- Email: [email protected]
Seeking urbanization security and sustainability: Multi-objective optimization of rainwater harvesting systems in China
Accepted Manuscript Research papers Seeking Urbanization Security and Sustainability: Multi-objective Optimization of Rainwater Harvesting Systems in China Yi Li, Quanliang Ye, An Liu, Fangang Meng, Wenlong Zhang, Wei Xiong, Peifang Wang, Chao Wang PII: DOI: Reference:
S0022-1694(17)30263-9 http://dx.doi.org/10.1016/j.jhydrol.2017.04.042 HYDROL 21973
To appear in:
Journal of Hydrology
Received Date: Revised Date: Accepted Date:
17 January 2017 16 April 2017 18 April 2017
Please cite this article as: Li, Y., Ye, Q., Liu, A., Meng, F., Zhang, W., Xiong, W., Wang, P., Wang, C., Seeking Urbanization Security and Sustainability: Multi-objective Optimization of Rainwater Harvesting Systems in China, Journal of Hydrology (2017), doi: http://dx.doi.org/10.1016/j.jhydrol.2017.04.042
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Seeking Urbanization Security and Sustainability: Multi-objective Optimization of Rainwater Harvesting Systems in China Yi Li a *, Quanliang Ye a, An Liu b, Fangang Mengc, Wenlong Zhang a, Wei Xiong a, Peifang Wang a, Chao Wang a a
Key Laboratory of Integrated Regulation and Resource Development on Shallow Lakes, Ministry of Education, College of Environment, Hohai University, Nanjing 210098, P.R. China
b
College of Chemistry and Environmental Engineering, Shenzhen University, Shenzhen 518060, P.R. China c
SYSU-HKUST Research Center for Innovative Environmental Technology
(SHRCIET), School of Environmental Science and Engineering, Sun Yat-sen University, Guangzhou 510275, PR China
*Corresponding author:
Dr. Yi Li
College of Environment, Hohai University, Nanjing 210098, P.R. China
Email: [email protected]
1
Abstract
Urban rainwater management need to achieve an optimal compromise among water resource augmentation, water loggings alleviation, economic investment and pollutants reduction. Rainwater harvesting (RWH) systems, such as green rooftops, porous pavements, and green lands, have been successfully implemented as viable approaches to alleviate water-logging disasters and water scarcity problems caused by rapid urbanization. However, there is limited guidance to determine the construction areas of RWH systems, especially for stormwater runoff control due to increasing extreme precipitation. This study firstly developed a multi-objective model to optimize the construction areas of green rooftops, porous pavements and green lands, considering the trade-offs among 24 hour-interval RWH volume, stormwater runoff volume control ratio (R), economic cost, and rainfall runoff pollutant reduction. Pareto fronts of RWH system areas for 31 provinces of China were obtained through nondominated sorting genetic algorithm. On the national level, the control strategies for the construction rate (the ratio between the area of single RWH system and the total areas of RWH systems) of green rooftops (ηGR ), porous pavements ( η PP ) and green lands (ηGL ) were 12%, 26% and 62%, and the corresponding RWH volume and total suspended solids reduction was 14.84 billion m3 and 228.19 kilotons, respectively. Optimal ηGR , η PP and ηGL in different regions varied from 1-33%, 654%, and 30-89%, respectively. Particularly, green lands were the most important RWH system in 25 provinces with ηGL more than 50%, ηGR mainly less than 15%, and η PP mainly between 10-30%. Results also indicated whether considering the 2
objective MaxR made a non-significant difference for RWH system areas whereas exerted a great influence on the result of stormwater runoff control. Maximum daily rainfall under control increased, exceeding 200% after the construction of the optimal RWH system compared with that before construction. Optimal RWH system areas presented a general picture for urban development policy makers in China.
Keyword: Rainwater harvesting; Multi-objective optimization; Stormwater runoff control; Water logging; Water scarcity
3
1. Introduction
Rapid urbanization has caused environmental problems by negatively affecting the hydrological cycles of urban water systems, which would exacerbate in the future (Roesner et al. 2001, Cadavid et al. 2013). Roads and buildings reduced pervious surface area, which limited natural water infiltration and increased rainfall runoff during rainstorm events. Excess rainwater deteriorated the quality and changed the flow patterns of the runoff leading to severe non-point source pollutions and effects on the ecosystem health of receiving water bodies (Damodaram et al. 2013, Moglia et al. 2016a, Montaseri et al. 2015). On the other hand, urbanization concentrated the population and other social resources in urban areas. This transformation led to an increasing demand of fresh water, and thus exacerbated the scarcity of water resources (Ghimire et al. 2014, Thomas et al. 2014). In this situation, rainwater has grown in popularity as an accessible water supply source (Li et al. 2000). Small onsite rainwater harvesting (RWH) technologies, such as green rooftops, porous pavements, and bioretention or green lands, have been successfully implemented as viable approaches to harvest rainwater in most regions (Melville-Shreeve et al. 2016, Moglia et al 2016b, Palla et al. 2011, Walsh et al. 2012). Moreover, RWH systems were intended to preserve and reinstate the pre-developed condition of urban lands. For this purpose, a number of approaches integrated RWH existed in developed and developing countries, such as Water Sensitive Urban Design (WSUD) in Australia (Coombes et al. 1999, Coutts et al. 2012); Best Management Practices (BMPs) and Low Impact Development (LID) in USA (Bhaskar et al. 2016, DER 1999); the 4
Building Research Establishment Environmental Assessment Method (BREEAM) and Sustainable Urban Drainage Systems (SuDS) in the UK (Andoh and Iwugo 2004, BREEAM 2011); the Sponge City Construction and Development in China (China Housing and Urban-Rural Deveolopment 2014).
RWH optimization techniques and evaluation models have been used extensively to design and optimize RWH systems. Silva et al. (2015) evaluated the technical feasibility and economic viability of domestic RWH systems in Porto and Almada, Portugal. Liang et al. (2011) analyzed the economic and financial performance of RWH systems in rural areas of Beijing using cost benefit analysis. Chen and Admas (2007) developed an urban stormwater quality model to evaluate rainwater pollutant buildup and washoff processes. On the other hand, stormwater runoff control has been a new challenge for urban rainwater management. Significant reduction of municipal peak flow may be provided by implementing RWH systems in urban areas (Jensen et al. 2010). Zhang et al. (2012) calculated the potential of collectable rainwater and the possibility of runoff volume reduction. Liu et al. (2015a) used a process-based stormwater runoff model to evaluate the runoff reduction effectiveness under various setting sizes of green infrastructure in Beijing, China. Sample and Liu (2014) used a rainwater simulation model to evaluate decentralized RWH systems in terms of water supply and runoff capture reliability across a wide range of land uses and locations in Virginia, USA.
Most of the previous studies either showed the contribution of RWH in water resource supplementation or in stormwater runoff control through some particular 5
criterions (such as storage size, water saving efficiency, economic analysis and environmental impacts) of interest. In spite of the achievements listed above, there were only a few studies integrating the two ultimate rainwater management components (i.e. RWH for water supplement and stormwate runoff control). Did it mean that rainwater harvesting objectives of water supplement and stormwater runoff control conflicted? To answer this question, many management efforts were requested throughout a watershed. However, there was no single RWH system that could be effective for the objectives in all regions (Lee et al. 2012). The characteristics of rainfall, economic development level, and pollution situations varied among different regions. The major problem was to select the best combination of RWH systems among different options available that obtained a practical, efficient, and sustainable strategy for the regions of different objectives, such as water resource augmenting, stormwater runoff control, economic feasibility and pollutants reduction. Thus, modeling methods and tools, like multi-objective optimization model, were needed to support the selection and assessment of feasible RWH systems, which could determine the construction area of RWH systems that achieved urbanization security and sustainability.
Multi-objective optimization model has been widely applied in solving water resource problems (Alzraiee et al. 2013, Arad et al. 2013, Reddy and Kumar 2007, Vasan and Raju 2006). Non-dominated Sorting Genetic Algorithm II (NSGA II) is one of the most efficient, multi-objective, evolutionary algorithms using the elitist approach (Kalyanmoy Deb 2002) and has gained popularity in finding the near 6
optimal solutions for optimization problems (A.Vasan and Simonovic 2010). Van Meter et al. (2014) recommended a holistic watershed-scale approach that accounts for trade-offs among water availability and socioeconomic wellbeing to explore the social, economic, and environmental dimensions of agricultural RWH ponds in India, and evaluate the viability of these centuries-old systems under current climate and population pressures. Zhang et al. (2014) developed a systematic framework with a multi-objective optimization model based on NSGA II considering the trade-off among wastewater reuse supplies and demands, costs and profits, as well as pollutants reduction. However, so far, no studies of multi-objective optimization model with evolutionary techniques have been available in RWH systems construction. Therefore, this study developed a multi-objective model for optimizing the construction areas of green rooftops, porous pavements, and green lands, considering the trade-offs among daily RWH volume, stormwater runoff control, economic cost, and pollutant reduction. Two more three-objective optimization models (one without daily RWH volume and another without stormwater runoff control) were built to compare the optimal results of three models. The optimal solution was obtained through NSGA II and selected on the basis of water resource situations and rainfall characteristics of each region. It was difficult to compare RWH system construction areas among regions directly due to different regional cover areas. Thus, an index called RWH system construction rate η k (the ratio between the area of single RWH system and the total areas of RWH systems) was defined to express three RWH system construction areas in different regions. The in-depth analysis of the results for 7
31 provinces could provide scientific instructions for the regional stakeholders and policy makers to design more feasible RWH systems and thus to improve urbanization security and sustainability of cities.
8
2. Material and Method 2.1 Study Areas The study areas are 31 provinces of China (except Hong Kong and Macao Special Administrative Regions and Taiwan). Particularly, Beijing and Wuhan cities are selected as case study to expound the optimization process (Fig. 1).
Beijing is located with latitude 39.4°-41.6° north and longitude 115.7°-117.4° east. It has a typical semi-humid monsoonal climate and the average annual precipitation is 532.1 mm (1981-2010). Beijing is suffering from severe water scarcity. The current available water resource per capital is only 94.9 m3 per year (Beijing Municipal Bureau of Statistics 2015), which is approximately 1/30 of the national average and 1/80 of the world average. In addition, the water pollution status has been increasing with the development of Beijing City. Therefore, water resource augmentation and rainfall runoff pollutant reduction have been the main objectives of RWH in Beijing (Beijing Municipal Planning Commission 2013).
Wuhan situates with latitude 29.9°-31.4° north and longitude 113.7°-115.0° east of Hubei province, and has a subtropical monsoon climate. Average annual precipitation in Wuhan is 1316 mm (1981-2010), with 2/3 (866.3 mm) of the total precipitation from May to August, which could easily cause water-loggings in flood season. Based on these austere situations, a large number of RWH projects have been carrying out in Wuhan to alleviate water-logging problems. In August 2015, the government published “Sponge City Planning and Design Guideline” (SCPDG) to instruct and improve RWH system construction. The document stated that the main 9
objective of RWH in Wuhan was stormwater runoff control (Wuhan Municipal Water Affairs Bureau 2015). Through the effort made by the local government as well as the support from the central government, Wuhan becomes the forerunner in the field of RWH system construction. The construction cases and experiences of RWH systems in Wuhan are regarded as references and instructions for other cities.
