Cheryl C. Swift and Jeffrey A. Henderson

Using GIS to Evaluate Land Use Change and Recreation Accessibility in the San Gabriel Valley: A Study in Environmental Justice

Abstract

This paper describes progress on a study using GIS to statistically determine the impacts of land use change, demographic change, and political conditions on the accessibility of quality recreation in San Gabriel Valley communities. Drawing upon a variety of data sources and methods from many academic disciplines, we are building a GIS database to describe land use changes (1928-1995), demographics, government structure and function, and park facilities and users. This database will be used in the curriculum at Whittier College to help educate students about GIS and environmental justice. It will also help politicians, planners, and developers in San Gabriel Valley communities understand more fully the inequitable impacts that urban development has had on the availability of quality recreation space, and will forward a predictive capability to estimate future impacts of current development proposals.

Introduction

The Los Angeles, California basin is the setting for a story that many land use planners and environmentalists wish they could forget. At the turn of the Century, southern California was a lush, natural landscape -- characterized by meandering streams, picturesque hillsides, and a variety of natural features and amenities. Some would have likened it to a garden. Today, Los Angeles is the second largest urban region in the United States. True, it is the home of a thriving, diversified economy and populace. However it is also the setting for air pollution, racial divisions, poverty, crime and other urban ills that regions elsewhere in the nation cannot rival. We have converted from an almost agrarian land ethic to an environment that is characterized by a striking lack of open and recreational space. Moreover, in Los Angeles -- perhaps more than anywhere -- these and other externalities that are inherent in urban development of this magnitude are borne inequitably by those who can least afford to mitigate them.

Given the drastic changes in the form of the city over time, investigations of environmental justice in Los Angeles should account for the impacts of historical change. Historical elements assembled for a study of environmental justice should encompass both the physical and demographic legacies of a region, as well as the social issues, perceptions, and trends that accompanied these changes. The former history assists in documentation of environmental impacts. The latter assists in interpretation of those impacts as positive or negative.

Premise

Environmental justice can be summed up most simply as the equitable distribution of the burden of externalities that are an inherent part of urban development. The most common examples of environmental "injustice" are the location of hazardous wastes, landfills, polluting industries or other noxious uses in areas populated predominately by lower income families or persons of color. These traditional NIMBY's are readily available examples of the kinds of injustice that take place routinely in land use planning practice.

Whittier College is currently completing a two-year, multidisciplinary study of environmental justice that treats the allocation of recreational and natural open spaces in the context of Los Angeles urban development, sponsored by the Willie Velasquez Research Institute (formerly Southwest Voter Research Institute). Specifically, this project addresses the manner in which park accessibility, size and character change relative to shifts in demography and land use over space and time. The selection of parks and recreation as dependent variables for the analysis signals a shift from the traditional focus of environmental justice studies mentioned previously. In fact, this approach is unique in its investigation of the equitable distribution of "positive" land uses across a region. By investigating the "flip side" of environmental justice, we wish to pose a question: Is it not just as important to provide access to positive influences as it is to restrict exposure to negative ones?

Scope

We describe herein the foundations for a two-year multi-disciplinary study of environmental justice in the history of Los Angeles urban development. The project's purpose is essentially three-fold: 1) to construct a database for the longitudinal study of environmental justice in Los Angeles, 2) to train students to understand GIS technology and the process of land use planning, and 3) to investigate recreational issues relative to environmental justice.

Based both upon the academic tradition of Whittier College and the political reality of grant administration at a small liberal arts college, the project has participating faculty from the Departments of Biology, Geology, Sociology, Political Science, Physical Education and Recreation, Economics, and Law. Each faculty representative has varying responsibilities within the project, which range from overseeing the work of student interns, to preparing data and reports for project submissions.

The foundation of the study is two consecutive summer internships. The internships are set aside for undergraduate students from Whittier College, who have expressed interest in pursuing study in urban affairs, environmental science or planning. Minority and underprivileged students are encouraged to apply for the internships, which are awarded through a competitive process. These summer programs are administered by the project manager and participating faculty. They are composed of two weeks of intensive training in the use of GIS, followed by eight weeks of intensive data collection in topic areas described later in this paper.

