Towards a method of rapid flood scenario mapping using hybrid approaches of hydraulic modelling and machine learning
- 1Geospatial Research Institute, University of Canterbury, Christchurch, New Zealand
- 2School of Earth and Environment, University of Canterbury, Christchurch, New Zealand
- 3National Institute of Water and Atmospheric Research (NIWA), Christchurch, New Zealand
- 4Departamento Ciencias y Técnicas del Agua y del Medio Ambiente, Universidad de Cantabria, Santander, Spain
Floods are the most common hazard in New Zealand, the second most costly and they will change rapidly in frequency and intensity, become more extreme as the impacts of climate change become realized. At the same time, we are undergoing an intense urban development and growing population lives in floodplains, increasing the risk for people’s households and wellbeing. Additionally, computers have limited power and capacity, so there is a limitation in the number of flood scenarios that can be assessed and in the complexity of the hydrodynamic modelling process. This research project, which is part of the 5-year multi-stakeholder research programme “Reducing flood inundation hazard and risk across Aotearoa/New Zealand”, supported by the New Zealand Government and led by the National Institute of Water and Atmospheric Research (NIWA); investigates the feasibility of using a hybrid hydrodynamic/machine learning model to reduce the numerical modelling load and enable probabilistic modelling. The study site is the Wairewa catchment (Little River, Canterbury, New Zealand), working closely with the Wairewa Rūnanga based there. A sample of flooding scenarios is constructed based on the characteristics of the main inundation driver (spatial and temporal characteristics of rainfall extreme events) and other inundation drivers (lake level and antecedent conditions in the catchment). Selected scenarios from this sample will be modelled through a previously calibrated hydrodynamic model and the resultant inundation maps (maximum water depth map for each event) will be used to train a machine learning algorithm to produce the maps for the remaining events. The hybrid model would provide for any flooding scenario (defined by a specific number of variables) the corresponding inundation map in a fast and accurate way, avoiding the hydrodynamic modeling process (very time and computationally expensive). Results from this research will be used to develop a Mātauranga Māori approach to flood resilience and flood related policies by the local and central governments.
How to cite: Pozo, A., Wilson, M., Lane, E., Méndez, F., and Katurji, M.: Towards a method of rapid flood scenario mapping using hybrid approaches of hydraulic modelling and machine learning, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-2462, https://doi.org/10.5194/egusphere-egu23-2462, 2023.