Integrated Urban Riverscape Planning: Spatial Prioritization for Environmental Equity

For our spatial MCDA application, we collaborated with the City of Charlotte and Mecklenburg County in North Carolina (Fig. 1), located in the southeastern piedmont region along the Charlanta megaregion. The western and southern portions of the county drains to the Catawba River Basin and the eastern streams are tributaries to the Yadkin River. The 500-year floodplain excludes the reservoirs along the western boundary (Lake Norman, Lake Wylie). The study focused on the portions of 33 watersheds and their subbasins within the county boundaries. Similar to other municipalities, Charlotte–Mecklenburg Storm Water Services (CMSWS) divides surface water management responsibilities among several main groups, including watershed planning, engineering and flood mitigation, and water quality. Some of the existing CMSWS approaches to prioritization include a watershed-scale water quality matrix (J. Hunt, personal communication, 2022), building-level flood risk assessment/risk reduction (RARR) tool (Charlotte–Mecklenburg Storm Water Services 2020), and reach-scale stream restoration ranking system (SRRS) (Mecklenburg County Storm Water Services 2021), all of which are supported by extensive spatial data. When planning stream improvements, CMSWS often partners with the Mecklenburg County Park and Recreation Department to incorporate greenway trails and other outdoor amenities. Mecklenburg County recently developed an Equity Action Plan, and CMSWS wanted to better understand how they could include social components with their surface water projects and initiatives. In general, our goal was to integrate the established CMSWS components (RARR, SRRS, etc.) while taking steps to identify environmental equity objectives. At the same time, our intent was to develop a generalized approach that could be broadly applied in other urban stream systems.

With a structured decision-making process, stakeholder groups typically work collaboratively to identify objectives and possible metrics (Bridges et al. 2015). This research was intended to support CMSWS with early planning and outreach initiatives by identifying areas for follow-up with neighborhoods and communities. Through conversations with CMSWS, we elicited baseline criteria and subcriteria for the spatial MCDA, which can be modified later as part of collaborative conversations. For analysis at the watershed and subbasin scales, the main criteria included three riverscape criteria (flood regulation, water quality regulation, ecosystem support) as well as amenity access and environmental justice (Table 1). We used a two-part approach to environmental justice by including measures of social vulnerability and historic injustice in a general landscape category, complemented by subcriteria in riverscape categories targeting specific aspects of social inequity, such as disproportionate exposure to flood risk or a lack of greenspace access. Following prioritization at the large (watershed) and medium (subbasin) spatial scales, the same objectives may be combined with additional feasibility criteria to evaluate local project alternatives. Preferences from multiple stakeholder groups can be separately elicited and then weighted together, and a trade-off matrix can be used to show the highest rated alternative for each group (Bridges et al. 2015).

Spatial prioritization approaches that incorporate social objectives typically use some type of demographic-based SVI (Meerow and Newell 2017), and environmental justice is especially important for the vulnerable communities identified by SVI. Environmental equity, on the other hand, relates to the distribution of risks and benefits, and metrics involve both social vulnerability (e.g., race, ethnicity, income) and either exposure or access, respectively. This is because environmental hazards to human health and well-being (e.g., flood damage, water pollution) also involve exposure, just as environmental benefits (e.g., greenway trails, outdoor education) require access opportunities. For example, a poor neighborhood is socially vulnerable in a general sense, whereas a poor neighborhood in a low-lying area floodplain area is both socially vulnerable and at risk of exposure to flooding.

Our approach to environmental equity built upon prior methods of analyzing inequitable flooding in the Charlanta megaregion (Debbage 2019). Starting with Mecklenburg County parcels and the 2019 NLCD, we selected parcels only where all or the majority of them were in the developed range between open space and high-density areas (NLCD classes 21–24). While the landcover raster was used to identify developed parcels (and can be easily accessed for other locations), there might be more precise ways to filter out parcels that are undeveloped open spaces or otherwise vacant.

