Quantifying multiple ecosystem services for adaptive management of green infrastructure

Abstract

Demands for ecosystem service assessments are growing around the world. However, management applications remain limited in part because we lack measurements linking ecosystem characteristics (ecosystem structure and processes) to final ecosystem services. Policymakers need marginal values, changes in final ecosystem services (direct link to human welfare) relative to additional units of ecosystem characteristics (∆final ecosystem services/∆ecosystem characteristics) to assess tradeoffs. Progress, however, has been slow due to confusion on ecological production functions (EPFs) in ecology. Here, we apply a new interdisciplinary approach to craft EPFs to evaluate four ecosystem services using the Yongding River Green Ecological Corridor as our case study. The Yongding Corridor is Beijing's largest, most expensive green infrastructure project (~ $2.5 billion USD), constructed as a network of seven artificial lakes and wetlands. The Beijing Government wants the Yongding Corridor to improve four ecosystem services: (1) water storage, (2) local climate regulation, (3) water purification, and (4) aesthetics. We first worked with stakeholders to determine final ecosystem service levels and then used the Variable Infiltration Capacity model to estimate key ecosystem processes from the designed lakes and wetlands. We coupled the modeling with ecological field data and social surveys to create EPFs. We evaluated the ecosystem services by calculating shortfalls and then determined synergies and tradeoffs to identify actions for reducing shortfalls. We found the Yongding Corridor is meeting desired levels for aesthetics, but incurred shortfalls on the remaining services. To obtain the desired services, we recommend managers: (1) maintain inflow rates and/or make the lakes deeper to reduce water loss rates; (2) improve water quality—wetlands have high nutrient retention, but nutrient loads must be reduced; and (3) plant shade trees since evaporative cooling from the lakes and wetlands is having no measurable impact on human comfort. Results indicate the absence of ecosystem functions in landscape design led to shortfalls, but solutions require coupling green and built infrastructure to obtain multi-functionality. Managers found marginal values useful for clarifying connections, which led to adaptive policy changes for improving green infrastructure.

Introduction

Cities are central to our pursuit of green growth and sustainability. The majority of us now call cities home, and forecasts show future generations will live in larger more populated cities (United Nations 2016). Of the world's thirty-one megacities (10 million or more residents), twenty-four are located in developing regions with China having the most of any country (Shanghai, Beijing, Chongqing, Guangzhou, Tianjin, and Shenzhen). Cities are concentrated with people, knowledge, and financial capital, making them centers of innovation and commerce. The density of development, however, causes them to experience higher levels of pollution, biodiversity loss, and social inequality compared to suburban or rural areas (Grimm et al. 2008). Cities depend heavily on ecosystem services from afar to meet their high material demands (U.S. National Research Council 2016). However, urban residents also need local ecosystem services for urban livability. If cities can increase urban ecosystem services while reducing their ecological footprints, they could significantly advance global sustainability. Hence, policymakers are pursuing strategies on urban sustainability, defined as: measurable improvements in human well-being through actions across environmental, economic, and social dimensions (U.S. National Research Council 2016). Urban sustainability is about creating cities that harness co-benefits among these dimensions. Hence, scientists are working to incorporate ecosystem functioning into urban design for more sustainable, resilient cities (Grimm et al. 2008, Bai 2018).

Green infrastructure is a core concept in using the ecosystem approach to advance urban sustainability. The European Commission (2013) defines green infrastructure as a strategically planned network of natural and seminatural areas with other environmental features designed and managed to deliver a wide range of ecosystem services. Green infrastructure includes parks, green roofs, rain gardens, constructed wetlands, and detention basins. In particular, decision-makers are using green infrastructure to manage freshwater ecosystems in cities. Ecologists often believe green infrastructure can produce a diversity of urban ecosystem services, such as aesthetics, flood mitigation, pollution reduction, and recreation (Andersson et al. 2014). Decision-makers see green infrastructure as a possible cost-effective means for resolving public health problems. For instance, “urban greening” has been proposed as a policy tool for mitigating urban heat island effects to reduce heat stress (Bowler et al. 2010). Also, the scenic beauty of green spaces can promote healthier communities by encouraging people to engage in physical activities while offering relaxing environments for leisure (Barton and Pretty 2010, Hipp and Ogunseitan 2011). However despite its clear purpose, the performance of green infrastructure is rarely evaluated (Haase et al. 2014). When projects are evaluated, scientists found green infrastructure often does not supply ecosystem services at desired levels (Pataki et al. 2011).

As government investments increase on ecosystem service projects, so are the needs for performance-based evaluations (Carpenter et al. 2009). However, the majority of ecosystem service projects lack monitoring and evaluation mechanisms (Naeem et al. 2015). For green infrastructure to become a viable alternative to gray infrastructure requires understanding why and when ecological efforts fall short on delivering social services. We need measurements of ecosystem services linking intermediate (ecosystem structure and functions) and final ecosystem services (ecological endpoints directly affecting human welfare, now referred to as final services) to determine actions for success (Palmer and Filoso 2009). The scientific challenge is creating practical, robust methods to overcome a data gap. We define the data gap as the lack of biophysical measurements linking ecosystem characteristics to final services (Wong et al. 2015). To address this data gap, experts suggest scientists develop ecological production functions (EPFs; U.S. National Research Council 2005, U.S. Environmental Protection Agency 2009) to quantify the link between functions and services in the service cascade (Haines-Young and Potschin 2009, 2018). Ecological production functions are regression models linking ecosystem characteristic metrics to final service indicators (Wong et al. 2015). Ecological production functions are useful because they allow scientists to generate marginal values to quantify synergies and tradeoffs among ecosystem services and different management actions. However, currently the vast majority of studies use secondary data to measure ecological and social indicators and then evaluate synergies and tradeoffs using correlation statistics (Eigenbrod et al. 2010, Martínez-Harms and Balvanera 2012, Wong et al. 2015). The few studies that create EPFs either evaluate a single service or provisioning services (Wong et al. 2015). We are finding the root cause to the problem is disciplinary confusion on EPFs. Hence, we synthesized the main concepts and data across ecology, economics, and environmental policy to develop a methodological framework on EPFs (presented in Wong et al. 2015). In this paper, we present the first application of our approach to estimate four ecosystem services using EPFs from the Yongding River Green Ecological Corridor in Beijing, China.

Our interdisciplinary approach builds upon the ecosystem service definitions of Boyd and Banzhaf (2007) and Fisher et al. (2008, 2009). Also, it supports the U.S. Environmental Protection Agency's Final Ecosystem Goods and Services (2013; Bell et al. 2017, O'Dea et al. 2017) and National Ecosystem Services Classification System (U.S. EPA 2015), and the European Environment Agency Common International Classification of Ecosystem Services (Haines-Young and Potschin 2018). These works lay the conceptual foundation for integrating ecosystem services into decision-making; however, they lack a methodological framework for applying the classifications to measure ecosystem services. In this paper, we explore the theoretical underpinnings of these frameworks by operationalizing EPFs to evaluate the Yongding River in Beijing. The Yongding Corridor is an important case study because currently it is Beijing's largest green infrastructure project (17.5 billion yuan; 2.5 billion USD) constructed explicitly to enhance ecosystem services for urban sustainability. The Yongding River was Beijing's largest river, but it has been dry for over thirty years due to damming and urbanization. Beijing officials believe the poor environmental quality on the Yongding River is preventing economic development in this region (Beijing Water Authority 2009). The government constructed a network of seven artificial lakes and wetlands as public parks to enhance ecosystem services. The four key ecosystem services (desired human benefits) are as follows: (1) water storage (water supply for artificial groundwater recharge—lakes act as reservoirs); (2) local climate regulation (summer cooling for human comfort to mitigate urban heat island effects); (3) water purification (drinking water quality); and (4) aesthetics (scenic beauty for economic development and recreation). The challenge for policymakers is evaluating the ecosystem service outcomes to manage for multi-functionality.

