Sustainable Urban Drainage System (SUDS) modeling supporting decision-making: A systematic quantitative review

Review Sustainable Urban Drainage System (SUDS) modeling supporting decision-making: A systematic quantitative review Pascual Ferrans a,d, ⁎, María N. Torres b,c , Javier Temprano a , Juan Pablo Rodríguez Sánchez e a Departamento de Ciencias y Técnicas del Agua y del Medio Ambiente, Universidad de Cantabria, Spain b Department of Civil, Structural and Environmental Engineering, University of Buffalo, USA c RENEW Institute, University of Buffalo, USA d Escuela de Ingeniería de Bilbao, Universidad del País Vasco UPV/EHU, Spain e Centro de Investigaciones en Ingeniería Ambiental (CIIA), Universidad de Los Andes, Bogotá, Colombia HIGHLIGHTS •This paper aims to quantitatively analyze how SUDS-DSS are being build and applied. •The question was “What is the role of SUDS models on the decision-making process?” •Database and snowballing searches methodswereusedtoappraisethe papers. •The research focus has shifted from simple representations to more sophisticated tools. •There are some aspects that require special attention for SUDSDSS development. GRAPHICAL ABSTRACT abstractarticle info Article history: Received 6 July 2021 Received in revised form 15 September 2021 Accepted 15 September 2021 Available online 25 September 2021 Editor: Fernando A.L. Pacheco Decision Support Systems (DSS) for Sustainable Urban Drainage Systems (SUDS) are a valuable aid for SUDS widespread adoption. These tools systematize the decision-making criteria and eliminate the bias inherent to expert judgment, abridging the technical aspect of SUDS for non-technical users and decision-makers. Through the collection and careful assessment of 120 papers on SUDS models and SUDS-DSS, this review shows how these tools are built, selected, and used to assist decision-makers questions. The manuscript classifies the DSS based on the question they assist in answering, the spatial scale used, the software selected, among other aspects. SUDS-DSS aspects that require more attention are identified, including environmental and social considerations, SUDS trains performance and criteria for selection, stochasticity of rainfall, and future scenarios impact. Suggestions for SUDS-DSS are finally offered to better equip decision-makers in facing emerging stormwater challenges in urban centers. © 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: Urban drainage modeling Water Sensitive Urban Design (WSUD) Optimized SUDS design Green infrastructure (GI) Decision support tools Sponge cities Science of the Total Environment 806 (2022) 150447 ⁎Corresponding author at: Escuela de Ingeniería de Bilbao, Universidad del País Vasco, Spain. E-mail addresses: [email protected] (P. Ferrans), [email protected] (M.N. Torres), [email protected] (J. Temprano), [email protected] (J.P. Rodríguez Sánchez). https://doi.org/10.1016/j.scitotenv.2021.150447 0048-9697/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv Contents 1. Introduction................................................................ 2 2. Previousreviews.............................................................. 3 2.1. SUDSmodeling........................................................... 3 2.2. SUDS-DSS............................................................. 3 3. Methodology............................................................... 4 4. Resultsanddiscussion........................................................... 4 4.1. Categorizationandhistoricaloverview ................................................ 4 4.2. Geographicalspread......................................................... 5 4.3. Casestudyscale,typologies,spatialandtemporalmodelingresolution.................................. 6 4.4. Modelingmethodologies:softwareandeventsselection ........................................ 8 4.5. Categories,processesmodeled,andstakeholders............................................ 8 4.6. DSSinputs,outputsandgeneralframework.............................................. 8 4.6.1. Inputsandoutputs..................................................... 8 4.6.2. SUDS-DSSframework................................................... 10 4.6.3. SUDS-DSSinvolvingoptimization.............................................. 10 5. Conclusionsandperspectives....................................................... 11 Disclosurestatement............................................................. 12 Funding................................................................... 12 Declarationofcompetinginterest........................................................ 12 AppendixA. Supplementarydata...................................................... 12 References.................................................................. 12 1. Introduction Sustainable Urban Drainage Systems (SUDS) are an integrated network of engineered vegetated areas and open spaces (i.e., green roofs, rain gardens, porous pavements, etc.) used to protect natural ecosystem principles and functions and to offer a wide variety of benefits to people and wildlife (Tang et al., 2021). SUDS are a complement to centralized conventional sewer systems infrastructure to minimize the hydrological urbanization impacts and increase resilience to extreme rainfall events in urban centers (Zhu et al., 2019). These structures have the ability to attenuate extreme rainfall events (Tang et al., 2021) and are known for providing multiple environmental benefits (Liao et al., 2013), including climate change impacts reduction (Coutts and Hahn, 2015;Jones and Somper, 2014;Ghodsi et al., 2020;Roseboro et al., 2021), along with ecological and social benefits and other potential monetizable benefits in the long term (Wolf, 2003;Hamann et al., 2020). SUDS are usually referred to using several other terms, including Best Management Practices (BMPs), Green Infrastructure (GI) (Benedict et al., 2006), blue-green systems (Bozovic et al., 2017), Low Impact Development (LID), source control (Hamel et al., 2013), sponge city (Xia et al., 2017), nature-based solutions (Kabisch et al., 2016;Oral et al., 2020), and Water Sensitive Urban Design (WSUD) (Wong, 2006), among others (Fletcher et al., 2015;Chatzimentor et al., 2020). This set of terms is not static since, as described in Fletcher et al. (2015),they respond to evolving technologies and the incorporation of other fields into the urban drainage practice, and have (some subtle, others drastic) differences in scope and principles. For the purpose of this review, terms referring to ‘nature-based stormwater management solutions’ will be unified under ‘SUDS’, and the difference in the scope they encompass will be overlooked. SUDS selection, design, and location is a high-level complexity problem that relies on tools that systematically introduce relevant information, usually based on the best available representation of the urban drainage system. In such a complex endeavor, modeling is necessary to predict the behaviour of SUDS configurations (type, design, and location) and appraise their impact in the urban system. Modeling, along with other tools like multi-criteria matrices and optimization tools are put together on frameworks to aid SUDS decision-making frequently referred to as Decision Support Systems (DSS). The applicability of DSS in diverse fields results on numerous definitions, which although well-established in a particular niche, are confused when applied to an interdisciplinary field. Provided that SUDS are part of both the urban and environmental systems, the term DSS tends to be used interchangeably with others such as Environmental Decision Support Systems (EDSS) (Poch et al., 2004;Matthies et al., 2007;Reichert et al., 2015) and Planning Support Systems (PSS) (Klosterman, 1997). For a discussion of the usage of these terms, refer to Kapelan et al. (2005) and Te Brömmelstroet (2013). In the field of SUDS, the above-mentioned terms are hardly separable since SUDS-DSS can be classified into both highly complex systems (EDSS - as defined in Poch et al. (2004)) and planning-actions-related (PSS). Building upon the DSS definition provided by Fox and Das (2000), in this study the term SUDS-DSS is used to refer to “an structured set of tools (e.g., optimization, artificial intelligence, numerical models, statistical methods, Geographical Information Systems (GIS)) to assist decision makers and provide recommendations on SUDS design and spatial deployment”. DSS is a valuable aid for SUDS widespread adoption. They systematize the decision-making criteria and eliminate the bias inherent to expert judgment. By making SUDS decision-making less technical, SUDS-DSS encourage their adoption and increase their impact at the local, regional, and global scale (Baptista et al., 2005). Available SUDSDSS generally aim to solve problems of two natures 1) SUDS design (preliminary or detailed) and 2) SUDS spatial location (selection and placement), seeking the best SUDS implementation scenarios in terms of, at least, water quantity/quality and having ideal margins of costbenefit(Veith et al., 2003). Because the primary SUDS objective is the attenuation of the hydrological cycle disturbances, SUDS-DSS rely on the best available hydrological representation of the study area and the SUDS structure. Urban Drainage Models (UDM) with SUDS modeling capabilities are commonly used for this purpose, (Krebs et al., 2013;Kong et al., 2017). While mechanistic approaches are generally preferred for hydrological representations, there are other simplified approaches used in SUDSDSS, chosen for practical reasons (e.g., fast convergence or straightforward coupling with other DSS modules). Calibrated UDMs are ideal for building a robust and reliable SUDS-DSS (Haris et al., 2016;Beck et al., 2017;Ellis, 2013;Iffland et al., 2021) but they are not always available. Furthermore, future scenario projections considering urbanization trends and climate change are desirable but not always included (Wang et al., 2020). Some SUDS-DSS emphasize the exploitation of SUDS environmental benefits, aiming to find the best configuration to maximize one or more objectives. Also, despite stakeholders' relevancy, these actors are P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 2 seldom included in the early steps of the decision-making process, resulting in SUDS designs and locations that do not adjust to the expectations of those who will benefit from the structure (Raei et al., 2019). Many of these limitations result from the reactive approach in which SUDS-DSS are conceived and built. When a DSS-SUDS responds to the particularities of the case study and its temporal necessities, the resulting DSS 1) fails in capturing a holistic and unbiased perspective and 2) its applicability is constrained to the case study that motivated its development (i.e., Torres et al. (2016);Kuller et al. (2017)). In a rapidly (regionally-focused) evolving field like the SUDS-DSS, it is necessary to make periodical assessments of the state-of-art to internationalize regional experiences and bring forward new perspectives for the development and use of SUDS-DSS. This paper presents a quantitative and critical discussion on how modeling-based SUDS-DSS are being used to support decision-making. Through a two-year-long (2019-[March]2021) appraisal of articles introducing or applying urban drainage models to assist SUDS-related planning actions, the current state of the art was quantitatively evaluated with the key objectives of: •Analyze and update the state of the art and latest-trends in modelingbased SUDS-DSS research. •Quantify and map the development and implementation of modelingbased SUDS-DSS. •Understand the (modeling) practices employed when using SUDSDSS. The next sections of this article are structured as follows. Section 2 summarises previous reviews. Section 3 describes each step of the systematic quantitative review. Section 4 is divided into six subsections that report the quantitative results and discuss implications. Finally, Section 5 provides a critical perspective and suggests future research directions. 