Leveraging Greenspace to Manage Urban Flooding: An Investigation of Nature-Based Solutions Implementation in U.S. Public Parks

Abstract

Many cities are looking to adopt nature-based solutions (NBS) in greenspace to manage urban flooding and provide diverse co-benefits. Yet little research exists to inform the planning and design of park NBS. This study investigated NBS adoption in 58 public parks across major U.S. cities, using a 2022 survey by the Trust for Public Lands and other secondary datasets. We developed a typology to conceptualize a wide range of park NBS into five high-level categories by size/capacity, location of the gray–green spectrum, and design objectives. We then employed this typology to explore how a park’s adopted NBS types may relate to its landscape and sociodemographic contexts. We found that the most used type of NBS in the studied parks was ECO (a typology we defined as conserving, restoring, or creating ecosystems to mitigate flooding through ecological processes and functions), while the least used NBS type was ENG (a typology we defined as imitating natural infiltration processes but having no living elements). Further, parks that adopted ECO had significantly higher percentages of greenspace in the surrounding, as well as higher flood risks. We also found notable—though not statistically significant—evidence of potential associations between the type of NBS implementation in a park and its nearby neighborhoods’ income level, poverty, and population racial and age compositions. Moreover, our findings indicated that park visitors were more privileged compared to residents living near a park. We concluded that park contextual factors deserve more explicit consideration in the planning and design of NBS and discussed key implications of this study for practice and future research around park NBS for urban flooding.

1. Introduction

1.1. Parks, Nature-Based Solutions, and Urban Flood Management

Parks are long considered a critical place for city dwellers to experience and interact with nature. A growing body of research indicates that parks can provide various environmental and social benefits, including reducing urban heat island effects, treating stormwater and mitigating flooding, supporting biodiversity, creating opportunities for aesthetic appreciation and recreation, and improving physical, mental, and social health [1,2,3,4,5,6]. Given these benefits, parks have been identified as playing significant roles for urban sustainability and resilience [7,8,9]. At the same time, parks are also a key community asset that needs to be sustained amidst adverse climate change impacts such as flooding [10,11]. For example, after Hurricane Sandy, the New York City Parks developed a design and planning guideline for resilient waterfront parks in the city, discussing how parks can offer high-quality experiences while simultaneously possessing the ability to stand and recover quickly from small and large storms [12].

Urban flooding occurs when an excessive inflow of stormwater—which comes from heavy rainfalls or from coastal storm surge and high tides—fails to be infiltrated or carried away by drainage systems in a timely manner [13]. Due to climate change, urban development, and aging water infrastructure, many cities are challenged by more frequent and extreme flooding events and suffer from considerable direct (e.g., buildings, infrastructure, life, and properties) and indirect damages (e.g., detriments to public health and cohesion of community or disruptions of jobs and transportation) [14,15,16,17]. Improving the ability of urban communities to withstand, adapt to, and recover from the impacts of flooding has become a major socioeconomic and environmental concern for national, state, and local governments [13].

Nature-based solutions (NBS) have great potential for enhancing urban flood resilience. NBS broadly refers to planning, design, environmental management, and engineering practices that mimic, integrate, and enhance living systems to address societal challenges such as climate change [18,19,20,21]. In the context of urban flood management, related terms green infrastructure and green stormwater infrastructure (GI/GSI) have been widely used in the U.S., with a focus on nature-inspired engineering techniques that manage stormwater runoff [22]. In this study, we use the term “NBS” because of its inclusion of more diverse types of practices and its emphasis on adapting to climate change impacts such as urban flooding that is limited to the source of stormwater. Specifically, we consider a rich portfolio of NBS practices ranging from restored or created ecosystems (e.g., flood plains or shorelines) to hybrid measures that integrate natural and engineered elements (e.g., retention ponds or rain gardens) and engineered structures that mimic natural water cycling (e.g., pervious pavement and underground water tanks) that can be used to manage urban flooding [23,24,25]. Unlike conventional infrastructure that resists and removes water as best as possible, NBS are designed to hold water and accommodate temporary inundation to protect more vulnerable built areas. This can help cities better withstand storms and more quickly recover from extreme events, thus enhancing the flood resilience [26,27,28]. Furthermore, NBS can provide co-benefits such as improving water quality, supporting biodiversity, and offering opportunities for recreation, aesthetic appreciation, and environmental education [29,30].

Guidelines for introducing NBS in park development and renovation projects to manage urban flooding and provide co-benefits have been developed by many national and municipal agencies in the U.S., such as the Environment Protection Agency (EPA), the National Recreation and Park Association (NRPA), the American Planning Association (APA), the Detroit Water and Sewerage Department, and the Chicago Metropolitan Agency for Planning [31,32,33,34,35]. As these guidelines indicate, the adoption of NBS can enhance park functions beyond aesthetic and recreational uses to improve neighborhoods, which helps to leverage funding and secure parks as protected lands against urbanization [32,35]. This is particularly valuable to underserved communities that have less access to high-quality green amenities but are disproportionately affected by climate change impacts including urban flooding [31,32,34]. In addition, parks often have sizable lands and plants and soils in place that are suitable for adopting NBS practices, particularly practices like detention basins and constructed wetlands that require relatively large areas to function [24,31]. Public spaces like parks also play a critical role in promoting NBS by offering real-life experiences to residents, businesses, and local governments where they can see and feel the benefits [36,37]. Pleasant experiences and positive perceptions of NBS can greatly benefit wellbeing [38] and encourage public acceptance and support for NBS [39].

However, despite much interest in implementing NBS in parks for flood resilience in practice, little research exists to inform planning and design decisions. Tong et al. (2022) examined how NBS implementation has been combined with park facilities (e.g., pathways, playgrounds, and lawns) and outdoor activities in 23 Sponge City parks in Shanghai, suggesting that larger parks can employ a wider range of NBS practices without compromising park recreational services [40]. Feldman et al. (2019) estimated the effectiveness of a park rain garden to manage an adjacent street’s runoff and observed an average retention rate of 78% across 26 storms and a full retention for storms under 10 mm [41]. They concluded that parks are an untapped opportunity to manage stormwater and would significantly contribute to municipal stormwater management goals by implementing NBS practices like rain gardens. These previous studies highlight the importance of integrating NBS into parks to achieve environmental and social co-benefits; however, they focused mainly on the techniques and feasibility of NBS implementation, offering limited insights into the planning and design rationale that guides the choice of a park to introduce NBS as well as the specific practice to use.