In this study, we examine how the new ecosystems on the Yongding River changed the flow of ecosystem services, and whether the changes are sufficient to supply the desired benefits. Our research questions are as follows: (1) What is the additionality in ecosystem service flows from the new ecosystems? (2) What are the synergies and tradeoffs among ecosystem services? (3) What are possible management actions for improving ecosystem service outcomes? We examine the hypothesis of whether urban green infrastructure can supply ecosystem services at required levels for human benefits. Furthermore, we evaluate whether estimating the linkage between ecosystem functions and final services can advance the application of ecosystem services in policy. To date, ecosystem services have been mainly used as a supporting concept with no explicit application in urban policy. Our objective is to present an example of how to conduct an ecosystem services assessment using EPFs, and how the science can inform management actions on green infrastructure in cities.

Materials and Methods

Evaluation approach

In the literature, there is growing consensus on defining ecosystem services as the direct or indirect contributions that ecosystems make to human well-being (Fisher et al. 2009, Haines-Young and Potschin 2009, U.S. Environmental Protection Agency 2009, TEEB 2010). Scientists are using the classifiers intermediate and final services to describe how ecosystem structure and functions (i.e., ecological mechanism) relate to human benefits. Intermediate services are ecosystem characteristics measured as ecosystem structure, processes, and functions that support final services (Wong et al. 2015). Final services are components of nature possessing an explicit connection to human well-being that have direct value to society (Boyd and Banzhaf 2007, Ringold et al. 2013). Clarifying the relationship between intermediate and final services is important for (1) minimizing double-counting; (2) bridging methodological divides between the natural and social sciences; and (3) clarifying the importance of ecosystem processes to the public. In the past decade, we have made progress on categorizing the link between ecosystem characteristics and final services, but quantification remains elusive due to disciplinary confusion on EPFs (U.S. National Research Council 2005, U.S. Environmental Protection Agency 2009, Bruins et al. 2016). From our experience, the primary problem is the lack of clarity among ecologists on: what is a production function. Currently, natural scientists are classifying any biophysical model as an EPF (Bruins et al. 2016, Bell et al. 2017). However, the term production function comes from economics where regression models are typically used to relate outputs (i.e., final services) to inputs (i.e., intermediate services) to estimate marginal values using regression coefficients. Social scientists use regressions to estimate how changes in causal mechanisms may change social outcomes. The focus is on quantifying the relationship between variables. Alternatively, natural scientists commonly use regression models to forecast outcomes to create predictive models. This disciplinary difference is important since it is inhibiting understanding of marginal values in ecology, which in turn is impairing progress on the data gap.

Fisher et al. (2008) defined marginality as ecosystem service values, regardless of the metric (biophysical or economic), expressed as a function of (small) changes in the flow of that service. Marginality is important in public policy because management decisions operate at the margins. For example, marginality can translate to assessing whether additional hectares of lake area on the Yongding River can improve human comfort in the summer. Does increased water area in a water-scarce region represent a beneficial tradeoff to Beijing? Despite the importance of marginal values, Fisher et al. (2008) and Wong et al. (2015) found that most studies estimate ecosystem services in terms of total ecosystem service provision or total economic value. We were unable to identify any studies that quantified marginal changes of multiple ecosystem services in particular regulating services. Hence, we developed an interdisciplinary approach to clarify the EPF process in ecology.

Our approach consists of three variables: (1) final services; (2) final service indicators (measured, proxy values); and (3) ecosystem characteristic metrics (main ecosystem structures/processes to produce final services, representing management options; Wong et al. 2015; Fig. 1). The final services are used to determine the final service indicators. Also, the final service levels provide the necessary biophysical units to select and measure key ecosystem characteristics via field data or biophysical models. Once the data are collected, we measure the ecosystem services by first calculating shortfalls, defined as the difference between final service indicators (i.e., observed values) and final service levels (i.e., required levels for benefits). Second, we create EPFs to calculate marginal ecosystem service values. We integrate (1) biophysical models to estimate ecosystem characteristics across space and time (ecology) with (2) regression models to link ecosystem characteristic metrics and final service indicators (economics).

Figure 1

Conceptual diagram of our framework on measuring and evaluating ecosystem services. Outlined are the methodological steps applied to the Yongding River Green Ecological Corridor: (1) identify desired human benefits using policy goals from official documents; (2) identify final ecosystem services working with stakeholders to determine appropriate policy targets; (3–4) indicator selection; (5–6) data collection; (7) ecological production functions using regression models to generate marginal ecosystem service values; and (8) evaluation of synergies and tradeoffs. Results from phase II inform phase I to refine management actions and targets. The letters correspond to different ecosystem services: (a) water storage; (b) local climate regulation; (c) water purification; and (d) aesthetics. Orange text represents the calculations for measuring ecosystem services. In step 3, we note the indicators used to calculate shortfalls in parentheses since we had to create a water loss factor to evaluate water efficiency.

Study site

Our study site is the Yongding River Green Ecological Corridor, which is located on the Yongding River in Beijing. In Beijing, the Yongding River is 170 km long with a total area of 1188 km², stretching between longitudes 115°70′ E–116°44′ E and latitudes 40°13′ N–39°39′ N (Fig. 2). The upstream northern section is mainly mountainous, consisting of secondary forests of deciduous broadleaf trees (elevation >1000 m). The downstream section is flat plains with urban and agricultural land uses (elevation <150 m). Beijing has a seasonal temperate, semiarid monsoonal climate with an annual average daily temperature of 11°C. The average annual precipitation is 550 mm where the majority of rainfall occurs from June to September.

Figure 2

(a) Study sites on the Yongding River Green Ecological Corridor local scale (red) and regional scale (blue). (b) Land-use/land-cover maps for local and regional scales for the pre- and post-Corridor periods with individual lakes and wetlands listed (Wong et al. 2017).

Construction of the seven lakes and wetlands started in 2010 (total length = 20 km; total area = 19.5 km²), and all the lakes and wetlands were complete in 2012: (1) Mencheng Lake, (2) Wetlands, (3) Lianshi Lake, (4) Garden Expo Lake, (5) Xiaoyue Lake, (6) Wanping Lake, and (7) Daning Reservoir (Fig. 2). It is estimated that 130 million m³ of water was added to the landscape with the majority being reclaimed water from wastewater treatment plants (Beijing Water Authority 2009). The lakes and wetlands have controlled inflow and outflow rates, and near-zero infiltration due to an impervious liner. Surface runoff into the lakes and wetlands is restricted because channels divert stormwater to retention basins. Pumping stations move water from north to south in a circular fashion (Wong et al. 2017). Annually, water is removed from the system to artificially recharge groundwater supplies outside the Yongding Corridor, and new water is returned to the system. Also, engineers constructed surface flow wetlands between Mencheng Lake and Lianshi Lake to buffer the lakes from nonpoint pollution. They selected emergent vegetation where the dominant vegetation is Phragmites (i.e., common reed) and Typha latifolia (i.e., cattail). The cultural function of the Yongding Corridor is to provide a network of parks for leisure and tourism. Policymakers want people to find beauty and enjoyment in the new ecosystems to motivate tourism and real-estate development (Yip 2008, Shi 2013).

Beijing is testing a new strategy using green infrastructure to guide development. This differs from conventional urban planning where green spaces are established after other land uses. The Yongding Corridor is an important component of Beijing's Master Plan (2004–2020) to move Beijing toward a green city. More than 290 billion yuan (47.3 billion USD) has been invested in the districts surrounding the Yongding River to fund the green development transition (Shi 2013). The focus is planning for sustainability centered on renewable energy, water efficiency, affordability, public transportation, and green space (Yip 2008). The ecosystem services from the Yongding Corridor are central to Beijing's vision on urban sustainability.

