Abstract
ABSTRACT
As awareness of climate change increases, U.S. cities are beginning to implement climate mitigation and adaptation initiatives to reduce population vulnerabilities to climate risks. This study contributes to a growing literature that quantitatively describes the relationships between sociodemographic variables and climate adaptation interventions in U.S. cities. Ordinary linear and simultaneous autoregressive models are used to evaluate early flood adaptation actions in Austin, Texas, to assess relationships between flood risk, green infrastructure, and measures of race and income. Findings of unequal exposure to flood risk and uneven access to flood resilience initiatives contribute to our understanding of color-blind urban planning responses to climate change and their potential to amplify inequitable protection from climate risks.
INTRODUCTION
Cities in the United States are experiencing climate change impacts ranging from more extensive flooding to increased heat waves and wildfires. 1 These ecological changes have extensive and varied social consequences, and there is wide agreement that many of these risks will disproportionately affect people already experiencing forms of social inequality. 2 As climate change impacts become increasingly evident, U.S. cities are pursuing climate adaptation plans to reduce risks through changes to land use, infrastructure, and service provision. How cities acknowledge existing racial inequalities in these plans varies widely, ranging from plans that fail to mention race 3 to cities that center racial equity in their planning process and associated projects. 4 The uneven attention to existing racial inequalities raises concerns that urban adaptation interventions may exacerbate patterns of unequal risk or create new forms of environmental injustice, especially for low-income communities of color. 5 As an emerging area of U.S. planning practice, climate adaptation may create or amplify racial inequalities while simultaneously presenting an opportunity for planners to enact racial equity praxis.
This article contributes to conversations on adaptation's potential to impact urban inequality by deploying Omi and Winant's 6 theory of racial projects and Hardy et al.’s 7 conception of color-blind adaptation to understand the role of planning in creating populations that are vulnerable or protected from climate change. In the Introduction section of this article, I share current research on adaptation planning and urban climate flood risks. In the Introduction section, I use theoretical concepts of racial projects and color-blind adaptation to situate climate planning within a lineage of racialized urban planning practices. In the Results section, I provide a case study of early climate adaptation efforts in Austin, Texas, to examine how color-blind adaptation planning practices in U.S. cities may (re)create racial inequalities.
Climate-driven flood risks in U.S. cities
Flooding is one of the deadliest and most frequent disasters in the United States. Over the past 30 years, flooding caused an average of 93 deaths per year and incurred ∼$8 billion in damages per year. 8 Loss of lives and property is compounded by flooding impacts on individual physical and mental health alongside disruptions to social infrastructure and services. 9 Climate change has intensified existing flood risks in many U.S. cities through shifts in precipitation patterns and increases in the frequency of extreme weather events. 10 Furthermore, U.S. coastal sea levels have already risen over the past century through a combination of climate and nonclimate related factors, with estimates that nuisance flooding rates are 10 times higher than historical averages in coastal cities. 11
Scientists predict that these trends will continue and by 2100, heavy precipitation events in the United States are expected to increase by 50%–300%, and sea levels are likely to rise between 0.3 and 2 m. 12 A 2017 study predicts that if current global emissions continue, sea level rise will likely displace 13–15 million U.S. coastal residents by 2100. 13 The combination of extreme precipitation events, sea level rise, population growth, and land use changes is expected to increase the frequency and severity of flood events in a large number of U.S. cities. 14
Environmental injustices associated with existing flood risks
Climate-driven flood risks are amplifications of existing environmental hazards faced by urban residents, which have well-documented inequalities associated with unequal exposure to flood-prone areas and uneven recovery after flood events. 15 In that literature, Rufat et al. conducted a meta-analysis of 67 case studies of urban flood disasters between 1997 and 2013 and found that race, gender, age, and income were associated with differential vulnerability to flood exposure, impact, and recovery. 16 Several studies support these findings and identify race, income, and age as the most significant predictors for flooding hardship and delayed recovery. 17 Notably, much of the scholarship examining flood recovery found similar trends of lower income, Black, and Latinx communities often experiencing delayed recovery because of unequal insurance coverage, disparate recovery funding, and uneven access to government recovery support. 18 Together, this body of literature highlights urban landscapes characterized by uneven flood risks, which often fall disproportionately on lower income, Black, and Latinx communities, and creates unequal impacts for those residents, which may be exacerbated by climate change.