1.2. Study Objective

This study aims to establish a better understanding of the status of NBS implementation in U.S. parks and how implementing NBS in a park may relate to its broader environmental and social contexts. A growing body of literature has researched how contextual factors can affect NBS’s benefits for local communities and identified important landscape and sociodemographic characteristics that should inform the planning and design of NBS at the city and neighborhood scales (e.g., [42,43,44,45]). For example, evaluating the multifunctionality of green infrastructure in Detroit, USA, Meerow and Newell (2019) found that GI measures can effectively provide co-benefits if they are sited in response to environmental and social contexts, even though the locations of current GI measures seemed to lack a basis on holistic considerations about these contexts [43]. To contribute to more systematic approaches to NBS implementation in urban parks, this study addressed the following questions: (1) What types of NBS practices have been implemented in parks to manage urban flooding? (2) How does a park’s implementation of NBS relate to its surrounding landscapes? Are there associations between a park’s landscape contexts and the type of NBS being used? (3) How does a park’s implementation of NBS relate to the sociodemographic characteristics of its users? Are there associations between a park’s sociodemographic contexts and the type of NBS being used? Our goal is to provide exploratory new insights by investigating some self-reported real-world NBS examples in specific parks, and not to provide a comprehensive examination.

2. Materials and Methods

2.1. Study Sample

We used the 2022 City Park Fact Survey by the Trust for Public Land to obtain parks that constituted the study sample. The Trust for Public Land is a U.S. nonprofit working to protect and create outdoor spaces for all. Since 2012, the organization has surveyed the 100 most populous U.S. cities every year to generate a comprehensive rating, the ParkScore index, for a city’s publicly accessible parks, trails, and open space [46]. This data source is used in many studies to examine various aspects of U.S. public parks such as public health outcomes [47] and greenspace equity [48]. The 2022 City Park Fact Survey, for the first time, included a section to inquire park and recreation agencies about examples of practices for building climate resilience that are employed within their park systems [49]. Specifically, five climate change impacts were considered: urban heat, stormwater flooding, coastal or riverine flooding, carbon emissions, and wildfire. This online survey was open from October 2021 to February 2022 [50]. We acquired the publicly available survey results from the Public Tableau (https://public.tableau.com/app/profile/will.klein/viz/ParksasCriticalClimateInfrastructure/Examples, accessed on 15 November 2022). In total, park and recreation agencies in 85 cities across 35 states reported 472 climate resilience-building practices that were implemented in parks and other public recreation facilities and greenspaces, and many of these examples were NBS practices [50].

The survey results dataset was organized by reported examples of climate resilience-building practices, with each row containing five types of descriptive text entries: (1) an overview, a category, and some additional information about the practice; (2) the name of the park where the practice was implemented; (3) the location of the practice (including geographical coordinates, city, and state); (4) the name of the agency that reported the practice to the survey; and (5) the type of climate change impact the practice aimed to address. For example, one practice included in the survey was using parks as a floodplain in Ann Morrison Park, Boise, ID. The entry for overview was “At Ann Morrison Park, Boise Parks and Recreation is reducing impacts from flooding through using parks as a floodplain”; the entry for additional information was “designed to accommodate seasonal river flooding”; the entry for category was “floodplain”; the entry for agency name was “Boise Parks and Recreation”; and the entry for climate change impact entry was “Flooding—Coastal/Riverine”.

Based on the survey data and our study objective, we conducted a systematic screening that involved three stages, each using clearly defined criteria to exclude irrelevant examples of climate resilience-building practices in the original dataset and to compile a sample of parks with NBS adoption for urban flood management (Figure 1).

In Stage 1, we first removed 270 examples of climate resilience-building practices from the survey that were not aimed at reducing stormwater or coastal/riverine flooding impacts. Second, we excluded 44 practices that were noted to be applied city-wide (the entry for park name of these practices was “city-wide”), rather than in a specific park. Third, we excluded 29 examples for which the overview entry did not contain any description about what practice was implemented. In Stage 2, we cross-validated a park’s name and location with Google Maps, as well as its adopted practices for reducing flooding impacts with the information on official websites (including those run by the park, the corresponding park and recreation department, or/and the associated landscape architecture design firm). We then excluded practices for urban flood management that met any of the following three criteria: (1) the practice was not located in an urban park but in entities such as golf courses, natural trails, nature reserves, and wildlife refuges; (2) the practice was reported for a park that was under construction, renovation, or planning and design when the survey was being conducted; or (3) the practice was from a greenspace network containing many parks. In Stage 3, we further excluded practices that are not nature-based (e.g., seawalls or levees) or involve general water management guidelines rather than specific NBS practices (e.g., minimizing grading during construction, recycling water for irrigation, or using native plants). The final study sample included 58 parks across the U.S., with a total of 65 NBS practices (Figure 2).

2.2. Variables for Landscape and Sociodemographic Contexts

We used secondary data to obtain variables for landscape and sociodemographic contexts. Variables were selected based on the existing literature as well as data availability from various sources. Table 1 lists the description and data source of each variable. All spatial data were processed in ArcGIS Pro 3.2.1.

Previous studies on the development of NBS at the city and neighborhood scales for urban flood management have highlighted landscape contextual factors, including the percentage of land covered by impervious surfaces and greenspace [42,43,51], and flood zones [42], and distance to waterways [45]. We used a 0.5-mile radius, a threshold for walking distance from a park [52,53], when examining the landscape contextual factors. We also examined flood risks within park boundaries. In addition, we included two variables for park characteristics that might also affect the implementation of NBS: park size, which was shown to be positively related to implementations of more diverse stormwater management practices [40]; and distance to the city center, which was shown to drive the abundance of water-sensitive urban design practices and is positively associated with larger practice sizes [45].

