Introduction
Modern cities face challenges from climate change, including increased severe rainfall and flooding, which damage infrastructure and affect quality of life. The sponge city approach is presented as a novel and sustainable solution for urban runoff management based on Low Impact Development (LID). LID uses small-scale green infrastructure like green roofs and permeable pavements to store, infiltrate, and treat rainwater locally. This method reduces pressure on traditional systems, improves groundwater quality, and decreases flood risk. Implementing the sponge city concept in District 2 of Tehran is important due to its high population and runoff challenges. Hydrological simulation models like SWMM are powerful tools; SWMM is widely used and capable of accurately modeling urban processes and the impact of LID. Effective LID design is complex and requires simultaneously considering multiple objectives (such as runoff volume and peak, required area, and cost), making it a Multi-Objective Optimization Problem (MOOP). Algorithms like NSGA-II are suitable for solving MOOP. This research aimed to determine the optimal location and design of LID systems in District 2 of Tehran, developing a multi-objective simulation and optimization model based on SWMM and NSGA-II to find the best combination and area allocation of LID to minimize total runoff, reduce peak flow, and optimize costs.
Methodology
This research focused on hydrological modeling and multi-criteria optimization of rainwater management using Low Impact Development (LID) methods to manage urban runoff and reduce flood impacts. District 2 of Tehran, a high-density residential area of about 230 hectares with high imperviousness (85%), was selected as the case study and divided into 17 sub-catchments. Hydrological modeling was performed using SWMM software, employing the Horton method for infiltration and the dynamic wave method for network flow. Five LID types (VS, PP, GR, BRC, RB) were considered for implementation in 20 urban blocks. The multi-objective NSGA-II algorithm was used to determine their optimal arrangement and combination. A comprehensive optimization model integrating SWMM and NSGA-II via MATLAB was developed. Decision variables were the layout ratios of the five LID types in the blocks (60 variables). The objective functions were minimizing total runoff, peak flow, and the total cost of LID facilities.
Results and Discussion
Multi-objective optimization aims to find the best balance between multiple conflicting objectives. While analyzing the Pareto front is common, visualizing solutions becomes challenging with more than three objectives. This paper proposed using 2D scatter plots between objectives to address this limitation. The Pareto optimal solution sets for LID layout ratios were analyzed under different short-term rainfall events, focusing on heavy rainfall. Each solution represents a specific LID layout plan defined by 60 decision variables. Figure 6 illustrates the relationships between total runoff cost, peak flow cost, and design rainfall conditions (5, 10, and 50-year return periods) using these plots, highlighting the best cost-benefit combinations and aiding decision-makers by showing trade-offs and diminishing marginal returns. Analysis of Figure 6 indicated that changes in LID system cost did not significantly affect peak flow variations, suggesting the NSGA-II algorithm successfully found the Pareto front. The scatter plots revealed key relationships: total LID system cost vs. total runoff volume formed a Pareto front with a clear trade-off where increased cost reduced runoff, showing diminishing marginal returns. Conversely, cost vs. peak flood flow plots showed limited variation in peak flow despite cost changes, particularly for 5 and 10-year storms. This suggests flood peak flow was less sensitive to cost changes than total runoff in the Pareto optimal solutions, indicating NSGA-II found a diverse set of non-dominated solutions effectively covering trade-offs. Two distinct optimal schemes were selected from the Pareto set based on the principle of diminishing marginal returns: Scheme 1 minimized cost for maximum total runoff reduction, and Scheme 2 focused on minimizing cost for maximum flood peak flow reduction. Tables 5 and 6 detail their performance for 5, 10, and 50-year events. For 5 and 10-year events, Scheme 1 was deemed desirable and economical, offering higher total runoff reduction (around 22.1% and 18.7%) with a lower budget compared to Scheme 2, and comparable peak flow reduction (around 15.4% and 11.2% vs around 15.6% and 11.8%). However, for the 50-year storm, both schemes showed significantly reduced effectiveness in total runoff (around 18%) and peak flow (around 7.5-8%) reduction. This suggests designed LID systems are less capable of managing excess water during very heavy rainfall, and addressing such events may require complementary or combined solutions with substantially higher investment, as simply increasing LID ratios or budget might not be sufficient.
Conclusion
This research evaluated the effectiveness of implementing optimal Low Impact Development (LID) methods for urban runoff management in District 2, Tehran. SWMM simulations showed significant peak flow reduction compared to the no-LID scenario. For Scheme 1 (10-year design), total runoff reduction rates were 31.8% (5-year), 20.7% (10-year), and 18.5% (50-year). Peak flow reductions were 20.3% (5-year), 17.85% (10-year), and 12.6% (50-year). These results indicate positive LID performance in controlling runoff and peak flow. The SWMM+NSGA-II model is effective for designing nature-based systems, reducing flood risks, improving groundwater, and serving as a valuable tool for urban planners. However, effectiveness against very severe rainfall may be limited.