A critical review on optimization and implementation of green-grey infrastructures for sustainable urban stormwater management

Abbreviations

Description

ACO

Ant colony optimization

AHP

Analytical hierarchy process

ANN

Artificial neural networks

BMPs

Best management practices

BR

Bioretention cell

DEMATE

Decision making trial and evaluation laboratory

DT

Detention tank

DYCORS

Dynamic coordinate search using response surface models

EESMB

Economic-environmental-social monetization benefit

GA

Genetic algorithm

GDE3

Generalized differential evolution 3

GPS

Generalized pattern search

GR

Green roof

IT

Infiltration trench

LSPC

Loading simulation program in C + +

MCDM

Multi-criteria decision-making

MOALOA

Multi-objective antlion optimization algorithm

MOSFLA

Multi-objective shuffled frog leaping algorithm

NBS

Nature based solutions

NSGA-II

Non-dominated sorting genetic algorithm II

PCSWMM

Personal Computer Stormwater Management Model

PICEA-g

Preference-inspired coevolutionary algorithm using goal vectors

PP

Permeable pavement

PROMETHE

Preference ranking organization method for enrichment evaluations

PSO

Particle swarm optimization

RB

Rain barrel

RD

Rooftop disconnection

RG

Roof garden

SA

Simulated annealing algorithm

SCM

Stormwater control measures

SUDS

Sustainable urban drainage system

SUSTAIN

System for urban stormwater treatment and analysis integration

SWMM

Stormwater management model

TOPSIS

Technique of order preference similarity to the ideal solution

TuRBO

Trust region Bayesian optimization

UDS

Urban drainage system

VS

Vegetative swale

WSUD

Water sensitive urban design

Traditionally, grey infrastructure (i.e., pipes/drains and detention tanks) was mainly used for collection, storage and conveyance of stormwater load from point sources to receiving water bodies (Duan et al. 2016; Li et al. 2019). Over the last two decades, green infrastructures, also known as stormwater control measures (SCM) and best management practices (BMPs) in the United States, low impact developments (LID) in China, water sensitive urban design (WSUD) in Australia and New Zealand, and sustainable urban drainage system (SUDS) and nature based solutions (NBS) in Europe, have been extensively studied, evaluated, and applied separately or in combination with existing UDS to mitigate stormwater management problems (Islam et al. 2021). The most commonly applied green infrastructures are bioretention cells (BC), green roofs (GR), rain gardens (RG), permeable pavements (PP), infiltration trenches (IT), vegetative swales (VS), and rooftop downspout disconnections.

Several recent studies conducted for performance evaluation of green infrastructures proved their significant contributions towards solving quantitative and qualitative stormwater management problems (Samouei & Özger 2020; Pei et al. 2021; Yang et al. 2021; Wang et al. 2021a; Mohammed et al. 2022). However, the accurate planning, design and construction of green and grey infrastructures in individual and combined settings within urban catchments is a difficult and complex process, as multiple influencing factors including topographical, land-use, weather patterns, groundwater conditions and economic resources need to be considered during construction. Urban planners need to carefully decide the type, size and location of green infrastructures, and size/volume and location of grey infrastructures (i.e., pipe/detention tank). The careful planning and implementation of both stormwater management systems boost their collective advantages and stormwater load management efficiencies.