For sociodemographic contexts, we used measures from the 2016–2020 American Community Survey 5-Year Estimates at the level of census block groups (CBGs) as proxies. A CBG is a geographic unit that typically has 600 to 3000 residents and constitutes a subdivision of a Census Tract with relatively homogeneous sociodemographic status and living conditions [54]. It is also the smallest geographic entity on which the American Community Survey 5-year Estimates are available. Specifically, we examined variables for race, household income, and poverty, based on the existing literature that considered social justice related to NBS implementation [51,55,56]. We also examined age to account for the elderly and underage populations whose wellbeing can especially benefit from parks, and car availability to account for the ability to access parks.

Table 1. Variables for park landscape and sociodemographic contexts included in this study.

	Name	Description	Data Source
Landscape Contexts	Land cover	Land cover at a 100 m resolution classified as two types: urban vs. vegetated greenspace [%]	Copernicus Global Land Service [57]
	High-flood-risk areas	Flood hazard severity of surrounding and within park areas was regrouped into high and low risks based on FEMA’s classification [%] 1	FEMA Flood Map Service Center [58]
	Distance to waterbody	Euclidean distance from the geometric center of the park to the nearest water body [km]	ESRI USA Parks [59], Copernicus Global Land Service [57]
	Distance to city center	Distance from a park to the city center as measured by the geographic location [km]	ESRI USA Major Cities [60]
	Park size	Area of a park [ha] 2	ESRI USA Parks [59]
Sociodemographic Contexts	Race	Black, Asian, Hispanic, and White population in a CBG [%]	American Community Survey [61]
	Household income	Median household income (MHI) of a CBG normalized by the average MHI of the city [in USD]
	Poverty	% of households below the poverty line
	Age	Age, coded into four groups: under 18 years old, 18–40 years old, 40–65 years old, and over 65 years old.
	Car availability	Average number of automobiles owned per household

¹ The high-flood-risk group includes FEMA flood zones that are coded as AE, A, AH, AO, VE, which are all areas with a 1% annual chance of flooding; the low-flood-risk group includes FEMA flood zones that are coded as D or X, which are all other areas considered with low risks or possible but undetermined flood hazards. ² Park boundary was adjusted manually in ArcGIS when deemed as inaccurate after a visual inspection of each park.

Table 1. Variables for park landscape and sociodemographic contexts included in this study.

	Name	Description	Data Source
Landscape Contexts	Land cover	Land cover at a 100 m resolution classified as two types: urban vs. vegetated greenspace [%]	Copernicus Global Land Service [57]
	High-flood-risk areas	Flood hazard severity of surrounding and within park areas was regrouped into high and low risks based on FEMA’s classification [%] 1	FEMA Flood Map Service Center [58]
	Distance to waterbody	Euclidean distance from the geometric center of the park to the nearest water body [km]	ESRI USA Parks [59], Copernicus Global Land Service [57]
	Distance to city center	Distance from a park to the city center as measured by the geographic location [km]	ESRI USA Major Cities [60]
	Park size	Area of a park [ha] 2	ESRI USA Parks [59]
Sociodemographic Contexts	Race	Black, Asian, Hispanic, and White population in a CBG [%]	American Community Survey [61]
	Household income	Median household income (MHI) of a CBG normalized by the average MHI of the city [in USD]
	Poverty	% of households below the poverty line
	Age	Age, coded into four groups: under 18 years old, 18–40 years old, 40–65 years old, and over 65 years old.
	Car availability	Average number of automobiles owned per household

¹ The high-flood-risk group includes FEMA flood zones that are coded as AE, A, AH, AO, VE, which are all areas with a 1% annual chance of flooding; the low-flood-risk group includes FEMA flood zones that are coded as D or X, which are all other areas considered with low risks or possible but undetermined flood hazards. ² Park boundary was adjusted manually in ArcGIS when deemed as inaccurate after a visual inspection of each park.

We employed two approaches to identifying the CBGs to focus on. One approach was proximity-based and captures CBGs within a park’s 0.5-mile, walking-distance radius. The other is visitation-based and captures the CBGs where the park visitors originate. The proximity-based CBGs (PBC) are easy to identify and represent people living in a park’s surrounding neighborhoods. In comparison, the visitation-based CBGs (VBC), though more challenging to identify, represent people visiting the park and are likely to more accurately reflect the sociodemographic characteristics of actual park users.

We obtained information about park visitors’ origin CBGs in 2022 from Advan Patterns+, a data product that tracks phone-based mobility for designated points of interest (POIs) [62]. A subset of our studied parks (n = 51) were designated POIs in Advan Patterns. For these 51 parks, we mapped the CBGs from where park visitors had traveled, which are also known as the “catchment” that represents the geographic scope of a POI’s visitors. In addition, we also explored the visitation patterns of parks in the subsample (n = 51). We considered three visitation variables available in the Advan mobility data: (a) the average weekly visitation counts of a park; (b) the average ratio of weekly visitation counts to the number of individual visitors, which gauges the frequency of repeated visits to a park; and (c) the median dwell time, which measures the average time visitors spend at a park.

2.3. Data Analysis

2.3.1. NBS Typology

To address our first research question about what types of NBS are implemented in parks to manage urban flooding, we examined descriptions of NBS practices from the survey and identified high-level categories. To achieve this, we first recorded key words in the overview and additional information describing each practice (e.g., pervious pavement, or coastline restoration). We then cross-validated the key words through a visual inspection and interpretation of a park’s aerial view on Google Maps, as well as reading through the statements about flood management on the park’s official websites. Based on common-themed key words and the existing literature, we categorized the NBS practices into five general types (Figure 3). For example, using only artificial components and no living element to mimic infiltration, ENG are highly “gray” practices that provide few social and ecological benefits beyond flood management and runoff treatment. In contrast, ECO are highly “green” practices that often conserve or restore wetland and floodplain ecosystems, supporting biodiversity while managing flooding. SGSI, TEM, and POND are hybrid practices that consist of both artificial (e.g., overflow structures and outlet pipes) and living elements. They can have varied aesthetic, recreational, and ecological benefits depending on the specific design and management decisions for plants and landforms.