GIS Database Elements

Figure 1: Schematic Diagram

To investigate the problem of environmental justice posed by land use conversion, the study was divided into four topical areas: Government Structure and Function, Parks and Recreation, Land Cover, and Demographics. These are illustrated as gold boxes in Figure 1. Each of these tasks could be investigated individually -- in fact within their respective disciplines, there are established techniques and methods for gathering and analyzing information that may seem drastically different. However, all of these tasks share one thing in common -- they all describe human activity that takes place at specific, discrete locations on the earth's surface. If this is the case, these activities can be mapped -- and if they can be mapped, a GIS will allow us to compare them both spatially and descriptively. Our study uses established research techniques within each discipline to investigate the problems that arise within each task outlined above. The results of these investigations are sorted into spatial and descriptive attributes and placed into the Geographic Information System to begin the process of drawing connections between them. Each layer of information is reduced to its least common denominator -- it takes place at a specific location in Los Angeles and can be correlated with other events that occur at that same location.

By constructing a four-dimensional interdisciplinary "view" of the San Gabriel Valley and the cities within it, students are exposed first hand to spatial interdisciplinary study, learning to link events to places. By re-constructing that "view" for previous years, students are encouraged to add a dimension of time to these connections. The goal of considering both interrelatedness and time is to forward a theory of cause-effect relative to environmental justice. This cause-effect question is a complex matter that has been widely debated and requires an historical perspective to resolve or disprove. If one can isolate conversions of land and aggregate them to census tracts it becomes possible to correlate land changes with demographic conditions. We are working toward completion of GIS data layers for historic census years that will allow us to correlate the changes with demographic shifts in a stepwise fashion. This, we believe, will allow for substantial determinations of initial presence of either population or use.

The Environmental Justice Project addresses a study area that encompasses much of the midsection of the Los Angeles River and San Gabriel River basins in Los Angeles County (See Figure 2). We have attempted to include cities (and unincorporated areas) that differ both in their predominant land use patterns, and in demographic profiles. The study area (present day) consists of many jurisdictions, listed in Table 1.

Methods

Each step in this interdisciplinary process may be characterized by methods chosen for data collection, automation and integration. A description of these methods as applied to the database elements for land use change, demographics, and park facilities follows. (1)

Land Cover Change

Land use change is being considered at five separate intervals over a sixty-seven year time period: 1928, 1945, 1965, 1975, and 1995. Resources for land cover analysis are largely drawn from the Fairchild Aerial Photography Collection at Whittier College. Full coverage of the study area is available for each of the previously mentioned years. Student interns were trained to interpret land cover from the photographs using a modified version of the Anderson system devised for interpretation of remotely sensed data. (2)

Students were encouraged to use the full range of the four-tier classification system when the scale of photography would afford a greater level of detail. Maps created from these aerial photos were digitized as vector GIS coverages, indexed to page boundaries for the Los Angeles County Thomas Brothers guide. Completed pages were then assembled together to create mosaics covering the entire study area. At the time of this publication, vector coverages have been completed for 1928 and 1995. Additional coverages for 1945, 1965, and 1975 will be completed during summer 1997.

For purposes of comparison, the land cover layers were converted to raster grids at 120m resolution. Additionally, the four-tier classification system was simplified to two-digit detail. (i.e. Single-Family Residential and Multi-Family Residential were collapsed to Residential, Service Industries and Retail Sales were collapsed to Commercial, Coastal Sage Scrub and Chaparral were collapsed to Brushland, etc.) The resulting grids were then COMBINEd in ArcInfo GRID, yielding a land cover change matrix. Grid cells indicating changes from natural vegetation or agriculture to developed uses were isolated for comparison to current demographic and park attributes.

Demographics

Demography is also being considered at multiple intervals, generally corresponding to intervals of land cover analysis. We are investigating demographic variables from the 1950, 1960, 1970, 1980, and 1990 Censuses. (3) We have selected census tracts and cities/jurisdictions as our units of analysis, primarily because of the difficulty associated with obtaining data for smaller areas prior to contemporary censuses. Student interns have been trained to construct historic census tract maps by transferring boundaries from paper manuscripts to the project's existing road base. Interns have also received basic training in extracting census data from CD-ROM and the World Wide Web.

A preliminary analysis associating per capita park space with a variety of demographic variables indicative of community demographic profiles was conducted at the city level using an extensive sample of cities within Los Angeles County. (4) A covariance matrix was run on 32 variables selected from the 1990 Census and other sources to trim the list to avoid redundancy and multicollinearity. As a result, five variables were run against per capita park space using the technique of multiple regression. Sixty-four (64) percent of the variation in park space per capita among the representative cities in LA county was explained using five primary variables. Several secondary variables were also retained from the covariance analysis to provide alternative or substititute values for primary variables that may not be reported at census tract level due to inadequate sample size. The primary and secondary variables are listed in Table 2. These variables have been used for our further analysis of census tract - level data.

