Enhancement of real-time flood inundation forecasting platform in taipei city - application of hot-start mechanism on continuous inundation modelling

Recently, urban floodings in Taipei city are mainly attributed to short-duration high-intensity rainfalls that drop massive water volume in a considerably short duration and result in excess runoffs beyond the capacity of drainage systems. For this situation, the Taipei City Government and the research team of the National Taiwan University worked together to establish a real-time flood inundation forecasting platform based on the developed National Taiwan University Cellular Automata Flood Inundation Model (NTU-CAFIM) of the research team. The platform sequentially and automatically conducts 2-hour urban flood forecasting in a 30-min frequency, and each flood inundation simulation comprises 4-hour observed rainfall and 2-hour forecasted rainfall by QPESUMS. To further increase the efficiency without losing the required accuracy of the platform, for each simulation, the 4-hour observed rainfall is replaced with the settings of initial conditions for the NTU-CAFIM and the 10-min observed rainfall. Correspondingly, the overland flow model (OFM) and the sewer flow model (SFM) of the NTU-CAFIM are both modified to incorporate the hot-start module to accomplish this task. At the beginning of a simulation, OFM and SFM (i.e., the SWMM) both read the formatted hot-start files that come from the previous simulation to set their initial conditions (i.e., water depths and water velocities) As the simulation time reaches the end of the 10-min observed rainfall, OFM and SFM both save the water depths and velocities at this time into their formatted hot-start files for the subsequent simulation. In this way, the frequency of the flood inundation simulation is shortened from 30 minutes to 10 minutes, which greatly enlarges the efficiency of the developed platform and more detailed measured data can be input into the simulation. The accuracy and efficiency of the advanced platform are evaluated and compared with the original platform through the 4 June 2021 extremely heavy rainfall event in Taipei. From the results, the advanced platform only takes 39% computational time of the original platform to perform the flood inundation simulation with almost the same accuracy as the original platform, which is a remarkable improvement. © 2022, Taiwan Agricultural Engineers Society. All rights reserved.