2.3.2. Statistical Analysis

Based on the developed typology, we analyzed the associations of the type of NBS practice used in a park and variables for its landscape and sociodemographic contexts to address our research questions about how contextual factors might affect NBS implementation in parks. We first calculated descriptive statistics for the five NBS types and the contextual factors by each type. We then employed one-way ANOVA to further test the statistical significance of the variations in landscape and socioeconomic contextual factors by NBS type. Further, we used the post hoc Tukey–Kramer test to better understand the pattern of pairwise differences among the five NBS types. The Tukey–Kramer test was chosen given the exploratory nature of this study (i.e., we did not have planned comparisons of certain NBS types based on current knowledge) as well as the unequal sample sizes [66]. Between-group and pairwise differences were considered statistically significant at p-value < 0.05. We also conducted a t-test to compare the distribution of socioeconomical characteristics of PBC and VBC across all parks, accounting for differences between visitors and adjacent residents. Significant differences were determined when p-value < 0.05. All statistical analysis was performed using Microsoft Excel 16.80 and Python 3.12.4, with descriptive statistics conducted in Microsoft Excel and ANOVA as well as the post hoc Tukey–Kramer test conducted using a Python package “Statsmodels” (Ver. 0.14.1) [67].

3. Results

3.1. Implementing NBS in Parks for Urban Flood Management

3.1.1. Park Geographic Scope and Flood Management Focus

The NBS practices in the 58 parks we focused on were reported by park and recreation agencies from 36 cities across 24 states in the U.S. These cities counted for 42.4% of the cities that responded to the survey and 36% of the most 100 populous cities in the U.S. that were invited to take the annual City Park Fact Survey. In addition, nearly half of these cities (n = 15) identified more than one example of parks that had adopted NBS practices. This suggests that the park and recreation agencies in many big cities are leveraging public park space to help manage urban flooding.

Cities that reported a high number of parks with examples of NBS included the following: Boise, ID (five parks); and Baton Rouge, LA, Houston, TX, San Francisco, CA, and Tucson, AZ (three parks each). At the state level, eight parks were in California, followed by Texas (six parks), and Arizona and Idaho (five parks each). Using the region designations in the Fifth National Climate Assessment [68], our study sample had more examples from the Southwest (fourteen parks), Southeast (twelve parks), and Midwest (eleven parks), and fewer examples from the Southern Great Plains (nine parks), the Northwest (six parks), and the Northeast (six parks). No example was reported from the Northern Great Plains (i.e., Montana, South and North Dakota, and Wyoming).

More often, NBS practices were reported for the management of stormwater flooding than for coastal/riverine flooding. Out of the 58 parks in the study sample, 74.1% (n = 43) implemented NBS practices to mitigate impacts from stormwater flooding, while 37.9% (n = 22) implemented NBS practices to mitigate impacts from coastal/riverine flooding; in addition, 12.1% parks (n = 7) used NBS for both flood management focuses.

3.1.2. Types of Park NBS Practices

The study sample presented a wide range of NBS practices for urban flood management (Figure 4 and Supplementary Materials). Based on the NBS typology we developed, overall, ECO were implemented most often in parks, followed by SGSI, POND, TEM, and ENG (Figure 4). Comparing the management focus of stormwater versus coastal/riverine flooding, parks addressing stormwater most often used SGSI, while parks addressing coastal/riverine flooding most often used ECO at a substantially higher rate than other types of practices. ENG were least often used across both management focuses, though it was most often used in parks addressing stormwater and coastal/riverine flooding simultaneously.

Seventeen parks in this study sample (29.3%) had multiple practices, including all seven parks that addressed both stormwater and coastal/riverine flooding. Among the remaining ten parks, seven addressed stormwater flooding and three addressed coastal/riverine flooding. Practice-wise, the five types of NBS were quite evenly used in these parks. TEM was most common, adopted in nine parks and often combined with POND and ECO. The top three combinations included TEM and POND (in five parks), TEM and ECO (in four parks), and ENG and SGSI (in four parks).

3.2. NBS Implementation and Landscape Contexts

Table 2. Descriptive statistics (means, with SD in parentheses) of landscape contextual factors of parks (N = 58) by NBS type.

Category	Variable	ECO	SGSI	POND	TEM	ENG
Surrounding land cover	% Greenspace	36.18 (19.54)	15.43 (14.64)	37.94 (25.77)	27.28 (17.58)	12.74 (11.27)
	% Built	50.94 (19.49)	80.62 (17.62)	58.67 (24.87)	70.14 (19.00)	72.98 (18.76)
	% Water	12.88 (20.65)	3.95 (7.14)	3.39 (6.51)	2.58 (9.03)	14.28 (15.13)
Surrounding and within-park flood risks	% High-risk area (surrounding)	33.79 (28.13)	10.09 (13.27)	13.55 (15.93)	11.54 (18.64)	20.81 (20.08)
	% High-risk area (within-park)	44.74 (43.27)	11.17 (21.32)	22.84 (37.19)	21.22 (36.94)	11.57 (17.72)
Park geographic characteristics	Area (ha)	62.03 (80.46)	11.91 (16.42)	43.75 (71.03)	42.47 (70.98)	10.30 (11.15)
	Distance to city center (km)	7.26 (4.99)	4.56 (3.40)	7.37(4.85)	5.38 (3.55)	4.38 (3.93)
	Distance to water bodies (km)	1.79 (2.43)	2.38 (2.40)	3.40 (3.74)	3.63 (3.36)	0.94 (1.57)

and

Table 3. ANOVA results for landscape contextual factors of parks by NBS type.

Category	Variable	Across NBS Types
		F-Value	p-Value
Surrounding land cover	% Greenspace	5.359	0.001 ***
	% Built	3.405	0.013 *
	% Water	2.446	0.054
Surrounding and within-park flood risks	% High-risk area (surrounding)	4.569	0.002 **
	% High-risk area (within-park)	2.797	0.032 *
Park geographic characteristics	Distance to center	1.734	0.152
	Area	2.189	0.079
	Distance to water body	1.958	0.110

* p < 0.05. ** p < 0.01. *** p < 0.001.

summarize the descriptive statistics for landscape contextual factors of parks and the ANOVA results, respectively.

Regarding surrounding land cover conditions, parks with SGSI on average had the most coverage of built areas, while parks with ECO had the least coverage of built areas. Parks with ENG on average had the least coverage of greenspace and the most coverage of surface water in their surroundings. In contrast, parks with POND had the most coverage of surrounding greenspace and parks with TEM had the least coverage of surrounding water. Parks with ECO also had relatively high coverage of greenspace in their surroundings. The differences in the surrounding built area and greenspace coverages between parks implementing varied NBS types were statistically significant as indicated by the ANOVA (Table 3). Further, post hoc tests using pairwise comparisons confirmed that parks with SGSI had statistically significant higher coverage of built areas than parks with ECO; and parks with ECO and POND both had statistically significant higher coverage of greenspace than parks with ENG and SGSI (Figure 5).