We conducted our ecosystem services assessment of the Yongding Corridor for the first year of operation (2012–2013). We define the temporal scale as (1) pre-Corridor (1 June 2009–30 June 2010) and (2) post-Corridor (1 June 2012–30 June 2013; Figs. 2, 3). To estimate the contribution of the new ecosystems to the desired benefits requires determining the change in ecosystem services from the baseline condition. The pre-Corridor levels were zero for water storage (no lakes), water purification (no wetlands), and aesthetics (no parks; Fig. 3). Local climate regulation, however, requires estimating the change in evapotranspiration (ET) between the pre- and post-Corridor periods. In this system, ET is the main ecosystem function regulating cooling on the Yongding River (Fig. 1; Wong et al. 2017). We have to relate the change in ET between both periods to human heat index values to estimate the contribution of the new ecosystems to local climate regulation. For the other ecosystem services, we simply estimate the ecosystem characteristics and final service indicators in the post-Corridor period. We also evaluated the ecosystem services at two spatial scales: (1) regional, defined as the full Yongding Corridor, 170 km long and 1188 km² in area; (2) local, defined as the lakes/wetlands in the urbanized section, 20 km long and 19.5 km² in area (spatial and temporal scales defined for each service in Table 1). We selected these spatial scales because climate regulation operates at a regional scale, and several drivers of ecological change operate at a regional scale (e.g., urbanization).

Figure 3

Photographs of the Yongding River (a) before (2009) and (b, c) after (2013) the green infrastructure project. Landscape architects and engineers transformed the dry riverbanks from landfills and sand pits to public parks with constructed lakes and wetlands.

Table 1. Final ecosystem services (endpoints), indicators, and metrics

Ecosystem services (ES)	Final ES (endpoints)	Final ES indicators	Ecosystem characteristic metrics
Water storage†	Water volume (million m^3) = 12.05 Water area (km^2) = 6.51	Water volume (million m^3) Water area (km^2) Fractional water loss (lake evaporation/volume)	Lake depth (m)
Local climate‡ regulation	Heat index values = 26 Sultry weather >26	Air temperature (°C) Relative humidity (%) Heat index values	Evapotranspiration (mm/h)
Water purification†	Drinking water quality (mg/L): Total nitrogen (TN) = 1.0 Total phosphorus (TP) = 0.2	TN (mg/L) TP (mg/L)	Wetland area (ha) Nutrient load (mg/L) Nutrient retention (mg/L)
Aesthetics†	Visitor preferences: Very beautiful Beautiful	Scenic beauty scores	Environmental quality: Climate Water quality

• † Spatial scale is local; temporal scale is 2012–2013.
• ‡ Spatial is local and regional; temporal scale is 2009–2010 and 2012–2013.

Stakeholder engagement to select the final ecosystem services

A key challenge when conducting an ecosystem services assessment is selecting legitimate final services. We worked with stakeholders (e.g., scientists, managers, park visitors, and residents) to determine the final services for the Yongding Corridor. We reviewed the official planning documents (i.e., policy and engineering plans) produced by the Beijing Water Authority (BWA), which oversees water resources management in Beijing. Policymakers outlined the specific policy objectives in the Green Yongding River: Construction Plan for an Ecological Corridor (2009). The desired benefits as written in the plan and confirmed by policymakers are as follows: (1) flood control; (2) water quality improvements to meet drinking water quality standards; (3) artificial groundwater recharge using the lakes as reservoirs; (4) functioning wetlands; and (5) aesthetic improvements for leisure and entertainment. From the policy objectives, we selected four ecosystem services using the Millennium Ecosystem Assessment categories. We identified three regulating services and one cultural service: sustaining water storage, local climate regulation, water purification, and aesthetics. Flood control is a gray infrastructure service in this system. We also surveyed park visitors and residents to determine public preferences for different ecosystem services (see subsection on aesthetics for details on the survey methods). The top five ecosystem services selected by visitors in terms of their social values were as follows: (1) leisure and travel, (2) air quality, (3) landscape preservation, (4) cooling, and (5) heritage value for future generations (Table 2). We also asked visitors to select dissatisfactions and comment on future concerns regarding the Yongding Corridor. The surveyed population identified three main problems impacting the functionality of the Yongding Corridor: summer discomfort due to few shade trees, water pollution, and water level. The final services approved by management were similar to the top ecosystem services and ecological concerns selected by residents/visitors.

Table 2. Visitor rankings of ecosystem service values from the Yongding River Green Ecological Corridor in Beijing, China (N = 193)

Social values (ecosystem services)	N	%
Leisure and travel (cultural)	149	77
Air quality (air pollution control)	85	44
Landscape preservation (cultural and biodiversity)	38	20
Cooling (local climate regulation)	33	17
Heritage value for future generations (cultural)	27	14
Water quality (water purification)	26	13
Water supply (water storage)	12	6
Other	1	1

We used the BWA policy documents, China's environmental standards, and visitor surveys to determine quantifiable levels for the final services. We worked with scientists and managers at the Beijing Water Science and Technology Institute (BWSTI) to determine appropriate levels. The BWSTI is a BWA public research institution who oversees the management of the Yongding River. For water storage, we used the BWA water supply targets, defined as (1) total lake volume ≥12.05 million m³; (2) total surface water area ≥6.51 km² (Table 1). For local climate regulation, the BWA described its target as the reduction in summer temperatures to improve human comfort. Government agencies have heat indices (derived from air temperature and humidity) to predict climate impacts on human comfort. The human heat index (HI) is a measure of how hot humans feel when relative humidity is included with air temperature. Heat index is used to determine the likelihood of humans experiencing heat disorders and discomfort via threshold levels for sultry weather. We used the Beijing Meteorological Bureau HI threshold levels: HI values of 27–28 represent sultry weather, 29–30 represent heavy sultry weather, and >30 represent extreme sultry weather (Wang and Gong 2010). Therefore for local climate regulation, we selected a final service level of HI ≤ 26, which represents no sultry weather meaning an environment where on average humans feel comfortable. For water purification, managers want to meet China's Ministry of Ecology and Environment Protection Grade III drinking water quality standards (GB3838-2002); thus, the final service levels are as follows: total nitrogen (TN) ≤1.0 mg/L and total phosphorus (TP) ≤0.2 mg/L. Lastly, we evaluated landscape aesthetics using a questionnaire-based survey to solicit information on public perceptions. We asked visitors and residents to score landscape aesthetics as (1) very unattractive, (2) unattractive, (3) okay, (4) beautiful, or (5) very beautiful. We selected the final service levels as visitor scores of (4) beautiful and (5) very beautiful. The use of legitimate endpoints grounded our assessment in Beijing's institutional context, thereby making the assessment relevant to management needs.

Land-use and land-cover change data

We compared U.S. Landsat satellite images for two time periods: (1) 22 September 2009 (Landsat 5 TM) for pre-Corridor; (2) 1 September 2013 (Landsat 8) for post-Corridor. Image preprocessing, classification, and classification accuracy assessment were performed using ArcGIS 10.1 (ESRI, Redlands, California, USA) and ERDAS IMAGINE 9.3 (Hexagon Geospatial, Norcross, Georgia, USA). A hybrid unsupervised and supervised approach was performed on the four images to classify the land-use and land-cover (LULC) into seven categories: (1) deciduous trees, (2) water, (3) grass, (4) wetland, (5) cropland, (6) urban, and (7) bare soil. We evaluated the accuracy of the LULC using: (a) field verification via global positioning system (GPS) points matching randomly generated reference points, and (b) high-resolution images on Google Earth and Baidu. For local-scale maps, overall accuracies were 90% (pre-Corridor) and 92% (post-Corridor) with kappa statistics of 0.83 (pre-Corridor) and 0.90 (post-Corridor). For regional-scale maps, overall accuracies were 94% (pre-Corridor) and 96% (post-Corridor) with kappa statistics of 0.91 (pre-Corridor) and 0.94 (post-Corridor). We used the LULC data to parameterize the Variable Infiltration Capacity model and quantify ecosystem area changes.