2. Previous reviews From a dedicated revision of previous reviews, it was evidenced how the SUDS research concerns and directions have been shaped by the development of interdisciplinary research, the broadening of SUDS understanding as providers of multiple benefits, and the late inclusion of urban planning in the stormwater management field (Kuller et al., 2017). Among the many reviews available, those tackling specifically SUDS modeling or SUDS-DSS and published in the last two decades (1997–[March] 2021) were considered. 2.1. SUDS modeling First urban stormwater models lacked the ability to model SUDS (see Burton et al. (2001) and Zoppou (2001) for a review). SUDS modeling gained sharper attention after their reported success in managing runoff. In 2007, Elliott and Trowsdale (2007) presented the first SUDS modeling review and proposed a classification based on their purpose: planning, preliminary or detailed design. Their review pointed out the importance of temporal and spatial resolution (particularly the limited ability of stormwater models to predict the flow rates from small catchments), runoff generation, and pollutants transport modeling. Two of their main findings were that i) only half of the SUDS models had a groundwater/baseflow component, and ii) there was a deficiency of tools that operate effectively at a large spatial scales. Ahiablame et al. (2012) provided a detailed review of SUDS representation in computational methods. By focusing on 4 SUDS types, they identified two modeling approaches: a process representation (e.g., infiltration, sedimentation, settling) and a practice representation, which uses an aggregation method to model the practice as a unit. They identified as areas of future research the scaling of SUDS practices from lot to watersheds and regional scales. The further refinement of SUDS models' physical processes representation (Kaykhosravi et al., 2018) was impulsed by the widening of SUDS models' usage for urban planning and decision-making. Kaykhosravi et al. (2018) compared stormwater models' capabilities of representing the hydrological and hydraulic SUDS processes and pointed out the need to develop more comprehensive SUDS models allowing various applications (i.e., research, conceptual, preliminary and detailed design, and operational support). With Ahiablame et al. (2012), it was evidenced that research grew on SUDS models' applicability for urban planning and policy-making. The authors promote the development of easy-to-use SUDS-DSS that effectively support decision makers and involve stakeholders, regulators, and policy-makers. As discussed previously, the attractiveness of SUDS for stormwater management is their ability to provide environmental benefits beyond the hydrological dimension (Caparrós-Martínez et al., 2020). Consequently, many SUDS-DSS are developed to help decisionmakers incorporate additional criteria for placement and design. There are informative reviews that tackled a broader perspective of modern stormwater management and UDMs. For example, Salvadore et al. (2015) stated that many modeling approaches target specificobjectives and that the level of detail in representing physical processes is not consistent. Other examples are the works by Bach et al. (2014) and Maftuhah et al. (2018), who focused on integrated urban water systems modeling. While Bach et al. (2014) classified integrated UDMs at one of four degrees of integration, Maftuhah et al. (2018) performed a classification considering social aspects, institutional dynamics, technical innovation, and local contexts. These bigger-picture reviews focused on drainage systems integration and interaction with other urban systems rather than focusing exclusively on SUDS. 2.2. SUDS-DSS Lerer et al. (2015) classified the SUDS-DSS according to the question it assists in answering: “How Much”,“Where”,and“Which”.Torres et al. (2016) focused on the geographical distribution of SUDS-DSS and the stormwater dimensions considered (e.g., quantity, quality, ecosystem services). Both reviews found case study specificity and lack of flexibility were drawbacks of most SUDS-DSS. On the other hand, Zhang and Chui (2018) reviewed SUDS-DSS for spatial decision-making and concluded that its generic structure couples a detailed UDM and an optimization tool, which communicate iteratively until a stop criterion is met. Other reviews on SUDS-DSS include the works by Zhou (2014) and Jayasooriya et al. (2020).Zhou (2014) made a comparison of modeling approaches and decision-aid tools for assessing SUDS alternatives. The author classified DSS into types of assessment tools: i) Economic, ii) Social, iii) Environmental, iv) Life-Cycle Assessment, and v) Health. Additionally, Zhou (2014) highlighted the importance of climate change and urbanization impacts in SUDS design, and stated that the future of the field are solutions that pursue a balance between the cost of investment and efficient performance (Zhou, 2014). More recently, Jayasooriya et al. (2020) revisited the importance of balancing environmental and economic goals and showed that despite many studies have recognized stakeholders' involvement importance, none have extensively studied the relevancy of their participation. Finally, the authors listed SUDS implementation barriers, including land ownership and lack of interest in negotiating land areas for SUDS placement. A seminal review that showed the importance of SUDS as part of the urban form is the work by Kuller et al. (2017). The authors proposed that SUDS location should not