For a baseline flood scenario, we then identified developed parcels overlapping the 500-year flood hazard zone. We opted to use a parcel-based approach to area calculations following Selsor et al. (2023) rather than raster coverage within each census block used in an earlier study (Debbage 2019), because parcels provide more accurate spatial resolution. In the GIS attribute table for parcels, we added fields to include census block group number as well as selection categories (developed, 500-year flood, water quality exposure, etc.), with 1 (yes) or 0 (no). Using the Summary Statistics tool (ArcMap v. 10.5), we calculated the area sum and exposure risk factor for each census block group (N = 624) as shown here:

We then estimated the number of exposed individuals in each category by multiplying the risk factor with the group populations from the census block group demographic data (Non-Hispanic Black, non-Hispanic White, Hispanic, Non-Hispanic, Below Poverty, and Above Poverty). The total number of individuals for each category was summed together for a larger spatial unit, typically the census tract. However, we also assigned each census block group to a subbasin and watershed based on the centroid of the census block group, thereby creating demographic watersheds as an alternative to conventional topographic delineation. The overall topographic and demographic watersheds were similar but not identical, and we would not recommend using the latter for hydrologic calculations. However, these demographic boundaries enabled us to directly compute categorical populations and risks without needing to apply weighting based on spatial areas and variable population densities. As with prior flood inequity studies (Debbage 2019; Selsor et al. 2023), we calculated risk ratios as follows:

A risk ratio greater than 1 indicates environmental inequity, such as a predominantly Black community with a disproportionately high exposure to flood hazards. Approximately one-third of the developed and flooded parcels also had RARR scores greater than zero, which we used for a separate flood scenario based on an established CMSWS flood mitigation strategy. We used the RARR scenario for the MCDA, but the baseline 500-year scenario could easily be implemented elsewhere. We used a similar risk calculation method for water quality equity. To approximate exposure to surface water pollution (e.g., fecal coliform bacteria), we combined the 500-year flood zone with the stormwater buffers, which ranged in width between 35 and 100 ft, thereby including the smaller streams that also convey polluted water. Fig. 2 illustrates the different delineation methods used for environmental risk exposure. Risk factors were based on developed parcels: (1) with building flood risk assessment/risk reduction (RARR) scores greater than one; (2) overlapping the 500-year flood zone; and (3) overlapping the water quality zone (combined stormwater quality buffers and 500-year flood zone).

We calculated benefit ratios for amenity access in a similar fashion, except that the data for population near public outdoor recreation had its own neighborhood spatial units (N = 464), and we did not use parcel areas or the NCLD raster. The equation for a benefit ratio looks nearly identical to a risk ratio, like the following example:

However, opposite from risk ratios, a benefit ratio less than 1 indicates social inequity, such as a low-income neighborhood with less access to public parks. For each scenario, after summing up the at-risk and total populations for the demographic groups at the tract, subbasin, and watershed scales, we obtained risk ratios using R statistical software. While the ratios can be calculated easily using a spreadsheet or GIS attribute table, the fmsb package in R also provided p-values to describe statistical significance (<0.05). However, the p-values were not used for spatial MCDA prioritization purposes.

Following the method of Selsor et al. (2023) to create a single combined equity metric for each spatial unit (census tract, watershed, subbasin), we first reassigned values of 1 for all risk ratios less than 1 and benefit ratios greater than 1 (i.e., no inequity) and then added together the ratios for race, ethnicity, and income level. For example, the poverty benefit ratio [Eq. (5)] was rolled together with the race and ethnicity benefit ratios to generate the overall amenity equity scores. In addition to calculating risk ratios for the Mecklenburg County census tracts (N = 305), we assigned each census block group to both a watershed (N = 33) and a subbasin (N = 113) and calculated equity at the larger scales to better support spatial prioritization. Nine subbasins, all located at the county boundaries, did not contain centroids of any census block groups and therefore received the lowest priority for the various equity subcriteria.

The basic procedure was to develop scores for all of the subcriteria at the watershed and subbasin scales, convert to a common priority scale (e.g., 0 to 1), and then apply weighted averages to calculate combined criteria and overall scores. For the purposes of this study, we assigned equal weights to all subcriteria to calculate the overall priority scores for the criteria, and then we assigned equal weights to all of the criteria to calculate the combined priority score. It is possible to use a Weighted Sum tool in ArcMap, but compiling the data in a spreadsheet with inputs for variable weights is more user-friendly for stakeholders (Fig. 3). The combined criteria and overall MCDA priority scores are automatically calculated for watersheds and subbasins.

