Leveraging Mapping Tools to Characterize and Operationalize Environmental Justice

The modern environmental justice movement was born in 1982, when a protest in North Carolina sparked the confluence of the environmental and civil rights movement, said Charles Lee (EPA). The state of North Carolina designated a predominantly African-American community in Warren County as the site for a new hazardous waste landfill. Protests against the decision led to 500 arrests, and Lee described that though the state’s decision stood, this incident led to the first national study on demographics related to hazardous waste sites in 1987; the first National People of Color Environmental Leadership Summit in 1991, where the delegates codified the principles of environmental justice; and President Clinton’s 1994 executive order on environmental justice. Building on Woychik’s earlier discussion of the exposome, Lee noted that there is great alignment between that concept and the drivers of environmental health impacts in terms of how the built, social, and natural environments interact with one another to affect the health of communities in different ways.

Lee explained how the research community came to understand the disproportionate effects that low-income communities, people of color, and Indigenous communities experience from their greater exposure to pollution. The key development was being able to quantify those effects using geographic information system technology and then map them to define, articulate, and operationalize the concept of disproportionate impacts. Environmental justice mapping, said Lee, triggered environmental justice initiatives in California, Illinois, New York, and the federal government through the Justice40 Initiative.

There has been a proliferation of what Lee called second-generation environmental mapping tools developed by states and municipalities, with community involvement being critical for the development of these tools. Also important has been the emergence of approaches that, for example, allow CalEnviroScreen’s methodology to be used with EJSCREEN’s data elements to look at the cumulative impacts of environmental exposures and the effects of past social policies on current environmental justice issues. For example, overlaying redlining maps in Oakland, California, from the 1930s onto the results from CalEnviroScreen shows the correlation between historical redlining and the location of current urban heat islands. Furthermore, Lee noted that historical surveyor descriptions detail environmental conditions in some redlined areas, and how this illustrates that redlining—and subsequent policy decisions—were done with full knowledge of the negative environmental impacts that they would cause. Lee added that understanding the impact of these past policy decisions is important in order to achieve environmental justice in the future.

Going forward, Lee suggested that the answers to the following questions would help determine where the historical arc of environmental justice mapping will lead.

1.

How can we promote information sharing, education, and cross-fertilization regarding data resources, lessons learned, and best practices?

2.

Given the interdependent and innovative nature of developing and using environmental justice data and platforms, how can we develop strategies that maximize efficiency and effectiveness?

3.

Given the multiple analytic and data-intensive requirements of environmental justice applications, how can we achieve environmental justice mapping platforms that have a fit-for-purpose capacity?

4.

How can mapping be used to elucidate systemic racism and current environmental conditions?

Using Geospatial Techniques and Technologies to Improve Environmental Justice Mapping and Analysis

Jayajit Chakraborty (The University of Texas at El Paso) discussed some of the strengths and weaknesses of different environmental justice mapping techniques. He began by discussing classified choropleth mapping, a widely used method for visualizing the distribution of environmental risks for a given affected population. Though straightforward and easy to understand, choropleth mapping provides a subjective visual assessment given that the patterns depend in part on how researchers classify the data. Therefore, he said, it is not a precise method for identifying problem areas that need attention or mitigation.

Geospatial cluster detection using a local indicator of spatial association is a more precise and statistically reliable approach. Clusters of contiguous counties are classified as either hot or cold spots, depending on whether they have significantly higher or lower exposure to air pollutants relative to the entire country, for example. Hot spots, explained Chakraborty, are areas needing attention, and they can be used for a distributive environmental justice analysis by comparing their population characteristics with those of cold spots. In the case of exposure to air pollution, this approach reveals a highly unequal social pattern in which hot spots contain a significantly higher percentage of Black, Hispanic, below-poverty, and uninsured residents compared to cold spots. He has used this approach to identify the sociodemographic characteristics of counties with both a higher incidence of COVID-19 infections and respiratory risks from exposure to air pollution.⁴

Recently, explained Chakraborty, environmental justice studies have found that heat waves and heat islands disproportionately affect minorities, lower income groups, and other socially disadvantaged populations. This research led to a project in which community members used geospatial technologies in a coordinated data collection and mapping campaign to produce high-resolution heat maps for El Paso, Texas. The communities then developed collaborative solutions for mitigating the unequal effects of heat exposure. The project’s goal was to incorporate innovations in sensor technology, spatial analytics, and community climate action to explore relationships between urban microclimates, local infrastructure, and human well-being and use that information to inform policy interventions. For example, the data are being used to identify areas where tree canopy expansion would have the most direct benefit to environmental and social conditions. Chakraborty is combining these data with the Centers for Disease Control and Prevention’s (CDC’s) social vulnerability index to analyze environmental justice impacts and understand the needs of local communities facing the most acute impacts.

Identifying and Prioritizing Neighborhood-Level Environmental Justice Drivers

Black and Brown communities living near industrial facilities bear a disproportionate environmental, climate, and health burden, said P. Grace Tee Lewis (Environmental Defense Fund, EDF), with the devastation wrought by Hurricane Harvey in Houston serving as an example. In Hurricane Harvey’s aftermath, the Houston Endowment funded the Data to Action (D2A) Project, which aims to support the development of environmental justice community action plans that will address underlying environmental and health risks for vulnerable, high-priority populations in the Houston area. As a partner in this project, EDF, together with researchers at Texas A&M University and North Carolina State University, developed a data-driven tool—HGBEnviroScreen—to identify neighborhoods in Houston that have the greatest cumulative vulnerability and, as Tee Lewis described, that D2A should prioritize.