U.S. adaptation planning is emerging
Investigating how city planners address uneven socioecological risks in their climate adaptation practices may provide insights into the role of city-led adaptation in addressing or perpetuating urban climate inequalities. Although U.S. cities have been working on climate change issues since the 1990s, adaptation planning is a relatively new field that gained momentum during the early 2000s. In general, the literature indicates a low uptake of adaptation planning and implementation in the United States, with a 2017 study finding only 50 U.S. cities with fully developed climate adaptation plans. 19 As adaptation planning grew as a field, scholars noted the potential for climate adaptation interventions to continue historical patterns of disparate risk or to create new forms of environmental injustice. 20 Within this area of investigation, studies have examined how social equity or justice is discussed in climate plans and generally found that although plans attend to these issues, few have goals or specific actions aimed at addressing existing or future climate inequalities. 21 In response, researchers have begun to investigate whether physical adaptation projects create “ecological enclaves” that protect higher income and predominantly white communities while ignoring or displacing communities of color. 22
Racialized urban planning
Urban planning and social inequality issues are deeply embedded in both planning practice and scholarship with discussions on racialized planning impacts and counter-praxis emerging across multiple planning domains. 23 In conversation with these scholars, I engage with theories of racial projects and color-blind adaptation to historically situate the racialized distribution of climate risks and frame climate adaptation as a burgeoning domain that offers planning an opportunity to not repeat practices that contribute to urban inequalities. 24 Racial projects account for the distribution of resources and power in ways that create or oppose racial formation. 25 Color-blind adaptation builds upon Bonilla-Silvia's study on color-blind racism 26 and is defined as “climate adaptation planning projects that altogether overlook racial inequality—or worse dismiss its systemic causes and explain away racial inequality by attributing racial disparities to non-racial causes.” 27 Racial projects and color-blind adaptation planning emerge from critical race studies, which frames racism as foundational, complex, and an ongoing structural process that morphs and changes over time, but is nevertheless omnipresent and continuously produces racial inequalities. 28
I draw from these concepts to theorize cities as urban landscapes formed by both the legacies of historical racism and continuous racialization processes, which result in social inequalities. 29 Segregation, racial covenants, redlining, urban renewal, disinvestment, and land use zoning are racial projects that distributed resources along racial lines and contributed to significant gaps in wealth accumulation, home ownership, educational attainment, health, and environmental conditions. 30 City planning was indirectly complicit or directly responsible for many of these practices and the resulting social inequalities that range from unequal infrastructure to disparate health outcomes. 31 These processes are often intertwined with the distribution of environmental goods and harms along racial and class lines. Examples of this include early urban planning practices that segregated communities of color into flood-prone areas, disparate stormwater infrastructure, and the ongoing Flint water crisis. 32 These same racialized logics have already surfaced with proposed urban climate adaptation initiatives, making it essential for planners to understand racialized urban climate impacts; grapple with planning's role in creating social inequalities; and consider these when designing adaptation plans, processes, and actions. 33
This article conceptualizes climate adaptation actions as racial projects that can be racist or antiracist according to how they distribute environmental risks and protections in cities. Following Hardy et al., I situate climate adaptation practices that do not consider existing urban disparities as color-blind adaptation, which creates, maintains, or increases racial inequalities. 34 Using environmental justice methods, I explore the relationships between flood risk, adaptation actions, and measures of income and race in Austin, Texas, to ask whether color-blind adaptation initiatives continue uneven access to environmental amenities or exposure to environmental risks. 35
METHODS
Climate adaptation and vulnerability in Austin, Texas
