Enhanced low flow prediction for water and environmental management

Abstract

The ability to predict low river flows is critical to water resources planning to sustain a healthy river ecosystem. However, the estimation of reliable low flows is difficult for a variety of reasons including lack of its proper conceptualisation. Consequently, arbitrary flow data transformation is done to enhance the influence of lower flows to help improve the goodness-of-fit during parameter optimisation in rainfall-runoff modelling. We carried out systematic model calibration using a range of flow data transformations to identify the one that results in best goodness-of-fits in 595 catchments across different rainfall regions of Australia. Effects of transformation methods on the prediction of low flow surrogates, namely the Baseflow Index (BFI) and cease-to-flow days or the Zero Flow Days (ZFD) were also investigated and regionalised. We found that the square-root transformation performed the best in modelling flow time series in all rainfall regions of Australia and for different ranges of precipitation and forest cover. Model parameters from most flow transformations predicted the mean annual ZFD and mean annual BFI well. Parameters from the log- and reciprocal-transformed flow were best in estimating the annual ZFD, while square-root and no-transformation did well in predicting the annual BFI. The observed BFI and ZFD were strongly correlated with the mean annual precipitation at a 5% significance level in all regions. It is also the biggest influencing factor for ZFD and BFI among several catchment attributes. ZFD and BFI had a strong negative correlation with each other implying that they may be interchangeable as low flow surrogates. There was a clear and widespread reduction in the probability of baseflow occurrence during the Australian millennium droughts of 2001–2010. The probabilities were reduced by up to 70% compared to the before-drought baseflow probability.