Variable Infiltration Capacity model

We chose the Variable Infiltration Capacity model (VIC) version 4.2.c because it can simulate the energy and water balances of lakes and wetlands to estimate the metrics listed in Table 1. Variable Infiltration Capacity is able to generate dynamic estimates of lake and wetland areas, water volumes, lake depths, and ET under different LULCs and climates (Bowling and Lettenmaier 2010). Variable Infiltration Capacity is one of the most popular hydrology models in the world for managing water resources and conducting hydrologic studies. First, the user defines the LULC (fraction of each LULC type) and the spatial scale of the grid cells (VIC assumes uniform grid cells in terms of topography and soil type), and the simulation is run for each grid cell separately. At the regional scale, we used one grid cell to represent the full Corridor (lakes/wetlands modeled as one water body). At the local scale, we modeled each lake and wetlands as individual grid cells.

We parameterized VIC using daily meteorological data (max and min air temperature, relative humidity, wind speed, and precipitation) from the Mentougou meteorological station (<1 km from Yongding Corridor) for the pre- and post-Corridor periods. Furthermore, we measured hourly air temperature and relative humidity at five lakes and wetlands from 9 November 2012 to 30 June 2013. We installed Hobo U23 Pro V2 data-loggers (Onset Computer Corps, Bourne, Massachusetts, USA) at (1) Mencheng Lake, (2) Wetlands, (3) Lianshi Lake, (4) Xiaoyue Lake, and (5) Wanping Lake. No data-loggers were deployed at the Garden Expo Lake and Daning Reservoir due to prohibited access. We placed the Hobos inside radiation shields at a height of 2 m above the ground in shoreline trees along the lakes and wetlands. Data-loggers were set to record air temperature and relative humidity at 30-min intervals. Accuracy of the instrument as per the manufacturer is temperature ±0.2°C at 0–50°C with resolution of 0.02°C; RH: ±2.5% from 10 to 90% with a resolution of 0.03%. We calibrated the data-loggers before installation and downloaded the data each month. We combined the measured hourly air temperature and relative humidity data with the Mentougou meteorological data. Variable Infiltration Capacity allows the user to incorporate multiple meteorological datasets at hourly or daily time steps. We also used the hourly air temperature and relative humidity data to calculate hourly HI values (see Local climate regulation subsection for the HI equation).

Inflow rates and basin depth: Area values were determined using engineering values, and the global VIC dataset for vegetation parameters (Appendix S1: Table S1). Unexpected inflow fluctuations occurred in 2012–2013, but inflow data were not measured for 2012–2013. Thus, we calibrated the model by altering the ideal inflow rates using engineering values to obtain the measured lake areas for September 2013 (Appendix S1: Table S2). We evaluated model accuracy by comparing our modeled water temperatures to measured monthly water temperatures. We determined that 80% of the modeled water temperatures fell within the 95% confidence level, thereby meeting the necessary accuracy requirements (Wong et al. 2017).

Ecological production functions

Water storage

We used three final service indicators: (1) mean annual lake volumes, (2) mean annual water areas, and (3) water loss factor (Table 1). First, we calculated shortfalls taking the difference between VIC estimates of mean annual water volumes and areas from 1 June 2012 to 30 June 2013 and BWA targets. We define the water loss factor (dimensionless) as the ratio of the total annual lake evaporation (expressed as a volume) to mean annual lake volume. We establish an EPF linking the water loss factor to lake depth as an ordinary least squares (OLS) regression (N = 16). We determined a power law relationship between lake depth and lake evaporation where evaporation rates increase exponentially at very low depths (≤0.8 m; Wong et al. 2017). We log-transformed the water loss factor, so our EPF for water storage is as follows (Wong et al. 2017):

$$log\text{WaterLoss}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1\text{Depth}+\mathrm{\epsilon }$$(1)

where Depth is lake depth (m) and WaterLoss is water loss factor (total annual lake evaporation (m³)/mean annual lake volume (m³)).

Local climate regulation

We measured hourly HI values using the Beijing Meteorological Bureau HI equation (Wang and Gong 2010) to determine diurnal variations at each lake and wetland:

$$\text{HI}=T-0.55\left(1-\text{RH}\right)\left(T-14.5\right)$$(2)

where HI is the hourly heat index value, T is hourly air temperature (°C), and RH is hourly relative humidity (fraction) measured using data-loggers. We calculated hourly HI values from 9 November 2012 to 30 June 2013 to determine the number of sultry events (i.e., climate shortfalls) at (1) Mencheng Lake, (2) Wetlands, (3) Lianshi Lake, (4) Xiaoyue Lake, and (5) Wanping Lake.

We estimated the change in ET from the addition of the lakes/wetlands using four VIC simulations: (1) no lakes/wetlands using pre-Corridor climate (pre,nolakes); (2) no lakes/wetlands using post-Corridor climate (post,nolakes); (3) lakes/wetlands using pre-Corridor climate (pre,lakes); and (4) lakes/wetlands using post-Corridor climate (post,lakes). We estimate the addition of ET from the new lakes/wetlands at local and regional scales:

$$\text{TotalET}={\text{ET}}_{\mathrm{post},\mathrm{lakes}}-{\text{ET}}_{\mathrm{pre},\mathrm{nolakes}}$$(3)

$$\text{Climate}={\text{ET}}_{\mathrm{post},\mathrm{nolakes}}-{\text{ET}}_{\mathrm{pre},\mathrm{nolakes}}$$(4)

$$\text{Lake}=\text{TotalET}-\text{Climate}={\text{ET}}_{\mathrm{post},\mathrm{lakes}}-{\text{ET}}_{\mathrm{post},\mathrm{nolakes}}$$(5)

where TotalET is the total change in mean hourly ET between periods; Climate is the amount of hourly ET from climatic changes between periods (i.e., climate effect); Lake is the amount of hourly ET from the addition of the lakes and wetlands between periods (i.e., lake effect; see Wong et al. 2017 for further details on the VIC modeling).

Next, we created the local climate regulation EPFs as OLS regressions linking VIC ET results to measured HI values for June 2013 (June is surrogate for summer conditions). We estimated EPFs for (1) Mencheng Lake, (2) Wetlands, (3) Lianshi Lake, (4) Xiaoyue Lake, and (5) Wanping Lake. At each site, we estimated an EPF for hourly daytime values defined as 06:00–20:00 (N = 480 per site); EPF using hourly nighttime values defined as 21:00–05:00 (N = 240 per site). Our local climate regulation EPF is as follows:

$$\text{HI}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1\text{ET}+{\mathrm{\beta }}_2\text{Sensible}+{\mathrm{\beta }}_3\text{Pressure}+\mathrm{\epsilon }$$(6)

where HI is heat index (h⁻¹), ET is evapotranspiration (mm/h), Sensible is sensible heat (W·m⁻²·h⁻¹), and Pressure is pressure (kPa; Wong et al. 2017).