We found that inclusion of the landscape criteria (amenity and environmental justice) in the Combined MCDA substantially altered the spatial prioritization compared with the Riverscape MCDA based only on flood regulation, water quality, and ecosystem support, but the differences were evident only at the subbasin scale [Figs. 6(c and d)]. The subset of criteria used for the Riverscape MCDA scenarios most closely align with the primary management goals elicited from our CMSWS collaborators as well as potential landscape-scale interventions and natural infrastructure strategies. For the watershed scenario including both riverscape and landscape criteria, Lower Little Sugar and Irwin had the highest combinations of priority scores [Figs. 6(a) and 7], with Upper Little Sugar and Sugar closely tied for third place—spatially, these watersheds are directly adjacent to one another. The water quality, ecosystem support, and amenity access overall criteria scores were comparable among the top few watersheds, but Upper Little Sugar had the highest environmental justice metrics (social vulnerability, population density, historic redlining), whereas Sugar exhibited the highest average flood risk (RARR scores). The highest priorities for the Riverscape MCDA scenario [Figs. 6(b) and 7] again included Lower Little Sugar, Sugar, and Irwin watersheds. Not surprisingly, Upper Little Sugar dropped in priority without the general environmental justice criteria. However, the Riverscape MCDA scenario still incorporated context-specific environmental risk equity. With side-by-side comparisons of the two priority results, we found similar watershed priorities between the two scenarios (Fig. 7). Lower Little Sugar scored highest for both scenarios, and the other watersheds with the highest priorities include Irwin, Sugar, and Upper Little Sugar. The Riverscape MCDA watershed priorities for Lake Norman and Twelve Mile were zero, because they had no developed parcels with environmental risk exposure (Fig. 2) or stream reaches with SRRS scores.

The subbasin prioritization highlighted the spatial hotspots combining social–ecological system vulnerabilities and deficiencies across multiple dimensions. We found a distinct contrast between the subbasin priorities for the two MCDA scenarios [Figs. 6(c and d) and 8], although subbasin hotspots are evident with both scenarios. For the combined scenario based on all five criteria, subbasins within the same watershed showed a degree of spatial similarity [Fig. 6(c)], and there were many more high-priority areas, which were probably due to the inclusion of landscape-based amenity access [Fig. 5(d)] and environmental justice [Fig. 5(e)]. In contrast, the riverscape scenario generated a distinct spatial pattern of isolated hotspots [Fig. 6(d)], a splotchy appearance corresponding to just a few outliers with the highest priority scores. These results make sense, given that the additional criteria (amenity access, environmental justice) used in the first scenario are distributed across the entire landscape rather than only in the arterial waterways. In Fig. 8, the Combined MCDA scenario including both riverscape and landscape criteria often generated higher mean values of basin priorities compared with the Riverscape MCDA.

A part of the purpose of a spatial MCDA is to help stakeholders leverage potential synergies among multiple objectives. An individual watershed or subbasin with a much higher overall priority score compared with others has greater combined effects from its constituent criteria. For example, the Lower Little Sugar watershed rose to the top of both MCDA scenarios (Fig. 7) with upper quartile scores for flood risk, water quality, ecosystem support, and environmental justice criteria (Fig. 4).

When we looked for positive or negative correlations between criteria across all watersheds, we found little statistical power to support conclusions about system-wide synergies or trade-offs. However, for watersheds with monitoring data, we did find evidence of a positive correlation (0.60, p = 0.0027) between the overall water quality and general environmental justice criteria [Fig. 9(a)]. This result intrigued us, especially because none of the water quality equity subcriteria scores (race, ethnicity, income, combination) were correlated with any of the environmental justice subcriteria. However, the lack of relationship between the equity scores and the general environmental justice criteria underscores the value of including both types of metrics and not just social vulnerability. The broader metrics are still valuable for capturing neighborhood characteristics beyond environmental risk exposure because neighborhoods may not spatially correspond to topographically delineated watersheds and subbasins.

When we analyzed the subcriteria, we also found a positive correlation between the equity scores for flood risk (RARR) and water pollution exposure at both subbasin (0.63, p < 0.0001) and watershed scales (0.51, p = 0.0022), as shown in Fig. 9(b). The environmental risk equity relationship makes sense, given the similar methods we used to develop the two metrics, with the only difference being the extent of exposure. This correlation might also have contributed to the pattern of subbasin priority outliers visible in Figs. 6(d) and 8 if there was an amplifying effect by including separate flood and water pollution risk ratios in the waterway-focused MCDA scenario.