be considered a one-way process, but rather a two-sided problem. By defending that “WSUD (SUDS) needs a place as much as a place needs a WSUD”, they proposed the first-of-akind suitability framework for SUDS planning. Kuller et al. (2017) went beyond in classifying PDSS into their approach towards SUDS, as a part of (a) the urban water cycle, (b) the urban form, and (c) the water governance. P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 3 Elliott and Trowsdale (2007),Ahiablame et al. (2012) and Kaykhosravi et al. (2018) extensively explored key aspects of SUDS modeling while Zhou (2014),Lerer et al. (2015),Torres et al. (2016), Zhang and Chui (2018),Jayasooriya et al. (2020) and Kuller et al. (2017) focused on SUDS-DSS taxonomy and good practices for SUDSDSS development. This review does not attempt to cover in detail topics already discussed in previous reviews, but to build upon these recommendations to quantitatively analyze how modeling-based SUDS-DSS are being build and applied. For example, what questions are more frequently being answered with SUDS-DSS? How models are being used in practice (scale of the cases of study, time steps, modeling windows, calibration procedures, etc.). How SUDS-DSS development and usage are spread geographically? 3. Methodology A systematic quantitative literature review locates, appraises, and synthesizes evidence of a specific issue limiting bias by deciding specific criteria to include and exclude studies (Petticrew, 2001). The two most widely used techniques to systematically collect publications were used: database and snowballing searches (Badampudi et al., 2015). In the first, a combination of keywords was used to search in different databases (Scopus, Web of Science -WOS, and Google Scholar); and in the latter, new pertinent papers were identified through the reference list (backward) and citations (forward) of a seed-set of influential papers (Jalali and Wohlin, 2012;Fontecha et al., 2021). The review question addressed in this study was “What is the role of SUDS models on the decision-making process?”The objectives were to i) understand which SUDS models/software are more frequently used and how they are deployed for decision-making, ii) determine which questions the SUDS-DSS assist in answering (e.g., Which SUDS? Where? How many?), iii) analyze the DSS capabilities (e.g., optimization, stakeholders inclusion, uncertainty analysis). Table 1 lists the search terms used in the databases. Papers whose title have at least three words in different keyword sets were included in the review. In this way, the inclusion of the key eligibility criteria was guaranteed: “decision/tool”,“SUDS”,“modeling”,and“stormwater”. The best effort was carried out to include a comprenhensive set of key search terms for the papers appraisal, but it is not guarantee that all terms have been included given the proliferation and volatility of local terminology. Similarly, it is acknowledged that much of the literature on SUDS-DSS applications is written in languages different to English, leaving out of the review applications of non-English-speaking countries. Once a first potential seed-set of significant papers was gathered, the inclusion criteria of the search were that the paper must be i) useful for decision-making (i.e., a tool or case study, not a framework, review, or experience report), ii) specific for stormwater (although other urban water cycle elements may be present), and iii) include SUDS modeling. All papers that answer the review question and fulfill the inclusion criteria were collected. If the review question was not answered after reading the whole document or/and the inclusion criteria were not fulfilled, the paper was withdrawn. The data extracted from each paper was stored by filling fields in a review tool developed in Excel Visual Basic (VB) to ease the information withdrawal. Approximately 270 papers were collected using the keywords in the search engines and subsequent snowball forward and backward procedures. Only 120 articles met the inclusion criteria and were studied in depth. The following subsections summarize the results extracted from these 120 manuscripts, but only some will be referenced as part of the bibliography of this document. For a complete list of the papers, please refer to Appendix 1. 4. Results and discussion 4.1. Categorization and historical overview Two broad categories were identified in the preliminary assessment of the articles, each creating differentiated research outcomes: 1) a SUDS model or 2) a DSS relying on a SUDS model. There are inputs and outputs in both categories, but the first refers to isolated modeling generally to assess SUDS performance, while the latter couples the SUDS model with other modules to include costs, stakeholders, and secondary benefits, for example. Another essential difference is the nature of the outputs. While the stand-alone SUDS model delivers runoff series, pollutants reductions, or any other performance measure, the DSS provides answers to decision-makers questions (i.e., Which SUDS is recommended or the best? What locations are suitable/optimal for SUDS? Which SUDS meets the pollutant reduction target?). Fig. 1 illustrates the relation between the two article categories and shows the count for SUDS models (category 1) and SUDS-DSS (category 2), the number of articles that address the questions Which?,Where?,How many?,orassists the design of individual SUDSand trains. A single paper can assist in answering more than one question, so a manuscript can be counted several times in Fig. 1 (once per question assisted). The total reviewed articles spanned the period comprised between 1997 and 2021. The number of articles had an increasing trend, starting with just a couple of publications per year, from 1997 to 2012, and then continued increasing from 2015 to 2020. The type of output was diverse (i.e., the question the tool assists in answering). Notice in Fig. 2(a), that SUDS-DSS commonly answered a single question, while since 2012, there is more output diversity; observe that since 2015, the outputs include 5 categories. These observations reflect both the diversification of SUDS-DSS users and the broadening of perspective from SUDS “units”to SUDS “systems”. SUDS trains were only included in DSS from 2015, which can be explained by the evolved capability of models to simulate flow among connected SUDS structures. Similarly, the “SUDS design”question predominated the early development of the tools, but with time this Table 1 Sets of keywords used for search. Papers whose title have at least three words in different keyword sets were considered for review. Set 1 Set 2 Set 3 Set 4 Decision/tool keywords SUDS keywords Modeling keywords Stormwater keywords Assess* Best management practice* (BMP) Model* Runoff Effective* Sustainable urban drainage systems (SUDS) –Storm* Cost* Green infrastructure (GI) –Urban Heuristic* Low impact development (LID) –Flood* Management Water sensitive urban design (WSUD) –Pluvial Optim* Nature based solutions –Rainfall Objective* Blue green systems –– Planning Sponge cities –– Support* Bioretention –– Tool* Infiltration –– Decision* Retention –– –Detention –– *Any word containing the root-word signaled by * also makes part of the set. For example, “assess*”includes the words “assesing”,“assessed”,and“*objective”includes “multi-objective”or “multiobjective”.Fig. 1. SUDS models embedded in a SUDS Decision Support System (DSS). P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 4 interest declined, giving space to other aspects, such as “which”, “where”or “how many”SUDS were suitable or optimal. This observation is tied to the change in the study scale interest and dimensions under study. Fig. 2(b) shows that larger study areas (i.e., subcatchment and catchment) gained attention over the last years in comparison with smaller scales, while the number of studies focusing on household and neighborhood scales has decreased over the last 6 years. The city-scale decision-making appeared for the first time in 2005 with the study developed by Makropoulos and Butler (2005), which used non-structural SUDS for potable water consumption reduction. This shift in the spatial scale can be explained by the exponential growth of computational capabilities (Burger et al., 2014), which allowed the modeling softwares to include bigger spatial scales over time without incurring in longer processing times. When appraising the papers, it was consistently found as a recommendation for future studies the development of SUDS-DSS capable of assisting decisionmaking at the city-scale (Makropoulos and Butler, 2005;Chen et al., 2017;Zubelzu et al., 2020). Furthermore, these articles pay special attention to the importance of including optimization and stakeholders bargaining models for decision-making at watershed and city-scales, and also make a special highlight on the importance of including economic, social and environmental dimensions. Similarly, Fig. 2 (c) shows that the dimensions considered in decision-making diversified with time. In 2006, SUDS-DSS were already considering the economic aspect along with the runoff quantity and quality, while the environmental and social dimension appeared more recently and continue gaining importance (Alves et al., 2020). 4.2. Geographical spread Fig. 3 shows that the majority of the studies were developed in Asia, with 58% of the articles, followed by North America (27%), Europe with (14%), Oceania with (6%) and South America (6%). The countries with the largest contributions were China, United States of America (USA), Iran and Australia, with 28%, 20%, 11%, and 5% respectively. The rest of the countries had a lower count (less than 3% from the total) and 36% when aggregated. Based on these numbers, it was possible to identify the urgent need for less-developed countries to increase the SUDSDSS scientificproductivity(Ferrans et al., 2018), considering there is a high potential for existing models to be implemented in these countries (McClymont et al., 2020). Developed countries can contribute to closing this gap through collaborative international projects (e.g. Resource Brandia, euPOLIS, euPOLIS, 2020, etc.), where other countries' expertise can accelerate their learning curve. As expected, the question the tool assists in answering and the SUDS aspects considered in each country are diverse. While more-developed countries, like the USA, Canada, and Australia included more aspects besides runoff quantity and quality, less-developed countries focused almost exclusively on these two. An exception is China, with several included aspects, and Brazil, which put special attention to the social Count 45 40 35 30 25 20 15 10 5 0 Year (a) Queson addressed over me (count). 1997 1999 2004 2005 50 2006 2007 2010 40 2011 2012 2013 30 2014 2015 2016 2017 20 2018 2019 2020 10 2021 0.0 0.2 0.4 0.6 0.8 1.0 Fraction 0 Year (b) Spaal scale modeled over me (propor- on) (c) Dimensions over me (count) Design How Many Trains Where Which Catchment City Household Neighborhood Subcatchment Economic Analysis Environmental benefits Quality Modelling Quantity Modelling Social benefits Year Count 1997 1999 2004 2005 2006 2007 2009 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 1999 2004 2005 2006 2007 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Fig. 2. Research interest evolution over time. Categories in panel (a) defined as follows: Design (preliminary or detailed design), How Many (number of suitable/optimal structures), Trains (order of interconnected typologies), Where (spatial allocation of SUDS), Which (optimal/suitable SUDS typologies). Panel (c) shows the aspects (dimensions) considered for SUDS decision-making using a timeline. Observe that early years focused exclusively in quantity and quality modeling. P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 5 aspect. This finding is explained by the urgent challenges (e.g., urban flooding and receiving water bodies quality impairment) that are still to overcome in less-developed countries. Once again, these results reinforce the need for developed countries to generate collaborative environments in which there is room to share experiences and gather information. 