Development and sensitivity analysis of a tool to generate synthetic urban drainage models globally

Continual improvements in the quality, coverage, and accessibility of global geospatial datasets now mean that it is feasible to derive hydraulically plausible urban drainage networks and simulation models of these networks in cities worldwide. Privacy concerns, coupled with the cost and uncertainties in developing traditional network models, have fuelled the demand for such synthetic alternatives. We present SWMManywhere, which can create a hydraulically plausible Storm Water Management Model (SWMM) simulation model for a city using only the boundary coordinates of the target area. The datasets used in SWMManywhere are global, although their quality varies from country to country. We assess the utility of a SWMManywhere model by comparing pluvial flooding, in-pipe flows, and drainage network outflows in known networks. A previously unexplored difficulty with the use of synthetic network generation in urban environments is delineating the network’s boundaries when there are multiple and competing plausible outfalls, which is typical of most large cities. By using a sensitivity analysis approach, we explore how changing parameters associated with the network topology and boundaries can alter simulations. Assessing the uncertainty in our method helps to understand whether synthetically generated network models can produce meaningful simulations in their presumed most common use case: a dense urban environment where little is known about the network’s boundaries or outfalls.