Dominant physical controls on hourly flow predictions and the role of spatial variability: Mahurangi catchment, New Zealand

Abstract

We present a systematic approach to the development of models for making hourly flow predictions, where each step provides insight into the relative importance of catchment and climatic properties (rainfall, soil properties, vegetation, topography), their spatial variability, and their influence on temporal flow variability at the outlet. Our modelling approach uses a simple conceptual model design requiring minimal calibration and physically meaningful input parameters estimated a priori from landscape data or from analyses of streamflow recession curves. The model structure allows direct measurement of parameter uncertainty and the ability to investigate its propagation through the model to produce bounds of predictive uncertainty via the Monte Carlo method. This method was applied to a number of model designs to assess the tradeoff between model complexity, accuracy and predictive uncertainty, and to identify the most appropriate model design under specific climatic conditions. With the preferred model, sensitivity analysis was used to identify the dominant controls on streamflow variability at the hourly timescale. In summary the aim is not to present a distributed model of universal applicability, but to generate insights into the climate, soil and vegetation controls on streamflow variability at the hourly timescale, and on predictive accuracy and uncertainty, that can be used in future modelling efforts.