Recently, heuristic optimization approaches for green and grey infrastructures optimization have become widely popular for exploring optimal designs of both infrastructures due to their capability to solve highly non-linear, complex and discrete optimization problems (Zhang & Jia 2023). The synthesis of recent review studies represents their main contributions to analyze the uncertainties for optimal design of stormwater systems (Duan et al. 2016), discuss different optimization methods for stormwater management (Shishegar et al. 2018), explain multi-objective optimization of green infrastructures considering their functional evaluations (Wang et al. 2020) and critical analysis of different obstacles, issues and opportunities recognized during their applications in the industrial region (Jayasooriya et al. 2020), evaluate spatial allocation processes with evaluation of multiple optimization tools and approaches (Zhang & Chui 2018), develop a comprehensive optimization framework with synthesis of existing research gaps and future research directions (Zhang & Jia 2023), and present a systematic literature review analysis considering the optimization and resilience of green infrastructures (Islam et al. 2021). Although a few recently published review studies synthesize green infrastructure optimization considering multiple factors (Jayasooriya et al. 2020; Wang et al. 2020; Zhang & Jia 2023), a comprehensive critical synthesis of green and grey infrastructures considering planning and designing as preliminary stages, and their impacts on optimization and implementation was left unexplored. Moreover, the current study explained generalized optimization procedures, and synthesized previous research studies considering their objective functions, retrofitting scales, publication time-scales and countries. Further, this review paper mainly contributes to synthesize existing literature on planning, design and optimization of green-grey infrastructures, highlights pros and cons, and discusses required improvements and future recommendations needed for improving green-grey optimization approaches.

The preliminary comprehensive planning and design of green-grey infrastructures is a complex and difficult task as it involves consideration of several factors including local climate and urbanization patterns, land-use spatial dynamics, short-term and long-term climate change projections, waterlogging conditions, identification of flood hot-spots, stakeholders’ participations, and availability of space and economic resources (Tanyanyiwa et al. 2023). Kuller et al. (2017) presented a reciprocal WSUD planning concept in terms of ‘opportunities’ and ‘needs’ by stating that ‘WSUD needs a place’ or ‘a place needs WSUD’. The WSUD planning-support tool was developed for spatial placement of green infrastructures considering biophysical, socio-economic, planning and governance as ‘opportunities’ and ecosystem services as ‘needs’, and their baseline factors and indicators are thoroughly explained (Kuller et al. 2019). Zhang & Chui (2018) presented a strategic planning cycle of green infrastructures which mainly emphasizes the determination of location, selection of type, and designing and sizing of green infrastructures based on pre-determined installation objectives and target areas. They presented a schematic representation of three main elements of the strategic cycle, i.e., location, type and size and their mathematical relationships with three main factors including composition, aggregation, and distribution pattern. The critical analysis of the relationship between these main elements and influencing factors explains different combinations, arrangements, and possible patterns of green infrastructure placements and their impacts on surface and groundwater hydrological dynamics before and after applications of green infrastructures.

Generally, the first step of planning and designing green infrastructures is to identify appropriate locations for installation and then appropriate types and retrofitting sizes carefully designed considering stormwater quantitative and qualitative management objectives. The urban planners and engineers need to include local stakeholders' interests and participation in the initial planning phase considering their concerns about locations and attitudes towards the retrofitting process. The socio-economic constraints (including, e.g., stakeholders’ concerns and land acquisition/availability) need to be fully resolved at the initial phase for the successful implementation of projects. Moreover, a preliminary technical study of catchment characteristics is necessary to understand local land-use spatial dynamics, soil characteristics, slope, groundwater conditions, drainage and impervious areas, as well as evaluation of existing green infrastructures.