Overall, high-flood-risk areas constituted 24.67% within the park boundaries and 18.83% in the park surroundings across parks in our study sample. Parks with ECO had the highest flood risks on average on both the surrounding and within-park levels. In contrast, parks with SGSI had the lowest surrounding as well as within-park flood risks. In addition, parks with TEM had relatively low surrounding flood risks and ENG had relatively low within-park flood risks. The following ANOVA test shows that the differences in flooding risks between parks implementing varied NBS types were statistically significant for both surrounding and within-park conditions (Table 3). Further, post hoc tests confirmed that parks with ECO had statistically significant higher surrounding flood risks than parks with SGSI, TEM, and POND. They also had statistically significant higher within-park flood risks than parks with SGSI (Figure 5).

None of the associations between park geographic characteristics and the type of NBS implementation was statistically significant, as indicated by the ANOVA.

3.3. NBS Implementation and Sociodemographic Contexts (RQ3)

Our analysis of sociodemographic contextual factors involved the following proximity-based and visitation-based CBGs: (1) PBC-N: proximity-based CBGs using the complete study sample (N = 58); (2) VBC-n: visitation-based CBGs using a subset of parks identified as points of interests in the Advan mobility data (n = 51); and (3) PBC-n: proximity-based CBGs using the subset of parks identified as points of interests in the Advan mobility data (n = 51), for comparisons with VBC-n.

Table 4. Means and standard deviations (in parentheses) of the sociodemographic contextual variables by NBS type. Results are reported to compare three groups: PBC-N (proximity-based CBGs using the study sample, N = 58), PBC-n (proximity-based CBGs using the sample subset with available mobility data, n = 51), and VBC-n (visitation-based CBGs using the same sample subset, n = 51).

	Variable	ECO	SGSI	POND	TEM	ENG
Economic status	MHI (PBC-N)	77,910 (37,127)	79,598 (34,715)	79,484 (31,953)	56,185 (21,548)	99,425 (43,161)
	MHI (PBC-n)	66,341 (31,879)	79,454 (32,112)	76,784 (30,118)	72,134 (37,700)	91,177 (48,745)
	MHI (VBC-n)	68,188 (17,294)	80,452 (31,629)	73,081(29,884)	60,340 (10,274)	74,310 (27,035)
	Adjusted MHI 1 (PBC-N)	1.39 (0.74)	1.00 (0.51)	1.24 (0.47)	0.90 (0.39)	1.30 (0.59)
	Adjusted MHI (PBC-n)	1.30 (0.82)	1.07(0.42)	1.17 (0.47)	0.82(0.46)	1.03 (0.32)
	% Poverty (PBC-N)	16.58 (12.47)	19.71 (12.63)	12.33 (10.16)	21.92 (9.89)	17.65 (12.58)
	% Poverty (PBC-n)	19.92 (12.81)	12.61 (12.06)	12.21 (7.72)	21.68 (14.51)	23.10 (16.17)
	% Poverty (VBC-n)	14.06 (6.09)	12.53 (6.54)	13.55 (6.72)	16.56 (4.00)	15.23 (5.71)
Race and ethnicity	% White (PBC-N)	53.12 (25.46)	41.39 (23.66)	63.61 (34.59)	48.99 (25.54)	49.64 (22.98)
	% White (PBC-n)	42.08 (23.89)	64.53 (39.32)	37.65 (28.23)	42.87 (26.05)	47.48 (25.28)
	% White (VBC-n)	67.84 (16.28)	63.01 (21.43)	64.48 (18.06)	60.75 (16.32)	62.02 (13.50)
	% Black (PBC-N)	22.05 (25.47)	23.33 (24.66)	21.03 (28.76)	21.61(22.83)	13.82 (19.57)
	% Black (PBC-n)	30.02 (28.32)	19.13 (30.19)	36.47 (34.22)	26.29 (23.53)	12.56 (14.29)
	% Black (VBC-n)	16.57 (14.50)	15.59 (18.16)	20.82 (20.70)	18.26 (16.39)	13.81 (7.93)
	% Asian (PBC-N)	4.46 (4.60)	16.43 (20.94)	8.51 (16.96)	4.62 (3.97)	21.77 (20.58)
	% Asian (PBC-n)	4.59 (5.17)	11.20 (20.13)	5.71 (6.71)	14.73 (20.34)	29.53 (21.62)
	% Asian (VBC-n)	5.26 (3.33)	11.59 (16.30)	4.75 (2.59)	4.45 (1.81)	11.75 (10.43)
	% Hispanic (PBC-N)	18.89 (25.21)	15.33 (13.24)	12.24 (14.04)	24.55 (26.57)	17.81 (16.19)
	% Hispanic (PBC-n)	20.25 (27.73)	15.46 (16.08)	19.78 (25.18)	11.95 (7.66)	21.30 (14.86)
	% Hispanic (VBC-n)	22.22 (19.28)	16.19 (12.92)	10.89 (6.60)	29.84 (22.99)	18.54 (12.14)
Age	% Senior (65+) (PBC-N)	12.79 (3.74)	10.99 (4.37)	14.23 (6.42)	11.93 (3.58)	12.16 (5.03)
	% Senior (65+) (PBC-n)	11.45 (3.58)	13.33 (7.40)	14.05 (3.05)	9.56 (3.23)	9.64 (4.68)
	% Senior (65+) (VBC-n)	12.34 (1.95)	12.84 (2.21)	12.46 (1.07)	11.60 (1.64)	14.02 (2.61)
	% Under 18 (PBC-N)	21.10 (5.82)	18.88 (9.76)	21.07(6.18)	18.98 (7.85)	17.42 (8.24)
	% Under 18 (PBC-n)	22.20 (6.54)	21.77 (6.97)	19.36 (5.88)	21.37 (11.68)	17.55 (12.37)
	% Under 18 (VBC-n)	24.64 (2.93)	23.01 (3.99)	25.56 (2.17)	25.40 (2.62)	22.84 (2.54)
Car	# Cars/household (PBC-N)	1.56 (0.42)	1.44 (0.45)	1.67 (0.31)	1.47 (0.36)	1.26 (0.57)
	# Cars/household (PBC-n)	1.61 (0.37)	1.73 (0.33)	1.53 (0.70)	1.53 (0.38)	1.26 (0.41)
	# Cars/household (VBC-n)	1.75 (0.30)	1.75 (0.19)	1.79 (0.23)	1.75 (0.17)	1.38 (0.49)

¹ Adjusted MHI was not included for visitation-based CBGs as the origin of visitation sometimes was beyond one city. As a result, the comparison of CBGs-level MHI to city-level MHI was not meaningful.

summarizes and compares the descriptive statistics of sociodemographic contextual variables among these three groups.