2.2 The Framework This study established a framework for optimizing the construction areas of RWH systems using a multi-objective optimization model based on the nondominated sorting genetic algorithm II (NSGA II) under physical constraints. The flowchart of the calculation process is presented in Figure 2. Three representative urban RWH systems, i.e. green rooftops, porous pavements and concave green lands, were selected as research objects, which were recommended in previous studies and Sponge Cities Construction Technology Guides (China Housing and Urban-Rural Deveolopment 2014, Wang et al. 2013). In fact, RWH systems involve the storage and distribution of rainwater besides collection, they are not considered in the present study since the process of rainwater collection is the focus of this study and the applicability of rainwater is rather limited to ecosystem utilization and environment consumption in China. 2.3 Objective Function The multi-objective optimization of RWH systems consisted of four objectives: maximization of the harvesting volume of 24 hour-interval rainwater; maximization of the stormwater runoff volume control ratio; minimization of the RWH system cost; 10
and maximization of pollutant reduction in rainfall runoff. RWH volume included two categories: 24 hour-interval RWH volume (QR, million m3) and RWH volume of maximum daily rainfall event (QH, million m3). The development of a model of RWH systems should be on a continuous daily rainfall time series for a minimum of 3 years and preferably 5 years (Lash et al. 2014). Therefore, average daily rainfall data of 30 years (1981-2010) was used to calculate QR, and maximum daily rainfall data (1981-2010) was used to calculate QH. The RWH volume from green rooftops was calculated by the infiltration processes (Wang et al. 2013). The calculation can be expressed as follows:
Pi ⋅ A1; Q1 = SGR ⋅ A1 ;
Pi < S GR
(1)
Pi ≥ SGR
where Q1 (mm·km2) was the RWH volume from green rooftops. Pi was the rainfall (mm), i was equal to 1 or 2, P1 was the 24 hour-interval average daily rainfall (19812010), and P2 was the volume of daily maximum rainfall (1981-2010). A1 (km2) was the construction area of green rooftops. The total storage of green rooftops (S GR, mm) was the infiltration into soil (FGR, mm). Infiltration has been a focus of agriculture and water research for a long period (Milla and Kish 2006). A number of mathematical models have been developed to evaluate the computation of infiltration which could be classified into physically based models and semi-empirical and empirical models. The semi-empirical and empirical models such as Horton models (Horton 1938), always in the form of simple
11
equations, were usually derived from either field or laboratory experiment data. The equation formulated as: f = f e + ( f 0 − f e )e − kt
where f e was the steady state value of f , f 0 was the value of f at t=0, and k was the infiltration decay factor. However, the semi-empirical and empirical models could not provide the detailed information of infiltration process and their physical meaning was not robust. Compared to the semi-empirical and empirical models, the physically based modes could substantially describe the detailed infiltration process, such as Green and Ampt model (Green and Ampt 1911):
B( H c + H ) f = A 1 + F where A and B were parameters which depended on soil characteristics, H c was the capillary potential at the wetting front (L), H was the head of water on the surface (L), and F was the cumulative infiltration. In this study, F was calculated referred from Green and Ampt model (Mein and Lerson 1973): F =K st + S f ∆θ ln(1 + F S f ∆θ )
(2)
where KS was the saturated hydraulic conductivity (mm/min), t was the time intervals of rainfall (min). Sf was suction at wetting front (mm), and ∆θ was the soil water deficit (unitless), which was the difference between saturated water content and initial water content. F0 was initial cumulative depth of infiltration (mm).
12
Similarly, the RWH volume from porous pavements was calculated by the infiltration processes. The calculation can be expressed as follows: Pi ⋅ A2 ; Q2 = S PP ⋅ A2 ;
Pi < S PP
(3)
Pi ≥ S PP
where Q2 (mm·km2) was the RWH volume from porous pavements. A2 (km2) was the construction area of porous pavements. The total storage of porous pavements ( S PP , mm) was equal to the infiltration into soil (the FGR). The volume of RWH from green lands was calculated by the interception and infiltration processes and the depression storage (Liu et al. 2015b, Wang et al. 2013). The calculation can be expressed as follows:
Pi < SGL Pi ⋅ A3 ; Q3 = SGL ⋅ A3 ; Pi ≥ SGL
(4)
where Q3 (mm·km2) was the volume of RWH from green lands. A3 (km2) was the construction area of green lands, and the total storage of green lands ( SGL , mm) was the sum of the interception by the infiltration into soil (FGR), the vegetation canopy (L, mm) and the depression storage of green lands (D, mm). L and D were calculated as Equations (5) and (6). (5)
L = S L × LAI
PC D = Sd max 1 − exp − Sd max
(6)
where S L denoted specific leaf storage (mm), and LAI was canopy leaf area index. Sd max was depression storage capacity of the green lands (mm), and PC was 13
accumulated residual rainfall (mm), which represented rainfall minus interception and infiltration.
Accordingly, QR and QH can be expressed as: 3
Max QR = ∑ Qk × 10 −3 ;
(7)
where i = 1
k =1
3
QH = ∑ Qk × 10 −3 ;
(8)
where i = 2
k =1
The stormwater runoff volume control ratio (R, %) was determined to describe the control efficiency of stormwater runoff by the following function:
Max R =
100QH W
(9)
where W (million m3) was the potential total runoff volume. The function of the third objective, i.e. the economic cost ( EC , billion RMB Yuan), was shown in Equation 10: 3
m
k =1
t =1
Min EC = ∑ ( EC k Ak × 10 −3 + ∑
Okt ) (1 + η ) t
(10)
where ECk (RMB Yuan/m2) was the unit construction cost of each RWH system. Okt (billion RMB Yuan) was the operation and maintenance cost occurring in year t of each RWH system, η was the discounting rate and m was the evaluation period (number of years). According to the publication Chinese Economic Evaluation Parameters on Construction (2006), the discount rate (η ) used in China was 8%. The evaluation period (m) was assumed to be 10 years (Liang et al. 2011). 14
For the last objective, pollutant reduction (PR, Tons), previous researches showed that total suspended solids (TSS) and chemical oxygen demand (COD) were the main pollutants in urban runoff (Huang et al. 2007, Vasan and Raju 2006). Taking Beijing as an example, two years of rainwater runoff samples were collected in 10 rain events from 2013 to 2015. The main characteristics and pollutant loads (TSS, COD, total nitrogen (TN) and total phosphorus (TP)) in the monitored rain events were summarized in Table S1 and Table S2, respectively, and the correlations between TSS and COD, TN, TP were illustrated in Fig. S1. Different degrees of correlation could be noticed between TSS and COD, TN, TP. Particularly, there was a high degree correlation between TSS and COD (R2=0.8579). Furthermore, the concentration of TSS in runoffs was more than 10 times higher than that of natural rainwater (Zhang et al. 2010). Therefore, TSS was chosen as the representative pollutant index and pollutants in natural rainwater were ignored. The function of pollutant reduction was shown in Equation 11: 3
Max PR = ∑ Qk ⋅ψ k ⋅ CTSS k
(11)
k =1
where ψ k was the runoff coefficient of different underlying surfaces and CTSS k (mg/l) was the concentration of TSS of different underlying surfaces.
2.4 RWH System Construction Rate The utilization of RWH systems would be an efficient measure for water supplement and water logging alleviation. The rational construction areas of RWH systems were the essential factor for sustainable rainwater management. However, it was difficult to 15
compare construction areas of RHW systems directly among regions due to different regional cover areas. Thus, an index, RWH system construction rate η k , was defined to express three RWH system construction areas in different regions.
ηGR =
η pp =
ηGL =
AGR AGR + APP + AGL
(12)
App
(13)
AGR + APP + AGL AGL AGR + APP + AGL
(14)
where AGR, APP, and AGL were the optimal construction area of green rooftops, porous pavements and green lands, respectively. 2.5 Three Multi-objective Optimization Scenarios Three multi-objective optimization model scenarios were utilized to optimize the RWH systems:
Scenario A: comprehensively considering the four objectives (i.e. Max QR, Max R, Min EC, and Max PR expressed above) to establish a four-objective optimization model for the purposes of sustainable stormwater management and daily water resources augmentation;
Scenario B: considering Max R, Min EC, and Max PR to establish a three-objective optimization model, which focused on stormwater runoff control;
16
Scenario C: considering Max QR, Min EC, and Max PR to establish another threeobjective optimization model, which concentrated on water resource augmentation.
2.6 Physical Constraints The major constraint for the first two objectives was the precipitation volume, while the objective of pollutants reduction depended on RWH volume. The precipitation volume in this study was an input constant. The precipitation data is available on: http://data.cma.cn/site/index.html for a 30-year period (1981-2010). In addition, the construction area of each RWH system played an important role in the calculation of this model. According to the Engineering technical code for rain utilization in building and sub-district (GB 50400-2006) and Ministry of Housing and Urban-Rural Development of each province, the construction area of green rooftops should be less than 20%-35% of the gross building rooftops under construction and more than existing green rooftops. The construction area of porous pavements should be less than 20%-30% of the urban trunk road and more than existing porous pavements referred from the Ministry of Transport of each province. The construction of the green lands should be less than the urban green lands and park construction planning, and more than the existing green lands area according to China Housing and UrbanRural Development and the Thirteen Five-Year Plan.
2.7 Multi-objective Optimization through NSGA-II NSGA II is an effective method for multiobjective optimization developed by Deb et al (Kalyanmoy Deb 2002). In NSGA II, the concept of Pareto-dominance is used to
17
rank the individuals (control strategies) of a population. As shown in Figure 2, the implementing process of NSGA II follows the following pseudocodes:
Randomly initialize the population P (0) of size N;
Fast nondomination sorting on P (0);
For every generation t;
Select a parent population Pp (t) from P(t) using a binary tournament selection;
Create a child population Pc (t) from Pp (t) through crossover and mutation operators;
Combine P (t) and Pc (t) into an intermediate population Pi(t);
Fast nondomination sorting on Pi (t);
Place the best N individuals from Pi (t) to P (t+1);
End loop
2.8 Data Overall, 31 provinces in mainland China were included in this study, excluding Hong Kong, and Macao Special Administrative Regions, and Taiwan. The meteorological data (i.e. precipitation and temperature data) of each region is available on: http://data.cma.cn/site/index.html for a 30-year period (1981-2010). Supplementary data, including EC1, EC2, EC3, AP, AG, ψ 1 , ψ 2 , and ψ 3 , were obtained from statistical reports (Beijing Statistical Yearbook, Wuhan Municipal Bureau of Statistic, Sponge 18
Cities Construction Technology Guides, Engineering technical code for rain utilization in building and sub-district, Ministry of Water Resources, Ministry of Housing and Urban-Rural Development and National Bureau of Statistics of China). Pollutant concentration data were obtained from rainfall events in Beijing and Wuhan. Other model parameters were taken from previous studies (Table S3). The study sites parameters were listed in Tables S4-S9.
19
3. Results and Discussion
3.1 RWH Systems of Beijing City Figure 3a and Table S11 present the Pareto solutions for RWH systems construction in Beijing. Figure 3a is a three-dimensional representation of the four objectives, where QR, R and PR are on the three-dimensional axes and EC is represented by the different size balls on the Pareto optimal surface. A smaller size ball represents a lower EC value and a larger size ball for a higher value. From this Pareto front, there was a clear trade-off between the objectives EC and PR. Since Beijing has been suffering from both water scarcity and water pollution, QR and PR were considered as the external sets of criteria (Table S10). The solution located at the back top corner in Figure 3a had the largest QR (473.32 million m3) and PR (3.50 kilotons), but it also had the largest EC (85.13 billion RMB Yuan). Moreover, it was found that the PR of the red point (Fig. 3a) was 1.84% (36.16 Tones) less than that of the back top one, while the corresponding total EC of the red point was 6.57% (4.68 billion RMB Yuan) less than that of the back top one. It implied that the solution represented by the red point was more economically feasible than that of the back top one with the same pollutants reduction level. Thus, the solution represented by the red point was selected as the control strategy, with values for QR, R, EC and PR equaling 239.63 million m3, 76.7% million m3, 66.53 billion RMB Yuan and 1931.18 tones, respectively.
Comparing the total optimal RWH areas in this study and the data investigated in Beijing (Fig. 3b), it was noticed that the percentage of constructed green rooftops was less 20%, while the percentage of porous pavements and green lands were both above 20
70%. The results provided the phenomenon that green rooftops were rarely chosen as rooftop for buildings due to the complex technologies and costly maintenance, which were requested in “Technical Code for Roof Engineering” published by the Ministry of Housing and Urban-Rural Development in China. In addition, the defects in construction technology and the single selection of covering vegetation made green rooftops develop leaking problems. In the U.S., mature technology and advanced processes guaranteed the practicality and reliability of green rooftops construction (Ghimire et al. 2014, Mun and Han 2012). Therefore, the green rooftops construction technology in Beijing should be improved to solve leaking problems, and select a more proper vegetation to form a practical and sustainable RWH system. The construction percentage of porous pavements was the largest of the three RWH systems (78.2%). The effect for runoff control of porous pavements was obvious through infiltration for soil storage or groundwater recharge. Moreover, porous pavements demonstrated the highest pollutant reduction of the three alternatives (Table S4).