Austin has a history of racial and economic segregation, significant inland flood risk, and substantial city-led efforts to increase flood resilience. The use of racial deed covenants in Austin proliferated between the 1890s and 1950s. The resulting segregation was further codified through the 1928 “Master Plan,” which restricted public goods and services for Black residents to the city's east side. 36 City ordinances and permit regulations were enacted to support the master plan, alongside the use of eminent domain, which forced Black residents to sell noneastside land and home assets to white investors at undervalued rates. 37 As the Austin Latinx population grew, they were also segregated into east and south Austin through a combination of private and public mechanisms. 38
In Austin, early residential segregation practices were often intertwined with environmental amenities. For example, the Hyde Park subdivision was developed in northwest Austin and advertised as “exclusively for White people” with the benefit of sitting “185 feet above the river.” 39 These racialized practices were further institutionalized in the 1930s through redlining and segregated public housing. 40 By the 1950s and 1960s, city-led policies included concentrating industrial land use and promoting urban renewal in the east side, which increased environmental and economic burdens, especially for Black and Latinx residents. 41 Explicit and de facto segregation has produced numerous social inequalities ranging from inequitable schools to disparate accumulations of generational wealth and subsequent economic insecurity. 42 Together these practices contributed to a racialized landscape that continues to shape the distribution of social outcomes across the city.
The racialization of space has often been intertwined with an unequal exposure to flooding, especially for Latinx residents who reside in areas closer to floodplains and flood-prone creeks. 43 Unequal exposure to flooding is especially problematic in a city with significant flash flood risks. Austin is located in flash flood alley, an area in central Texas where topography, soil conditions, development patterns, and rainfall intensity combine to create one of the most flood-prone regions of the United States. In tandem with existing flood risks, Hayhoe's climate change forecast for Austin predicts limited changes in average annual precipitation but increases in extreme rainfall events, which could intensify the frequency and scale of flooding in the city. 44
Austin began working on climate change issues in 2007 but focused exclusively on climate mitigation efforts. As climate change impacts became more evident, the City of Austin started to characterize existing and future adaptation projects to improve flood resilience. 45 These projects incorporate traditional flood mitigation efforts like zoning and building regulations, storm drainage infrastructure, and channelization. They also include more unique approaches including flood buyouts, early warning systems, and green infrastructure. 46 Climate adaptation is more explicitly addressed in the 2018 “Climate Resilience Action Plan” and the 2014 “Towards a Climate Resilient Austin” document. 47 The 2018 report focuses on protecting city assets from increased flood, drought, wildfire, and heat risks. The plan addresses structural integrity and redundancy concerns for critical infrastructure such as water treatment plants, power substations, and evacuation centers. The plan does not address the protection of noncity infrastructure but does propose developing “resilience hubs” in low-income areas to promote disaster preparation and response. 48 Noncity assets were addressed in the 2014 “Towards a Climate Resilient Austin,” which used downscaled climate models to identify future climate risks and potential adaptation efforts. Regarding future flood predictions and resilience projects, the document specifies existing programs the City of Austin can expand upon, namely flood buyouts, structural flood proofing, land use regulations, early warning systems, and urban forestry. 49 The plan emphasizes using green infrastructure to reduce runoff intensity from flood events. 50
Although Austin is still developing a comprehensive approach to climate adaptation, the initial efforts already mentioned offer an opportune reflection point to evaluate the relationships between income, race, and the spatial location of existing flood-resilience interventions. Within Austin's Hazard Mitigation Plan, Imagine Austin Comprehensive Plan, Watershed Master Plan, and Towards a Climate-Resilient Austin documents, the city shows evidence of considering differential actions for low-income communities, but issues of racial segregation and racialized flood exposure are not mentioned. This signals the potential for a color-blind adaptation approach where the city does not officially or explicitly indicate attention to an already racialized landscape in deciding where to locate their flood adaptation actions.