At time of this publication, census tract coverages containing boundary data and the variables listed above have been completed for 1990, 1980, and 1950. Additional coverages for 1960 and 1970 will be completed shortly, affording us the ability to correlate land use changes and demographic shifts in a stepwise manner. For purposes of comparison to tracts contained within incorporated cities (other than Los Angeles), tracts in unincorporated Los Angeles County and within the City of Los Angeles have been assigned "boroughs" or neighborhoods. These subareas are listed along with the incorporated cities within the study area in Table 1. Demographic data remains in vector format at the tract and city level for comparison with other data layers and previous census statistics.

Park Facilities

Data collected for park facilities provide (primarily) a dependent variable for our analysis. We are primarily interested in park availability (both size and distribution) and park quality throughout our project study area. Availability can be determined using variations of National Park and Recreation Association (NRPA) size and distribution requirements. Quality can be determined using secondary data sources such as city and county records, and verified using on-site surveying of facilities present and their condition. Student interns have been trained to review secondary data sources, perform on-site facility surveys, interview park users and activity leaders, and incorporate these data as attributes for a vector park coverage spanning the extent of the project study area.

Parks within the study area have been classified using a modified version of the NRPA standards. Park classifications used within this study are listed in Table 3. It should be noted that these first classifications are based exclusively upon acreage, and that more thorough classification schema would rely also upon distribution (how far such a park should be from each household) and upon facilities (the extent to which smaller parks could actually serve the same function as larger ones based on present facilities). However, acreage plays a much stronger (and indeed a deterministic) role in defining park classifications, justifying the use of these preliminary classifications in estimating park availability throughout the study area.

At time of this publication, all parks within the study area have been classified using this methodology, and refinement of the classification scheme and reclassification of some parks based on observed use is occurring. Facility surveys and reviews of secondary inventory sources continues. Dedication dates for all parks in the study area are under investigation. Once complete, this set of data will allow us to subtract parks in five intervals to match demographic changes between censuses -- in effect creating multiple historic park coverages.

Analysis

Data collected in the formats described above can then be compared against each other to determine both interrelatedness at one point in time, and relatedness at multiple points in time. At time of this publication, we have begun to preliminarily assess these relationships between layers of our database that are complete. Below is a sampling of our preliminary findings.

Land Use Change (1928 - 1995)

The first analysis we conducted was a change comparison between land cover inferred from 1928 aerial photography and land cover inferred from 1995 aerial photography. Due to scale discrepancies between the source data for each year, and the differing complexity of land cover and land use between the two time periods, we selected to compare land uses using 120m grid cells at two-digit land cover classification under the Anderson system. Also, we chose for this analysis to isolate the impacts of the loss of native vegetation and agricultural lands over time, rather than to consider the entire extent of land use changes. Our goal in researching this issue is to determine what significant correlations exist between conversion of native vegetation and present demographic characteristics of areas impacted by such conversion. Spatial distribution of land cover changes are shown in Figures 2 and 3. Correlation coefficients comparing conversion of native vegetation to current demographic characteristics are presented in Table 4.

Figure 2: Land Cover Change 1928 - 1995 -- Loss of Native Vegetation

Figure 2 presents distribution of native vegetation lost between 1928 and 1995 within our project study area. Shaded areas indicate the location of changes in use from shrublands and riparian habitat to current use. These areas include significant populations of oak woodland, chaparral and coastal sage scrub, in addition to riparian corridors surrounding major rivers and alluvial fans extending from the foothills of the San Gabriel Mountains in Monrovia and Sierra Madre.

Colors shown in the legend indicate the current (1995) land cover for converted areas. It is significant to note the predominance of such conversion in areas of high elevation (Puente Hills, Montebello Hills, and Verdugo Mountains in particular), and surrounding watercourses (Los Angeles River, headwaters of San Gabriel River, and former course of Tujunga Wash in particular). Converted lands are primarily in current residential use, indicating the historical desirability of residential development in areas of high elevation. Correlations between such conversion and current demographic characteristics are put forth in Table 4 and subsequently discussed.

