Search

Two-dimensional hydrodynamic models are computationally expensive. This drawback can limit their application to solving problems requiring real-time predictions or several simulation runs. Although the literature presented improvements in using Deep Learning as an alternative to hydrodynamic models, Artificial Neural Networks applications for flood prediction cannot satisfactorily predict floods for areas outside the training datasets with different boundary conditions. In this paper, we used a conditional generative adversarial network (cGAN) aiming to generalize flood predictions in catchments not included in the training process. The proposed method, called cGAN-Flood, uses two cGAN models to solve a rain-on-grid problem by first identifying wet cells and then estimating the water depths. The cGANs were trained using HEC-RAS outputs as ground truth. cGAN-Flood distributes a target flood volume (vt) in a given catchment, which can be calculated via water balance from hydrological simulations. Our approach was trained on ten and tested on five urban catchments with distinct characteristics. The cGAN-Flood was compared to HEC-RAS for different rainfall magnitudes and surface roughness. We also compared our approach to the Weighted Cellular Automata 2D (WCA2D), a rapid flood model (RFM) used for rain-on-grid simulations. Our method successfully predicted water depths in the testing areas, showing that cGAN-Flood could generalize to different locations. However, cGAN-Flood tended to underestimate depths in channels in some areas for events with a small peak of precipitation intensity. cGAN-Flood was 50 and 250 times faster than WCA2D and HEC-RAS, respectively. Due to its computational efficiency and accuracy, we suggest that cGAN-Flood can be applied when fast simulations are necessary, and it can be a viable modeling solution for flood forecasts in large-scale watersheds., Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Sanitary Engineering

69A3551747106, National Transportation Library (NTL) Curation Note: As this dataset is preserved in a repository outside U.S. DOT control, as allowed by the U.S. DOT's Public Access Plan (https://doi.org/10.21949/1503647) Section 7.4.2 Data, the NTL staff has performed NO additional curation actions on this dataset. The current level of dataset documentation is the responsibility of the dataset creator. NTL staff last accessed this dataset at its repository URL on 2022-11-11. If, in the future, you have trouble accessing this dataset at the host repository, please email NTLDataCurator@dot.gov describing your problem. NTL staff will do its best to assist you at that time., Flooding are likely to increase worldwide due to climate change. Large storms, referred here as superstorms, defined as events with return period equal or larger than 100 years, can lead to an increase of property damages and loss of life. The ability to predict and plan for the impacts of superstorms on transportation infrastructure is key to mitigate future damages and losses. This study analyzed 51 combinations of future projections for representative concentration pathways (RCP) 4.5 and 8.5 scenarios, which were used to calculate future 1st and 3rd quartiles, median, minimum and maximum intensity-duration-frequency curves (IDF). A HEC-HMS and GSSHA models were built for Leon Creek and Upper San Antonio watersheds. HEC-RAS 1D and 2D were used to evaluate flooding in 20 bridges and the extent of flooded area and roads in both watersheds and to test flood control scenarios. Land use modification with 5, 10 and 15% of LID areas in the watersheds were simulated. The use of levees and altering channels were evaluated. In addition, we evaluated how an increasing the storage capacity of the Olmos Dam would contribute to reduce flood impacts downstream. Results show that the 3rd quartile of projected IDF is closest to the one originated with observed precipitation, which is likely to increase in the future. The near future (2025-2049) under RCP 4.5 scenario presented the greatest increase in intensity. HEC-HMS models showed that discharge peak will increase for all future periods under both scenarios, for the 100- and 500-years storms. Flood projections generated by GSSHA for 100- and 500-years and future precipitation showed that flooded area can increase significantly. For instance, the increase in flooded roads can be more than 80% in near future for 500-year storm in Leon Creek watershed. HEC-RAS analysis showed that all 20 analyzed bridges can be flooded with 500-years storm with climate change and 15 with the 100-year storm. Simulation showed that LID implementatio