Methods
Dependent variables
To understand color-blind adaptation planning implications, I completed regression analyses on five dependent variables: 100- and 500-year flood risks, small-scale green infrastructure, stormwater retention ponds, and city-led flood buyouts. Exposure to flood risk is measured by FEMA's 2016 digital floodplain maps for Austin as the percentage of a census block group in the 100- and 500-year floodplain. Flood adaptation initiatives were downloaded from Austin's open data portal and capture existing flood resilience efforts identified in four City of Austin planning documents. 51 Austin's comprehensive plan, “Imagine Austin,” calls for the use of green infrastructure including urban forestry, rain gardens, green roofs, and rainwater collection to serve as climate-related flood adaptation mechanisms. 52 Austin's “Watershed Master Plan” emphasizes the use of buyouts, engineering solutions, and green infrastructure to address existing and future climate risks. 53 The “Watershed Plan” highlights the use of rain gardens and pervious pavement as small-scale green infrastructure features that can have important flood mitigation contributions if deployed at a significant scale across the city. 54 The updated “Hazard Mitigation Plan” also cites the use of buyouts, structural elevation, and flood walls as ways to reduce flood impacts. 55
This article focuses on buyouts, green infrastructure, and stormwater ponds based on their explicit connection to city-identified climate-related flood resilience projects, their localized impacts, and data availability. Austin has a long-term flood buyout program, which began as a joint project with the Army Corps of Engineers (ACOE). The City of Austin has expanded that program without the ACOE to buy homes that experience severe and repetitive flooding. The data used for this project exclude pre-2014 buyouts to separate the ACOE project from newer city actions. Stormwater ponds were limited to city-maintained flood control ponds to exclude noncity efforts and regional flood detention ponds. Green infrastructure includes rain gardens, bioswales, green roofs, and porous pavement but excludes existing large-scale green infrastructure such as parks and the tree canopy to focus explicitly on flood resilience projects designed to have localized impacts.
Independent variables
This case study compares the colocation of flood risk, adaptation actions, and sociodemographic variables used in environmental justice and flood disaster literature. Data were gathered from the American Community Survey (ACS) 5-year data at the block group level. ACS data have documented measurement error issues, which are typically more pronounced at the census block group scale. 56 As an initial approach to address data quality issues, I used 5-year estimates and removed observations with fewer than 500 residents. 57 I chose census block groups instead of tracts because they are similar in size to an average neighborhood and are at a more appropriate scale given the localized impacts of small-scale green infrastructure initiatives. 58
For the environmental justice variables, median household income was selected because lower income households often lack financial means to mitigate risk (e.g., flood proofing homes) and have longer and more difficult flood recoveries. 59 Variables that account for race and ethnicity are imprecise proxies for issues of systemic racism in the United States, in addition to established research showing that historic and de facto racial segregation resulted in communities of color living in areas with higher environmental and hazard risks. 60 Racism is also a factor in prolonged and unequal recovery after a flood event. 61 A combined measure of the percentage of residents >65 and <18 years of age was included, along with the percentage of people experiencing one or more physical or mental disabilities. These variables reflect established disaster literature that documents increased flood vulnerability for these groups related to independence and mobility issues, larger incidents of trauma during flood events, and slower recoveries after flooding. 62 Households without a vehicle were included in response to literature connecting low flood evacuation rates with transportation access issues. 63 I incorporated the percentage of impervious cover drawn from the 2016 National Land Cover Database based on studies connecting impervious area, runoff, and flooding intensity. 64 Control variables for the adaptation models include exposure to flood risk and population density (Table 1).
Descriptive Statistics
This analysis would have benefited from data on elevation, soil conditions, property values, and hard infrastructure adaptation efforts. West Austin is higher and hillier with more permeable soils, whereas east Austin has lower and flatter topography and less permeable clay soils. 65 In general, west Austin's topography and geological conditions result in higher runoff until the water reaches the denser soils on the flatter east side. 66 Given these conditions, the city may target actions in the west side to slow or capture runoff before it reaches more flood-prone parts of the city. Future research could incorporate this ecological data and use qualitative approaches to explore siting decisions by the city. Hard infrastructure tends to require significant urban planning and public finance and may offer greater flood protection than small-scale interventions such as bioswales. Property values may inform city decision making for adaptation siting, especially around decisions between interventions such as buyouts or floodwalls. Data sets on hard infrastructure and property values were not easily accessible and, therefore, not included in this study, but would strengthen future analyses.