The different objectives of RWH systems exerted a great influence on RWH system areas and the corresponding results (Fig. 4). When only considering rainwater as a water supplement source (Scenario C), QR in Scenario C could reach 218.87 million m3 per year. Focusing RWH systems on stormwater runoff control (Scenario B), the optimal R was 78.6% provided by Scenario B. Comparing these two scenarios, PR in Scenario C was approximately 5 times of that in Scenario B. The environmental impacts of stormwater runoff depended heavily on the frequency and intensity of 21
stormwater events. The frequency (1.7 times, Table S9) and intensity of stormwater events in Beijing were both low, thus leading to a low pollutants reduction. In addition, green lands were preferred for stormwater runoff control due to its large water storage contribution. However, the economic cost was a considerable factor for green lands, which reflected in Scenario B that the value of EC was 36.7% more than that in Scenario C. In order to achieve water resource augmentation and water-logging alleviation simultaneously (Scenario A), the QR in Scenario A was 9.5% more than that in Scenario C while the R was at the same level with that in Scenario B. It suggested that the optimal RWH construction areas in Scenario A could obtain a better water saving effect without influencing the efficiency of stormwater runoff control. Furthermore, the PR in Scenario A was the highest of the three scenarios. The noticeable deficiency of Scenario A was the huge economic cost. However, with current economic and societal development in China, economic investment would not be the main limiting force considered with the security and environmental friendliness of urbanization progress.
3.2 RWH Systems of Wuhan City Figure 5a and Table S12 present the Pareto solutions for RWH system construction in Wuhan. R and EC were considered as the external sets of criteria because Wuhan has severe water logging disasters in flood season which cause enormous economic losses. Comprehensively considering the trade-off between R and EC (the same selected standards for Beijing), the solution represented by the red point was selected as the control strategy (with QR 230.31 million m3 , R 51.03%, EC 45.08 billion RMB 22
Yuan, and PR 2303.27 tones). Moreover, comparing the total optimal RWH system areas in this study and RWH system areas in Wuhan (Fig. 5b), the percentage of constructed green rooftops was extremely low while that of constructed porous pavements was the highest. These were consistent with the situations in Beijing. Since Wuhan has been suffering from severe water logging problems, R value was a critical variable for the design and development of local RWH systems. Taking RWH systems into account for stormwater runoff control in Scenario B (Fig. 6), the optimal R was 52.57% under the average annual maximum daily rainfall (198.5 mm, 1980-2010). It indicated that the probability of water-loggings happening could decline more than 50% under daily rainfall less than 198.5 mm after the optimal RWH system of Scenario B were constructed. However, the corresponding EC reached 49.57 billion RMB Yuan. This was a high pressure for public expenditure of local or central government. Even though Wuhan does not have issues with quantityrelated water scarcity, the average annual precipitation (1316 mm) is a huge potential fresh-water source. Adding objective Max QR into the model (Scenario A), the optimal QR was 230.31 million m3. Meanwhile, the R value in Scenario A was only 1.54% less than that in Scenario B. This showed that water supply and stormwater runoff control were not two competing objectives; both of them could be achieved by the construction of optimal RWH system. Moreover, economic investment in Scenario A was 9.1% less than that in Scenario B. The reason could be explained that when RWH focusing on runoff control, the optimal green lands construction area was 8.73 km2 more than that in Scenario A thus extending the cost. Furthermore, PR in 23
Scenario A was much higher than that in Scenario B. Therefore, the optimal RWH areas in Scenario A gave consideration to both water logging alleviation and water resource augmentation, meanwhile it was a cost-effective and low-impact management strategy. 3.3 RWH Systems of China Potential stormwater abatement and water saving were found to strongly correlate to local annual precipitation quantities and patterns. The objectives of RWH systems were formulated according to water resource status and rainfall characteristics in different regions. However, it must be noted that water-logging disasters have been a nationwide urbanization problem in China (Zhang et al. 2014, Xia 2013). Therefore, stormwater runoff control was an essential concern of RWH. Through the multiobjective optimization model, Pareto solutions of RWH system for 31 provinces of China were obtained (data not shown). According to the external sets of criteria (Table S10), the control strategies of RWH system were selected and shown in Table S13. Figure 7 illustrates the spatial distributions of QR, R, EC and PR in China. On the national level, QR and PR were 14.84 billion m3 and 228.19 kilotons, respectively. The corresponding reduction of COD, TN, and TP were 113.87, 6.97, and 1.23 kilotons, respectively. In Figure 7a, the QR in China decreased from the southeast to the northwest because annual precipitation decreased from east to west. Guangdong province had the maximum QR (1341.92 million m3) according to its large daily rainfall and RWH system construction areas. The reducing trend of R was opposite to QR but for the same reason. High frequency and intensity of stormwater in southeast 24
China made a great pressure for RWH to control runoff volume. Meanwhile, the serious water-logging problems required a higher standard for RWH systems. It was not surprising that the great attention paid by the central government to accelerate RWH construction in eastern China. Major EC and TSS reductions were found to be concentrated in the east of China. There was also a decrease from the east to west in accordance with the reducing trend of GDP per capita and development level of cities. Geographical distributions of ηGR , η PP , and ηGL were presented in a general picture (Fig. 8). It was found that there were 25 provinces with ηGL more than 50% (50-89%), 25 provinces with η PP more than 15% (18%-54%), and 22 provinces with
ηGR less than 15% (1-15%). Green lands and porous pavements predominated for runoff control and water supplementation due to their large water storage volume. The
ηGR of provinces in southern China was larger than that in northern provinces. Furthermore, there was a decreasing trend of ηGR from northeast (3%-33%) to northwest (1%-6%, except 22% in Xinjiang). The major function of rooftops in northwestern China was drying crops and clothing rather than RWH. Traditional storage tanks would be a better choice for household water supplementation in these regions. In Southeast coastal provinces, η PP values ranged from 20-25% except for the Jiangsu province (54%). The government paid high attention to the construction of porous pavements in Jiangsu province. According to the requirement of “Jiangsu Province Sponge City Construction Guideline” (Jiangsu Municipal Water Affairs Bureau of Jiangsu Province 2015), the coverage ratio of porous pavements for new city trunk roads should be up to 50-60%. . Furthermore, η of three alternatives did 25
not show much variation in southeastern China due to same main purpose of waterloggings alleviation. Whether considering objective MaxR made a non-significant difference on the control strategies of RWH area whereas exerted a great influence on the corresponding results of the optimization model. Comparing the η k of three alternatives obtained in Scenario A and C (Fig. 9), it was noticed that 25 provinces (except for Hunan, Jilin, Jiangsu, Shandong, Tianjin, and Chongqing) had variation amplitude of ηGR less than 5%, other 25 provinces (except for Anhui, Guizhou, Jiangsu, Liaoning, Tianjin, and Zhejiang) had variation amplitude of η PP no more than 5%, and another 24 provinces (except for Anhui, Jilin, Jiangsu, Shandong, Shanghai, Tianjin and Chongqing) had variation amplitude of ηGL less than 10%. The variation amplitude of each η k in Tianjin ( ηGR : 7%; η PP : 11%; ηGL : 17%) and Jiangsu (ηGR : 7%; η PP : 12%; ηGL : 18%) were large due to relatively high stormwater frequency and intensity leading R as a critical variable. In addition, it was also found that the η PP value of 28 provinces (except for Gansu, Hunan and Guangxi) reduced while the ηGL value of 29 provinces (except for Gansu and Hunan) increased when optimization model with objective MaxR. Green lands became the first choice for RWH with an increasing ηGL in 94% of the provinces in Scenario A. However, a lower ηGL in Scenario A of Gansu province was because Gansu suffered from serious quantity-related water scarcity with the result that water resource in comparison to water logging problem was a historical limit for development. When water resource augmentation turned into major objective in Scenario C, ηGL showed a positive 26
change due to large storage volume and available area in Gansu. For Hunan province, the lower ηGL in Scenario A could be explained by the fact that fluvial landform areas in Hunan accounted for 64.76% of the total regional area. In addition, available land areas for green lands in Hunan was less than that of other neighboring provinces (such as Jiangxi, Guangdong, Hubei). Large fluvial landform areas greatly relieved the pressure for runoff control, while limited available construction areas for green lands caused the optimal construction area to always be at the maximum level. For all 31 provinces, maximum daily rainfall under control after the construction of optimal RWH system areas in Scenario A would increase by exceeding 200% more than that before construction (as shown in Eq. 8). Choosing Hubei province with highfrequency water-logging events as an example, daily rainfall volume under control increased from 43.3 mm before the optimal RWH systems constructed to 121.2 mm after construction. 3.4 Sensitivity Analysis Climate change is expected to change the volume and pattern of precipitation (IPCC 2014). In order for a RWH system to be adapted to climate change, it is argued that it should be able to perform at the same level under the future climate (Youn et al. 2012). Thus, sensitivity analysis was carried out to assess the robustness of the results in response to the climate variables such as temperature and precipitation (Fig. 10). It could be seen that the impact of AGR was noticeable in response to 20% variance of 24 hour-interval average daily precipitation, 20% variance of average daily maximum precipitation and 10% variance of average temperature. The AGR showed a positive 27
change at 20% variance of 24 hour-interval average daily precipitation (Fig. 10a), and in contrast, a negative change at 20% variance of average daily maximum precipitation (Fig. 10b) in most provinces. Particularly, the positive change on AGR was more than 50% in Anhui and Chongqing provinces, while the negative change more than 50% in Fujian and Ningxia provinces. The changes of AGL were opposite to AGR at the variance of precipitation. In spite of a weak stability for the results of AGR under different variances of precipitation, it would not influence the RWH systems greatly in China, since the proportion of green rooftops predominately ranged between 1-15% as discussed before. Besides, the change of APP and AGL were both less 20% at variance of average temperature (Fig. 10c). It was implied that the variance of temperature is not the domain factor for the performance of RWH systems in the future. 4. Conclusion
The present study firstly developed a systematic multi-objective optimization model to optimize construction areas of green rooftops, porous pavements and green lands, considering the trade-offs among 24 hour-interval RWH volume, stormwater runoff volume control ratio, construction cost, and rainfall runoff pollutants reduction, for the purpose of stormwater runoff control and water resource augmentation in China. Results showed that on the national level, the control strategies of construction rate of green rooftops (ηGR ), porous pavements (η PP ) and green lands (ηGL ) were 12%, 26% and 62% with corresponding QR and PR for 14.19 billion m3 and 234.79 kilotons, respectively. Optimal ηGR , η PP and ηGL in different regions varied from 1-33%, 628
54%, and 30-89%, respectively, owing to local water resource status and precipitation characteristics. Particularly, green lands was the most important RWH system in 25 provinces with ηGL more than 50% while ηGR was mainly less than 15%, and η PP was mainly between 10-30%. It was also found whether considering objective MaxR made a non-significant difference on RWH system areas whereas exerted a great influence on the result of stormwater runoff control. Maximum daily rainfall under control would increase by exceeding 200% after optimal RWH system areas construction compared with that before construction.
29
Associated Content
Supporting information One figure and thirteen tables are provided to present detailed information for this study. Author Information
Corresponding Author *Phone: 86-2583786251. Fax: 86-25-83786090. E-mail: [email protected] (Y.L.) Acknowledgment
This study was financially supported by the Foundation for Innovative Research Groups of the National Natural Science Foundation of China [No. 51421006], the National Natural Science Foundation of China [No. 91547105 and No. 51479066], the Foundation Research Funds for the Central Universities [No. 2016B10614], the Priority Academic Program Development of Jiangsu Higher Education Institutions, and the Top-Notch Academic Programs Project of Jiangsu Higher Education Institutions [No. PPZY2015A051].
