' was identified; for example, increasing the total water balance volume to accommodate what effectively may well be errors in streamflow gauging. Calibration of hydrological models against all gauging and other data simultaneously in a way that considers the error in observations will help reduce such compensating errors, reduce the unattributed component in water accounts, and provide consistent uncertainty estimates for each term. Show more