Water purification

We measured water quality at the lakes and wetlands once a month at 20 sites from March 2013 to August 2013 (Fig. 4). At each site, one water sample was taken from the shoreline using common water grab sampling techniques, and one water sample was taken 6 m from the shoreline using a telescoping pole. The water samples were placed in ice chests and stored in freezers. All samples were analyzed 2–3 d from the sampling date. We used spectrophotometric methods and UV PharmaSpec 1700 Shimadzu (Shimadzu Corp, Kyoto, Japan) to assess TN and TP. We followed the standard water quality methods of China's Ministry of Ecology and Environment in order to compare our results to the national water quality standards (GB 3838-2002). First, we calculated shortfalls per lake and wetlands comparing mean monthly nutrient levels to the drinking water quality endpoints. Nutrients enter directly into the wetlands; thus, we separated sites into (1) upstream (N = 6 sites); (2) wetlands (N = 7 sites); and (3) Lianshi Lake (N = 2 sites; downstream of wetlands; Fig. 4). We estimated nutrient loading as the difference in mean nutrient levels between the upstream and wetlands. We measured wetland nutrient retention as the difference in mean nutrient levels between the Wetlands and Lianshi Lake.

Figure 4

Water quality monitoring sites. (a) US Landsat 8 image of the Yongding River Green Ecological Corridor on 1 September 2013, illustrating the Beijing Water Authority (BWA) sections for water quality monitoring (no monitoring sites at Garden Expo Lake and Daning Reservoir due to construction): (1) Upper Mencheng Lake, (2) Lower Mencheng Lake, (3) Upper Lianshi Lake, (4) Lower Lianshi Lake, (5) Xiaoyue Lake, and (6) Wanping Lake. (b) Map of our 20 sampling sites to measure lake water quality, nutrient loading, and nutrient retention from March 2013 to August 2013.

The water purification EPF is an OLS regression linking Variable Infiltration Capacity wetland area estimates and measured nutrient loading to measured Lianshi Lake water quality (N = 6):

$$\text{LianshiLake}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1\text{Area}+{\mathrm{\beta }}_2\text{Loading}+\mathrm{\epsilon }$$(7)

where Area is average monthly wetland area (ha), Loading is average TN and TP loading concentrations (mg/L), and LianshiLake is average TN and TP (mg/L) in Lianshi Lake.

Aesthetics

From April 2013 to September 2013, we conducted visitor surveys each month at either Mencheng Lake or Wanping Lake (most popular parks). Our survey consisted of 22 questions (Appendix S1), spanning demographics (e.g., gender, income level), visitation (e.g., trip frequency, visit purpose), ecosystem service preferences, and perceptions of the Yongding Corridor (e.g., scenic beauty, environmental quality). Survey dates were selected randomly, covering weekdays and weekends to obtain a representative sample of visitors. We conducted the surveys in the morning and afternoon (before 6 pm) at Mencheng Lake for April and June–August 2013. For May 2013, we conducted the surveys in the afternoon (before 6 pm) at Wanping Lake. For September 2013, we conducted the surveys in the evening (after 6 pm) at Mencheng Lake. Peak park usage is in the summer and fall on the weekends, and nighttime on weekdays and weekends because most families visit the parks after dinner. Arizona State University's Institutional Review Board approved of all materials, methods, and questions. Each survey visit included at least two surveyors over a period of 2–4 h. Surveyors invited visitors over 12 yr old to participate. Informed consent was obtained after the nature of the study was explained to each participant. Participants were given the option of completing the questionnaire as self-administrated or surveyor-administrated. Those who chose surveyor-administrated listened to questions in Mandarin, and surveyors offered no guidance to participants. A total of 193 respondents participated in the population survey. Respondents were first asked to score landscape aesthetics and then environmental quality and overall trip satisfaction. The order in which questions were presented is important to minimize bias in answers.

We first calculated shortfalls as the difference in visitor scores of scenic beauty from the desired levels (Table 1). Second, we created the aesthetics EPF linking the final service of water purification and local climate regulation to aesthetics as intermediate services because studies have empirically shown these factors can influence people's experiences of parks (Nassauer et al. 2001, Steinwender et al. 2008, Hipp and Ogunseitan 2011). Landscape aesthetics is the interaction between biophysical features and the perceptual and judgmental processes of the human viewer (Daniel 2001); thus, we cannot simply use environmental monitoring data. We have to consider how people perceive environmental quality, thereby making visitor scores of water quality and climate the ecosystem characteristic metrics for this cultural service. Perceptions of water quality and climate are possible mechanisms influencing scenic beauty (Steinwender et al. 2008). Researchers often use ordinal logistic regressions to estimate the impact of people's perceptions of environmental quality on scenic beauty and relaxation. Hipp and Ogunseitan (2011) showed the probability of park visitors experiencing a relaxing and pleasurable environment increased by 78% when they perceived healthy or very healthy water. Similar to Hipp and Ogunseitan (2011), we created the aesthetics EPF as an ordinal logistic regression, linking perceived environmental quality (regulating services) to perceived scenic beauty (N = 193):

$$\begin{array}{cc}\hfill ln\left(\frac{\text{prob(Aesthetics)}}{1-\text{prob(Aesthetics)}}\right)=& {\mathrm{\beta }}_0+{\mathrm{\beta }}_1\text{Water}\hfill \\ & +{\mathrm{\beta }}_2\text{Cooling}\hfill \\ & +{\mathrm{\beta }}_3\text{Satisfaction}+\mathrm{\epsilon }\hfill \end{array}$$(8)

where Aesthetics is visitor scores of scenic beauty. Water is visitor scores of water quality: very unhealthy (value = 1), unhealthy (value = 2), moderate (value = 3), healthy (value = 4), and very healthy (value = 5). Cooling is visitor scores of climate described as very hot (value = 1), hot (value = 2), warm (value = 3), cool (value = 4), and cold (value = 5). Satisfaction variable is overall trip experience described as very unpleasant (value = 1), unpleasant (value = 2), okay (value = 3), enjoyable (value = 4), and very enjoyable (value = 5). We used Satisfaction because people's feelings about their trip experience have shown to influence ratings of scenic beauty.

Statistical analysis

We created OLS regression models using Stata 12.1 (StataCorp, College Station, Texas, USA) to create EPFs for water storage, local climate regulation, and water purification. All OLS models were tested for their explanatory power using the Ramsey RESET test, link test, and variance inflation factor. For aesthetics, we used an ordinal logistic regression model since the variables from the surveys are ordinal. We tested the proportional odds assumption using the omodel test in Stata and assessed the goodness of fit to ensure our aesthetics model was suitable. The regression models listed showed the best predictive power given these statistical tests.

Results

Ecosystem services

We found the Yongding Corridor met target levels for landscape aesthetics but fell short on the other ecosystem services. Policymakers want the green infrastructure to enhance regional environmental quality; however, we determined the designed ecosystems are providing local benefits with near negligible regional impacts. The beneficiaries are mainly local residents who visit the parks, and the majority of them are low income. From 2009 to 2013, we determined the largest LULC change was a 935% (5.33 km²) increase in water area on the Yongding River in Beijing. At the regional scale, urban area increased by 52% (93.02 km²), deciduous trees increased by 9% (52.85 km²), and water area increased by 328% (6.53 km²). Despite the rapid rise in water area, we found the water class was <1% of total regional area while the urban class was 23% of total regional area in 2013. The LULC changes in the four-year period highlight the importance of evaluating the ecosystem services across spatial scales. Managers need to consider how to establish and protect the designed ecosystems given rapid regional urbanization.

We identified substantial shortfalls for water storage, shown in Fig. 5a: (1) −6.42 million m³ total mean annual lake volume (more than 50% reduction in water volume); (2) −1.46 km² total mean annual water area. We created a water storage production function linking our water loss factor to lake depth (m). We calculated a statistically significant coefficient for the Depth variable β₁ = −0.48 (P < 0.01), which translates to $(e^{-0.48}-1)\times 100=-38\%$ (Table 3, Fig. 6; Appendix S2: Table S1). We used the regression coefficient to estimate the water storage service as a one meter increase in lake depth leads to a 38% decrease in fractional water loss in this system. From our modeling, we also determined that the water loss factor for the lakes was 0.81 for 2012–2013 (Wong et al. 2017), but to obtain the desired total lake volume >12.05 million m³ (final service level) requires reducing the water loss factor to 0.37. Using our water storage EPF results, we estimate managers have to make mean lake depth ≥1.6 m (i.e., create deeper lakes) to obtain a water loss factor of 0.37. This equates to a 54% reduction in water loss to reach the desired water storage targets.