30
Reference
A.Vasan and Simonovic, S.P., 2010. Optimization of Water Distribution Network Design Using Differential Evolution. Journal of Water Resources Planning and Management 136(2), 279-287. Alzraiee, A.H., Bau, D.A. et al., 2013. Multiobjective design of aquifer monitoring networks for optimal spatial prediction and geostatistical parameter estimation. Water Resources Research 49(6), 3670-3684. Andoh, R.Y.G. and Iwugo, K.O., 2004. Sustainable Urban Drainage Systems: A UK Perspective. Urban Drainage. Arad, J., Housh, M., Perelman, L. et al., 2013. A dynamic thresholds scheme for contaminant event detection in water distribution systems. Water Res 47(5), 1899-1908. Beijing Municipal Bureau of Statistics, 2015. Beijing Statistical Yearbook. China Statistics Press, Beijing, China (in Chinese). Beijing Municipal Planning Commission, 2013. Rainwater Control and Use Project Guideline. (in Chinese). Bhaskar, A.S., Hogan, D.M. and Archfield, S.A., 2016. Urban base flow with low impact development. Hydrological Processes 30(18), 3156-3171. BREEAM, 2011. http://www.breeam.org/page.jsp?id=181. (accessed 15 May 2013). Cadavid, Londoño Catalina Ando et al., 2013. Valuing preferences over stormwater management outcomes including improved hydrologic function. Water Resources Research 49(7), 4114-4125. 31
China Housing and Urban-Rural Deveolopment, 2014. Sponge Cities Construction Technology Guides. China (in Chinese). Coombes, P.J., Argue, J.R. et al., 1999. Figtree Place: a case study in water sensitive urban development. Urban Water Journal, 335-343. Coutts, A.M., Tapper, N.J., et al., 2012. Watering our cities: The capacity for Water Sensitive Urban Design to support urban cooling and improve human thermal comfort in the Australian context. Progress in Physical Geography 37(1), 2-28 Damodaram, Chandana Zechman et al., 2013. Simulation-Optimization Approach to Design Low Impact Development for Managing Peak Flow Alterations in Urbanizing Watersheds. Journal of Water Resources Planning and Management 139(3), 290-298. DER, 1999. Low-Impact Development Design Strategries: An Integrated Design Approach. DER, Largo, Maryland, USA. Ghimire, S.R., Johnston, J.M., et al., 2014. Life cycle assessment of domestic and agricultural rainwater harvesting systems. Environ Sci Technol 48(7), 40694077. Green, W.H. and Ampt, G.A., 1911. Studies on soil physics, I. Flow of air and water through soils. Journal of Agricultural Science 28, 1-11. Helmreich, B. and Horn, H., 2009. Opportunities in rainwater harvesting. Desalination 248(1-3), 118-124. Horton, R.E., 1938. The interpretation and application of runoff plot experiments with reference to soil erosion problems. Soil Science Society of America Proceeding, 32
340-349. Huang, J.-l., Du, P.-f., et al., 2007. Characterization of surface runoff from a subtropics urban catchment. Journal of Environmental Sciences 19(2), 148-152. IPCC, 2014. Climate Change 2014: Synthesis Report. Geneva, Switzerland. Jensen, M.A., Steffen, J., et al., 2010. Do rainwater harvesting objectives ofwater supply and stormwater management conflict? Low Impact Development. Jiangsu Municipal Water Affairs Bureau of Jiangsu Province, 2015. Jiangsu Province Sponge City Construction Guideline, China (in Chinese). Jieyun, C. and J., A.B., 2007. A derived probability distribution approach to stormwater quality modeling. Advances in Water Resources 30(1), 80-100. Kalyanmoy Deb, A.P., Sameer Agarwal, T. Meyarivan, 2002. A Fast and Elitist Multiobjective Genetic Algorithm NSGA-II. IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION 6, 182-197. Lash, D., Ward, S., et al., 2014. Robust rainwater harvesting: probabilistic tank sizing for climate change adaptation. Journal of Water and Climate Change 5(4), 526. Lee, J.G., Selvakumar, A., et al., 2012. A watershed-scale design optimization model for stormwater best management practices. Environmental Modelling & Software 37, 6-18. Li, F., Cook, S., et al., 2000. Rainwater harvesting agriculture: An integrated system for water management on rainfed land in China's semiarid areas. Ambio 29(8), 477-483. Liang, X., Van Dijk and Pieter, M., 2011. Economic and financial analysis on 33
rainwater harvesting for agricultural irrigation in the rural areas of Beijing. Resources, Conservation and Recycling 55(11), 1100-1108. Liu, W., Chen, W. and Peng, C., 2015a. Influences of setting sizes and combination of green infrastructures on community’s stormwater runoff reduction. Ecological Modelling 318, 236-244. Liu, W., Chen, W., et al., 2015b. A water balance approach to assess rainwater availability potential in urban areas: the case of Beijing, China. Water Science & Technology: Water Supply 15(3), 490. Mein, R.G. and Lerson, C.L., 1973. Modeling Infiltration during a Steady Rain. Water Resources Research Vol. 9, No. 2, 384-394. Melville-Shreeve, P., Ward, S. and Butler, D., 2016. Rainwater Harvesting Typologies for UK Houses: A Multi Criteria Analysis of System Configurations. Water 8(4), 129. Milla, K. and Kish, S., 2006. A low-cost microprocessor and infrared sensor system for automating water infiltration measurements. Computers and Electronics in Agriculture 53(2), 122-129. Moglia, M., Gan, K. and Delbridge, N., 2016a. Exploring methods to minimize the risk of mosquitoes in rainwater harvesting systems. Journal of Hydrology 543, 324-329. Moglia, M., Gan, K., et al., 2016b. Investigation of pump and pump switch failures in rainwater harvesting systems. Journal of Hydrology 538, 208-215. Montaseri, M., Hesami Afshar, et al., 2015. Development of Simulation-Optimization 34
Model (MUSIC-GA) for Urban Stormwater Management. Water Resources Management 29(13), 4649-4665. Mun, J.S. and Han, M.Y., 2012. Design and operational parameters of a rooftop rainwater harvesting system: definition, sensitivity and verification. J Environ Manage 93(1), 147-153. Palla, A., Gnecco, I. and Lanza, L.G., 2011. Non-dimensional design parameters and performance assessment of rainwater harvesting systems. Journal of Hydrology 401(1-2), 65-76. Reddy, M.J. and Kumar, D.N., 2007. Multiobjective Differential Evolution with Application to Reservoir System Optimization. Journal of computing in civil engneering 21(2), 136-146. Roesner Larry A. and Bledsoe, B.P., 2001. Are best-management-practice criteria really environmentally friendly? ASCE Journal of Water Resources Planning and Management 127, 150-154. Sample, D.J. and Liu, J., 2014. Optimizing rainwater harvesting systems for the dual purposes of water supply and runoff capture. Journal of Cleaner Production 75, 174-194. Silva, C.M., Sousa, V. et al., 2015. Evaluation of rainwater harvesting in Portugal: Application to single-family residences. Resources, Conservation and Recycling 94, 21-34. Thomas, R.B., Kirisits, M.J., et al., 2014. Rainwater harvesting in the United States: a survey of common system practices. Journal of Cleaner Production 75, 166-173. 35
Van Meter, K.J., Basu, N.B., et al., 2014. Monsoon harvests: the living legacies of rainwater harvesting systems in South India. Environ Sci Technol 48(8), 42174225. Vasan, A. and Raju, K.S., 2006. Application of Differential Evolution for Irrigation Planning: An Indian Case Study. Water Resources Management 21(8), 13931407. Walsh, C.J., Fletcher, T.D. and Burns, M.J., 2012. Urban Stormwater Runoff: A New Class of Environmental Flow Problem. Plos One. Wang, R., Eckelman, M.J. and Zimmerman, J.B., 2013. Consequential Environmental and Economic Life Cycle Assessment of Green and Gray Stormwater Infrastructures for Combined Sewer Systems. Environ Sci Technol 47(19), 11189-11198. Wuhan Municipal Water Affairs Bureau, 2015. Sponge City Planning and Design Guildeline of Wuhan City. Wuhan,China, (in Chinese). Xia, W., 2013. The Cause Analysis and Countermeasures of Waterlogging in Wuhan. Applied Mechanics and Materials 357-360, 1962-1965. Youn, S.-g., Chung, E.-S., et al., 2012. Probabilistic estimation of the storage capacity of a rainwater harvesting system considering climate change. Resources, Conservation and Recycling 65, 136-144. Zhang D.D., Yan D.H. and Y Wang., 2014. Research Progress on Risk Assessment and Integrated Strategies for Urban Pluvial Flooding. Journal of Catastrophology 29, 144-149. 36
Zhang, M., Chen, H., Wang, J. and Pan, G., 2010. Rainwater utilization and storm pollution control based on urban runoff characterization. Journal of Environmental Sciences 22(1), 40-46. Zhang, W., Wang, C., Li, Y., et al., 2014. Seeking sustainability: multiobjective evolutionary optimization for urban wastewater reuse in China. Environ Sci Technol 48(2), 1094-1102. Zhang, X., Hu, M., et al., 2012. Urban Rainwater Utilization and its Role in Mitigating Urban Waterlogging Problems—A Case Study in Nanjing, China. Water Resources Management 26(13), 3757-3766.
37
Fig. 1. - Marked The study areas and the geographic location of Beijing city and Wuhan city. 1: Heilongjiang, 2: Jilin, 3: Liaoning, 4: Hebei, 5: Beijing, 6: Tianjin, 7:Shandong, 8: Jiangsu, 9: Shanghai, 10: Zhejiang, 11: Anhui, 12: Jiangxi, 13: Fujian, 14: Guangdong, 15: Hainan, 16: Hunan, 17: Guangxi, 18: Yunnan, 19: Guizhou, 20: Sichuan, 21: Chongqing, 22: Hubei, 23: Henan, 24: Shanxi, 25: Inner Mongolia, 26: Shaanxi, 27: Gansu, 28: Ningxia, 29: Qinghai, 30: Xinjiang, 31: Tibet.
38
Fig. 2. - Marked Methodology Flowchart. QR was 24 hour-interval harvesting rainwater volume. R was stormwater runoff volume control ratio. EC was RWH systems construction cost. PR was total suspended solids reduction in rainfall runoff. P1 was the 24 hour-interval average daily rainfall (1981-2010). P2 was the rainfall of the daily maximum rainfall (1981-2010).
39
a
b
Fig. 3. – marked (a) Pareto front obtained by the nondominated sorting genetic algorithm using the four-objective optimization problems study for rainwater harvesting (RWH) systems areas in Beijing. (b) The construction areas of three selected RWH systems this study and the data investigated in Beijing.
40
Fig. 4. – Marked The 24 hour-interval RWH volume (QR), stormwater runoff volume control ratio (R), RWH systems construction cost (EC) and TSS reduction (PR) in Scenario A (Four-objective optimization model), Scenario B (Three-objective optimization model without QR) and Scenario C (Three-objective optimization model without R).
41
a
b
Fig. 5. – Marked (a) Pareto front obtained by the nondominated sorting genetic algorithm using the four-objective optimization problems study for RWH systems areas in Wuhan. (b) The construction areas of three selected RWH systems in this study and the data investigated in Wuhan. 42
Fig. 6. - Marked The 24 hour-interval RWH volume (QR), stormwater runoff volume control ratio (R), RWH systems construction cost (EC) and TSS reduction (PR) in Scenario A (Four-objective optimization model), Scenario B (Three-objective optimization model without QR).
43
a
b
million m3 0-200 200-400 400-600 600-1000 1000+
c
d
Tones
billion RMB Yuan
0-2000 2000-4000 4000-8000 8000+
0-50 50-80 80-100 100+
Fig. 7. - Marked Spatial distributional of (a) harvesting volume of 24 hour-interval rainwater, (b) stromwater control ratio, (c) economic cost of RWH systems construction, (d) TSS reduction in rainfall runoff in China.
44
Fig. 8. - Marked Geographical distribution of RWH systems construction rate and the average annual rainfall. Green rooftop (blue), porous pavements (red) and green lands (green). 45
Fig. 9. – Marked Comparison of the three RWH systems contruction areas obtained by the multi-objective optimization model with (left) and without (right) the objective MaxR, where the less ones are shown by the red dotted line box.
46
a.
b.
47
c.
Fig. 10. Results of the sensitivity analysis on the construction area of green rooftops (AGR), porous pavements (APP), and green lands (AGL), respectively, in 31 provinces for precipitation and temperature. (a) 20% variance of 24 hour-interval average daily precipitation; (b) 20% variance of average daily maximum precipitation; (c) 10% variance of average temperature. 1: Heilongjiang, 2: Jilin, 3: Liaoning, 4: Hebei, 5: Beijing, 6: Tianjin, 7:Shandong, 8: Jiangsu, 9: Shanghai, 10: Zhejiang, 11: Anhui, 12: Jiangxi, 13: Fujian, 14: Guangdong, 15: Hainan, 16: Hunan, 17: Guangxi, 18: Yunnan, 19: Guizhou, 20: Sichuan, 21: Chongqing, 22: Hubei, 23: Henan, 24: Shanxi, 25: Inner Mongolia, 26: Shaanxi, 27: Gansu, 28: Ningxia, 29: Qinghai, 30: Xinjiang, 31: Tibet.
Note for the change of Fig. 10.: The Fig. 10 was redrawn with a high resolution.