Figure 3: Land Cover Change 1928 - 1995 -- Loss of Agricultural Lands

Figure 3 shows the distribution of agricultural lands converted between 1928 and 1995 within our project study area. Shaded areas indicate the location of changes in land cover from agricultural production to current use. These areas incorporate such uses as cropland, pasture, orchards, groves, vineyards, confined livestock and grazing.

Again, colors shown in the legend indicate the current (1995) land cover for converted areas. Predominant areas for agricultural conversion over this time period include areas in Burbank, the San Fernando Valley, most of the corridor encompassing the Rio Hondo and San Gabriel River, the present day Cities of Industry and Commerce, and much of the residential sections of East Whittier, West Covina, La Puente and La Habra Heights. Converted lands are predominately in current residential use, but the amount of conversion to industrial use is also quite substantial, illustrating some of the physical impacts of these predominately single-use contract cities. Demographic characteristics associated with these changes and those illustrated by vegetation change are presented in Table 4.

The correlation coefficients listed in Table 4, while small, illustrate some very interesting connections between land conversion and current demographic composition of 1990 Census Tracts. (5) Both median household income and the percentage white population are negatively correlated with conversion of land from these "open" uses to commercial use. Income and education are both negatively correlated with conversion of open land to industrial use, but the percentage of population of Hispanic origin is positively correlated with such change. This finding supports conclusions reached by others researching the disproportionate exposure of persons of color to toxic releases. Income, education, owner-occupation rates of dwelling units, and percentage white population are all positively correlated to conversion of open land to predominately residential use. Percentage of Hispanic origin population is negatively correlated to such conversion. When combined with the results determined for percentage industrial conversion, this implies a disproportionate tendency for Hispanic populations to settle in areas characterized by predominately industrial conversions more than characterized by predominately residential conversions. This is a rather alarming characteristic, when considered in context with findings regarding the presence of parks in communities characterized by high Hispanic populations.

Demographic Variables (1990)

Based in large part upon the findings of our preliminary survey of park space per capita versus demographic statistics at the city level, we engaged in a further study of park distribution, size and characteristics at a census tract level throughout the study area. Our reasoning for conducting this analysis at the tract level, rather than block group or block, was to ensure our ability to compare the results with those for previous years for which such detailed information would be problematic. Our methodology was to compare the primary variables listed in Table 2 with park characteristics observed in each census tract. A spatial join was performed between a vector parks coverage for the study area and a 1990 census tract coverage containing the demographic variables of interest. An example of the spatial distribution of these combined variables is provided as Figure 4. Correlation coefficients comparing several of the 1990 primary demographic variables to characteristics of parks are provided in Table 5.

Figure 4: Median Household Income and Regional Parks

Figure 4 shows the distribution of median household income in 1990 within our project study area at the census tract level. Shaded tracts illustrate the wide range of income present throughout the area. Outlined areas show the presence of metropolitan (large regional) parks in correspondence to income.

Yellow and beige colored tracts indicate areas of lower income, while orange and red tracts indicate areas of high median household income. Predominant areas of low income include much of East Los Angeles, areas surrounding the midsection of the Los Angeles River, and isolated areas in the San Fernando Valley, Pasadena, El Monte, Rosemead and Whittier. Areas of high income correspond well with areas of high elevation, illustrated primarily in the Puente Hills, Santa Monica Mountains, Verdugo Mountains, and the foothills of the San Gabriel Mountains.

Tracts outlined by green lines indicate the presence of a large regional park where native vegetation has been maintained or re-introduced. These tracts can be found in areas surrounding Griffith, Ernest E. Debs, Elysian, Verdugo, Arroyo Seco, Whittier Narrows, Santa Fe Dam, and Schabarum Regional Parks, among others. Tracts outlined by yellow lines indicate the presence of a large regional parks composed predominately of introduced vegetation. These tracts can be found surrounding Brookside, Lincoln, Arcadia County and Bicknell Regional Parks, among others. Large regional parks are found predominately in tracts characterized by high median income. Furthermore, with the exception of Elysian Park, all parks characterized by predominately native vegetation are found in areas of higher income. Median income and other 1990 demographic characteristics of our study area tracts are correlated with park features in Table 5.