We then performed an ANOVA to investigate the difference in sociodemographic contexts by park NBS type. The ANOVA results indicated no statistically significant difference in any variable. However, we observed some noteworthy patterns in the descriptive statistics. First, some park NBS types had repeated highest and lowest means of variables in the two proximity-based groups (PBC-N and PBC-n), which provides evidence of potential correlations between the type of NBS adopted in a park and its nearby neighborhoods’ sociodemographic status that are worth further research. Second, the proximity-based versus visitation-based sociodemographic contexts showed notable differences in a few sociodemographic contextual variables.

3.3.1. Park NBS Adoption and Nearby Neighborhoods’ Sociodemographic Status

In both PBC-N and PBC-n, regarding economic status, the mean poverty rates were lowest in neighborhoods around parks with POND. In contrast, parks with TEM were in neighborhoods with the highest average poverty rate in PBC-N and the second highest in PBC-n. Similarly, the mean adjusted MHI was lowest in neighborhoods around parks with TEM in both PBC-N and PBC-n, while parks with ECO were in neighborhoods with the highest average adjusted MHI. For race and ethnicity, the mean Asian population percentage was highest in neighborhoods around parks with ENG and lowest in neighborhoods around parks with ECO in both groups. In addition, the mean Asian population percentage varied greatly among parks with different NBS types, ranging from below 5% to nearly 30%. The mean Black population percentages were lowest in neighborhoods around parks with ENG. In contrast, parks’ surrounding neighborhood White and Hispanic population percentages showed some opposite trends in PBC-N and PBC-n, suggesting potentially higher variations in White and Hispanic populations in neighborhoods around POND, SGSI, and TEM. Regarding age and car availability, the mean percentage of seniors was highest in neighborhoods around parks with POND in both PBC-N and PBC-n. In contrast, parks with TEM were in neighborhoods with the lowest mean percentage of seniors in PBC-n and the second lowest in PBC-N. The mean percentage of populations under 18 years old was highest in neighborhoods around parks with ECO in both PBC-N and PBC-n. In addition, parks with ENG were in neighborhoods with the lowest mean percentage of populations under 18 years old and number of cars owned in both groups.

3.3.2. Differences between Proximity- and Visitation-Based Sociodemographic Contexts

Comparing the proximity-based and visitation-based results (PBC-n vs. VBC-n), there are some notable differences between people who visit the parks versus those living close to the parks. First, the mean poverty rates in VBC-n were lower than those in PBC-n across parks with all types of NBS except POND. Second, the mean White population percentages in VBC-n were higher than those in PBC-n across parks with all types of NBS except SGSI. In contrast, the mean Black population percentages in VBC-n were lower than those in PBC-n across parks with all types of NBS except ENG. Third, the mean average number of cars owned per household was higher in VBC-n than in PBC-n across parks with all types of NBS. Fourth, the mean percentages of populations under 18 years old were higher in VBC-n than in PBC-n across parks with all types of NBS and showed a smaller variability by NBS type.

To further investigate the differences between VBC-n and PBC-n, we performed a follow-up t-test for % poverty, % White, % Black, average number of cars per household, and % under 18. In summary, the differences in grand means across parks (n = 51) were statistically significant for all five variables but % Black (

Table 5. T-test results for the grand means across the sample subset with available mobility data (n = 51), comparing between visitation-based CBGs (VBC-n) and proximity-based CBGs (PBC-n). Cohen’s d values indicate small (0.2–0.5) and medium (0.5–0.8) effect sizes for the four variables.

Variables	Group	Mean	SD	t-Value	p-Value	Cohen’s d
% Poverty	Visitation-based	14.28	5.84	−2.06	0.043	−0.409
	Proximity-based	18.30	12.62
% White	Visitation-based	64.19	17.05	3.34	0.001	0.661
	Proximity-based	48.14	29.80
# Cars owned per household	Visitation-based	1.71	0.293	2.12	0.037	0.420
	Proximity-based	1.56	0.435
% Populations under 18 years old	Visitation-based	24.37	3.07	3.29	0.002	0.652
	Proximity-based	20.32	8.22

). These results might be explained by that wealthier, White populations comprise more users of parks, even when many low-income, Black populations may live close by. In addition, many park users might be families with kids visiting from beyond the 0.5-mile walking distance.

3.3.3. Descriptive Results for Park Visitation Patterns by NBS Type

We also examined Advan mobility data-based visitation variables for 51 parks to provide additional insights for visitation patterns of parks with varied types of NBS. On average, these parks served users from a broad geographic scope, ranging from 263 origin CBGs (parks with SGSI) to 538 CBGs (parks with ENG). Interestingly, the median dwell time showed an opposite trend—visitors to parks with ENG on average spent the shortest time (27 min, SD = 8.57) while visitors to parks with SGSI spent the longest time (44 min, SD = 36.24). Given the result that ENG and SGSI were both relatively close to the city center, where CBGs are often smaller and denser, this difference is more likely explained by park design characteristics or park user preferences than the location of the park. In addition, the mean ratios of visits to visitors were very similar across all parks, except a higher ratio for parks with TEM, indicating a possibility of visitors coming repeatedly to these parks. In contrast, parks with TEM had the lowest weekly visitation on average. These results should be treated as exploratory and demand further investigations since the differences in visitation patterns by park NBS type were not statistically significant, as indicated by the ANOVA.

