Large-Scale, High-Resolution Urban Pluvial Flood Dataset for training Deep Learning Emulators

Flood inundation simulations can help in planning preventive measures against flood damage, howeverconventional simulation methods can be computationally expensive and time-consuming. An alternative tospeed up the calculations is to use data-driven emulators. In this work, we present two contributions: (1)the development of a large-scale, high-resolution flood dataset and (2) the development of deep learning(DL)-based emulators for flood prediction trained using our dataset. We show that in comparison toprevious works, our emulators are able to generalize to previously unseen test locations and achievecomparable performance metrics in terms of RMSE. In comparison to a GPU-accelerated simulator, aninference time speed-up of approximately 1000 times is achieved using these emulators. The dataset andcode is available at https://github.com/dinesh-k-natarajan/urban-flood-emulator/. Folder structure: The zip file contains the following contents: "city_level_data" folder contains 200 subfolders named after each city in our dataset. Each city subfolder contains two further subfolders: "dem": Contains the elevation map as (TIFF file) of each city of dimensions 2048 x 2048 "simulations": contains the maximum water depth maps for 5 simulated rainfall events. These are the outputs of the physics-based flood simulator and are the ground truth labels for our deep learning emulators. "rainfall_events" folder contains the 5 synthetic rainfall events as text files containing the time-series of spatially constant, temporally varying rainfall intensities" "sets.csv" contains the list of cities and which subset (training/validation/test) they belong to. Note: The elevation data was obtained from Geobasis NRW at https://www.opengeodata.nrw.de/produkte/geobasis/hm/dom1_tiff/dom1_tiff/ Related work: For more information, please refer to our paper: Natarajan, D.K., Stricker, M., Mukherjee, R., Charfuelan, M., Nuske, M., Dengel, A. (2026). Deep Learning Emulators for Large-Scale, High-Resolution Urban Pluvial Flood Prediction. In: Braun, T., Paaßen, B., Stolzenburg, F. (eds) KI 2025: Advances in Artificial Intelligence. KI 2025. Lecture Notes in Computer Science(), vol 15956. Springer, Cham. https://doi.org/10.1007/978-3-032-02813-6_17