Advancing the efficient development and deployment of hydrologic and hydraulic models for large scale and real-time applications

Abstract

Hydrologic and hydraulic models are tools that may be typically applied, respectively, to simulate streamflow and to determine depths and locations of flooding. While these tools are crucially important for predicting flood events, they require niche expertise and a high degree of effort to be developed and deployed effectively.

This thesis aims to streamline the level of effort required to develop and deploy quality models within the typical workflows that support the simulation of flood events. First, the selection of optimal model structures within hydrologic models is addressed. The blended hydrologic model, which allows the selection of mathematical equations to represent processes in the hydrologic cycle to occur as part of a calibration exercise, is tested and shown to provide benefit to both model performance as well as scientific utility. Secondly, the blended model is improved through an extensive empirical experiment which delivers a high performing blended version 2 model. This model achieves high performance scores across the contiguous USA without a need to adjust the model structure, which will greatly reduce the time-consuming step in practice of manually selecting the optimal model structure for a given watershed.

Finally, a novel method for hydraulic modelling and flood mapping is introduced. Improved geospatial methods are paired with a one-dimensional hydraulic model solver and then benchmarked against conventional methods. The result is shown to provide improved accuracy of flood maps while maintaining a computational runtime that is suitable for large-scale and real-time applications. Overall, it is anticipated that this research benefits the development of crucial tools for predicting and simulating flooding.