The environmental equity aspect of our study produced several important results and applications. First, we found that spatial trends in flood inequity were in agreement with previous findings about the areas in Charlotte with socioeconomic disparities, and the magnitudes of risk ratios above 1, despite the differences in how we defined flood exposure and aggregated risk areas using parcels. For example, Debbage (2019) found that the Landsowne neighborhood (Census Tract 37119002004) had some of the highest risk ratios that were statistically significant, the worst being 3.28 for the Non-Hispanic Black population, and we found that the corresponding RARR and 500-year flood risk ratios were both 3.40 (p < 0.0001) for the same tract. However, the spatial resolution of the analysis matters, because Debbage (2018) found only one significant risk ratio (below poverty versus above poverty) for Mecklenburg County as a unified whole. Like Debbage, however, we found a different story when zooming into the census tract level for equity comparisons by demographic groups (Fig. 10). The Hispanic, Non-Hispanic Black, and Below Poverty demographic categories all had more census tracts with inequitable risk ratios (>1) for both flood scenarios. While the presence of any risk ratio greater than 1 shows that there are equity concerns that need to be addressed, the number of tracts with statistically significant risk scores greater than 1 exceeded that with risk scores less than 1, suggesting overall inequities across race, ethnicity, and income level. In contrast, among subbasins with statistically significant risk ratios, only the poverty category continued to show a negative disparity, and the larger watershed scale altogether erased this tendency through overall aggregation.

In addition to confirming overall trends in flood risk, we were able to take a step toward operationalizing flood risk equity through spatial prioritization. For starters, we combined multiple minority and income categories (Hispanic, Non-Hispanic Black, Below Poverty) to create a single equity score for each subbasin and watershed, thereby including intersectionality of these socially vulnerable classes (e.g., low-income Black and Hispanic people). Leveraging multiple spatial scales was also useful when characterizing the distributions of environmental risks and benefits. Even if risk ratios at the large county scale failed to demonstrate overall inequities, the medium and small spatial scales enabled us to determine which watersheds, subbasins, and tracts were the areas of greatest concern. Therefore, we applied the environmental equity scores as part of the criteria for a spatial MCDA.

Moreover, we found that the environmental equity techniques were both relatively simple and flexible enough for a range of practical applications. For example, we were able to use the 500-year flood zone as a proxy for flood risk similar to previous work (Debbage 2019), and the same basic technique worked with alternative definitions of environmental risk (e.g., surface water pollution exposure) and a specialized approach that incorporated risk frequency and probability (i.e., RARR scores). Access to environmental benefits (i.e., public outdoor recreation) was also readily incorporated in a comparable fashion through the use of benefit ratios. Furthermore, it was relatively straightforward to estimate risk and benefit factors (i.e., exposure and access), combine them with census data, and then perform risk and benefit ratio calculations with tools widely used by municipalities and utilities (GIS, spreadsheets).

Properties and people suffering from flood damage, underfunctioning stream reaches that fail to support aquatic life, and impaired watersheds that fall short of regulatory goals are just the tip of the urban riverscape iceberg encountered by organizations and stakeholders. Our spatial MCDA incorporates all of these specific concerns in an approach that further addresses some of the underlying challenges and system drivers. For example, these problems involve multiple spatial scales with variable degrees of interaction, and our MCDA approach includes both the large (watershed) and the medium (subbasin) landscape context, with further potential to modify metrics for evaluating local project alternatives while sharing the same overarching criteria. The sheer scope of responsibilities held by stream and watershed managers has led to a natural division of labor, which can function as a barrier. Our spatial MCDA bridges departmental boundaries and includes input from CMSWS individuals tasked with multiple missions, which is reflected in the range of criteria and incorporation of various established management tools (i.e., RARR, SRRS, Water Quality Matrix). Engineers and scientific specialists might also struggle with questions related to human dimensions outside of their technical expertise, such as how to incorporate social equity in a meaningful way. For example, this study agreed with previous findings of flood inequity in the Charlotte metropolitan area (Debbage 2019) with regard to locations and magnitudes of socioeconomic disparities, which is most evident at smaller spatial scales. Our dual approach to environmental equity in the urban stream spatial MCDA uses widely recognized metrics of social vulnerability in tandem with equitable distributions of specific environmental risks and benefits, similar to prior spatial planning for green infrastructure (Meerow and Newell 2017). Finally, the spreadsheet serving as the user interface for the spatial MCDA helps address the inherent challenge of multiobjective prioritization in a transparent and flexible fashion through weighted criteria and subcriteria that can be easily modified by multiple stakeholder groups to explore alternative riverscape scenarios.