The preliminary analysis before optimization involves considerations of multiple factors including target/required scale and appropriate locations where retrofitting of green-grey infrastructures is planned. The technical survey, planning and evaluation of existing land-use and topographical characteristics of the catchment provide reasonable information for informed decision-making to decide the magnitude/degree of retrofitting scale and suitable locations for implementation (Zhou & Wu 2020). The primary land-use assessment quantifies impervious and pervious areas of each sub-catchment in the catchment/watershed and calculates the maximum level of imperviousness of each sub-catchment which can be practically retrofitted with different types of green infrastructures (Yang et al. 2023b). The specific land-use analysis can also provide information about the maximum retrofit size of each type of green infrastructure that can be defined as the maximum retrofit limit in the optimization algorithm. Similarly, the preliminary assessment of space availability can be conducted for detention tank placement considering existing and future construction plans of local stakeholders (Haghighatafshar et al. 2019; Latifi et al. 2023). Furthermore, the selection of the retrofitting scale also depends on the availability of resources like planning and implementation time, development budget, retrofitting stages, computational model, and stakeholders' willingness and participation (Xu et al. 2020; Zhang et al. 2023). In case of limited time and budget, the retrofitting scale of green-grey infrastructure reduces and can be shifted into different planning and implementation stages. However, it is important to highlight that the understanding of the functional performance and implications of green infrastructures at smaller scales to mitigate target objectives are well-understood by many recently conducted studies, however, the optimization of green infrastructures at the larger or different scales is relatively unexplored and complex (Zhang & Jia 2023). In addition, selection of an appropriate hydrological-hydraulic simulation model is one of the crucial steps by considering the modelling scale and required hydrological and water quality modelling outputs as objectives. Furthermore, multiple hydrological-hydraulic simulation models are available for stormwater modelling, exemplified as Stormwater Management Model (SWMM) (Rossman 2010), Personal Computer Stormwater Management Model (PCSWMM) (James et al. 2010), InfoWorks ICM (Autodesk Inc. 2024) and MIKE URBAN (Dhi 2008). Each of them has its pros and cons, so that selection of an appropriate model considering required modelling objectives and coupling/interface compatibility with optimization algorithms is critical (Ferrans et al. 2022).

Multiple optimization algorithms are available for the optimization of water resources systems' problems, like evolutionary algorithms, genetic algorithm (GA) (with its different extensions as non-dominated sorting genetic algorithm II & III) (Kumar et al. 2022), ant colony optimization (ACO) (Hou et al. 2019), generalized pattern search (GPS) (Audet & Dennis 2002), simulated annealing algorithm (SA) (She et al. 2021), particle swarm optimization (PSO) (Jean et al. 2021), artificial intelligence-based methods like artificial neural networks (ANN) (Chu et al. 2020), and machine learning algorithms. These optimization algorithms have different capabilities including the potential for solving optimization problems, computational efficiencies, and coupling/interface flexibilities with numerical models. The selection of an appropriate optimization algorithm considering the factors as mentioned earlier makes the optimization process more efficient and practically implementable, and the synthesis of recent optimization algorithms (Dong et al. 2021; Liang et al. 2023; Zamani et al. 2023; Zhu et al. 2023) employed for green-grey infrastructures proves NSGA-II is the most extensively applied optimization algorithm compared to others because of its configuration structure and comparatively easy integration with different numerical models and availability in various programming languages.

The optimization problem is mainly based on three components: objective functions, decision variables and constraints. Generally, the objective functions are calculated based on simulated outputs of numerical models (i.e., surface runoff, peak runoff, flood volume, pollutant concentration, etc.), or mathematical models (i.e., life cycle retrofitting cost, flood damage calculation, etc.). Meanwhile, objective functions are usually defined with the maximization of benefits versus the minimization of costs. In terms of decision variables (such as types, sizes and locations of green infrastructures, and storage volume and locations of detention tanks), these variables are determined by local land-use and topographical assessments (Saniei et al. 2021; Leng et al. 2022). In the optimization process, initial sets of decision variables are randomly generated or assigned with original values/ranges; then each set of decision variables is iteratively simulated by a numerical model; and afterwards, model responses are employed to calculate objective functions with defined constraints. This iterative process is repeated until the specified number of iterations/stopping criteria is reached. Multiple optimal solutions are iteratively calculated and improved based on a non-dominating optimization process, and the final optimal solutions are calculated based on the last iteration against optimal designs of decision variables (Leng et al. 2021).