4.3. Case study scale, typologies, spatial and temporal modeling resolution In recent years, there has been a rapid development of computational capabilities, which allowed the representation of detailed processes at larger scales. For SUDS models, these advances permitted refinements in process representation and finer spatial and temporal resolution. Previously, it was shown that the spatial scale focus has shifted in time, partially because the modeling capabilities have allowed the inclusion of more detail and more complexity, but also because decision-makers needs have evolved. Every day, decision-makers rely more upon software to make decisions, while expert judgment has been gradually replaced by systemic procedures that reduce the amount of bias and manual work, generally speeding up the process (Hattab et al., 2020). From the appraised papers, 55% used a catchment-scale, 22% a subcatchment-scale, followed by neighborhood-, household-, and cityscale, with 12%, 6%, and 5%, respectively. Fig. 4(a) shows that the study area has a range of 6 orders of magnitude, with a minimum value of 0.01 hectares (ha), a maximum value of 650,000 ha, and a standard deviation of 80,000 ha. The catchment-scale has the largest range and number of outliers (comprising 6 orders of magnitude). Fig. 4 (a) shows that the area is not determinant of the spatial unit of analysis, since the same study area (e.g., 100 ha) can be classified into city, catchment, sub-catchment, or neighborhood. The smallest scale found was the household, with a mean area of 0.1 ha and the largest was the city-scale, with a mean of 1000 ha. Fig. 4(b) shows that most DSS answered the question “Which SUDS”, disregarding the scale. In general, all questions can be answered at a neighborhood-scale or bigger, while for the household-scale, the only questions addressed were “Design”(detailed dimensioning and location) and “Which SUDS”. The household-scale has a larger proportion of “Design”, which was expected considering that smaller areas makes it more attainable to reach a higher level of detail. Fig. 4(c) shows the frequency of the most common SUDS typologies found in the review. Curiously, the most used typology is one that does not use green areas directly (permeable pavements), followed by grassed swales, bio-retention cells, and rain barrels. In total, nearly 30 articles included SUDS in their models, but do not specify which type (e.g. Jia et al., 2012;Raei et al., 2019;Rodríguez-Sinobas et al., 2018). It was found that all typologies had similar proportions with respect to the question they assist in answering (Fig. 4(c)), showing that the tools do not differentially address the questions depending on the typologies they include. Fig. 4(e) shows that (as expected) some of the typologies are more frequently used in large scales: constructed wetlands, detention and infiltration basins, dry detention, and bio-retention ponds; while some other typologies have more flexibility to be used in large, medium, and small scales: storage tanks, rain barrels, infiltration trenches, and grassed swales. Specifically for the household-scale, the most popular typologies were green roofs, rain gardens, permeable pavements, and bio-retention strategies. Contrary to the scale, the land use (e.g., commercial, industrial, residential, recreational open space) showed notrends regarding typologies; all typologies were present in similar proportion. Only 4% of the articles (5 papers) allowed the inclusion of SUDS trains (typologies sequentially interconnected) despite the literature recommends trains to increase the structures' efficiency in managing runoff (e.g., Bastien et al., 2010). Those articles that did consider SUDS trains, selected the train components based on experts knowledge and the reported efficiency of individual SUDS to control target pollutants (e.g., Xu et al., 2017;Jayasooriya et al., 2016; Zafra et al., 2017), instead of following an standard procedure. The time step used for the calculations showed a high variability among the different studies, ranging from 1 min to 2 h. Fig. 4 (d) shows a scatter plot of the size of the study area and the modeling time step, using color codes for the softwares. The outlier in Fig. 4 (d) (30 days time step) corresponds to Chang et al. (2011), who performed an own-developed model based on water balance equations with a simulation period of 50 years. Disregarding the software, there is no evident area-time trend in Fig. 4(d). This can be attributed to differences in the models complexity even when the same software is used. The figure shows that the SWMM software is scattered along the two axis, as opposed to other softwares that are clustered in areas of the plot (e.g. L-THIA-LID is found in large areas and large time steps only). A previous study, Salvadore et al. (2015) reviewed 100 UDMs to compare the space-temporal resolutions. The authors identified two clusters: catchment-scale applications (larger temporal resolutions) and small-size cases of study. The authors found the finest temporal Fig. 3. Literature geographical distribution. P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 6 Design How Many Trains Where Which Permeable Pavements Grassed Swales Bioretention Cells Rain Barrels Green Roofs Bioretention Ponds Infiltration Trenches Raingardens Not Differentiated Constructed Wetland Storage Tanks Infiltration Basins Dry Ponds Green Space Detention Basins Tree