An application of a log-transformed low-flow (LTLF) model to baseflow separation

A logarithmic transformation of flow is applied to the estimation of baseflow. Such a transformation has been applied in several hydrological studies since the discovery that flow is well described by a log-normal distribution. However, the main purpose of previous applications was normalisation of the error between simulated and observed flows. The new approach proposed here consists of an application of log-transform that leads to a formulation of the rainfall-flow process as a rate of change, instead of a typical mass balance problem. It applies a stochastic transfer function approach to log-transformed flow, enabling the problem to be stated in a statistically rigorous form. The difficulty in obtaining good hydrological predictions in extreme conditions lies both in the nonlinear character of processes governed by extreme hydrological signals and in the scarcity and inaccuracy of extreme flow measurements, particularly in early records. It has previously been shown that a log-transformed low-flow (LTLF) model allows for the decomposition of the simulated flow into slow and fast components that can be interpreted as baseflow and runoff. This paper compares the baseflow estimates obtained from the LTLF model with those obtained from other flow separation techniques, i.e. the Wittenberg nonlinear storage model and the Chapman linear filter. Also, the conditions for the applicability of the LTLF formulation are stated and the discussion is illustrated using some "monstrous", i.e. difficult to model, catchments from the UK and Poland. A rainfall-flow model based on the logarithm of flow gives a good representation of baseflow for catchments with a well-defined baseflow component.