2D AdH simulation of riverine hydrodynamics in a section of the Red River in Louisiana, USA, parameterized by the bottom friction coefficient. , in A neural operator emulator for coastal and riverine shallow water dynamics

This dataset curates a series of 2D Adaptive Hydraulics (AdH) suite simulations of riverine hydrodynamics in a section of the Red River in Louisiana, USA. The simulation uses an unstructured triangular mesh consisting of 12291 computational nodes and 23316 elements, has a natural inflow velocity condition upstream, a tailwater elevation boundary downstream, and no flow boundary conditions along the river banks. The bathymetry data were obtained from USACE Hydrographic surveys, while the inflow and tailwater boundary conditions were estimated based on observations from USGS gage 073556009 over 60 days starting at 01/01/2017:0800 hrs. The bed topography measurements and water surface elevations are referenced with respect to the bathymetry data, following the NAVD88 convention. The simulation is run for a 60-day period with a 5-minute time step, while the state variables (H, U, V) are saved every 15 minutes. The simulation data is parametrized by varying the scalar bottom friction coefficient uniformly between 0.02375 and 0.02625 to select 13 values.

This dataset is utilized to train and evaluate a novel neural emulator framework called Multiple Input Temporal Operator Network (MITONet), as has been described in the paper titled "A neural operator emulator for coastal and riverine shallow water dynamics". A preprint is available at https://doi.org/10.48550/arXiv.2502.14782.