Exploring suitable domain size for high-resolution urban rainfall simulation

Abstract

High-resolution urban rainfall simulation is useful for understanding the interaction between urbanisation and local weather as well as the impact of climate change over cities and the impact of adaptation measures such as urban greening. Previous studies on mesoscale Numerical Weather Prediction (NWP) modelling have largely focused on spatial resolution and other aspects (e.g., spin-up time and model parameterisations), with limited investigations on how to determine a suitable domain size. Domain size is likely an important factor when the spatial resolution of modelling is within convection-permitting regime (less than 5 km). In this study, 64 summer domain tests are simulated with the Weather Research and Forecasting (WRF) model, over Newcastle upon Tyne, with ERA5 as input data and a radar product from the UK Met Office for validation. Using an integrated evaluation indicator, alongside spatial distribution maps, it has been found that too large or too small domain sizes both have negative impacts on the simulation results and an optimal domain size for the events here is identified. We find that domain size has a stronger influence on the event simulation than changing grid resolution within the range 1–4.5 km and thus should be a primary consideration. We also find that, for more accurate simulation, smaller domain sizes are better suited to heavy rainfall events than to lighter ones. In a similar way, smaller domain sizes perform better for rainfall events that cover larger spatial areas. Although the optimal domain size identified here is specific to the region/season, the sensitivities and relative influences are expected to be more generally applicable and show the importance of testing domain sizes before embarking on production simulations.