The statistical model of recreation preferences is estimated from survey data that characterizes outdoor recreation trips taken by a sample of recreators from five counties in New Jersey and is described in Kinnell et al. (2006) . The survey data are from 358 recreators located in Bergen, Essex, Hudson, Passaic, and Union County, New Jersey who recorded 1499 outdoor recreation trips to 181 sites. For the statistical model, the data are augmented with site characteristics and distances from each of the 358 recreators to all 181 sites.

The data was collected using a telephone survey and a mail survey. The telephone survey began in May 2000. It asked the respondents questions regarding their outdoor recreation preferences, past visitation from the beginning of the year through the date of the screener and expected visitation through the end of the year. The screener served as a recruitment tool for mail survey participants and collected information for developing survey weights. Of 6312 telephone numbers for residential households, 4525 individuals were available to be interviewed and 2139 completed the screener. Of these approximately 76% (1629) indicated that they visit parks in New Jersey and 651 (39%) participated in the mail portion of the survey.

The mail portion of the survey collected information on the screened participants’ June and July 2000 outdoor recreation trips. Of the 651 respondents who returned a mail questionnaire, 358 (56%) took trips in June and/or July and provided data on 1499 trips to 181 sites throughout New Jersey. The mail surveys collected actual trip data on outdoor recreation trips, including location visited, distance traveled, duration of the trip, and activities engaged in.

The underlying model used to estimate the recreational preference function is the conditional logit model. This model, introduced by Daniel McFadden in 1974, is used to explain the choices individuals make among a finite set of alternatives, based on the characteristics of the options and the individuals. Outdoor recreation applications are inherently spatial in that they must consider the travel costs of accessing recreation sites. These have been applied in a variety of contexts including recreational fishing (Bingham et al., 2011) , beach visitation (Lew & Larsen, 2008) , forest recreation (Agimass et al., 2017) , and urban park visits as described in Kinnell et al. (2006) which is the model that is enhanced to evaluate environmental justice in this current application.

In the outdoor recreation context, the model estimates the probability of a survey respondent visiting each site based on the characteristics of the sites and distances from their home to each site. The utility index of site j for respondent i is expressed in Equation (1).

$U_j^i=V_j^i+{\epsilon }_j^i=X_j\beta +{\epsilon }_j^i$ (1)

In this expression, X is the matrix of data that represents sites and travel distances, and the vector β is a set of estimated coefficients. The error term represents unobserved factors and averages to zero. The dependent data are represented as [0 or 1] and grouped by choice occasions with a (1) for the site visited on a choice occasion and (0) for sites that weren’t visited. The statistical model estimates coefficients for travel costs and park attributes that are consistent with the conditional logit mathematical structure in Equation (2) and best predict (i.e., maximizes the likelihood of observing) the trip data. As seen in Equation (2), the probability of visiting any site arises from the utility of that site divided by the total utility of all sites.

$P_{ij}=\frac{\exp \left(V_j^i\right)}{{{\displaystyle \sum }}_{j=1}^J\exp \left(V_j^i\right)}$ (2)

This process results in estimated coefficients (β) for distance and park attributes. Coefficients transferred from that model and descriptions are included in Table 1 .

The term environmental justice was popularized when, in 1982, Warren County, the poorest county in North Carolina, was selected by the state as the location for a PCB landfill. Subsequent protests did not stop the landfill but did draw attention to the connections between poverty, race, and environmental conditions

*Significant at the 10% level; **Significant at the 5% level; ***Significant at the 1% level.

(Newkirk II, 2016) . Quantitative relationships between race and proximity to toxic waste were demonstrated several years later (Commission for Racial Justice, 1987) , and in 1992 the United States Environmental Protection Agency established the Office of Environmental Justice which intended to address environmental issues that affect poor and minority communities.

In recent characterizations, environmental justice refers to the fair treatment and meaningful involvement of all people, inclusive of race, color, national origin, or income, with respect to the development, implementation, and enforcement of environmental laws, regulations, and policies (USEPA, 2024) . It advocates for the fair distribution of environmental benefits and burdens across all communities and emphasizes the meaningful participation of all stakeholders in decision-making processes. Environmental justice also seeks to ensure that all members of society, “have equitable access to a healthy, sustainable, and resilient environment in which to live, play, work, learn, grow, worship, and engage in cultural and subsistence practices.” (USEPA, 2024)

This effort adopts the perspective that in the provision of public goods, when one group is at a disadvantage, causing a relative improvement for the disadvantaged group is an improvement in fairness. Given the linkage between fair treatment and environmental justice, it follows that improvements in fairness result in improvements in environmental justice.

Measuring the value of spatially distributed public goods requires spatially explicit measures of value. Although site choice survey data and an estimated model, such as that of Section 2, can characterize the satisfaction of the surveyed individuals, this information generally cannot be directly used to evaluate value spatially. Although the data are suitable for estimating a preference function, they are not spatially representative because many neighborhoods and parks do not appear in the data used to estimate the statistical model.

For example, the data employed to estimate the preference function used here is from a sample of 358 recreators who recorded 1499 outdoor recreation trips to 181 sites. As seen in Table 1 , this number of trips is sufficient for estimating a preference function with many significant coefficients. However, the 358 recreators are drawn from a population of 3.3 million people who live in thousands of different neighborhoods, and the 181 parks represent a small portion of more than 1000 parks in the study area.²

This shortcoming is addressed by conducting a geospatial fusion that connects the estimated preference function to the entire region of interest. Doing so requires collecting location and characteristic data for parks that could be visited by people who live in the modeled region. Although sites and their characteristics are sometimes available in inventories created by recreation professionals, this is not usually the case. Moreover, any area will contain multiple jurisdictions such as state, county, and local.

