• Four SWMM models with various spatial discretization resolutions were constructed. • The scale effects on the LID layout optimization are discussed. • Solutions optimized by the finest model performed best under small rainfall events. • Modelling scale had little impact on the layout optimization under a heavy rainstorm. • Layouts optimized by a coarse model overshadowed some types of LID practices. Climate change and increasing urbanization have worsened urban floods and other water problems, and it is generally accepted that the optimization of the spatial layout of low impact development (LID) practices is a promising solution. Existing studies have focused on coupling the hydrological model with optimization methods, but little is known about the effects of the spatial discretization of the model on the optimization results. In this study, the scale effects were examined in a case study in Guangzhou, China. Four models at various spatial discretization levels, namely models R1, R2, R3, and R4, with an average unit size of 0.022*10 4 , 0.405*10 4 , 2.43*10 4 , and 4.865*10 4 m 2 , respectively, were constructed to investigate the performance difference caused by delineation scales. The results show that the highest-resolution model R1 (one unit represented one type of land cover) provides more cost-effective layout schemes, but during a heavy rainstorm, the solution set provided by a coarser model are very similar to that generated by the finest model R1. A coarser model can provide wider solution sets, but most of these schemes overshadow some types of LID practices such rain garden and bio-retention cell, which also have a critical role in urban heat mitigation, air quality improvement, and so on. Due to the cumbersome work of catchment subdividing and time-consuming optimization process, a coarse model would be a good substitute to a large area. These findings provide new insights on how to achieve better performance by subdividing the catchment at a proper scale when optimizing the spatial layout of LID practices.