Complex Workflow for National Flood Assessment
❄
| ABSTRACT
A national-scale understanding of flood inundation hazard is vital for disaster risk reduction. The MBIE Endeavour programme Mā te haumaru ō te wai aims to develop consistent flood hazard maps for all of Aotearoa New Zealand. To achieve this, a series of nested workflows has been developed. The overall workflow is composed of 4 main components to (1) prepare DEMs and roughness maps, (2) create design rainfall events to use for forcing (3) model hydrological flows in the upper catchment and (4) hydrodynamically model inundation on the flood plain. The workflow iterates through a geopackage of floodplains around Aotearoa and their upstream catchments to simulate the flood hazard under current conditions and future temperature increases. It ensures all necessary steps are systematically completed for each floodplain, managing the flow of data and model outputs between the different models. The software Cylc is used to orchestrate the workflow. An overarching control workflow is used to manage the different options which then spawn sub-workflow for each of the four components. By splitting the overall workflow into sub-workflows, this allows for modular development of each of the component workflows and models, allowing updates and improvements to be implemented and tested before being pushed to the wider workflow. The sub-workflow controlling the hydrological simulations and hydrodynamics simulation are the most complex. Their functionality if outlined in the next two paragraphs. Simulation of runoff and flow routing of streams and rivers on the “steep” part of catchments is simulated using the NIWA TopNet model. This hydrological model was modified to be consistent between gauged and ungauged catchments and provides boundary conditions for the hydrodynamics model. The hydrological calculations are significantly faster than the hydrodynamics and rely on climatology-scale simulation of the catchment to identify representative base flow and soil/groundwater conditions. The model is forced with either a designed storm (for hazard assessment) or a historical storm (for running past scenarios). Flooding is simulated using the BG_FLOOD Hydrodynamics model with forcing from the hydrological model, rainfall event and tide. The model uses a quadtree type mesh that is well suited for GPU computation and allows iterative refinement of the mesh. The hydrodynamics model is run in two phases. Firstly, at a uniform, relatively low resolution to coarsely, but quickly, capture the expected inundation extent. This coarse assessment of expected flow depth and speed is then used to identify target areas where finer resolution is required (e.g. flow constrictions, potential breaches). Additional information such as stop bank locations can also be used in this step. The model is then run with an adapted mesh that captures these features at high resolution, allowing the key mechanisms and features of flooding in the developed parts of catchments to be captured in a computationally efficient way. The cascading model system is demonstrated and validated using recent and historical flood in Waikanae, Westport and Hawke’s Bay. The hydrological model is validated against recorded flow and the hydrodynamics model is validated against post-flood high water marks collected by the local council. |