TG-PFT benchmark: Fine-grid topographic features and the rainfall and storm tide of Cyclone Gabrielle for the Gisborne catchment, New Zealand

Abstract

This dataset contains inputs of the BG-Flood simulations in the paper “TG-PFT: A Topography-Guided Progressive Focused Transformer for Flood Map Downscaling”. It covers the entire Waipaoa catchment of Gisborne (Tairāwhiti) region of Aotearoa New Zealand (2652 square kilometers), and holds the fine-grid topographic features and the rainfall and storm tide of Ex-Tropical Cyclone Gabrielle.The fine-grid topographic features are built from the national 8 m hydrologically conditioned DEM and hydraulic roughness map (Pearson et al., 2026). They are the DEM, the roughness length, the slope, the sine and the cosine of the aspect, and the topographic wetness index (TWI) on the 8 m fine grid. The catchment boundary is also included, as it defines the area of interest (AOI).The rainfall and the storm tide of Cyclone Gabrielle, from 11 to 15 February 2023, were described in Pelmard et al. (2026). A copy of the BG-Flood results of Cyclone Gabrielle on the fine and coarse grids is also included. The simulations use the ManyRivers branch of BG-Flood (Bosserelle et al., 2022), version 0.9, available at https://github.com/CyprienBosserelle/BG_Flood.The DEM, the hydraulic roughness map, and the rainfall and storm tide of Cyclone Gabrielle are data of Earth Sciences New Zealand (ESNZ). This dataset is released under the CC BY-NC-ND 4.0 license and may not be used for commercial purposes or shared in modified form. Users should cite this dataset and the sources listed in README.txt.input_files.tar holds the inputs, and results_gabrielle.tar holds the BG-Flood outputs (BGout.nc) of Cyclone Gabrielle on every grid. README.txt describes each file. All spatial data are in the New Zealand Transverse Mercator 2000 coordinate system (EPSG:2193), and elevations are referenced to the New Zealand Vertical Datum 2016 (EPSG:7839). The paired coarse- and fine-grid flood maps, the design rainfall and the tide of the flood scenarios, and the input files for running BG-Flood are available at https://doi.org/10.17608/k6.auckland.34803144, and the code at https://github.com/WilliamLUO0/TG-PFT. ReferencesBosserelle, C., Lane, E. M., & Harang, A. (2022). BG-Flood: A GPU adaptive, open-source, general inundation hazard model. In Australasian Coasts & Ports 2021: Te Oranga Takutai, Adapt and Thrive (pp. 152–158). New Zealand Coastal Society.Pearson, R., Harang, A., Bosserelle, C., Dean, S., Cattoën, C., Carey-Smith, T., Srinivasan, R., Shiona, H., Wilkins, M., Smart, G., Wadhwa, S., Reeves, G., & Lane, E. M. (2026). Nationwide 8m hydrologically conditioned digital elevation and hydraulic roughness maps for Aotearoa New Zealand. Geoscience Data Journal, 13(2), e70061.Pelmard, J., Lane, E. M., Harang, A., Bosserelle, C., Xu, Z., Pearson, R., Carey-Smith, T., Smart, G., Bind, J., Kumar, R. S., Cheng, E., Bui, G., Paulik, R., Zorn, C., & Wotherspoon, L. (2026). Regional flood hazard mapping for February 2023 Ex-Tropical Cyclone Gabrielle, New Zealand. Earth System Science Data Discussions, 2026, 1–29.