Analysis
I used two-tail t-tests to compare the means of independent variables with flood risk and adaptation actions (buyout, small-scale green infrastructure, and stormwater ponds). I built five multivariate regression models to understand the relationships between flood risks, adaptation actions (buyout, small-scale green infrastructure, and stormwater ponds), and measures of race, ethnicity, income, age, disability, and vehicle access in Austin. I first constructed ordinary least squares (OLS) models and assessed those for multicollinearity issues. Condition indices were <30, and variance inflation factors were <6 for all models, which are within acceptable levels. I tested model residuals for independence using Breusch-Pagan and spatial autocorrelation using Moran's I tests. 67 I ran Moran's I tests using multiple spatial weights matrices, including queen first- and second-order contiguity, and K-nearest neighbors (K = 2–6), and selected the weight matrix with the highest autocorrelation coefficients. 68 Spatial autocorrelation was present in each OLS model (p < 0.05), which violates assumptions of independence.
I addressed spatial autocorrelation through creating simultaneous autoregressive (SAR) models, which incorporate spatial effects through either a spatial lag of the dependent variable (spatial lag model) or the error terms (spatial error model). 69 I used the Lagrange Multiplier and Robust Lagrange Multiplier tests to determine whether a spatial lag or spatial error approach was more appropriate. Spatial lag models were indicated for each regression. SAR models with queen first-order contiguity spatial weights matrices had lower Akaike information criterions and spatial autocorrelation was no longer significant (p > 0.05), indicating that the SAR models were more appropriate than the OLS models. SAR models were tested for heteroscedasticity using a spatial Breusch-Pagan test, and robust standard errors were applied when indicated.
RESULTS
Statistical analysis of the relationships between flood risk, proximity to an adaptation action, income, race, ethnicity, age, vehicle access, and disability yielded three major findings. First, two-tailed t-tests comparing adaptation actions with sociodemographics indicated that census block groups with buyouts have significantly lower income levels, a higher percentage of Latinx residents, and a lower percentage of white residents. Census block groups with stormwater ponds have a higher concentration of Asian American residents (Table 2).
Two-Tailed t-Statistics of Difference Flood and Adaptation Means
p < 0.05, *p < 0.1.
GI, green infrastructure.
Second, spatial lag regression results for flood risks (Models 1 and 2) show that after controlling for impervious cover and population density, the percentage of Latinx residents is positively correlated with both the 100- and 500-year floodplains. I estimate that a 1% increase in Latinx residents increases 100- and 500-year flood risks by 6.8% and 7.3%, respectively. Census block groups with higher concentrations of youth and seniors are negatively correlated with the 100- and 500-year floodplains, but block groups with more residents experiencing a disability have a positive relationship with the 500-year floodplain. The remaining variables are statistically insignificant. The flood risk results may be complicated by the fact that Austin has widespread exposure to flood risk, with >60% of the block groups in this study having at least some overlap with the 100- and 500-year floodplains (Table 3).
Simultaneous Autoregressive Analysis of 100- and 500-Year Flood Risks
p < 0.01, **p < 0.05, *p < 0.1.
AIC, Akaike information criterions.
Third, spatial lag regression results for adaptation actions (Models 3–5) are presented in Table 4. Small-scale green infrastructure models indicate a negative relationship with Latinx residents and a positive relationship with households without a vehicle. There are positive correlations between stormwater ponds and Latinx and Asian American residents, but the coefficient for Asian Americans is larger. I estimate that a 1% increase in Latinx and Asian American residents in a census block group increases stormwater ponds by 0.7 and 2.8 units, respectively. There are negative relationships between buyouts and households without a vehicle and Black residents, but a positive correlation with people experiencing a disability. Model outcomes estimate that a 1% increase in Black residents decreases buyouts by 13.2 homes. Whereas a 1% increase in people experiencing a disability increases buyouts by 17.4 homes. Together, these results point toward varying environmental justice concerns, especially for Latinx residents who are more likely to live in an area with exposure to the 100- and 500-year floodplains, and have less access to small green infrastructure, while conversely being more likely to live in proximity to stormwater ponds. Outcomes also indicate that Black and Asian American residents may not face statistically disproportionate flood risk exposure, but Asian American residents are more likely to live in areas with greater access to stormwater ponds, and Black residents are less likely to live in census block groups with city-led buyouts.