Figure 5

Graphs showing the ecosystem services shortfalls: (a) total mean annual lake volume (million m³) and total mean annual lake area (km²) shown in blue and water storage deficits shown in red; (b) total number of sultry events shown in red from 9 November 2012 to 30 June 2013; (c) mean total nitrogen concentration (mg/L) and mean total phosphorus concentration (mg/L) at Lianshi Lake shown in tan and drinking water quality shortfalls shown in red; (d) scenic beauty scores shown in purple as the percent of respondents who scored the Yongding Corridor as beautiful or very beautiful there is no shortfall on scenic beauty. Abbreviation is FS, final ecosystem service levels.

Table 3. Summary of regression statistics for ordinary least squares regressions

Ecosystem service (ES)	R ^2	Final ES indicators	Ecosystem characteristic metrics	Coefficients	Marginal ES
Water storage	0.97	Water loss	Lake depth	−0.48**	1 m increase in lake depth relates to 38% decrease in fractional water loss
Local climate regulation	Daytime: 0.76–0.87	Heat index	Evapotranspiration	−2.31 to −7.07**	1 m increase in hourly ET relates to 2.3 to 7.1 decrease in hourly HI values in daytime
	Nighttime: 0.80–0.88	Heat index	Evapotranspiration	−3.76 to −4.78**	1 m increase in hourly ET relates to 3.8 to 4.8 decrease in hourly HI values in nighttime
Water purification	Total nitrogen: 0.86	Total nitrogen	Wetland area & nutrient loading	Wetland area: −0.10** Loading: 0.41*	1 ha increase in wetland area reduces TN by 0.10 mg/L; 1 mg/L increase in nutrient load increases TN by 0.4 mg/L
	Total phosphorus: 0.93	Total phosphorus	Wetland area & nutrient loading	Wetland area: −0.01* Loading: 0.04	1 ha increase in wetland area reduces TP by 0.01 mg/L

• *P < 0.05, **P < 0.001.

Figure 6

Scatterplot of the lakes/wetlands at both the local and regional scales under ideal and altered inflow rates (N = 16) showing log fractional water loss (mean total annual lake evaporation/mean annual lake volume) to mean lake depth (m; Wong et al. 2017).

We measured climate shortfalls using Beijing's official HI to determine the number of sultry events (HI > 26) at each site (Fig. 5b). The mean total number of shortfalls is 70 sultry events per park from 9 November 2012 to 30 June 2013. For 2012–2013, we estimated total annual ET was 1115 mm (0.13 mm/h) at the local scale and 941 mm (0.11 mm/h) at the regional scale. Between the pre- and post-Corridor periods, we calculated ET increased by 677 mm/yr (0.08mm/h) at the local scale and 484 mm/yr (0.06 mm/h) at the regional scale (Wong et al. 2017). Next, we performed a series of model simulations to isolate the ecosystem effect from the climate effect. We calculated the new lakes and wetlands increased mean local ET by 0.03 mm/h with no increase in regional ET (Wong et al. 2017). The majority of the increase in ET between both periods was due to a 70% increase in precipitation in Beijing in 2012–2013.

To determine the impact of the new ecosystems on human comfort, we created local climate regulation EPFs linking daytime and nighttime HI and ET values for June 2013 (June is surrogate for summer conditions). Our local climate regulation EPFs were statistically significant (P < 0.01) at all sites. Marginal values are interpreted as follows for daytime HI, if hourly ET increases by 0.01 mm then HI decreases by 0.02–0.07 (Table 3, Fig. 7; Appendix S2: Table S2). For nighttime HI, if hourly ET increases by 0.01 mm then HI decreases by 0.04–0.05 (Table 3, Fig. 8; Appendix S2: Table S2). We determined the cooling rates from ET are statistically significant, but practically insignificant for improving human comfort. For these ecosystems to impact human comfort (reduce summer HI by more than 1 unit) requires an estimated 33-fold increase in ET given the lakes and wetlands only increased local ET by 0.03 mm/h.

Figure 7

Scatterplots relating mean hourly daytime heat index and mean hourly evapotranspiration (mm) for June 2013 for (a) Mencheng Lake, (b) Wetlands, (c) Lianshi Lake, (d) Xiaoyue Lake, and (e) Wanping Lake (Wong et al. 2017).

Figure 8

Scatterplots relating mean hourly nighttime heat index and mean hourly evapotranspiration (mm) for June 2013 for (a) Mencheng Lake, (b) Wetlands, (c) Lianshi Lake, (d) Xiaoyue Lake, and (e) Wanping Lake (Wong et al. 2017).

We examined water purification by measuring TN and TP levels across the lakes and wetlands to quantify nutrient retention and nutrient loading. First, we found the nutrient levels at Lianshi Lake were far greater than the TN and TP final service levels for drinking water quality (Grade III standard) since they were on average higher than Grade V (no permitted uses; Fig. 5c). During our fieldwork, we identified the pollution enters the Yongding Corridor directly as domestic sewage from nearby shoreline homes. We calculated a mean nutrient load concentration of 18.9 mg/L for TN and 1.8 mg/L for TP; however, average wetland nutrient retention was 61% for TN and 66% for TP. Despite the high nutrient retention from the wetlands, Lianshi Lake (located immediately downstream of the wetlands) had an average TN concentration of 8.4 mg/L and TP concentration of 0.6 mg/L. Therefore under the current wetland area, we estimate managers need to reduce TN by an additional 88% and TP by 68% to meet the final service levels. We created two EPFs for water purification linking Lianshi Lake TN and TP levels to wetland area and nutrient load. Our water purification coefficients are statistically significant (P < 0.05), suggesting a one-hectare increase in wetland area reduces TN by 0.10 mg/L and TP by 0.01 mg/L at Lianshi Lake. Alternatively, a one mg/L increase in nutrient load increases TN by 0.4 mg/L at Lianshi Lake (Table 3, Fig. 9; Appendix S2: Table S3). We estimate managers need to increase wetland area by 50% (40 ha increase) to obtain the required TP level while also reducing the nutrient load by 35% to reach the TN level (6.6 mg/L TN decrease).

Figure 9

Scatterplots relating Lianshi Lake water quality to: (a) mean total nitrogen loading (mg/L); (b) mean total phosphorus loading (mg/L); (c, d) wetland area (ha).

We assessed the relationship between the regulating ecosystem services and aesthetics by linking visitor perceptions of environmental quality to scenic beauty scores. We determined the majority of visitors found the Yongding Corridor as beautiful (46%) or very beautiful (36%); thus, we identified no shortfall in scenic beauty (Fig. 5d). Our aesthetics EPF is statistically significant (P < 0.01) where results suggest the probability of visitors ranking the Yongding Corridor as very beautiful is significantly greater when water quality is considered very healthy (probability is 64%, holding other variables at their means) and climate cold (probability is 52%, holding other variables at their means; Table 4, Fig. 10).

Table 4. Summary of statistics for ordinal logistic regression using predicted probabilities for an aesthetics score of “very beautiful.”

