Comprehensive assessment of decentralized stormwater systems in urbanized areas under climate change and watershed characteristics

Abstract

This PhD thesis comprehensively assesses decentralized stormwater management systems within a highly urbanized and flood-prone watershed in Istanbul. Although decentralized systems occupy a vast range of measures in different types, this research focuses on the performance of three kinds of these systems including Rain Barrels, Cisterns, and Drywells in urban runoff mitigation under the climate change and physical characteristics of the subjected watershed. The aforementioned systems are denoted in this thesis with capital initial letters as RB, CI and DW respectively. The study is structured into two main phases: climate assessment and model development, each contributing vital insights into sustainable urban water management. The climate assessment phase analyzed precipitation and temperature trends in Istanbul using data from a ground-based meteorological station, MERRA-2 temperature reanalysis, and IMERG precipitation estimations. Three non-parametric methods including the Mann-Kendall (MK) test, Sen's slope estimator, and Innovative Trend Analysis (ITA) were employed to identify the magnitude and direction of the possible trends. The reason for using different data sources and methods is to provide a comprehensive comparison between ground-based meteorological data with space-based weather observatory products and different methods that are used in trend analysis of hydrometeorological variables. This comparison is believed to provide valuable insights toward trend analysis of hydrometeorological variables and assist the transition process from traditional toward more novel data sources and methods in this field of study. Results of the first phase of the study revealed significant upward trends in temperature on seasonal and annual scales, with summer exhibiting the strongest increase followed by the spring, fall, and winter seasons. Conversely, multi-duration analysis using the ITA method identified intervals of decreasing trends, particularly between 1980 and 1999, except during summer. A distinctive upward trend re-emerged in the 2000–2019 interval. Precipitation trends were less consistent compared to the temperature, with single-duration analysis indicating significant increases during fall and summer but insignificant decreases during winter and spring. Annual analysis also resulted in significant increases in the precipitation trend. The multi-duration analysis uncovered more nuanced trends, including significant seasonal variations in specific intervals. The compliance against ground-based data was at a higher level in MERRA-2 data compared to the IMERG data. Especially weak results were observed in the trend analysis of precipitation in the spring season and high-value group of the fall season based on the IMERG data. In the same vein, the ITA method proved to be particularly effective for detecting sub-trends in temperature and precipitation data across different value groups, offering a graphical advantage over MK and Sen's slope methods. This capacity is critical for analyzing short-term variations and hydrometeorological extremes in densely populated megacities like Istanbul. In the model development phase, the highly urbanized Ayamama watershed in Istanbul which is known for its dreadful flood history in 2009, was selected as the study area. In this phase of the study, based on the climate assessment findings, Rain Barrel (RB), Cistern (CI), and Drywell (DW) units were selected for implementation in the hydrologic model of the study area due to their retention and infiltration capabilities. These systems can assist in addressing the water shortage due to increasing temperature and evaporation and also act as an auxiliary tool for mitigating runoff due to increasing precipitations in the city. A robust framework then was developed to integrate the design, implementation, and simulation of the aforementioned decentralized stormwater management units in the Ayamama watershed, using a combination of geospatial data, hydrologic-hydraulic modeling, and expert-driven decision-making. First, a physically based model was constructed using the Storm Water Management Model (SWMM), calibrated and validated through an auto-calibration process with OSTRICH optimization software. For calibration, water level data from the main drainage channel in the study area for different rainfall events were obtained from the Istanbul Metropolitan Municipality (IBB). These data were compared with the channel water levels generated by the simulation of the same rainfall events in the SWMM model. The automatic calibration process minimized the difference between the water level obtained from the simulation and the actual water level measured in the channel, thereby significantly improving the model's accuracy. This process was fully automated and carried out using the OSTRICH optimization software. As a result of the calibration, the Nash-Sutcliffe Efficiency Coefficient (NSE) values of 0.77 and 0.67 were achieved for the calibrated and validated models respectively, indicating a good agreement between the simulated and measured water depth in the channel. In this study, blind implementations without any logical reason are avoided. This means that a strict strategy is employed to determine the number, location, and type of decentralized stormwater management units to be implemented in the study area. To identify suitable locations for RB, CI, and DW implementation, a multi-criteria decision analysis was conducted by using the Analytical Hierarchy Process (AHP). In this process a survey was conducted with 10 participating experts to fill out three pairwise comparison matrices for RB, CI, and DW units based on a 1-9 Likert scale. This approach established practical thresholds for the number of units, balancing spatial, hydrological, and environmental constraints. Consequently, the initially defined maximum number of units to be implemented in the study area which was calculated only based on the available residential and industrial areas was reduced by 30.4%, 20%, and 57.2%, respectively for RB, CI, and DW units, ensuring optimal functionality and alignment with site conditions. The key findings of the second phase of this study can be summarized below: In the sensitivity analysis, model calibration, and validation stage, the most five sensitive parameters were found to be %imperviousness, conduit roughness, width of the subcatchment, N-imperviousness, and slope of the subcatchment. Which were introduced to the OSTRICH along with 5 other sensitive parameters to be calibrated automatically. The integration of SWMM and OSTRICH demonstrated a strong capability for handling large, complex models with numerous parameters, affirming OSTRICH's utility in calibration and optimization tasks. In decision making process about determining the location and number of RB, CI, and DW units to be implemented in the study area, amoung various criteria, subcatchment runoff, occupied area by residential and industrial areas, sub-soil permeability, and proximity to seawater were identified as the most critical factors influencing the suitability. The obtained results at this stage highlight the importance of initial hydrologic modeling in prioritizing subcatchments for the implementation of decentralized stormwater management units. In the assessment of the hydrologic response of the watershed under various designed storm events, implementing RB, CI, and DW units resulted in significant reductions in peak runoff (23.63%–54.03%) and runoff volume (11.17%–27.43%) under storm events with 2-hour durations and return periods ranging from 2 to 50 years. The time to peak also increased, indicating improved watershed response. However, it is noteworthy to mention that the reduction rate in urban runoff was inversely related to the storm's intensity or return interval. The developed model also was assessed under a real case long-duration storm event as well. For prolonged or intermittent storms, such as the storm event experienced in September 2009 in the Ayamama watershed, the performance of RB, CI, and DW units in runoff reduction declined due to limited storage capacities and slow infiltration because of saturation of DW units, emphasizing the need for periodic recovery mechanisms or bigger storage volume of the units. Finally, the water harvesting potential of the implemented decentralized stormwater management units was assessed. A year-long simulation using 2018 meteorological data showed that implemented RB and CI units could capture over 12 million m³ of rainwater, equivalent to 5.15% and 7.46% of the storage capacities of the Ömerli and Terkos dams, respectively which are the first two biggest dams of Istanbul. This underscores their potential to address water scarcity challenges in the city which is intensified due to climate change and population growth. This study contributes significantly to the body of knowledge in urban water management, offering scalable solutions for mitigating the dual threats of urbanization and climate change in megacities like Istanbul. The findings of this research provide valuable guidance for urban planners and decision-makers in designing sustainable runoff mitigation strategies tailored to local climate conditions. The proposed framework demonstrates how decentralized stormwater systems can be effectively integrated into urban environments, particularly in areas with limited groundwater resources and available space for large-scale implementations. Future research should focus on evaluating the impacts of evolving climate conditions on the performance of LID systems and optimizing their implementation strategies under constraints such as budget, space, and competing land-use priorities. Expanding the application of advanced calibration tools and exploring hybrid approaches that combine centralized and decentralized stormwater solutions could further enhance urban resilience against climate-induced challenges.