In addition to park information, the approach requires recreator origin and population information. Census designations are ideal for this. The smallest geographic areas defined by the United States Census Bureau are blocks. The primary reason for using blocks is that they allow the best spatial resolution. This is particularly important for urban parks where small scale distance is a critical feature of utility. Census block data carries the latitude and longitude of the centroid of the block. With this, and park latitude and longitude, travel distances are calculated from all origins to all parks.³

Census blocks vary in size and shape, but are designed to be relatively uniform in population size. Census blocks are delineated based on a variety of factors, including population density, physical features such as roads and bodies of water, and administrative boundaries. Although census blocks do not specifically represent neighborhoods, they often contain groups with similar preferences. Certain information, such as detailed income breakdowns, are not available at this level. However, it is possible to transfer information from higher levels down to census blocks.

Having park and distance information, it is possible to calculate the value of parks with and without consideration of their location. This assessment focuses on Hudson County, New Jersey. Considering county and local parks there are a total of 214 sites in the county. With notation for surveyed sites (j) changed to reflect all parks (p), the summed product of the site characteristics and coefficients yields the utility for each site (U_p) independent of its location and can be calculated as in Equation (3).

$U_p=\exp \left(V_p\right)$(3)

The results for each modeled park are depicted in Figure 1 . In this figure, the size of the circle indicates its quality based on the site-related attributes of Table 1 . For example, the largest circle represents Lincoln Park in Jersey City, which has 273 acres, 6.8 miles of trails, sports facilities, a waterbody, bathrooms, and playgrounds. There are many low value parks, indicated by little circles. These are typically small acreage open spaces without any facilities or water.

Parks provide value based on both their quality and location. Given the quality of parks and locations of recreators, the disutility of distance for each block and park is added to get the distance adjusted utility for each block. Replacing the i with b to represent the shift from surveyed individuals to blocks, and summing over all parks returns a single number representing the utility of all parks for each block.

$\underset{p=1}{\overset{P}{{\displaystyle \sum }}}\exp \left(V_p^b\right)$(4)

Figure 2 depicts the relative per-trip value for census blocks in Hudson County. The value is made relative by dividing the results of Equation (4) by the maximum of Equation (4). In Figure 2 , the darkest blocks receive the highest per-trip value from park visits. Lighter shading indicates less relative value. As this figure indicates, value is highest in the center of the county. Values are scaled by dividing by the highest value so that the highest value block receives one hundred percent value. The lowest value block receives approximately 13 percent of the per-trip value received by the block with the highest value.

Under Executive Order (EO) 14008, federal agencies are required to consider actions to address disproportionately high and adverse health, environmental, climate, and other cumulative impacts on disadvantaged communities. These agencies include federal trustees who are parties to National Resource Damage (NRD) claims.⁴ Under the Comprehensive Environmental Response, Compensation, and Liability Act, 42 U.S.C. § 9601, et seq., responsible parties are required to restore offset losses. EO 14008 requires that environmental justice should be considered when conducting settling and crediting activities for Natural Resource Damage Assessments (NRDA).

A 2020 New Jersey Department of Environmental Protection (NJDEP) report that evaluates Environmental Justice communities identifies 3,154 Census block groups with at least 35 percent of households classified as low income, or at least 40 percent minority, or at least 40 percent having limited English proficiency. East Newark is noted as having 100 percent of its population meeting the low

income criteria and 100 percent of its population meeting the minority criteria, resulting in New Jersey designating East Newark as an Environmental Justice Community (NJDEP, 2020) .⁵

In July of 2022, the U.S. Department of Justice finalized an agreement with the

BASF Corporation (BASF) to settle part of the damages from the Diamond Alkali Superfund Site and the Berry’s Creek Study Area (NOAA, 2024) . The agreement includes the creation of a five-acre park along the shoreline of the Passaic River in East Newark. East Newark Riverfront Park will include walking paths, forested areas, pollinator gardens, a wetland, and green spaces for recreation. The project announcement notes that New Jersey has designated East Newark as an Environmental Justice Community. It also states that East Newark has very limited access to public parks and open greenspace options, has insufficient tree canopy and areas where the public can walk or bike, and that correcting such disparities is the goal of environmental justice. An additional notable feature of this project is that it is being developed before remediation is complete (Federal Register, 2022) . Although accelerating restoration activities in this manner has recognized advantages, these benefits have historically been challenging to realize (Goldsmith, 2023; Stahl et al., 2023) .

The environmental justice implications of East Newark Riverfront Park can be evaluated by conducting counterfactual simulations with the park added to the set of parks from which recreators choose. This is accomplished by adding East Newark Riverfront Park to the model, identifying its appropriate geographic location, and specifying its attributes: five acres, a quarter mile of trails, a picnic area, sport facilities, a waterbody, bathrooms, and a playground.

Environmental justice can be evaluated spatially through the block level value changes.⁶ These are depicted in Figure 3 . Comparing Figure 2 and Figure 3 indicates the degree to which previously disadvantaged areas are being improved by the addition of East Newark Riverfront Park. As seen in Figure 2 , the area near the new park historically has urban park-related value that is among the lowest in Hudson County. Figure 3 shows that the Census blocks near the park have the highest value improvements. The value changes presented in Figure 3 are percentage improvements relative to baseline and range from about one percent (lightest yellow) to six percent (darkest brown). Comparing Figure 2 and Figure 3 indicates that areas that previously had low value are seeing it improved.

Direct evaluation of location effects in East Newark is possible by comparing value in baseline to counterfactual value. To evaluate value change by demographics, the per-trip value by block is multiplied by demographic membership by block and then divided by the sum of total per-trip value for blocks in East Newark, resulting in the percent of the total value change in East Newark by demographic group. Table 2 summarizes these results by the US Census’ demographic groups.