Ranking	Water	Cooling	Satisfaction	Marginal ES
1	0.10 (0.05)*	0.09 (0.05)	0.02 (0.02)	Logistic regression is statistically significant where the probability of visitors ranking the Yongding Corridor as very beautiful is 64% more likely if visitors perceived water quality as very healthy; 52% more likely if visitors perceived climate as cold; or 52% more likely if visitors perceived the trip as very enjoyable
2	0.18 (0.05)**	0.16 (0.05)**	0.05 (0.03)
3	0.31 (0.04)**	0.25 (0.04)**	0.12 (0.04)**
4	0.47 (0.06)**	0.38 (0.04)**	0.28 (0.04)**
5	0.64 (0.10)**	0.52 (0.08)**	0.52 (0.06)**

Notes

• Water variable rankings for water quality are 1, very unhealthy; 2, unhealthy; 3, moderate; 4, healthy; 5, very healthy. Cooling variable rankings for climate are 1, very hot; 2, hot; 3, warm; 4, cool; 5, cold. Satisfaction variable rankings for trip satisfaction are 1, very unpleasant; 2, unpleasant; 3, ok; 4, enjoyable; 5, very enjoyable.
• *P < 0.05, **P < 0.01, standard error shown in parentheses.

Figure 10

Predicted probabilities for obtaining aesthetics score of very beautiful under (a) different water quality rankings with 95% confidence intervals; (b) different climate rankings with 95% confidence intervals; (c) different trip satisfaction rankings with 95% confidence intervals. Asterisks indicate statistically significant probabilities at P < 0.05.

Despite the shortfalls, the overall environmental condition on the Yongding River improved from the previous condition because the Yongding River was highly degraded. However, Beijingers are concerned about whether or not managers can increase (or maintain) enhancements since stated concerns mirror shortfalls. Next, we use the marginal ecosystem service values to evaluate synergies and tradeoffs to identify actions for enhancing multi-functional performance (Fig. 11).

Figure 11

We used the statistically significant regression coefficients from the ecological production functions (P < 0.05) to estimate the marginal service values to assess synergies and tradeoffs in terms of management feasibility for meeting final services (i.e., reduce shortfalls).

Synergies and tradeoffs

A biophysical tradeoff between reducing water loss to sustain water supplies and increasing evaporative cooling to improve human comfort exists; however, we determined the tradeoff is insignificant in practice (Fig. 11). Managers must either maintain inflow at ideal rates and/or increase lake depth by more than 1.6 m to reach the final service levels. Water loss from ET must increase by over 3300% to improve human comfort. This translates to ~168 km² increase in water area (water area = 5.1 km² for 2012–2013), which is not feasible due to regional water scarcity. Wetlands are providing high nutrient retention, but the nutrient load is far greater than the wetland capacity to meet the drinking water quality standards. Given space limitations, it is difficult to double wetland area, and thus, effluent must be dramatically reduced. We estimate no expansion of wetland area would require ~75% decrease in nutrient load to obtain the target levels on water quality. Lastly, we found significant relationships between visitor perceptions of water quality and climate and perceptions of scenic beauty where the probability of people perceiving very beautiful scenery increased by 64% when they perceive very healthy water or 52% when they perceive cold climate. Furthermore, when we surveyed visitors the problems of greatest concern were the three regulating services. There is likely a synergistic relationship between water storage, local climate regulation, and water purification and scenic beauty. These findings suggest the public perceive the shortfalls on the regulating services, and based on our aesthetics EPF results, this could threaten favorable perceptions of scenic beauty over time.

From these relationships, we identify actions for improving the effectiveness of the Yongding Corridor: (1) maintain consistent inflow and/or make the lakes deeper to sustain the lakes and wetlands; (2) reduce the nutrient load to improve water quality; and (3) plant shade trees and/or construct shade structures to improve human comfort. Each recommendation addresses a specific shortfall on the regulating services, which we determined is likely important for maintaining and enhancing aesthetics over time. Policymakers identified multiple ecosystem services, yet stakeholders clearly were targeting aesthetics. The ecosystem services assessment illustrates the importance of managing all four ecosystem services, considering for the first time ecological functionality. We presented the results to managers who felt the ecosystem services information was very useful because it clarified connections. Our ecosystem services information was incorporated into Beijing's Five-Year Water Resources Sustainable Use Plan (2016–2020), and the Beijing Municipal Commission of Development and Reform is using the information to formulate guidelines for other green infrastructure projects. This is significant because Beijing plans to double its use of green infrastructure by 2030, which is quickly becoming a popular municipal trend throughout China.

Discussion

Green infrastructure performance

Populous countries like China sit on the frontlines of crafting sustainable cities where ecological footprints are reduced and local ecosystem services are enhanced. Green infrastructure is seen as useful for helping nations achieve the United Nations Sustainable Development Goals (SDG), such as clean water and sanitation (SDG 6), sustainable cities and communities (SDG 11), and ecosystem conservation (SDGs 14 and 15). Cities around the world are pursuing green infrastructure projects like the Yongding Corridor for the same ecosystem services examined in this study. Examples include Singapore's vertical gardens and river projects, Berlin's riverbanks, Melbourne's urban forests, Philadelphia's network of rain gardens, and Tianjin's constructed wetlands (Liu and Bergen 2018). Cities, however, lack monitoring systems to measure ecosystem service outcomes to determine whether green infrastructure can deliver multiple ecosystem services at required levels.

Currently, there is an ongoing debate in science and policy about the potential of green infrastructure to provide essential services like clean drinking water and climate regulation (Muller et al. 2015, Palmer et al. 2015). From our analysis, we found the Yongding Corridor was meeting the required levels for aesthetics but fell short on the regulating services. The overall environment on the Yongding River in Beijing improved from the green infrastructure because the Yongding River was highly degraded. However despite improvements in aesthetics, the Yongding Corridor was not supplying ecosystem services at the required levels for water supply, a comfortable summer environment, and drinkable water. From working with managers, we determined the shortfalls occurred because ecosystem structure and functions were not considered in the design of the ecosystems. We found the lakes and wetlands were highly inefficient for storing water in Beijing's arid environment. They were too shallow, which made them vulnerable to drying. Landscape architects want a water landscape that looks expansive with minimal water investment. However, we determined they could save more water overtime while reducing vulnerability to drying by making the lakes deeper given Beijing's climate (Wong et al. 2017). Lake drying also threatens the visible condition of the landscape noted as a top concern by visitors. We also determined the lakes and wetlands were having no measurable impact on human comfort in the summer; thus, the biophysical tradeoff on reducing water loss and increasing ET was not translating into a social tradeoff. This finding is important because it helps managers prioritize services and actions for improvement. The Yongding Corridor was designed to have minimal shoreline trees. Designers intended for the main source of cooling to be evaporation from the lakes and wetlands. Given our results, we believe shade trees as suggested by visitors likely would have a far greater cooling effect than relying on the evaporation of the lakes and wetlands alone. In the summer, we noticed most visitors stayed under the bridges for shade because there were few shade trees; thus, they were unable to engage with the landscape. Lastly, we found the wetlands were providing high nutrient retention, but the nutrient load was far greater than the wetland capacity to purify the water for drinking water quality. Hence, we determined a strong statistical relationship between visitor perceptions of water quality and scenic beauty because the poor water quality was creating algal blooms and odors.

Adaptive management

From our experience, we found ecosystem service assessments can help promote adaptive management—a means of improving system knowledge, by testing assumptions against experience (Lee 1993)—an institutional necessity for sustainable development. Previously, policymakers assumed the desired benefits would materialize with the addition of the seven lakes and wetlands. A common assumption is that green infrastructure equals sustainable development. Our control of nature is limited even with designed ecosystems; thus, we need monitoring to understand how to support ecosystem processes for final services (Lindborg et al. 2017). Our results show the importance of scale since policymakers and planners felt certain these ecosystems would produce regional benefits. However, we found the ecosystem services from the Yongding Corridor are occurring locally. Furthermore, the ecosystem services information helped managers understand the importance of the regulating services where ecosystem functionality is considered in green infrastructure design and management. Lastly, managers stated the ecosystem services information helped them understand how the other services support aesthetics.