The correlation coefficients listed in Table 5 illustrate some compelling connections between demographics and park characteristics within a census tract. (6) Both the number (frequency) of parks and the number of types of parks (diversity) present in a census tract are positively correlated with the percentage white population. Average park type (a proxy for park size), park frequency, and park diversity all increase proportionately with the percentage of owner-occupied housing units in the study area. Park size, park diversity, and park frequency are all positively correlated with median household income. However, Park diversity, park frequency, and the standard deviation of park type (indicating relative size and diversity) are all negatively correlated with the percentage of population of Hispanic origin. This finding supports our initial hypothesis that park space and quality are not equally distributed across our study area with respect to the location of populations of color or low-income populations. When combined with the results from analysis of land cover change, this suggests that persons of low income and minorities have settled in areas characterized predominately by industrial conversion, and that these areas do not share the same park frequency and quality of other areas characterized by higher incomes and white populations.

Conclusions

Questions of environmental justice consider the relative equity of distribution of both positive and negative externalities of development across space and time. While many studies have considered the distribution characteristics of negative externalities, such as toxic releases, there is a relative lack of conclusive evidence regarding the relative distribution of positive externalities, such as parks and recreation opportunities and the preservation of open space and native vegetation. This study seeks to provide such evidence in the context of teaching future professionals to recognize the signs of environmental injustice, and to avoid them when advising development decisions or advocating social or economic policy. Through the use of GIS, it becomes possible to investigate these questions from many disciplinary approaches, and to integrate findings in meaningful ways that compliment undergraduate education.

Our preliminary findings suggest that less affluent people and persons of color tend to live in areas that have undergone disproportionate changes from natural or agricultural states to industrial use. This is particularly true for persons of Hispanic origin. The converse is also true -- persons of low income or of Hispanic origin tend not to live in areas that have undergone disproportionate changes from natural or agricultural states to residential use. These same people live in areas with a striking lack of available park space, and furthermore a lack of diversity in terms of what space is offered to them. These same characteristics are not demonstrated in communities characterized by high income or predominately white populations, indicating that these particular externalities are not distributed equally between communities on the basis of race or income.

Further investigation into issues of environmental justice from an historical land use perspective is a particularly worthwhile endeavor, especially if such studies address both positive and negative externalities of land conversion. If conducted in a stepwise fashion comparing conversions to corresponding demographic changes, it will contribute much toward the cause-and-effect question surrounding environmental justice: "Which came first, the nuisance or the population?" Better yet, an understanding that responsibility for these circumstances cannot be placed wholly upon development or populations will assist us to make this question itself irrelevant and allow policy makers to focus on the problem at hand -- mitigating environmental injustice.

Acknowledgements

The authors would like to acknowledge the following parties for their assistance in this ongoing project: Environmental Protection Agency, Willie Velasquez Research Instititue (formerly SVRI), Whittier College Environmental Justice Research Committee, Los Angeles County Urban Research Division, and Occidental College.

End Notes

(1) Data collection, automation and integration methods for Park Users and Government Structure and Function will be discussed in the Environmental Justice Project's final report, forthcoming in November, 1997. This paper focuses primarily on the role of land use change and demographics as dependent variables, and the relationships of park characteristics to them.

(2) Anderson, et. al., 1976. It is anticipated that the land use categories will be converted to conform to the American Planning Association's Land Based Classification System (LBCS) when that project is completed.

(3) Historic census data collection is being conducted in collaboration with Jim Sadd at Occidental College. Upon completion of the Whittier and Occidental projects, tract-level data will be available in digital format for all of Los Angeles County for these years. Current historical research has not surfaced boundary maps for Los Angeles County for the 1930 and 1940 Censuses. Should these data become available, we wish to incorporate them in our analysis. Please contact the author if you have access or leads to these resources.

(4) Overturf, et. al., 1996.

(5) (6) The low values of these correlation coefficients suggest that there are several other significant factors that may assist in describing current demographic characteristics or park characteristics within an area. It is not primarily the value of the coefficients we are concerned with, but rather the signs.

References

Anderson, et.al., "A Land Use and Land Cover Classification System for Use with Remote Sensor Data," USGS Professional Paper 964, USGPO, 1976.

Overturf, Steven, Dianna Barrantes and Jennifer Veridiano, "Socioeconomic and Demographic Analysis", Environmental Justice Project Working Paper.

Author Information

Dr. Cheryl C. Swift
Assistant Professor of Biology

Whittier College
13406 Philadelphia Street
Whittier, California 90608
Phone: (562) 907-4200 x4273
email: cswift@whittier.edu

Jeffrey A. Henderson, AICP
Geographic Information Systems Project Manager

Whittier College
13406 Philadelphia Street
Whittier, California 90608
Phone: (562) 907-4200 x4514
email: jeffh@keck.whittier.edu