Breaking Down the Computational Barriers to Real‐Time Urban Flood Forecasting

Abstract

Flooding impacts are on the rise globally, and concentrated in urban areas. Currently, there are no operational systems to forecast flooding at spatial resolutions that can facilitate emergency preparedness and response actions mitigating flood impacts. We present a framework for real-time flood modeling and uncertainty quantification that combines the physics of fluid motion with advances in probabilistic methods. The framework overcomes the prohibitive computational demands of high-fidelity modeling in real-time by using a probabilistic learning method relying on surrogate models that are trained prior to a flood event. This shifts the overwhelming burden of computation to the trivial problem of data storage, and enables forecasting of both flood hazard and its uncertainty at scales that are vital for time-critical decision-making before and during extreme events. The framework has the potential to improve flood prediction and analysis and can be extended to other hazard assessments requiring intense high-fidelity computations in real-time.Plain Language Summary Currently, we cannot forecast flooding depths and extent in realtime at a high level of detail in urban areas. This is the result of two key issues: detailed and accurate flood modeling requires a lot of computing power for large areas such as a city, and uncertainty in precipitation forecasts is high. We present an innovative flood forecasting method that resolves flood characteristics with enough detail to inform emergency response efforts such as timely road closures and evacuation. This is achieved by performing complex analysis of information on flooding impacts well before a future storm event, which subsequently allows much faster predictions when flooding actually happens. This approach completely changes the demand for required resources, replacing the nearly impossible burden of computation in real-time with the easy problem of data storage, feasible even with a low-end computer. Example results for Hurricane Harvey flooding in Houston, TX, show that predictions of both flood hazard and uncertainty work well over different areas of the city. This approach has the potential to provide timely and detailed information for emergency response efforts to help save lives and reduce other negative impacts during major flood events and other natural hazards.

Real-Time Streamflow Forecasting Framework, Implementation and Post-Analysis Using Deep Learning

“…In this study, we worked towards realizing Ivanov et al (2021) design in operational forecasting using deep learning and implemented a framework for real-time predictions. Our results from real-time operations reveal that our framework is capable of effectively leveraging all available real-time data.…”

“…These traditional hydrologic models were designed by state or federal agencies, and they are physical process-based models that are computationally intensive and rely on high-performance computations. Recently, a blueprint for real-time flood forecasting was developed by Ivanov et al (2021). The blueprint suggests that pre-event training surrogate models should be developed and that real-time data such as the forecasted rainfall data and flood stage should be used for real-time forecasting with surrogate models.…”

“…Based on the blueprint by Ivanov et al (2021), we proposed a framework for real-time operational streamflow forecasting using deep learning algorithms. We also implemented the framework with the deep learning models in April 2021 in the state of Iowa.…”

Flood Risk in Urban Areas: Modelling, Management and Adaptation to Climate Change. A Review

“…In our opinion, it is in applications of these kinds, where the main concern is to make fast and accurate predictions with limited interest in the physical analysis of the hydrological processes taking place, that ML-based models constitute a good alternative to physically based flood-inundation models that solve the 2D-SWEs. Several examples can be found in the recent literature, including surrogate models trained for a specific case study [116][117][118][119], and more ambitious approaches that try to develop generalized learning techniques that are capable of making predictions in case studies different from those that were used to train the algorithms [120,121]. There have also been attempts to integrate ML techniques within physically based numerical models by replacing the computationally expensive parts of the solver with simple and more efficient data-driven approximations [122].…”

An Impacts-Based Flood Decision Support System for a Tropical Pacific Island Catchment with Short Warnings Lead Time

“…Implementing a regular programme of flow gaugings, in particular during and after high flow events to improve existing rating curves and unit hydrograph calibrations used in the FDSS predictive models; • Statistical downscaling of NWP models for Samoa to finer grid resolutions, as well as more frequent real-time updates consistent with the GFS/ECMWF forecast frequencies (e.g., [44,45]). This includes the refinement of rainfall IDF's at key sites and ingestion of additional forecast models where available for Samoa; • Application of new impacts/exposure forecasting techniques using statistical, precomputed, metamodeling approaches (e.g., [46,47]); and • Assessing community perspectives in relation to warnings communications and response behaviour to progress towards the integration of a more people-centred approach to early warnings systems (e.g., [9,48]).…”