48
A four-objectives optimization method for RWH systems in China was
established. Green lands are dominant with the construction rate varying from 30~89% in
China. Green rooftops, porous pavements and green lands were comprehensively
considered. Daily rainfall under control increased by 2 times after optimal RWH construction.
49
50
S0022-1694(17)30263-9 http://dx.doi.org/10.1016/j.jhydrol.2017.04.042 HYDROL 21973
To appear in:
Journal of Hydrology
Received Date: Revised Date: Accepted Date:
17 January 2017 16 April 2017 18 April 2017
Please cite this article as: Li, Y., Ye, Q., Liu, A., Meng, F., Zhang, W., Xiong, W., Wang, P., Wang, C., Seeking Urbanization Security and Sustainability: Multi-objective Optimization of Rainwater Harvesting Systems in China, Journal of Hydrology (2017), doi: http://dx.doi.org/10.1016/j.jhydrol.2017.04.042
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Seeking Urbanization Security and Sustainability: Multi-objective Optimization of Rainwater Harvesting Systems in China Yi Li a *, Quanliang Ye a, An Liu b, Fangang Mengc, Wenlong Zhang a, Wei Xiong a, Peifang Wang a, Chao Wang a a
Key Laboratory of Integrated Regulation and Resource Development on Shallow Lakes, Ministry of Education, College of Environment, Hohai University, Nanjing 210098, P.R. China
b
College of Chemistry and Environmental Engineering, Shenzhen University, Shenzhen 518060, P.R. China c
SYSU-HKUST Research Center for Innovative Environmental Technology
(SHRCIET), School of Environmental Science and Engineering, Sun Yat-sen University, Guangzhou 510275, PR China
*Corresponding author:
Dr. Yi Li
College of Environment, Hohai University, Nanjing 210098, P.R. China
Email: [email protected]
1
Abstract
Urban rainwater management need to achieve an optimal compromise among water resource augmentation, water loggings alleviation, economic investment and pollutants reduction. Rainwater harvesting (RWH) systems, such as green rooftops, porous pavements, and green lands, have been successfully implemented as viable approaches to alleviate water-logging disasters and water scarcity problems caused by rapid urbanization. However, there is limited guidance to determine the construction areas of RWH systems, especially for stormwater runoff control due to increasing extreme precipitation. This study firstly developed a multi-objective model to optimize the construction areas of green rooftops, porous pavements and green lands, considering the trade-offs among 24 hour-interval RWH volume, stormwater runoff volume control ratio (R), economic cost, and rainfall runoff pollutant reduction. Pareto fronts of RWH system areas for 31 provinces of China were obtained through nondominated sorting genetic algorithm. On the national level, the control strategies for the construction rate (the ratio between the area of single RWH system and the total areas of RWH systems) of green rooftops (ηGR ), porous pavements ( η PP ) and green lands (ηGL ) were 12%, 26% and 62%, and the corresponding RWH volume and total suspended solids reduction was 14.84 billion m3 and 228.19 kilotons, respectively. Optimal ηGR , η PP and ηGL in different regions varied from 1-33%, 654%, and 30-89%, respectively. Particularly, green lands were the most important RWH system in 25 provinces with ηGL more than 50%, ηGR mainly less than 15%, and η PP mainly between 10-30%. Results also indicated whether considering the 2
objective MaxR made a non-significant difference for RWH system areas whereas exerted a great influence on the result of stormwater runoff control. Maximum daily rainfall under control increased, exceeding 200% after the construction of the optimal RWH system compared with that before construction. Optimal RWH system areas presented a general picture for urban development policy makers in China.
Keyword: Rainwater harvesting; Multi-objective optimization; Stormwater runoff control; Water logging; Water scarcity
3
1. Introduction
Rapid urbanization has caused environmental problems by negatively affecting the hydrological cycles of urban water systems, which would exacerbate in the future (Roesner et al. 2001, Cadavid et al. 2013). Roads and buildings reduced pervious surface area, which limited natural water infiltration and increased rainfall runoff during rainstorm events. Excess rainwater deteriorated the quality and changed the flow patterns of the runoff leading to severe non-point source pollutions and effects on the ecosystem health of receiving water bodies (Damodaram et al. 2013, Moglia et al. 2016a, Montaseri et al. 2015). On the other hand, urbanization concentrated the population and other social resources in urban areas. This transformation led to an increasing demand of fresh water, and thus exacerbated the scarcity of water resources (Ghimire et al. 2014, Thomas et al. 2014). In this situation, rainwater has grown in popularity as an accessible water supply source (Li et al. 2000). Small onsite rainwater harvesting (RWH) technologies, such as green rooftops, porous pavements, and bioretention or green lands, have been successfully implemented as viable approaches to harvest rainwater in most regions (Melville-Shreeve et al. 2016, Moglia et al 2016b, Palla et al. 2011, Walsh et al. 2012). Moreover, RWH systems were intended to preserve and reinstate the pre-developed condition of urban lands. For this purpose, a number of approaches integrated RWH existed in developed and developing countries, such as Water Sensitive Urban Design (WSUD) in Australia (Coombes et al. 1999, Coutts et al. 2012); Best Management Practices (BMPs) and Low Impact Development (LID) in USA (Bhaskar et al. 2016, DER 1999); the 4
Building Research Establishment Environmental Assessment Method (BREEAM) and Sustainable Urban Drainage Systems (SuDS) in the UK (Andoh and Iwugo 2004, BREEAM 2011); the Sponge City Construction and Development in China (China Housing and Urban-Rural Deveolopment 2014).
RWH optimization techniques and evaluation models have been used extensively to design and optimize RWH systems. Silva et al. (2015) evaluated the technical feasibility and economic viability of domestic RWH systems in Porto and Almada, Portugal. Liang et al. (2011) analyzed the economic and financial performance of RWH systems in rural areas of Beijing using cost benefit analysis. Chen and Admas (2007) developed an urban stormwater quality model to evaluate rainwater pollutant buildup and washoff processes. On the other hand, stormwater runoff control has been a new challenge for urban rainwater management. Significant reduction of municipal peak flow may be provided by implementing RWH systems in urban areas (Jensen et al. 2010). Zhang et al. (2012) calculated the potential of collectable rainwater and the possibility of runoff volume reduction. Liu et al. (2015a) used a process-based stormwater runoff model to evaluate the runoff reduction effectiveness under various setting sizes of green infrastructure in Beijing, China. Sample and Liu (2014) used a rainwater simulation model to evaluate decentralized RWH systems in terms of water supply and runoff capture reliability across a wide range of land uses and locations in Virginia, USA.
Most of the previous studies either showed the contribution of RWH in water resource supplementation or in stormwater runoff control through some particular 5
criterions (such as storage size, water saving efficiency, economic analysis and environmental impacts) of interest. In spite of the achievements listed above, there were only a few studies integrating the two ultimate rainwater management components (i.e. RWH for water supplement and stormwate runoff control). Did it mean that rainwater harvesting objectives of water supplement and stormwater runoff control conflicted? To answer this question, many management efforts were requested throughout a watershed. However, there was no single RWH system that could be effective for the objectives in all regions (Lee et al. 2012). The characteristics of rainfall, economic development level, and pollution situations varied among different regions. The major problem was to select the best combination of RWH systems among different options available that obtained a practical, efficient, and sustainable strategy for the regions of different objectives, such as water resource augmenting, stormwater runoff control, economic feasibility and pollutants reduction. Thus, modeling methods and tools, like multi-objective optimization model, were needed to support the selection and assessment of feasible RWH systems, which could determine the construction area of RWH systems that achieved urbanization security and sustainability.
Multi-objective optimization model has been widely applied in solving water resource problems (Alzraiee et al. 2013, Arad et al. 2013, Reddy and Kumar 2007, Vasan and Raju 2006). Non-dominated Sorting Genetic Algorithm II (NSGA II) is one of the most efficient, multi-objective, evolutionary algorithms using the elitist approach (Kalyanmoy Deb 2002) and has gained popularity in finding the near 6
optimal solutions for optimization problems (A.Vasan and Simonovic 2010). Van Meter et al. (2014) recommended a holistic watershed-scale approach that accounts for trade-offs among water availability and socioeconomic wellbeing to explore the social, economic, and environmental dimensions of agricultural RWH ponds in India, and evaluate the viability of these centuries-old systems under current climate and population pressures. Zhang et al. (2014) developed a systematic framework with a multi-objective optimization model based on NSGA II considering the trade-off among wastewater reuse supplies and demands, costs and profits, as well as pollutants reduction. However, so far, no studies of multi-objective optimization model with evolutionary techniques have been available in RWH systems construction. Therefore, this study developed a multi-objective model for optimizing the construction areas of green rooftops, porous pavements, and green lands, considering the trade-offs among daily RWH volume, stormwater runoff control, economic cost, and pollutant reduction. Two more three-objective optimization models (one without daily RWH volume and another without stormwater runoff control) were built to compare the optimal results of three models. The optimal solution was obtained through NSGA II and selected on the basis of water resource situations and rainfall characteristics of each region. It was difficult to compare RWH system construction areas among regions directly due to different regional cover areas. Thus, an index called RWH system construction rate η k (the ratio between the area of single RWH system and the total areas of RWH systems) was defined to express three RWH system construction areas in different regions. The in-depth analysis of the results for 7
31 provinces could provide scientific instructions for the regional stakeholders and policy makers to design more feasible RWH systems and thus to improve urbanization security and sustainability of cities.
8
2. Material and Method 2.1 Study Areas The study areas are 31 provinces of China (except Hong Kong and Macao Special Administrative Regions and Taiwan). Particularly, Beijing and Wuhan cities are selected as case study to expound the optimization process (Fig. 1).
Beijing is located with latitude 39.4°-41.6° north and longitude 115.7°-117.4° east. It has a typical semi-humid monsoonal climate and the average annual precipitation is 532.1 mm (1981-2010). Beijing is suffering from severe water scarcity. The current available water resource per capital is only 94.9 m3 per year (Beijing Municipal Bureau of Statistics 2015), which is approximately 1/30 of the national average and 1/80 of the world average. In addition, the water pollution status has been increasing with the development of Beijing City. Therefore, water resource augmentation and rainfall runoff pollutant reduction have been the main objectives of RWH in Beijing (Beijing Municipal Planning Commission 2013).
Wuhan situates with latitude 29.9°-31.4° north and longitude 113.7°-115.0° east of Hubei province, and has a subtropical monsoon climate. Average annual precipitation in Wuhan is 1316 mm (1981-2010), with 2/3 (866.3 mm) of the total precipitation from May to August, which could easily cause water-loggings in flood season. Based on these austere situations, a large number of RWH projects have been carrying out in Wuhan to alleviate water-logging problems. In August 2015, the government published “Sponge City Planning and Design Guideline” (SCPDG) to instruct and improve RWH system construction. The document stated that the main 9
objective of RWH in Wuhan was stormwater runoff control (Wuhan Municipal Water Affairs Bureau 2015). Through the effort made by the local government as well as the support from the central government, Wuhan becomes the forerunner in the field of RWH system construction. The construction cases and experiences of RWH systems in Wuhan are regarded as references and instructions for other cities.