Decision-making as a post-optimization step plays a significant role in the selection of appropriate/best optimal solutions among different optimal solutions/alternatives against their optimal designs which are calculated accordingly for practical implementations. Different multi-criteria decision-making (MCDM) tools such as analytical hierarchy process (AHP), technique of order preference similarity to the ideal solution (TOPSIS), decision-making trial and evaluation laboratory (DEMATE), preference ranking organization method for enrichment evaluations (PROMETHE), etc. have been developed over recent years (Tan et al. 2021), and AHP and TOPSIS among all MCDM techniques have been extensively applied for planning and decision-making of green-grey infrastructures. As an example, the detailed procedure of TOPSIS is discussed among all MCDM methods. Firstly, the specific evaluation indicators/criteria (e.g., cost, flood volume, urban drainage system resilience, pollutant concentration, etc.) are defined based on which different optimal solutions/alternatives would be evaluated. Secondly, a decision matrix is constructed based on optimal solutions and criteria to evaluate the performance of each optimal solution against each defined criterion. Thirdly, the decision matrix is standardized by considering the equal/determined importance of all criteria. During this normalization process, weights are assigned to all criteria based on their relative significance estimated through stakeholder preferences, expert opinions, and analytical methods. The next step is to determine the ideal and negative ideal solutions and their Euclidean distances from each alternative and between each alternative, their relative closeness to ideal solutions. Finally, the alternatives/optimal solutions are ranked based on relative closeness to ideal solutions (TOPSIS score), with the highest value classified as the best option (She et al. 2021; Tansar et al. 2023).

The above-mentioned findings of previous optimization studies highlighted some facts and research gaps that need to be emphasized in the future. The previous studies mainly focused on specific quantitative and qualitative objectives of green-grey infrastructures (as mentioned in Figure 6(a)), however, green-grey infrastructures optimization objectives need to be broadened to consider social and ecological optimization objectives. A limited number of studies recently investigated these objectives (Jia et al. 2022; Liu et al. 2023; Zhu et al. 2023) (as shown in Table 1); however, more research is needed to clearly understand green infrastructures' holistic benefits and development of more socio-ecological indexes (Pugliese et al. 2022; Zhang & Jia 2023).

Table 1

Summary of green-grey infrastructures optimization characteristics found in the selected 31 recent studies on/after 2019

View Large

Besides, the local stakeholders are key players in the practical implementation of these stormwater control measures, the consideration and quantification of their opinions at the planning, design and optimization stages as social acceptability indicators can provide insights into project success, and addressing their concerns can improve the probability of success to achieve target design goals. In addition to the earlier mentioned benefits of green-grey infrastructures implementation, it is important to highlight how these stormwater management approaches contributed to improving the resilience of UDS, reducing damages to local infrastructures, changing local groundwater table quality and quantity, as very few studies recently considered them (Bakhshipour et al. 2021a, 2021b; Wang et al. 2021b; Tansar et al. 2023; Wang et al. 2023a, 2023c), and further work is needed for deeper understanding. Furthermore, stormwater management practices have been significantly studied and implemented in a few countries, however, more awareness about their comprehensive benefits is needed to expand their sustainable stormwater management benefits to other countries.

Most previous optimization studies have focused on smaller scales, however, their impact at the larger scale is rarely explored for multiple reasons including but not limited to computational and economic resource limitations, technical expertise and experiences, hydro-meteorological data unavailability, stakeholder's concerns, etc. However, the findings of optimization studies at the smaller scale may be generalized for larger scales, assuming similarities in their regional topographical and hydro-meteorological conditions, land-uses, flood risks, drainage infrastructures, urban developments, climate change impacts, etc. (Chui 2017; Wang et al. 2020; Yang et al. 2023a). Meanwhile the benefits of stormwater management practicesin reality may be significantly different for larger scales like at watersheds and city scales because of their regional topographical and drainage infrastructures dynamics. So, multiscale optimization studies are needed in order to compare their benefits at smaller and larger scales by fulfilling existing research gaps, however, its implementation may face challenges because of variations in types, sizes and locations of stormwater management practices. Furthermore, another planning, design and optimization approach can be adopted in which some specific area (i.e., the area having higher flood risk) among larger scale areas (i.e., watersheds and city) is prioritized for retrofitting, followed by retrofitting by other regions based on the availability of resources (Wang et al. 2017; Xu et al. 2018; Xu et al. 2020; Zhang et al. 2023). This approach is also known as multi-stage planning and implementation based on priority levels of different regions.