Pits Sand Filters Area (ha) 10 5 10 3 10 1 10 −1 Subcatchment Neighborhood Household City Catchment 0.0 0.2 0.4 0.6 0.8 1.0 Fraction Spatial Scale (a) Area (ha) vs. spaal scale (b) Spaal scale vs. queson addressed 100 80 10 5 60 10 4 40 10 3 20 10 2 0 10 1 10 0 Software L -THIA-LID SWMM SUSTAIN Own-developed MIKE URBAN WetSpa-Urban SewerGEMS 10 −1 Typology 10 0 10 1 10 2 10 3 10 4 Time Step (Minutes) (c) SUDS typologies frequency (d) Area (ha) vs. modeling me step (min) color-coded by Soware Tree Pits Storage Tanks Sand Filters Raingardens Rain Barrels Permeable Pavements Not Specified Infiltration Trenches Infiltration Basins Green Space Green Roofs Grassed Swales Dry Ponds Detention Basins Constructed Wetland Bioretention Ponds Bioretention Cells 0.0 0.2 0.4 0.6 0.8 1.0 Fraction (e) Spaal scale per SUDS typology Design How Many Trains Where Which Catchment City Household Neighborhood Subcatchment Count Area (Ha) Tipology Spatial Scale Fig. 4. Case study (land use and typologies) and temporal resolutions per spatial scale. Categories in panels (b, c) defined as follows: Design (preliminary or detailed design), How Many (number of suitable/optimal structures), Trains (order of interconnected typologies), Where (spatial allocation of SUDS), Which (optimal/suitable SUDS typologies). P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 7 and spatial resolutions to be 1 s and 10 m 2 , but these UDMs did not have the capabilities of modeling SUDS. When comparing our results to Salvadore et al. (2015), it was evidenced that category 2 articles (studies developing/applying a DSS) were only being implemented in what these authors call “catchment-scale applications,”meaning that SUDSDSS are still in the larger scale of urban drainage modeling in terms of temporal and spatial modeling granularity. As will be discussed in Section 4.4, the decision on the temporal and spatial resolution is also related to SUDS model capabilities and the selection of event-based or continuous simulation. 4.4. Modeling methodologies: software and events selection The Storm Water Management Model (SWMM) (Rossman, 2010) developed by the US Environmental Protection Agency (EPA) was the most frequently used in the papers appraised (46% of the studies). The second most common software were own-developed non-commercial models, with 16% of the studies, followed by L-THIA-LID (PurdueUniversity, 2016) (4%) and SUSTAIN (Shoemaker et al., 2009)(2%). The rest of the studies, which represent the 20% of the total articles, used other software (e.g., MIKE URBAN (DHI, 2008), MUSIC (eWater, 2020), SUDSLOC (Viavattene et al., 2011), GISP (Meerow and Newell, 2017), ReVISIONS (Hargreaves et al., 2019), SSANTO (Kuller et al., 2019), UrbanBEATS (Bach et al., 2013), WSCT (Zhang et al., 2020), etc.) each representing less than 2% of the total number of publications. SWMM, MIKE URBAN, and L-THIA-LID were the only software capable of modeling SUDS trains (see Fig. 5(a)). SWMM and the owndeveloped models were used in similar proportions to address all questions, while the other softwares were used to tackle more targeted questions. The most basic approaches used in own-developed models consisted of water balance calculations, statistical analyses, or GIS-based tools (Wang and Wang, 2018;Zhen et al., 2006;Lee et al., 2010;Yang and Best, 2015). Others opted for numeric algorithms to solve hydrological and hydraulic differential equations (e.g., Beck et al., 2017;Perez-Pedini et al., 2005;Gülbaz and Kazezyılmaz-Alhan, 2017;Wright et al., 2018). Finally, some studies relied on linear programming to embed the hydrological equations in the optimization model. This last approach requires a simplified representation of the processes occurring in the SUDS structure. For example, Sebti et al. (2016) estimated SUDS impact using a variant of the Improved Rational Hydrograph method (IRH), and Chang et al. (2011) and Torres et al. (2020) used water balance equations. Regardless of the complexity in these studies, it was identified that the necessity of developing own models and tools is driven by the lack of flexibility of the available software, in particular when data is limited and its format incompatible with the required inputs. Regarding the type of temporal simulation, 58% of the studies used event-based simulations, 28% used continuous simulation, and 14% performed a comparison analysis using both approximations. Fig. 5 (b) shows the proportion of continuous/event-based approaches for each spatial scale; as the spatial scale decreases, the proportion of studies performing continuous simulation grows. Disregarding whether event-based or continuous, 88% of the studies appraised used a deterministic approach, 9% a stochastic approach, 3% did a comparative analysis of both approaches. These results evidence that the computational resources needed to perform timeand resource-consuming simulations (continuous and stochastic approaches) are available in seldom cases. Fig. 5(c) and (d) presents box-plots with the number of events (event-based) or the number of years (continuous-simulation) analyzed for each DSS question addressed. For event-based, the number of events ranged between 1 and 10, with outliers up to 20 and a maximum value of 53. It was found that 56% of the studies used design rainfall events with return periods (ranging from 5 to 50 years), 37% used representative historical rainfall events, 6% employed synthetically generated events, and only 3% based their analyses on forecasted events. On the other hand, Fig. 5(d) shows that the majority of the studies using continuous-simulation analyzed from 1 to 25 years. Fig. 5(e) presents a stacked bar plot differentiating the type of simulation performed per question addressed. The “Design tools”used continuous simulation in a higher proportion than the rest of the categories, with 46% of the studies using a continuous simulation. The myriad of modeling settings and precipitation events selection evidence that there is still no consensus on good practices for modeling-based SUDS decision-making. 