2.2 The Framework This study established a framework for optimizing the construction areas of RWH systems using a multi-objective optimization model based on the nondominated sorting genetic algorithm II (NSGA II) under physical constraints. The flowchart of the calculation process is presented in Figure 2. Three representative urban RWH systems, i.e. green rooftops, porous pavements and concave green lands, were selected as research objects, which were recommended in previous studies and Sponge Cities Construction Technology Guides (China Housing and Urban-Rural Deveolopment 2014, Wang et al. 2013). In fact, RWH systems involve the storage and distribution of rainwater besides collection, they are not considered in the present study since the process of rainwater collection is the focus of this study and the applicability of rainwater is rather limited to ecosystem utilization and environment consumption in China. 2.3 Objective Function The multi-objective optimization of RWH systems consisted of four objectives: maximization of the harvesting volume of 24 hour-interval rainwater; maximization of the stormwater runoff volume control ratio; minimization of the RWH system cost; 10
and maximization of pollutant reduction in rainfall runoff. RWH volume included two categories: 24 hour-interval RWH volume (QR, million m3) and RWH volume of maximum daily rainfall event (QH, million m3). The development of a model of RWH systems should be on a continuous daily rainfall time series for a minimum of 3 years and preferably 5 years (Lash et al. 2014). Therefore, average daily rainfall data of 30 years (1981-2010) was used to calculate QR, and maximum daily rainfall data (1981-2010) was used to calculate QH. The RWH volume from green rooftops was calculated by the infiltration processes (Wang et al. 2013). The calculation can be expressed as follows:
Pi ⋅ A1; Q1 = SGR ⋅ A1 ;
Pi < S GR
(1)
Pi ≥ SGR
where Q1 (mm·km2) was the RWH volume from green rooftops. Pi was the rainfall (mm), i was equal to 1 or 2, P1 was the 24 hour-interval average daily rainfall (19812010), and P2 was the volume of daily maximum rainfall (1981-2010). A1 (km2) was the construction area of green rooftops. The total storage of green rooftops (S GR, mm) was the infiltration into soil (FGR, mm). Infiltration has been a focus of agriculture and water research for a long period (Milla and Kish 2006). A number of mathematical models have been developed to evaluate the computation of infiltration which could be classified into physically based models and semi-empirical and empirical models. The semi-empirical and empirical models such as Horton models (Horton 1938), always in the form of simple
11
equations, were usually derived from either field or laboratory experiment data. The equation formulated as: f = f e + ( f 0 − f e )e − kt
where f e was the steady state value of f , f 0 was the value of f at t=0, and k was the infiltration decay factor. However, the semi-empirical and empirical models could not provide the detailed information of infiltration process and their physical meaning was not robust. Compared to the semi-empirical and empirical models, the physically based modes could substantially describe the detailed infiltration process, such as Green and Ampt model (Green and Ampt 1911):
B( H c + H ) f = A 1 + F where A and B were parameters which depended on soil characteristics, H c was the capillary potential at the wetting front (L), H was the head of water on the surface (L), and F was the cumulative infiltration. In this study, F was calculated referred from Green and Ampt model (Mein and Lerson 1973): F =K st + S f ∆θ ln(1 + F S f ∆θ )
(2)
where KS was the saturated hydraulic conductivity (mm/min), t was the time intervals of rainfall (min). Sf was suction at wetting front (mm), and ∆θ was the soil water deficit (unitless), which was the difference between saturated water content and initial water content. F0 was initial cumulative depth of infiltration (mm).
12
Similarly, the RWH volume from porous pavements was calculated by the infiltration processes. The calculation can be expressed as follows: Pi ⋅ A2 ; Q2 = S PP ⋅ A2 ;
Pi < S PP
(3)
Pi ≥ S PP
where Q2 (mm·km2) was the RWH volume from porous pavements. A2 (km2) was the construction area of porous pavements. The total storage of porous pavements ( S PP , mm) was equal to the infiltration into soil (the FGR). The volume of RWH from green lands was calculated by the interception and infiltration processes and the depression storage (Liu et al. 2015b, Wang et al. 2013). The calculation can be expressed as follows:
Pi < SGL Pi ⋅ A3 ; Q3 = SGL ⋅ A3 ; Pi ≥ SGL
(4)
where Q3 (mm·km2) was the volume of RWH from green lands. A3 (km2) was the construction area of green lands, and the total storage of green lands ( SGL , mm) was the sum of the interception by the infiltration into soil (FGR), the vegetation canopy (L, mm) and the depression storage of green lands (D, mm). L and D were calculated as Equations (5) and (6). (5)
L = S L × LAI
PC D = Sd max 1 − exp − Sd max
(6)
where S L denoted specific leaf storage (mm), and LAI was canopy leaf area index. Sd max was depression storage capacity of the green lands (mm), and PC was 13
accumulated residual rainfall (mm), which represented rainfall minus interception and infiltration.
Accordingly, QR and QH can be expressed as: 3
Max QR = ∑ Qk × 10 −3 ;
(7)
where i = 1
k =1
3
QH = ∑ Qk × 10 −3 ;
(8)
where i = 2
k =1
The stormwater runoff volume control ratio (R, %) was determined to describe the control efficiency of stormwater runoff by the following function:
Max R =
100QH W
(9)
where W (million m3) was the potential total runoff volume. The function of the third objective, i.e. the economic cost ( EC , billion RMB Yuan), was shown in Equation 10: 3
m
k =1
t =1
Min EC = ∑ ( EC k Ak × 10 −3 + ∑
Okt ) (1 + η ) t
(10)
where ECk (RMB Yuan/m2) was the unit construction cost of each RWH system. Okt (billion RMB Yuan) was the operation and maintenance cost occurring in year t of each RWH system, η was the discounting rate and m was the evaluation period (number of years). According to the publication Chinese Economic Evaluation Parameters on Construction (2006), the discount rate (η ) used in China was 8%. The evaluation period (m) was assumed to be 10 years (Liang et al. 2011). 14
For the last objective, pollutant reduction (PR, Tons), previous researches showed that total suspended solids (TSS) and chemical oxygen demand (COD) were the main pollutants in urban runoff (Huang et al. 2007, Vasan and Raju 2006). Taking Beijing as an example, two years of rainwater runoff samples were collected in 10 rain events from 2013 to 2015. The main characteristics and pollutant loads (TSS, COD, total nitrogen (TN) and total phosphorus (TP)) in the monitored rain events were summarized in Table S1 and Table S2, respectively, and the correlations between TSS and COD, TN, TP were illustrated in Fig. S1. Different degrees of correlation could be noticed between TSS and COD, TN, TP. Particularly, there was a high degree correlation between TSS and COD (R2=0.8579). Furthermore, the concentration of TSS in runoffs was more than 10 times higher than that of natural rainwater (Zhang et al. 2010). Therefore, TSS was chosen as the representative pollutant index and pollutants in natural rainwater were ignored. The function of pollutant reduction was shown in Equation 11: 3
Max PR = ∑ Qk ⋅ψ k ⋅ CTSS k
(11)
k =1
where ψ k was the runoff coefficient of different underlying surfaces and CTSS k (mg/l) was the concentration of TSS of different underlying surfaces.
2.4 RWH System Construction Rate The utilization of RWH systems would be an efficient measure for water supplement and water logging alleviation. The rational construction areas of RWH systems were the essential factor for sustainable rainwater management. However, it was difficult to 15
compare construction areas of RHW systems directly among regions due to different regional cover areas. Thus, an index, RWH system construction rate η k , was defined to express three RWH system construction areas in different regions.
ηGR =
η pp =
ηGL =
AGR AGR + APP + AGL
(12)
App
(13)
AGR + APP + AGL AGL AGR + APP + AGL
(14)
where AGR, APP, and AGL were the optimal construction area of green rooftops, porous pavements and green lands, respectively. 2.5 Three Multi-objective Optimization Scenarios Three multi-objective optimization model scenarios were utilized to optimize the RWH systems:
Scenario A: comprehensively considering the four objectives (i.e. Max QR, Max R, Min EC, and Max PR expressed above) to establish a four-objective optimization model for the purposes of sustainable stormwater management and daily water resources augmentation;
Scenario B: considering Max R, Min EC, and Max PR to establish a three-objective optimization model, which focused on stormwater runoff control;
16
Scenario C: considering Max QR, Min EC, and Max PR to establish another threeobjective optimization model, which concentrated on water resource augmentation.
2.6 Physical Constraints The major constraint for the first two objectives was the precipitation volume, while the objective of pollutants reduction depended on RWH volume. The precipitation volume in this study was an input constant. The precipitation data is available on: http://data.cma.cn/site/index.html for a 30-year period (1981-2010). In addition, the construction area of each RWH system played an important role in the calculation of this model. According to the Engineering technical code for rain utilization in building and sub-district (GB 50400-2006) and Ministry of Housing and Urban-Rural Development of each province, the construction area of green rooftops should be less than 20%-35% of the gross building rooftops under construction and more than existing green rooftops. The construction area of porous pavements should be less than 20%-30% of the urban trunk road and more than existing porous pavements referred from the Ministry of Transport of each province. The construction of the green lands should be less than the urban green lands and park construction planning, and more than the existing green lands area according to China Housing and UrbanRural Development and the Thirteen Five-Year Plan.
2.7 Multi-objective Optimization through NSGA-II NSGA II is an effective method for multiobjective optimization developed by Deb et al (Kalyanmoy Deb 2002). In NSGA II, the concept of Pareto-dominance is used to
17
rank the individuals (control strategies) of a population. As shown in Figure 2, the implementing process of NSGA II follows the following pseudocodes:
Randomly initialize the population P (0) of size N;
Fast nondomination sorting on P (0);
For every generation t;
Select a parent population Pp (t) from P(t) using a binary tournament selection;
Create a child population Pc (t) from Pp (t) through crossover and mutation operators;
Combine P (t) and Pc (t) into an intermediate population Pi(t);
Fast nondomination sorting on Pi (t);
Place the best N individuals from Pi (t) to P (t+1);
End loop
2.8 Data Overall, 31 provinces in mainland China were included in this study, excluding Hong Kong, and Macao Special Administrative Regions, and Taiwan. The meteorological data (i.e. precipitation and temperature data) of each region is available on: http://data.cma.cn/site/index.html for a 30-year period (1981-2010). Supplementary data, including EC1, EC2, EC3, AP, AG, ψ 1 , ψ 2 , and ψ 3 , were obtained from statistical reports (Beijing Statistical Yearbook, Wuhan Municipal Bureau of Statistic, Sponge 18
Cities Construction Technology Guides, Engineering technical code for rain utilization in building and sub-district, Ministry of Water Resources, Ministry of Housing and Urban-Rural Development and National Bureau of Statistics of China). Pollutant concentration data were obtained from rainfall events in Beijing and Wuhan. Other model parameters were taken from previous studies (Table S3). The study sites parameters were listed in Tables S4-S9.
19
3. Results and Discussion
3.1 RWH Systems of Beijing City Figure 3a and Table S11 present the Pareto solutions for RWH systems construction in Beijing. Figure 3a is a three-dimensional representation of the four objectives, where QR, R and PR are on the three-dimensional axes and EC is represented by the different size balls on the Pareto optimal surface. A smaller size ball represents a lower EC value and a larger size ball for a higher value. From this Pareto front, there was a clear trade-off between the objectives EC and PR. Since Beijing has been suffering from both water scarcity and water pollution, QR and PR were considered as the external sets of criteria (Table S10). The solution located at the back top corner in Figure 3a had the largest QR (473.32 million m3) and PR (3.50 kilotons), but it also had the largest EC (85.13 billion RMB Yuan). Moreover, it was found that the PR of the red point (Fig. 3a) was 1.84% (36.16 Tones) less than that of the back top one, while the corresponding total EC of the red point was 6.57% (4.68 billion RMB Yuan) less than that of the back top one. It implied that the solution represented by the red point was more economically feasible than that of the back top one with the same pollutants reduction level. Thus, the solution represented by the red point was selected as the control strategy, with values for QR, R, EC and PR equaling 239.63 million m3, 76.7% million m3, 66.53 billion RMB Yuan and 1931.18 tones, respectively.
Comparing the total optimal RWH areas in this study and the data investigated in Beijing (Fig. 3b), it was noticed that the percentage of constructed green rooftops was less 20%, while the percentage of porous pavements and green lands were both above 20
70%. The results provided the phenomenon that green rooftops were rarely chosen as rooftop for buildings due to the complex technologies and costly maintenance, which were requested in “Technical Code for Roof Engineering” published by the Ministry of Housing and Urban-Rural Development in China. In addition, the defects in construction technology and the single selection of covering vegetation made green rooftops develop leaking problems. In the U.S., mature technology and advanced processes guaranteed the practicality and reliability of green rooftops construction (Ghimire et al. 2014, Mun and Han 2012). Therefore, the green rooftops construction technology in Beijing should be improved to solve leaking problems, and select a more proper vegetation to form a practical and sustainable RWH system. The construction percentage of porous pavements was the largest of the three RWH systems (78.2%). The effect for runoff control of porous pavements was obvious through infiltration for soil storage or groundwater recharge. Moreover, porous pavements demonstrated the highest pollutant reduction of the three alternatives (Table S4).