Table 1 provides summarized information of 31 green-grey infrastructures optimization studies along with their optimization objectives, decision variables, retrofitting scales, green-grey infrastructure types, stormwater management approach, and simulation model with optimization algorithm. Only nine out of 31 reported studies in Table 1 have used a combined stormwater management approach (i.e., green and grey infrastructures optimization), while other studies considered only green infrastructures optimization. Future research studies should consider/prefer a combined stormwater management approach in order to fully understand the potential and limitations of both systems against extreme rainfall conditions considering future uncertainties (i.e., urban development and climate change). Furthermore, the majority of reported studies in Table 1 have employed metaheuristic algorithms (i.e., NSGA-II, PSO, etc.) whose computational time is one of the main challenge reported in stormwater management optimization studies. The computational time to converge optimization problem depends on many factors, mainly on the number of objective functions, number of decision variables, study area size, number of iterations, population size, cross-over, mutation rates, etc. Considering this limitation, many previous optimization studies have considered smaller study areas to perform optimization of green-grey infrastructures, and have discussed this as one of the limitations in their research findings. Recently, the surrogate-based/data-driven optimization approaches have grabbed significant attention and have reduced substantial computational time in comparison to traditional optimization approaches, in which this method replicates the optimization problem by developing surrogate models based on input-output datasets and objective functions, the developed surrogate model takes significantly less simulation time in comparison with the original model (i.e., hydrological-hydraulic simulation model). As a result, the convergence of optimization problem to find an optimal solution with a surrogate-based model collectively reduces computation time in comparison with the traditional optimization method.

A critical review study is conducted on green-grey infrastructures planning, design and optimization, analyzed its recent trends of development and implementation at different regions around the globe, its current status, research gaps and possible future research direction and recommendations. The statistical analysis of previous research studies that mainly focused on green-grey infrastructures optimization represents that the trend of such studies started from 2012 onwards, and the major portion of optimization research was conducted from the top ten countries, as China, USA and Iran ranked in the first three places, respectively.

The comprehensive planning and designing as a preliminary step before green-grey infrastructures optimization should consider multiple factors including local climate and urbanization patterns, land-use spatial dynamics, short-term and long-term climate change projections, waterlogging conditions, identification of flood hot spots, stakeholders' participation, and availability of space and economic resources, which play critical roles in the practical implementation of the designed project. Following that, the optimization problem is based on defined objective functions, decision variables (i.e., types, sizes and locations of green infrastructures, and storage volume and locations of detention tanks), and constraints which are solved by coupling any hydrological-hydraulic simulation models with optimization algorithms.

Furthermore, the previous optimization studies mostly considered specific quantitative and qualitative objective functions for optimization, while socio-ecological optimization objectives were rarely explored. Therefore, future research is recommended to quantify comprehensive social and ecological benefits to expand the horizon of current green-grey infrastructures' advantages. Lastly, future optimization studies should consider a larger retrofitting scale (i.e., watershed or city scale) to understand green infrastructures' holistic benefits on a bigger horizon by using surrogate-based optimization methods to reduce computational time. Furthermore, the concept of multi-stage planning and implementation of stormwater management practices needs to be applied in practice in which high flood-risk areas are prioritized for retrofitting compared to others, providing support to cope with economic constraints.

This work was supported by the Hong Kong Research Grants Council (RGC) (project no. C5002-22Y), the NSFC/RGC under the Joint Research Scheme (project no. N_PolyU559/22) and the Hong Kong Polytechnic University (project no. 4-ZZNF). We also appreciate comments and suggestions given by anonymous reviewers for the improvement of this manuscript.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.