Simultaneous Autoregressive Analysis of Flood Adaptation Interventions
p < 0.01, **p < 0.05, *p < 0.1.
DISCUSSION
This case study provides an initial example of a color-blind adaptation approach to flood resilience in Austin, Texas, and raises three potential lessons for adaptation or resilience planning. First, this case study contributes to a growing body of research connecting adaptation planning with environmental justice implications related to inequitable access to physical climate resilience projects. 70 My statistical models draw attention to ways flood resilience practices might exacerbate climate justice issues for Latinx and Black communities. Study results indicate unequal flood risk and access to small-scale green infrastructure projects for Latinx residents, and that Black residents are less likely to live in a neighborhood with city-led flood buyout programs, even when controlling for flood risk.
These findings align with previous research by Anguelovski et al. that described how adaptation planning efforts could increase sociospatial inequalities through interventions that disadvantage lower income communities of color or protect whiter and wealthier neighborhoods. 71 In this article, household income was statistically insignificant in each regression model, which is contrary to what we might expect given previous environmental justice and flood hazard studies. 72 Further analysis on these relationships in Austin should include property values to get a more refined picture of the relationships between income and flood risks. In addition, a separate longitudinal evaluation of relationships between flood buyouts, sociodemographics, and resident resettlement would increase our understanding of the impacts of Austin's flood buyout program.
Second, I suggest that study outcomes are related to color-blind adaptation planning, and future climate resilience planning efforts in Austin may benefit from understanding how color-blind adaptation planning may recreate racial inequalities. Content analysis of the City of Austin's flooding and climate resilience planning documents showed that despite Austin's existing racial and class inequalities and history of racialized urban planning practices, these documents do not acknowledge or attempt to address racialized flood risks. 73 I argue that bypassing discussions of racialized flood exposure is an example of Hardy et al.'s conceptualization of color-blind adaptation planning, defined as “projects that altogether overlook racial inequality.” 74 I am not implying that the City of Austin nor individual planners are being intentionally discriminatory. Instead, I am suggesting that color-blind approaches appear to be politically neutral, but that neutrality at best maintains the status quo, which in Austin is deeply unequal. Therefore, regardless of intent, planning is conscripted into maintaining racial inequalities.
A third planning lesson aligns with long-standing calls to disabuse the concept of planning as neutral, apolitical, or color blind. 75 Like many cities in the United States, Austin is fundamentally a racialized space characterized by systemic and pervasive social inequalities that have been created and maintained through multiple racist ideologies and practices. 76 Therefore, planning in these spaces is never neutral and should be framed as either intentionally working to enact an antiracism praxis or maintaining racial inequalities.
CONCLUSION
As planners are increasingly confronted with climate change impacts and tasked with developing responses that protect residents, ecosystems, and infrastructure, we must both acknowledge and address existing racial inequalities in our climate planning processes, plans, and adaptation actions. Failure to do so is likely to maintain or magnify those inequalities. This argument holds for all domains of planning. However, climate adaptation is still emerging across the United States and offers a timely point for reflection and opportunity to chart a more just path forward.
Footnotes
ACKNOWLEDGMENTS
The author thanks Drs. Robert Paterson, Bjørn Sletto, Pavithra Vasudevan, and two anonymous reviewers for their generous support and feedback during the development of this article.
AUTHOR DISCLOSURE STATEMENT
No competing financial interests exist.
FUNDING INFORMATION
No funding was received for this article.