The different objectives of RWH systems exerted a great influence on RWH system areas and the corresponding results (Fig. 4). When only considering rainwater as a water supplement source (Scenario C), QR in Scenario C could reach 218.87 million m3 per year. Focusing RWH systems on stormwater runoff control (Scenario B), the optimal R was 78.6% provided by Scenario B. Comparing these two scenarios, PR in Scenario C was approximately 5 times of that in Scenario B. The environmental impacts of stormwater runoff depended heavily on the frequency and intensity of 21
stormwater events. The frequency (1.7 times, Table S9) and intensity of stormwater events in Beijing were both low, thus leading to a low pollutants reduction. In addition, green lands were preferred for stormwater runoff control due to its large water storage contribution. However, the economic cost was a considerable factor for green lands, which reflected in Scenario B that the value of EC was 36.7% more than that in Scenario C. In order to achieve water resource augmentation and water-logging alleviation simultaneously (Scenario A), the QR in Scenario A was 9.5% more than that in Scenario C while the R was at the same level with that in Scenario B. It suggested that the optimal RWH construction areas in Scenario A could obtain a better water saving effect without influencing the efficiency of stormwater runoff control. Furthermore, the PR in Scenario A was the highest of the three scenarios. The noticeable deficiency of Scenario A was the huge economic cost. However, with current economic and societal development in China, economic investment would not be the main limiting force considered with the security and environmental friendliness of urbanization progress.
3.2 RWH Systems of Wuhan City Figure 5a and Table S12 present the Pareto solutions for RWH system construction in Wuhan. R and EC were considered as the external sets of criteria because Wuhan has severe water logging disasters in flood season which cause enormous economic losses. Comprehensively considering the trade-off between R and EC (the same selected standards for Beijing), the solution represented by the red point was selected as the control strategy (with QR 230.31 million m3 , R 51.03%, EC 45.08 billion RMB 22
Yuan, and PR 2303.27 tones). Moreover, comparing the total optimal RWH system areas in this study and RWH system areas in Wuhan (Fig. 5b), the percentage of constructed green rooftops was extremely low while that of constructed porous pavements was the highest. These were consistent with the situations in Beijing. Since Wuhan has been suffering from severe water logging problems, R value was a critical variable for the design and development of local RWH systems. Taking RWH systems into account for stormwater runoff control in Scenario B (Fig. 6), the optimal R was 52.57% under the average annual maximum daily rainfall (198.5 mm, 1980-2010). It indicated that the probability of water-loggings happening could decline more than 50% under daily rainfall less than 198.5 mm after the optimal RWH system of Scenario B were constructed. However, the corresponding EC reached 49.57 billion RMB Yuan. This was a high pressure for public expenditure of local or central government. Even though Wuhan does not have issues with quantityrelated water scarcity, the average annual precipitation (1316 mm) is a huge potential fresh-water source. Adding objective Max QR into the model (Scenario A), the optimal QR was 230.31 million m3. Meanwhile, the R value in Scenario A was only 1.54% less than that in Scenario B. This showed that water supply and stormwater runoff control were not two competing objectives; both of them could be achieved by the construction of optimal RWH system. Moreover, economic investment in Scenario A was 9.1% less than that in Scenario B. The reason could be explained that when RWH focusing on runoff control, the optimal green lands construction area was 8.73 km2 more than that in Scenario A thus extending the cost. Furthermore, PR in 23
Scenario A was much higher than that in Scenario B. Therefore, the optimal RWH areas in Scenario A gave consideration to both water logging alleviation and water resource augmentation, meanwhile it was a cost-effective and low-impact management strategy. 3.3 RWH Systems of China Potential stormwater abatement and water saving were found to strongly correlate to local annual precipitation quantities and patterns. The objectives of RWH systems were formulated according to water resource status and rainfall characteristics in different regions. However, it must be noted that water-logging disasters have been a nationwide urbanization problem in China (Zhang et al. 2014, Xia 2013). Therefore, stormwater runoff control was an essential concern of RWH. Through the multiobjective optimization model, Pareto solutions of RWH system for 31 provinces of China were obtained (data not shown). According to the external sets of criteria (Table S10), the control strategies of RWH system were selected and shown in Table S13. Figure 7 illustrates the spatial distributions of QR, R, EC and PR in China. On the national level, QR and PR were 14.84 billion m3 and 228.19 kilotons, respectively. The corresponding reduction of COD, TN, and TP were 113.87, 6.97, and 1.23 kilotons, respectively. In Figure 7a, the QR in China decreased from the southeast to the northwest because annual precipitation decreased from east to west. Guangdong province had the maximum QR (1341.92 million m3) according to its large daily rainfall and RWH system construction areas. The reducing trend of R was opposite to QR but for the same reason. High frequency and intensity of stormwater in southeast 24
China made a great pressure for RWH to control runoff volume. Meanwhile, the serious water-logging problems required a higher standard for RWH systems. It was not surprising that the great attention paid by the central government to accelerate RWH construction in eastern China. Major EC and TSS reductions were found to be concentrated in the east of China. There was also a decrease from the east to west in accordance with the reducing trend of GDP per capita and development level of cities. Geographical distributions of ηGR , η PP , and ηGL were presented in a general picture (Fig. 8). It was found that there were 25 provinces with ηGL more than 50% (50-89%), 25 provinces with η PP more than 15% (18%-54%), and 22 provinces with
ηGR less than 15% (1-15%). Green lands and porous pavements predominated for runoff control and water supplementation due to their large water storage volume. The
ηGR of provinces in southern China was larger than that in northern provinces. Furthermore, there was a decreasing trend of ηGR from northeast (3%-33%) to northwest (1%-6%, except 22% in Xinjiang). The major function of rooftops in northwestern China was drying crops and clothing rather than RWH. Traditional storage tanks would be a better choice for household water supplementation in these regions. In Southeast coastal provinces, η PP values ranged from 20-25% except for the Jiangsu province (54%). The government paid high attention to the construction of porous pavements in Jiangsu province. According to the requirement of “Jiangsu Province Sponge City Construction Guideline” (Jiangsu Municipal Water Affairs Bureau of Jiangsu Province 2015), the coverage ratio of porous pavements for new city trunk roads should be up to 50-60%. . Furthermore, η of three alternatives did 25
not show much variation in southeastern China due to same main purpose of waterloggings alleviation. Whether considering objective MaxR made a non-significant difference on the control strategies of RWH area whereas exerted a great influence on the corresponding results of the optimization model. Comparing the η k of three alternatives obtained in Scenario A and C (Fig. 9), it was noticed that 25 provinces (except for Hunan, Jilin, Jiangsu, Shandong, Tianjin, and Chongqing) had variation amplitude of ηGR less than 5%, other 25 provinces (except for Anhui, Guizhou, Jiangsu, Liaoning, Tianjin, and Zhejiang) had variation amplitude of η PP no more than 5%, and another 24 provinces (except for Anhui, Jilin, Jiangsu, Shandong, Shanghai, Tianjin and Chongqing) had variation amplitude of ηGL less than 10%. The variation amplitude of each η k in Tianjin ( ηGR : 7%; η PP : 11%; ηGL : 17%) and Jiangsu (ηGR : 7%; η PP : 12%; ηGL : 18%) were large due to relatively high stormwater frequency and intensity leading R as a critical variable. In addition, it was also found that the η PP value of 28 provinces (except for Gansu, Hunan and Guangxi) reduced while the ηGL value of 29 provinces (except for Gansu and Hunan) increased when optimization model with objective MaxR. Green lands became the first choice for RWH with an increasing ηGL in 94% of the provinces in Scenario A. However, a lower ηGL in Scenario A of Gansu province was because Gansu suffered from serious quantity-related water scarcity with the result that water resource in comparison to water logging problem was a historical limit for development. When water resource augmentation turned into major objective in Scenario C, ηGL showed a positive 26
change due to large storage volume and available area in Gansu. For Hunan province, the lower ηGL in Scenario A could be explained by the fact that fluvial landform areas in Hunan accounted for 64.76% of the total regional area. In addition, available land areas for green lands in Hunan was less than that of other neighboring provinces (such as Jiangxi, Guangdong, Hubei). Large fluvial landform areas greatly relieved the pressure for runoff control, while limited available construction areas for green lands caused the optimal construction area to always be at the maximum level. For all 31 provinces, maximum daily rainfall under control after the construction of optimal RWH system areas in Scenario A would increase by exceeding 200% more than that before construction (as shown in Eq. 8). Choosing Hubei province with highfrequency water-logging events as an example, daily rainfall volume under control increased from 43.3 mm before the optimal RWH systems constructed to 121.2 mm after construction. 3.4 Sensitivity Analysis Climate change is expected to change the volume and pattern of precipitation (IPCC 2014). In order for a RWH system to be adapted to climate change, it is argued that it should be able to perform at the same level under the future climate (Youn et al. 2012). Thus, sensitivity analysis was carried out to assess the robustness of the results in response to the climate variables such as temperature and precipitation (Fig. 10). It could be seen that the impact of AGR was noticeable in response to 20% variance of 24 hour-interval average daily precipitation, 20% variance of average daily maximum precipitation and 10% variance of average temperature. The AGR showed a positive 27
change at 20% variance of 24 hour-interval average daily precipitation (Fig. 10a), and in contrast, a negative change at 20% variance of average daily maximum precipitation (Fig. 10b) in most provinces. Particularly, the positive change on AGR was more than 50% in Anhui and Chongqing provinces, while the negative change more than 50% in Fujian and Ningxia provinces. The changes of AGL were opposite to AGR at the variance of precipitation. In spite of a weak stability for the results of AGR under different variances of precipitation, it would not influence the RWH systems greatly in China, since the proportion of green rooftops predominately ranged between 1-15% as discussed before. Besides, the change of APP and AGL were both less 20% at variance of average temperature (Fig. 10c). It was implied that the variance of temperature is not the domain factor for the performance of RWH systems in the future. 4. Conclusion
The present study firstly developed a systematic multi-objective optimization model to optimize construction areas of green rooftops, porous pavements and green lands, considering the trade-offs among 24 hour-interval RWH volume, stormwater runoff volume control ratio, construction cost, and rainfall runoff pollutants reduction, for the purpose of stormwater runoff control and water resource augmentation in China. Results showed that on the national level, the control strategies of construction rate of green rooftops (ηGR ), porous pavements (η PP ) and green lands (ηGL ) were 12%, 26% and 62% with corresponding QR and PR for 14.19 billion m3 and 234.79 kilotons, respectively. Optimal ηGR , η PP and ηGL in different regions varied from 1-33%, 628
54%, and 30-89%, respectively, owing to local water resource status and precipitation characteristics. Particularly, green lands was the most important RWH system in 25 provinces with ηGL more than 50% while ηGR was mainly less than 15%, and η PP was mainly between 10-30%. It was also found whether considering objective MaxR made a non-significant difference on RWH system areas whereas exerted a great influence on the result of stormwater runoff control. Maximum daily rainfall under control would increase by exceeding 200% after optimal RWH system areas construction compared with that before construction.
29
Associated Content
Supporting information One figure and thirteen tables are provided to present detailed information for this study. Author Information
Corresponding Author *Phone: 86-2583786251. Fax: 86-25-83786090. E-mail: [email protected] (Y.L.) Acknowledgment
This study was financially supported by the Foundation for Innovative Research Groups of the National Natural Science Foundation of China [No. 51421006], the National Natural Science Foundation of China [No. 91547105 and No. 51479066], the Foundation Research Funds for the Central Universities [No. 2016B10614], the Priority Academic Program Development of Jiangsu Higher Education Institutions, and the Top-Notch Academic Programs Project of Jiangsu Higher Education Institutions [No. PPZY2015A051].