Since 2016, managers started implementing the ecosystem services information to modify management actions. We have not conducted any recent measurements to assess ecological and social changes from the new modifications. However, we identified visual changes in the lakes and wetlands using high-resolution satellite images from Google Earth. We observed substantial increases in vegetation (trees and wetlands) from 2013 to 2016 (Fig. 12). In 2013, shorelines had very few shade trees and they were mainly bare ground. We found a substantial visual increase in shoreline trees at all sites in 2016. Prior to our assessment, managers thought the poor water quality was due to engineering challenges on inflow rates leading to low water levels (Lü et al. 2012). They did not account for nutrient sources outside the Yongding Corridor, and how nutrient loads from stormwater may impact nutrient retention. When comparing Google Earth images in summer and fall from 2013 to 2015 and 2016, we saw substantial reductions in algae where nutrients enter the system (Fig. 12). Managers are working to reduce nutrient loading and increase wetland area. Lastly, the Beijing Government is acknowledging the importance of maintaining inflow rates for water storage by stating: “it is essential to secure the ecological water demand of the [Yongding River] in the restoration, during which Beijing will annually pour into the river 75 million cubic meters of reclaimed water for replenishment” (Gu 2017).

Figure 12

Satellite images of management changes to improve effectiveness. (a) Mencheng Lake with few shoreline trees on 28 October 2013. (b) Mencheng Lake with increased shoreline trees on 30 October 2016. (c) Wetlands on 17 October 2014 showing algae covering surface water area due to high nutrient levels. (d) Wetlands on 30 October 2016 showing no apparent algae likely due to efforts to reduce nutrient loading. Google Earth 7.1. 2017. Yongding River and Beijing 39°57′25.59″–39°55′12.18″ N, 116°05′37.92″–116°08′07.06″ E, elevation 105–88M. 3D Buildings data layer. Map data: Google, CNES/Airbus, DigitalGlobe, viewed 26 August 2018. http://www.google.com/earth/index.html.

Stakeholders identified regulating services, but like most greening projects, the focus was aesthetics. The ecosystem service information explained the importance of managing water storage, water purification, and local climate regulation, and their influence on aesthetics. On 30 March 2017, the Beijing Government publicly announced new policy goals for the Yongding River: “Gradually, the Yongding River will have the functions of flood control, water resources storage and purification, local climate regulation, leisure and entertainment” (Gu 2017). For the first time, the government is making ecological functions (i.e., water resources storage, water purification, and local climate regulation) explicit policy goals on par with engineering (i.e., flood control) and cultural functions (i.e., leisure and entertainment). We believe the ecosystem service analysis helped elevate the importance of these services and the value of ongoing management (monitoring, learning, and action). It is far too soon, however, to determine the degree of the institutional changes and whether these policy changes can produce significant improvements on the Yongding River. But a notable institutional shift is the recent decision to craft a longer-term plan (10 yr) where provincial (Beijing, Tianjin, Hebei, and Shanxi) and municipal governments establish: “comprehensive treatment and ecological restoration of the Yongding River [to formulate the] first cross-provincial program for river treatment in northern China” (Gu 2017).

Constraints of green infrastructure

Our findings also support the conclusions of other studies, suggesting the most effective approach for cities is to use a mixture of gray and green infrastructure to produce the required ecosystem service flows (Muller et al. 2015, Palmer et al. 2015). In 2013, several residential communities surrounding the Yongding River in Beijing lacked sewage systems. This portion of Beijing was underdeveloped in roads and plumbing. Hence, people dumped their waste directly either into the lakes or in the streets. Managers created the wetlands to provide tertiary treatment for recycled water from wastewater treatment plants, but they did not account for nutrient loading from shoreline homes. Fixing this problem requires improving sewage systems, creating retention basins, and community involvement. Similar considerations have been observed on the Thames River in London where green infrastructure alone was deemed technically infeasible for controlling water pollution (Muller et al. 2015). The challenge is improving how cities combine gray and green infrastructure to support ecological functionality for ecosystem services. Lastly, an important future research topic is evaluating the overall sustainability of green infrastructure projects. The Yongding Corridor relies heavily on built infrastructure to import water from wastewater treatment plants and pumps to circulate water. However, the Yongding Corridor also provides much-needed green space to impoverished communities whose local environments were highly degraded. The communities living along the Yongding River endured decades of pollution, dust, and trash with few green spaces compared to wealthier districts in Beijing. We need sustainability analyses of green infrastructure considering the various social, economic, and ecological dimensions to improve its utility.

Lessons learned

Several challenges limit full implementation of ecosystem service assessments in policy, but progress on generating marginal values is possible. Currently, most urban studies quantify ecosystem services by measuring either ecological processes or final services. As we learned in this study, policymakers need measurements connecting ecosystem characteristics to final services. However, using fragmented, disciplinary knowledge is insufficient to address the data gap (Norgaard 2008). Integrated thinking is required—knowledge of how to connect issues, and skills to identify strategic actions on connections. Relationships between ecological processes, environmental quality, and human benefits are complex. However, we are finding we can create EPFs with relatively simple datasets. The difficulty is practicing integrated thinking to technically combine methods and data to formulate analytical frameworks like Fig. 1. From our experience, we identify two main challenges for creating EPFs. First is selecting appropriate final services by translating policy goals (qualitative statements) to final services. This is difficult since it requires training policymakers on final services and natural scientists on social science criteria. Second is the lack of technical expertise to perform the modeling and acquire the necessary data to create EPFs for multiple services across scales. Evaluating multiple services is very difficult as highlighted in this study. One first has to understand a diversity of methods, and then, one has to craft an approach for unifying them. In this study, we show how we created simple regression models as a first step on establishing EPFs for multiple services; however, we believe EPFs will improve as understanding of the concept improves in ecology.

Lastly, we learned the importance of working with stakeholders to enhance the legitimacy and usability of the science. We worked collaboratively with managers and surveyed visitors to ground our analysis in the institutional context of the project. Our biophysical results on the regulating services were supported by resident and visitor concerns. The inclusion of public opinion into the assessment acted as an important bridge between managers and residents. Visitor insights helped to legitimize the selection of the ecosystem services and our management recommendations, which led to their adoption. Managers had not consulted with residents or visitors on the design or implementation of the green infrastructure project. Hence, we found an ecosystem services assessment can provide an important social function as a boundary object between scientists, policymakers, and local communities.

Conclusion

The Yongding River Green Ecological Corridor represents the new types of greening initiatives being tested around the world to address urban sustainability challenges. Ecosystem service information is highly needed to help guide these ambitious urban projects. We found the ecosystem services approach useful for clarifying the importance of ecological functionality, which led managers to prioritize ecosystem functions in our designed ecosystems. However, we also learned about the limits of green infrastructure. Green infrastructure alone often is unable to generate ecosystem services at required levels. We need to better couple green and gray infrastructure to obtain multi-functional landscapes. Obstacles to urban sustainability are substantial, but we are learning progress is possible if we can help our institutions practice integrated thinking.

Acknowledgments

We would like to thank Kai Lee, Dennis Lettenmaier, Ted Bohn, and Ma Dongchun for their advice and suggestions throughout the research process, which were critical to drafting this manuscript. We also would like to thank the survey participants, and managers and scientists at the BWA and BWSTI for participating in this study. Our research was made possible due to the generous financial support of the National Science Foundation of China (71533005); Ford Foundation Predoctoral Fellowship; US National Science Foundation Graduate Research Fellowship (NSF DGE-1311230); Philanthropic Educational Organization Scholar Award; State Key Laboratory of Urban and Regional Ecology Research Grant (SKLURE2012-2-3).