With our urban stream spatial MCDA, we wanted to help CMSWS identify areas in Mecklenburg County with opportunities to realize multiple potential benefits. In contrast to a green infrastructure study in Detroit (Meerow and Newell 2017), we found little evidence for synergies or trade-offs across criteria, perhaps because we were focused on riverscapes with different environmental risks and benefits. Although different weightings may result in different prioritizations (Meerow and Newell 2017), our case study did highlight specific watersheds and subbasins as promising locations for addressing multiple objectives. Similarly, the inclusion of alternative or additional metrics, or even the normalization of the metrics, may lead to different spatial priorities.

Identifying hotspots on the basis of more environmental risks, coupled with fewer existing benefits, was the primary task of our MCDA approach, but we envision a complementary spatial prioritization with benefits-related subcriteria such as future greenway trails, adopted streams, outdoor education, and other social connectivity. Furthermore, specific natural infrastructure solutions, such as those for flood mitigation, involve different types of spatial criteria (Hovis et al. 2021). If stream renovation (restoration, revitalization, naturalization, etc.) priorities feature ecological uplift, then upgrading urban aquatic and riparian ecosystems from poor to fair may involve the use of success indicators that are different from indicators such as improvement from fair to good. For example, a poorly vegetated stream buffer zone presents a potential opportunity to enhance riparian conditions and support terrestrial wildlife, but connectivity to a high-quality forested stream reach may increase the likelihood of restoration success for fish and aquatic insects. In short, what makes for a good opportunity for environmental benefits and co-benefits is highly context specific, so it may be most appropriate to conduct separate analyses at the subbasin, floodplain, and/or stream reach scales rather than lumping opportunities together with social–ecological system vulnerabilities and deficiencies. Feasibility criteria could also be incorporated to reflect constraints faced by managers, such as Clean Water Act compliance, water capture potential across the landscape, or available acreage for project development.

The inclusion of environmental equity in our case study corresponds to particularly valuable opportunities, providing a potential starting point for water managers struggling to incorporate social components. We demonstrated how to include social vulnerability and advance environmental equity with our spatial MCDA. Moreover, our approach is intended to be a collaborative decision support tool for community inclusion and diverse stakeholder groups. That being said, when adding new subcriteria with priority scores based on potential opportunities (e.g., specific NI solutions, alignment with other planned improvements, etc.), it will be especially important to review potential trade-offs with environmental equity criteria, such as risks associated with neighborhood displacement and gentrification (neighborhood change score). We are concerned about the potential for win–win–win scenarios based on some combination of good opportunities or a resilience narrative that inadvertently reinforces existing systemic injustices caused by unacknowledged social trade-offs (Eakin et al. 2017; Béné et al. 2018). However, the potential alignment that we found between the overall water quality and environmental justice criteria is a promising avenue for further study and action.

Working together with CMSWS on the urban stream MCDA case study demonstrated one practical application of our approach. However, our larger intent was to create a multifunctional spatial prioritization framework to support the larger body of managers, practitioners, and communities tackling urban riverscape challenges. Integrating existing management tools and strategies such as those used by CMSWS (RARR, SRRS) showed how the general approach could be tailored for specific organizations and departments. However, the spatial MCDA could easily transfer to other municipalities with alternative methods for characterizing flood risk, channel stability, water quality, and so on. For example, floodplain managers could apply our spatial MCDA in tandem with probabilistic mapping (Stephens and Bledsoe 2020) or frequency-based risk equity (Selsor et al. 2023), and use different methods for determining which parcels are developed and are at risk. Environmental scientists could evaluate and prioritize stream restoration using ecological potential based on the ratio of existing to predicted biotic scores (Paul and Allen 2022). Subcriteria related to fecal coliform and water quality could incorporate spatial data about basement backups (Alves et al. 2021). Census data, the CDC version of SVI, and historic redlining maps are readily available in the United States to support environmental equity objectives, and our methods for delineating demographic watersheds and calculating environmental risk and benefit ratios are highly adaptable. The spatial MCDA used software tools that are already familiar to technical specialists in our target audience, and GIS can be coupled with story maps to present information to the wider group of stakeholders (Meerow and Newell 2017; EPA 2020).