30
Reference
A.Vasan and Simonovic, S.P., 2010. Optimization of Water Distribution Network Design Using Differential Evolution. Journal of Water Resources Planning and Management 136(2), 279-287. Alzraiee, A.H., Bau, D.A. et al., 2013. Multiobjective design of aquifer monitoring networks for optimal spatial prediction and geostatistical parameter estimation. Water Resources Research 49(6), 3670-3684. Andoh, R.Y.G. and Iwugo, K.O., 2004. Sustainable Urban Drainage Systems: A UK Perspective. Urban Drainage. Arad, J., Housh, M., Perelman, L. et al., 2013. A dynamic thresholds scheme for contaminant event detection in water distribution systems. Water Res 47(5), 1899-1908. Beijing Municipal Bureau of Statistics, 2015. Beijing Statistical Yearbook. China Statistics Press, Beijing, China (in Chinese). Beijing Municipal Planning Commission, 2013. Rainwater Control and Use Project Guideline. (in Chinese). Bhaskar, A.S., Hogan, D.M. and Archfield, S.A., 2016. Urban base flow with low impact development. Hydrological Processes 30(18), 3156-3171. BREEAM, 2011. http://www.breeam.org/page.jsp?id=181. (accessed 15 May 2013). Cadavid, Londoño Catalina Ando et al., 2013. Valuing preferences over stormwater management outcomes including improved hydrologic function. Water Resources Research 49(7), 4114-4125. 31
China Housing and Urban-Rural Deveolopment, 2014. Sponge Cities Construction Technology Guides. China (in Chinese). Coombes, P.J., Argue, J.R. et al., 1999. Figtree Place: a case study in water sensitive urban development. Urban Water Journal, 335-343. Coutts, A.M., Tapper, N.J., et al., 2012. Watering our cities: The capacity for Water Sensitive Urban Design to support urban cooling and improve human thermal comfort in the Australian context. Progress in Physical Geography 37(1), 2-28 Damodaram, Chandana Zechman et al., 2013. Simulation-Optimization Approach to Design Low Impact Development for Managing Peak Flow Alterations in Urbanizing Watersheds. Journal of Water Resources Planning and Management 139(3), 290-298. DER, 1999. Low-Impact Development Design Strategries: An Integrated Design Approach. DER, Largo, Maryland, USA. Ghimire, S.R., Johnston, J.M., et al., 2014. Life cycle assessment of domestic and agricultural rainwater harvesting systems. Environ Sci Technol 48(7), 40694077. Green, W.H. and Ampt, G.A., 1911. Studies on soil physics, I. Flow of air and water through soils. Journal of Agricultural Science 28, 1-11. Helmreich, B. and Horn, H., 2009. Opportunities in rainwater harvesting. Desalination 248(1-3), 118-124. Horton, R.E., 1938. The interpretation and application of runoff plot experiments with reference to soil erosion problems. Soil Science Society of America Proceeding, 32
340-349. Huang, J.-l., Du, P.-f., et al., 2007. Characterization of surface runoff from a subtropics urban catchment. Journal of Environmental Sciences 19(2), 148-152. IPCC, 2014. Climate Change 2014: Synthesis Report. Geneva, Switzerland. Jensen, M.A., Steffen, J., et al., 2010. Do rainwater harvesting objectives ofwater supply and stormwater management conflict? Low Impact Development. Jiangsu Municipal Water Affairs Bureau of Jiangsu Province, 2015. Jiangsu Province Sponge City Construction Guideline, China (in Chinese). Jieyun, C. and J., A.B., 2007. A derived probability distribution approach to stormwater quality modeling. Advances in Water Resources 30(1), 80-100. Kalyanmoy Deb, A.P., Sameer Agarwal, T. Meyarivan, 2002. A Fast and Elitist Multiobjective Genetic Algorithm NSGA-II. IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION 6, 182-197. Lash, D., Ward, S., et al., 2014. Robust rainwater harvesting: probabilistic tank sizing for climate change adaptation. Journal of Water and Climate Change 5(4), 526. Lee, J.G., Selvakumar, A., et al., 2012. A watershed-scale design optimization model for stormwater best management practices. Environmental Modelling & Software 37, 6-18. Li, F., Cook, S., et al., 2000. Rainwater harvesting agriculture: An integrated system for water management on rainfed land in China's semiarid areas. Ambio 29(8), 477-483. Liang, X., Van Dijk and Pieter, M., 2011. Economic and financial analysis on 33
rainwater harvesting for agricultural irrigation in the rural areas of Beijing. Resources, Conservation and Recycling 55(11), 1100-1108. Liu, W., Chen, W. and Peng, C., 2015a. Influences of setting sizes and combination of green infrastructures on community’s stormwater runoff reduction. Ecological Modelling 318, 236-244. Liu, W., Chen, W., et al., 2015b. A water balance approach to assess rainwater availability potential in urban areas: the case of Beijing, China. Water Science & Technology: Water Supply 15(3), 490. Mein, R.G. and Lerson, C.L., 1973. Modeling Infiltration during a Steady Rain. Water Resources Research Vol. 9, No. 2, 384-394. Melville-Shreeve, P., Ward, S. and Butler, D., 2016. Rainwater Harvesting Typologies for UK Houses: A Multi Criteria Analysis of System Configurations. Water 8(4), 129. Milla, K. and Kish, S., 2006. A low-cost microprocessor and infrared sensor system for automating water infiltration measurements. Computers and Electronics in Agriculture 53(2), 122-129. Moglia, M., Gan, K. and Delbridge, N., 2016a. Exploring methods to minimize the risk of mosquitoes in rainwater harvesting systems. Journal of Hydrology 543, 324-329. Moglia, M., Gan, K., et al., 2016b. Investigation of pump and pump switch failures in rainwater harvesting systems. Journal of Hydrology 538, 208-215. Montaseri, M., Hesami Afshar, et al., 2015. Development of Simulation-Optimization 34
Model (MUSIC-GA) for Urban Stormwater Management. Water Resources Management 29(13), 4649-4665. Mun, J.S. and Han, M.Y., 2012. Design and operational parameters of a rooftop rainwater harvesting system: definition, sensitivity and verification. J Environ Manage 93(1), 147-153. Palla, A., Gnecco, I. and Lanza, L.G., 2011. Non-dimensional design parameters and performance assessment of rainwater harvesting systems. Journal of Hydrology 401(1-2), 65-76. Reddy, M.J. and Kumar, D.N., 2007. Multiobjective Differential Evolution with Application to Reservoir System Optimization. Journal of computing in civil engneering 21(2), 136-146. Roesner Larry A. and Bledsoe, B.P., 2001. Are best-management-practice criteria really environmentally friendly? ASCE Journal of Water Resources Planning and Management 127, 150-154. Sample, D.J. and Liu, J., 2014. Optimizing rainwater harvesting systems for the dual purposes of water supply and runoff capture. Journal of Cleaner Production 75, 174-194. Silva, C.M., Sousa, V. et al., 2015. Evaluation of rainwater harvesting in Portugal: Application to single-family residences. Resources, Conservation and Recycling 94, 21-34. Thomas, R.B., Kirisits, M.J., et al., 2014. Rainwater harvesting in the United States: a survey of common system practices. Journal of Cleaner Production 75, 166-173. 35
Van Meter, K.J., Basu, N.B., et al., 2014. Monsoon harvests: the living legacies of rainwater harvesting systems in South India. Environ Sci Technol 48(8), 42174225. Vasan, A. and Raju, K.S., 2006. Application of Differential Evolution for Irrigation Planning: An Indian Case Study. Water Resources Management 21(8), 13931407. Walsh, C.J., Fletcher, T.D. and Burns, M.J., 2012. Urban Stormwater Runoff: A New Class of Environmental Flow Problem. Plos One. Wang, R., Eckelman, M.J. and Zimmerman, J.B., 2013. Consequential Environmental and Economic Life Cycle Assessment of Green and Gray Stormwater Infrastructures for Combined Sewer Systems. Environ Sci Technol 47(19), 11189-11198. Wuhan Municipal Water Affairs Bureau, 2015. Sponge City Planning and Design Guildeline of Wuhan City. Wuhan,China, (in Chinese). Xia, W., 2013. The Cause Analysis and Countermeasures of Waterlogging in Wuhan. Applied Mechanics and Materials 357-360, 1962-1965. Youn, S.-g., Chung, E.-S., et al., 2012. Probabilistic estimation of the storage capacity of a rainwater harvesting system considering climate change. Resources, Conservation and Recycling 65, 136-144. Zhang D.D., Yan D.H. and Y Wang., 2014. Research Progress on Risk Assessment and Integrated Strategies for Urban Pluvial Flooding. Journal of Catastrophology 29, 144-149. 36
Zhang, M., Chen, H., Wang, J. and Pan, G., 2010. Rainwater utilization and storm pollution control based on urban runoff characterization. Journal of Environmental Sciences 22(1), 40-46. Zhang, W., Wang, C., Li, Y., et al., 2014. Seeking sustainability: multiobjective evolutionary optimization for urban wastewater reuse in China. Environ Sci Technol 48(2), 1094-1102. Zhang, X., Hu, M., et al., 2012. Urban Rainwater Utilization and its Role in Mitigating Urban Waterlogging Problems—A Case Study in Nanjing, China. Water Resources Management 26(13), 3757-3766.
37
Fig. 1. - Marked The study areas and the geographic location of Beijing city and Wuhan city. 1: Heilongjiang, 2: Jilin, 3: Liaoning, 4: Hebei, 5: Beijing, 6: Tianjin, 7:Shandong, 8: Jiangsu, 9: Shanghai, 10: Zhejiang, 11: Anhui, 12: Jiangxi, 13: Fujian, 14: Guangdong, 15: Hainan, 16: Hunan, 17: Guangxi, 18: Yunnan, 19: Guizhou, 20: Sichuan, 21: Chongqing, 22: Hubei, 23: Henan, 24: Shanxi, 25: Inner Mongolia, 26: Shaanxi, 27: Gansu, 28: Ningxia, 29: Qinghai, 30: Xinjiang, 31: Tibet.
38
Fig. 2. - Marked Methodology Flowchart. QR was 24 hour-interval harvesting rainwater volume. R was stormwater runoff volume control ratio. EC was RWH systems construction cost. PR was total suspended solids reduction in rainfall runoff. P1 was the 24 hour-interval average daily rainfall (1981-2010). P2 was the rainfall of the daily maximum rainfall (1981-2010).
39
a
b
Fig. 3. – marked (a) Pareto front obtained by the nondominated sorting genetic algorithm using the four-objective optimization problems study for rainwater harvesting (RWH) systems areas in Beijing. (b) The construction areas of three selected RWH systems this study and the data investigated in Beijing.
40
Fig. 4. – Marked The 24 hour-interval RWH volume (QR), stormwater runoff volume control ratio (R), RWH systems construction cost (EC) and TSS reduction (PR) in Scenario A (Four-objective optimization model), Scenario B (Three-objective optimization model without QR) and Scenario C (Three-objective optimization model without R).
41
a
b
Fig. 5. – Marked (a) Pareto front obtained by the nondominated sorting genetic algorithm using the four-objective optimization problems study for RWH systems areas in Wuhan. (b) The construction areas of three selected RWH systems in this study and the data investigated in Wuhan. 42
Fig. 6. - Marked The 24 hour-interval RWH volume (QR), stormwater runoff volume control ratio (R), RWH systems construction cost (EC) and TSS reduction (PR) in Scenario A (Four-objective optimization model), Scenario B (Three-objective optimization model without QR).
43
a
b
million m3 0-200 200-400 400-600 600-1000 1000+
c
d
Tones
billion RMB Yuan
0-2000 2000-4000 4000-8000 8000+
0-50 50-80 80-100 100+
Fig. 7. - Marked Spatial distributional of (a) harvesting volume of 24 hour-interval rainwater, (b) stromwater control ratio, (c) economic cost of RWH systems construction, (d) TSS reduction in rainfall runoff in China.
44
Fig. 8. - Marked Geographical distribution of RWH systems construction rate and the average annual rainfall. Green rooftop (blue), porous pavements (red) and green lands (green). 45
Fig. 9. – Marked Comparison of the three RWH systems contruction areas obtained by the multi-objective optimization model with (left) and without (right) the objective MaxR, where the less ones are shown by the red dotted line box.
46
a.
b.
47
c.
Fig. 10. Results of the sensitivity analysis on the construction area of green rooftops (AGR), porous pavements (APP), and green lands (AGL), respectively, in 31 provinces for precipitation and temperature. (a) 20% variance of 24 hour-interval average daily precipitation; (b) 20% variance of average daily maximum precipitation; (c) 10% variance of average temperature. 1: Heilongjiang, 2: Jilin, 3: Liaoning, 4: Hebei, 5: Beijing, 6: Tianjin, 7:Shandong, 8: Jiangsu, 9: Shanghai, 10: Zhejiang, 11: Anhui, 12: Jiangxi, 13: Fujian, 14: Guangdong, 15: Hainan, 16: Hunan, 17: Guangxi, 18: Yunnan, 19: Guizhou, 20: Sichuan, 21: Chongqing, 22: Hubei, 23: Henan, 24: Shanxi, 25: Inner Mongolia, 26: Shaanxi, 27: Gansu, 28: Ningxia, 29: Qinghai, 30: Xinjiang, 31: Tibet.
Note for the change of Fig. 10.: The Fig. 10 was redrawn with a high resolution.
48
A four-objectives optimization method for RWH systems in China was
established. Green lands are dominant with the construction rate varying from 30~89% in
China. Green rooftops, porous pavements and green lands were comprehensively
considered. Daily rainfall under control increased by 2 times after optimal RWH construction.
49
50