4.5. Categories, processes modeled, and stakeholders Table 2 shows that 82% of the studies included water quantity, 53% water quality, 28% economic analysis, and only 8% and 3% included environmental and social benefits. Within the environmental aspects, the most common are air quality and energy savings (e.g., Chang et al., 2011) and in the social aspect the recreation, acceptability, and amenity (e.g., Jia et al., 2012). Table 3 shows the processes more frequently modeled in the quantity and quality dimensions besides the rainfallrunoff process (which was included in all the articles). Not surprisingly, the processes more frequently modeled are infiltration and evapotranspiration since these two are responsible for the runoff volume reduction and the peak flow flattening. Of secondary importance were the groundwater flow and sedimentation processes. The first is frequently neglected, despite its proven importance in SUDS efficiency (Zhen et al., 2004;Xu et al., 2020b), because of the lack of local data, while the latter is less included in proportion to other processes given that it is irrelevant in runoff quantity assessments. The majority of the SUDS-DSS appraised had scenario modeling (65%) (i.e., comparing SUDS spatial configurations using performance metrics). However, most of the studies did not consider climate change nor urbanization trends projections (only 12% had this capability). It is highly recommended to re-direct efforts to include future projections in SUDS-DSS since it is expected that climate change and urbanization rates will play a major role in future urban hydrology, particularly in large urban centers (Xu et al., 2020a;Saldarriaga et al., 2020). Previous works identified that a key aspect to guarantee a successful decision-making process is the early inclusion of stakeholders perspectives and preferences (e.g., Jayasooriya et al., 2020;Ahiablame et al., 2012;Torres et al., 2020). However, it was found that 87% of the studies do not include stakeholders. Table 4 shows that from the 13% (16 articles out of the 120 reviewed) that did consider one or several stakeholders, the most common are local authorities (31%), utilities (13%), neighbors (13%), politicians (6%), and Environmental Agencies (EA) (6%); 31% of the articles include at least one stakeholder, but do not state which one. From the papers reviewed, none considered the opinion/preferences of the community members, who ultimately are impacted by the decisions. It is highly recommendable that the stakeholders' positions are included for decision-making, particularly the communities. 4.6. DSS inputs, outputs and general framework 4.6.1. Inputs and outputs This subsection is dedicated to the articles in category 2, DSS relying on a SUDS model and including additional dimensions besides the hydrological. A total of 83 articles lie in this category, compiling 4 types of input variables: hydro-meteorological (e.g., precipitation, runoff, temperature or evapotranspiration), study-site (e.g., land uses, impermeability, slope, infiltration rate, presence of conventional drainage systems), water quality (e.g., loads of pollutants like nutrients, organic matter, solids, and heavy metals), and economic (e.g., SUDS and land costs and monetary quantification of environmental services). The most frequent input variable to SUDS-DSS are the hydrometeorological variables, with a percentage of inclusion between 40 and 60%. The study site features also presented high percentages of inclusion (30-50%). The runoff quality (build-up and wash-off P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 8 parameters) variables had a lower frequency: 25%, 13%, 12% and 2% for total suspended solids, total phosphorus, total nitrogen, and heavy metals, respectively. Finally, among the economic input variables, SUDS costs were included in 40% of the studies, but other economic indicators (e.g., economic return, net present value, etc.) exhibit percentages of inclusion lower than 2%. Regarding the evolution over time of the input variables, it was noticed that from 2000 to 2010 the most frequently included category was the study site aspects, with some reduced usage of economic indicators. From 2010 forward, the water quality inputs started to be included, showing an increasing trend over the years. The economic aspects also showed an increasing trend over time. The hydro-meteorological variables showed a steady trend of inclusion over time. These trends reflect both the increased data availability and the driving interest on SUDS from a more holistic perspective. Design How Many Trains Where Which Years of the Data Base 90 75 60 45 30 15 0 City Catchment Subcatchment Neighborhood Household Software 0.0 0.2 0.4 0.6 0.8 1.0 Fraction (a) Queson addressed among the different soware (b) Temporal simulaon among the different spaal scales 50 40 10 1 30 20 10 10 0 Design Trains Which Where How Many Question Addressed 0 Where How Many Trains Which Design Question Addressed (c) Number of events simulated for (d) Years simulated for each quesons each quesons addressed addressed. Design How Many Trains Where Which 0.0 0.2 0.4 0.6 0.8 1.0 Fraction (e) Type of simulaon for the different quesons addressed. Continuos Event Based Count Number of Events AnnAGNPS CANOE Software Comparison GIFMod GIS-SWMM GSSHA L-THIA-LID MIKE URBAN MODFLOW MUSIC Own-developed PCSWMM SEWSYS SUDSLOC SUSTAIN SWMM SewerGEMS UrbanBEATS WABILA Water Sensitive Cities Toolkit WetSpa-Urban WinSLAMM Question Addressed Spatial Scale Continuos Event Based Fig. 5. Software, simulation, and temporal resolution for each question addressed. P. Ferrans, M.N. Torres, J. Temprano et al. Science of the Total Environment 806 (2022) 150447 9