Improving Hydrological Modelling: Properties of Scales and Processes

Abstract

In the last few decades hydrological modelling has seen great advances, yet the application of hydrological models is limited spatially (catchment), temporally (period under study), and by the model structure used. This means a single model does not work everywhere or in every condition. A few reasons can be identified; inconsistency of model structure, limited understanding of the individual catchment processes, and limited data predictability. In this research, hydrological modelling has been improved by focusing on a strategy to use more information from different data types (e.g. remotely sensed, reanalysis data), a strategy to improve the representation of the catchment system dynamics, and a strategy to use data-driven inference to strengthen model conceptualization. This research showed that using remotely sensed soil moisture as the only calibration variable can identify when and how soil moisture calibration influences streamflow prediction in a distributed modelling environment. This, in addition to the spatial scaling investigation with reanalysis data showed that scaling depends on the topography and vegetative cover of the catchment. This research has also developed a visualization tool for easy identification of model structural consistency. This can be applied within the proposed hydrological model calibration framework, whereby model structural consistency can be evaluated and improved by fine-tuning of model parameters. In addition, data based modelling has been explored to identify flow paths in a catchment-system. Investigation performed in this study enables modellers to identify plausible catchment stores. Practical implications of this research include robust modelling by developing flexible model structures, which integrates data knowledge and catchment physics. Therefore, water managers and resources planners can have better predictions for future water management and resource planning.