An efficient simulation-optimization approach based on genetic algorithms and hydrologic modeling to assist in identifying optimal low impact development designs

Abstract High rates of soil imperviousness, intensified by urbanization, have been contributing strongly to the occurrence of floods all over the world. To mitigate these impacts, Low Impact Development (LID) techniques seek to preserve the hydrology of urban catchments closer to pre-development conditions by using distributed stormwater control systems. Nevertheless, the application of these techniques is associated with a variety of challenges, including the design of the LID controls, due to the significant number of variables involved and the need to attend to multiple objectives simultaneously. In this context, the application of hydrologic simulation models integrated with optimization techniques has been recently explored as an alternative to assist in planning LID scenarios. This work aims to verify the applicability of an adaptation of the Genetic Algorithm NSGA-II, together with the hydrologic model SWMM, to assist the optimal design of a LID scenario seeking to reduce the stormwater runoff and the total costs on different return periods. This scenario has considered the combined implementation of permeable pavements, green roofs and bioretention cells. The results showed that the model was capable of finding a great variety of optimal solutions on various levels of runoff reduction, at different costs, for all return periods considered. Regarding the applicability of the optimization model as a LID design method, some limitations were found related to practical applications and possible oversizing of the subjacent layers of the LIDs. Therefore, suggestions on how to improve the model have been made to solve the identified problems.