4. Discussion

4.1. Park NBS and Urban Flood Management

This study showed that a high percentage of park and recreation agencies in U.S. big cities are adopting NBS in public parks to mitigate urban flooding and are explicitly linking this effort with climate resilience and adaptation. The source of urban flooding (e.g., riverine, coastal, stormwater, flash, and snowmelt) varies greatly in different regions and cities, and flood risks depend on many factors including the natural environment, land use history and pattern, and stormwater and sewer systems [13]. We found that nearly 75% of parks in the study sample have implemented NBS practices for managing stormwater-related flooding than coastal/riverine flooding. For example, park and recreation agencies reported many more parks that addressed stormwater-related flooding (fourteen stormwater examples versus one coastal/riverine flooding example out of the fourteen parks) even in the Southwest Region where many coastal urban communities are challenged by increasing flooding events due to extreme rainfall, storm surge, high tides, and sea level rises [68]. The dominant focus on stormwater in the current park NBS adoption might be attributed to the widespread idea of leveraging green stormwater infrastructure (GSI) to meet requirements of the Clean Water Act. Since the 1990s, the Environmental Protection Agency (EPA) has been promoting green infrastructure as a cost-effective approach to reduce non-point source pollution in urban stormwater runoff and improve local water quality [69].

Related to the result that notably more parks use NBS to address stormwater-related flooding, our findings indicated relatively low flood risks within and in the immediate surroundings of parks that have adopted NBS. Overall, high-flood-risk areas constituted only 24.67% within the park boundaries and 18.83% the park surroundings (based on a 0.5-mile radius), with percentages ranging from 11 to 45% and 10 to 34% by the NBS type, respectively. In our study, areas of high flood risk were determined according to the FEMA flood maps. These maps define flood hazard zones by the 100-year flood [49] and do not necessarily include urban areas prone to stormwater-induced flooding [13]. In addition, this study provides evidence of the risk of compound floods, which describes the occurrence of two or more independent events that produce flood hazards jointly in space and/or time [70], and the effort of leveraging park space to address compound flooding through various NBS. A total of 12.1% of parks in our study sample implemented NBS practices for both stormwater and coastal/riverine flooding management. We argue that the future planning, design, and management of urban parks need to pay more attention to coastal, riverine, and compound flooding and further explore how to expand the role of NBS in flood control. For example, some landscape architects are leading innovative planning and design strategies for shoreline parks that integrate protection berms, coastal ecosystem restorations, amenities, and recreational facilities to enhance flood resilience and biodiversity while creating opportunities for people to interact with nature [71,72].

4.2. Park NBS Type, Contextual Factors, and Implications of Their Relationships

Overall, the most used type of NBS in the studied parks was ECO (a typology we defined as conserving, restoring, or creating ecosystems to mitigate flooding through ecological processes and functions), while the least used NBS type was ENG (a typology we defined as imitating natural infiltration processes but having no living elements). This suggests that preserving and enhancing natural landscapes may be an important goal of adopting NBS in parks in the U.S. Some scholars argue that measurable benefits for biodiversity should be a necessity for NBS [73]. Urban parks are well-positioned to adopt NBS practices that can provide more ecological benefits because parks largely consist of natural landscapes and are already protected from land development. In addition, unlike Tong et al. (2022) [40], we did not find a notable correlation between park size and NBS type or the presence of multiple types of practice. In this study, nearly 30% of the parks adopted multiple types of NBS practices, and we observed a tendency of using multiple types of practices when the flooding management goal addressed both stormwater and coastal/riverine flooding or focused only on stormwater-related flooding.

Importantly, this study provides evidence of some noteworthy associations between the NBS type a park has adopted and its landscape and sociodemographic contexts (Figure 6), suggesting that the choice of NBS might be implicitly influenced by the surrounding environment and communities. Below we discuss key findings on the relationships between park NBS type and contextual factors and their implications.

ECO was the most used NBS practice across the parks included, especially for addressing coastal/riverine flooding. Further, parks with ECO adoption tended to have high flood risks within the park boundaries and in their surroundings, as well as more greenspace in their surroundings. These results indicate that parks with ECO are likely to be located on or near waterways and shorelines within a less built area. Choosing an ECO type of NBS for these parks may be a strategy to take advantage of better baseline environmental conditions and to further expand the existing greenspace through habitat restoration or construction. There has been an ongoing effort for daylighting buried urban channels and restoring the floodplain of channelized streams in the past decades to improve water quality and habitat, with a growing primary focus on flood control [74]. In coastal areas, the idea of using NBS such as living shorelines to enhance coastal resilience is also drawing more attention, including the adoption of living shorelines in urban parks [75]. Based on our findings, we argue that an equally critical direction is to develop novel, hybrid NBS that can effectively address increasing extreme precipitation and coastal weather events by leveraging both native ecosystems and new engineering techniques through interdisciplinary collaborations among marine ecologists, coastal engineers, and landscape architects [76]. In addition, our study provides some evidence that parks with ECO might tend to be located in neighborhoods with higher percentages of underage populations and lower percentages of Asian populations. Including educational programs that prioritize youth for parks with ECO may contribute to effective public outreach strategies and broader awareness and acceptance of NBS for flooding.

SGSI was most often used in parks that focused on stormwater-related flooding. Parks with SGSI adoption on average had the lowest flood risks within the park boundaries and in their surroundings. In addition, they also had the highest percentage of built areas in their surroundings as well as significantly lower percentages of greenspace compared to parks with ECO and POND. These results match the primary designed function of SGSI—to collect and treat stormwater runoff with soils and plants at or near where it is generated. This function is particularly suitable in urban upland locations of high imperviousness [77]. Reflecting on the dominant use of SGSI in parks to manage stormwater-related flooding, we caution on a planning and design approach to NBS that narrowly picks feasible options from a portfolio of techniques [22,44]. Rather, diverse NBS including detention basin and floodable landscapes (TEM), retention ponds (POND), and restored or constructed wetlands (ECO) should be considered and explored to maximize environmental and social co-benefits.

