IberSWMM+: A high-performance computing solver for 2D-1D pluvial flood modelling in urban environments

• Iber-SWMM + can speed up simulations by up to 200 times. • Run times are reduced from hours or days to minutes. • Considers fully distributed hydrological simulations and surface and sewer flows. • A dedicated tool for modelling building roofs is implemented. • SWMM simulation can become a computational bottleneck. Urban drainage modelling is essential for effective city planning and flood management. The increasing complexity of urban environments and the growing availability of high-resolution data have led to the need to develop more sophisticated and freely accessible urban drainage models. This paper presents the parallel implementation of Iber-SWMM, a freely distributed integrated 2D/1D urban drainage model for modelling surface and sewer flows and their interactions. Iber-SWMM constitutes an advance in the field by incorporating a fully distributed hydrological approach, advanced roof modelling tools, and GIS interoperability, offering a comprehensive solution for urban hydrodynamics. Originally designed for research in small urban drainage models due to CPU limitations, Iber-SWMM has now been enhanced with High Performance Computing (HPC) techniques. This allows for the simulation of high-resolution urban models with fine meshes comprising millions of elements, essential for accurate representation of complex urban geometries. We validated the model through laboratory-scale tests and two city-scale scenarios, providing detailed input data and demonstrating the applicability of the model in real-world situations. Our results show that the GPU-accelerated version achieves simulation speeds up to 200 times faster than the sequential version for large models. For instance, in a city-scale scenario with approximately 6 million cells, 3000 nodes, and 3000 links, simulation time was reduced from 72 h to just 20 min. To ensure result consistency and assess convergence, we conducted simulations using low, medium, and high-resolution computational meshes for each case study. Our findings indicate that both parallel and sequential versions produce consistent results, with convergence typically achieved at medium to high resolutions. Notably, we observed that for very large models, the computation of the drainage network in SWMM can become a bottleneck, suggesting an area for future optimization. By enabling the simulation of high-resolution urban models with millions of elements up to 200 times faster than sequential versions, this study bridges the gap between academic research and practical urban planning, empowering stakeholders to conduct more detailed, city-wide simulations, and ultimately contributing to faster urban flood risk management.