ENG was the least used type of NBS in this study. This result contrasted with Tong et al. (2022), who found that permeable pavement in pathways and open spaces—an ENG practice based on our NBS typology—was the most common stormwater management measure in the Sponge City Parks in Shanghai and were used in all 21 parks they examined [40]. Despite the lower use of ENG overall, ENG was more often used than all the other types of NBS in parks that addressed both stormwater and coastal/riverine flooding. This might be due to a tendency to adopt highly engineered NBS rather than natural systems when risks of compound floods are present. In addition, parks with ENG are located in neighborhoods with notably, though not statistically significant, higher percentages of Asian populations, lower percentages of Black and underage populations, and fewer cars owned per household. To the extent that ENG practices typically provide few social co-benefits, future research is needed to better understand whether the ENG type of NBS is more likely to be adopted in parks in underserved communities and the related implications for equity.

TEM was the top type used in parks with multiple NBS adoption, often combined with POND and ECO. For example, the combination of detention basin and retention ponds was present in Utah Park in Aurora, CO, and the combination of riverbank restoration and temporary flooding accommodation was present in Waterfront Park in Louisville, KY. With increasing extreme precipitation events under climate change, design strategies that create flexible uses and can accommodate both wet and dry weather needs are on the rise [78,79]. Urban parks are a suitable location to create relatively large detention and floodable landscapes since there is much open space for outdoor activities that take place only in dry weather. Such space can be leveraged to temporarily store runoff and flood water during and shortly after a storm hits, though it is important to tailor the design of TEM type of NBS with community members’ flooding experiences since they could be perceived as unsafe [80].

Parks with POND adoption on average had the highest percentage of greenspace in their surroundings. Pairwise comparisons showed that they had significantly higher percentages of greenspace than parks with SGSI and ENG, and significantly lower surrounding flood risks than ECO. In addition, they were located in neighborhoods with notably, though not statistically significant, lower poverty rates and higher percentages of senior populations. These results suggest that POND may likely be adopted by parks in communities that are wealthier and have a better coverage of greenspace, although these communities may not have very high flood risks. Given that retention ponds may contribute to the value increase in nearby homes [81], extra attention is needed to best enhance equity and avoid green gentrification when siting and designing the POND type of NBS.

In addition, this study indicates notable differences between visitation-based and proximity-based sociodemographic contexts. Specifically, we found individuals and families that are wealthier and White more often visited parks with NBS. Even when NBS practices are adopted in parks located in neighborhoods that are less wealthy and have more minorities, the residents may not receive as many NBS social benefits due to less use of the park. This finding suggests potential inequity in the delivery of social benefits by park NBS that is influenced by park visitation preference. It is important, therefore, to not only site NBS in underserved communities but also utilize inclusive processes to engage members of underserved communities in the planning, design, and implementation of NBS [82]. Further, disadvantaged communities need NBS types with more co-benefits beyond flood control.

4.3. Limitations and Future Research

This study has several limitations that demand caution when interpreting or generalizing the results. First, inherent limitations in the study sample may bias the results of this study. We used the 2022 City Park Fact Survey conducted by the Trust for Public Land to collect parks to be investigated and to obtain information about NBS practices adopted in these parks. Since the City Park Fact Survey targets the 100 most populous U.S. cities, and the 2022 Survey only asked park and recreation agencies to provide examples of climate resilience-building practices, the survey data cannot represent all public parks in the U.S. that have adopted NBS to address urban flooding. For example, compared to the summary of the Fifth National Climate Assessment [68], our study sample included few examples from the Northeast, a region that has experienced extreme precipitation and problematic flooding in recent years and undertaken notable efforts to build resilience using NBS. In addition, the survey combined coastal and riverine flooding—which are often viewed as two distinctive sources for urban flooding—into a single climate impact. While the use of the 2022 City Park Fact Survey served this study’s exploratory nature well, the study findings should be interpreted with caution and without generalizations. Second, we did not account for the time of various secondary datasets and assumed a relatively stable status of the parks’ landscape and sociodemographic contexts in the data analysis. For example, when determining the value of average weekly visitation-based sociodemographic contextual variables, mobility data from year 2022 and the 2016–2020 American Community Survey 5-Year Estimates data were combined. Third, the Adavn mobility data, from which we obtained the park visitation records, were extracted from cellphone-based geolocation data. This is subject to varied sampling rates among different geographic areas and population groups, with a potentially poorer representation of people with low socioeconomic status of minority groups; therefore, the mobility data may result in biased measures for the visitation-based CBGs in our study.

Future research can focus on specific cities faced with urban flooding challenges as case studies and examine NBS adoption in all parks across the city. Emphases should be put on, first, the relationships between NBS adoption, NBS type, and community sociodemographic status, based on the limited but notable evidence of inequity in social benefits delivery from this study; and second, how varied sources of urban flooding (e.g., stormwater, coastal, riverine, and flash) may affect NBS adoption in terms of siting location and practice type. Further, research on NBS and park contexts can be complemented by investigations of the NBS planning and design rationale through, for example, policy analysis and interviews with managers and designers. More research is also needed to evaluate the flooding mitigation efficacy as well as ecological and social co-benefits of park NBS given their widespread adoption and potentially significant role in enhancing urban resilience and affecting nature experiences in cities.

5. Conclusions

Nature-based solutions (NBS) are gaining increasing attention in discourse and actions for flood resilience. This study explored the adoption of NBS for urban flood management in public parks, a critical community asset to protect and expand nature in cities. We found that diverse NBS practices have been adopted in public parks in many U.S. big cities to address stormwater-related and coastal/riverine flooding under climate resilience and adaptation efforts. We developed a high-level typology of park NBS practices and provided new insights into the relationships between NBS type and a park’s landscape and sociodemographic contexts. Our findings suggest that the choice of what NBS type to implement in a park is associated with its surrounding land cover, surrounding and interior flood risks, and neighborhood sociodemographic status. We also found that, overall, park visitors were more privileged compared to residents living near a park. These observed trends have implications for the success of park NBS in delivering co-benefits in an equitable manner and warrant further research to confirm their generalizability and to explore effective planning and design strategies.

This study addresses the limited research on park NBS adoption and broadens the focus of NBS practices from flooding management techniques to landscape design and planning that approach urban flood resilience as a societal challenge. We argue that landscape and sociodemographic contextual factors deserve explicit consideration in the adoption of park NBS. To the extent that parks make up key multifunctional urban greenspace, the siting and design of park NBS must build on a comprehensive understanding of the local environment and community. Further, park managers and designers should consider a diverse portfolio of NBS and, when necessary, develop novel hybrid practices to effectively enhance urban flood resilience and deliver co-benefits.