The greater Houston area, including Galveston and Brazoria, Tee Lewis explained, houses the largest North American petrochemical complex and has been in nonattainment status for federal ozone standards since 1997.⁵ The region also has the highest number of people without health insurance in the nation, said Tee Lewis. To capture the unique elements that, taken together, can lead to disproportionate burdens on certain populations in the region, Tee Lewis’s team combined a variety of data in five domains—health, social vulnerability, flooding, environmental sources, and environmental exposure and risk—at the Census tract level and used EPA’s Toxicological Priority Index to create a score for ranking and comparing neighborhoods in terms of factors driving vulnerability.

D2A used the resulting maps to fund four community organizations and provide technical expertise to support them in developing and implementing their action plans. These action plans include air monitoring, sampling for lead in the soil, and distributing food and water to the residents. Tee Lewis said that data visualization proved to be critical for communicating HGBEnviroScreen’s results to community members. She also explained that neighborhoods can have similar rank scores arising from different combinations of components in the five domains, producing their unique vulnerabilities. This research suggests that distinct approaches can be useful at the Census tract level to achieve comparable neighborhood-level changes, and in fact, highly vulnerable Census tracts can improve their rankings by 10 to 20 percent by addressing just one key area of vulnerability.⁶

Tee Lewis also suggested that the highest-ranking Census tracts are not distributed randomly across the Houston region. She described that the most vulnerable areas are located in east Houston, where industrial facilities are concentrated and minority and low-wealth populations are greatest. Community organizations and local governments are now using HGBEnviroScreen to support decision-making and municipal planning in a manner that establishes a mechanism for community members to integrate health equity, climate, and environment into their activities to address environmental justice concerns in their neighborhoods.

Discussion

What the three presentations in this session illustrated, said Kristi Pullen Fedinick (Natural Resources Defense Council), is the importance of understanding not just exposures to pollution but also how other factors in the lived environment, various social stressors, and an individual’s biological and genetic makeup interact with those exposures to potentially increase susceptibility to harmful effects. Pullen Fedinick described these interactions as being akin to a three-legged stool, where the legs represent the differential burdens that people face arising from psychosocial stress, the absence or presence of health-promoting behaviors, and the totality of exposures and susceptibility factors.

Pullen Fedinick wondered how these tools can help EPA perform risk evaluations that will protect potentially exposed and susceptible subpopulations and how municipalities can use them when making local zoning decisions, for example. She noted that conducting this type of fit-for-purpose analysis requires creating room for conversations that can inform decisions to address the disproportionate impacts of systemic racism on environmental justice. She advised against waiting for “Big Data” to quantify every exposure and determine its causal relationship with a particular health outcome before taking action to protect communities. As she put it, if a community has a tool that can measure every single thing that they are exposed to, that will not change the fact that they are experiencing those exposures. Instead, she challenged researchers and policymakers to use these tools to identify the communities with disproportionate exposures and adverse health effects and then develop solutions to fix the existing problems.

In a discussion with all of the session’s speakers, Lee said that having better data is important for producing more accurate results, but he also cautioned against becoming consumed by the data. “No amount of data is ever going to accurately depict the complexity of environmental justice communities,” said Lee. Tee Lewis added that there needs to be a balance between choosing the most valuable data sources and considering how they work in combination with the other factors to reflect lived experiences and determining what is needed for advocacy and policymaking.

Wilson noted that some communities participating in this type of research have expressed fears that public exposure of problems facing their communities could further devalue their property or make it impossible to sell their homes. He wondered if anyone has examined the effect of publishing adverse environmental information on the mental health and the economic well-being of communities and asked how those concerns can be addressed. Lee acknowledged that these are important questions to which there are as yet no answers, and suggested that addressing them will require extensive involvement with the affected communities. Doing so will be particularly important, added Wilson, given that self-determination is a core principle of environmental justice.

Personal Air Pollution Exposure and Geolocation Monitoring for Precision Environmental Health

Assessing personal exposure to air pollution with high spatiotemporal resolution is incredibly challenging, said Rima Habre (University of Southern California). Habre explained that this is in part because air pollution consists of a complex mixture of chemicals from many sources, but also because humans move and engage in a variety of behaviors that influence their exposures to air pollution. As a result, estimates of outdoor residential particulate matter with an aerodynamic diameter of less than 2.5 microns (PM_2.5) correlate poorly with personal exposure measurements. However, advances in geospatial technologies, analytics, and personal monitoring provide opportunities to generate personalized recommendations and interventions to mitigate the harms from air pollution.

At the population level, Habre described, it is possible to integrate remote sensing data, ground measurement data, and even data from low-cost, crowdsourced sensor monitoring networks to generate machine learning-based models with high spatiotemporal resolution of air pollution exposure. These models can capture variations and discriminate sources of exposure, such as wildfire smoke. While valuable for health analyses, models with high spatiotemporal resolution can also reveal new exposure patterns. For example, plotting predicted wildfire PM_2.5 concentrations in California on a map of so-called “environmental justice communities”—communities where environmental and socioeconomic stressors may cumulatively affect health outcomes—generated by CalEnviroScreen showed that the highest median weekly wildfire-related PM_2.5 concentrations from 2008-2017 occurred in Census tracts with the highest CalEnviroScreen cumulative impact scores.⁷

Habre is using global positioning system (GPS) and air pollution monitoring devices to generate continuous data on a minute-by-minute timescale to determine personal PM_2.5 exposure and mobility patterns. When combined with ecological momentary assessment surveys that sample subjects’ current behaviors and experiences in their natural environments, these data can provide a detailed picture of how mobility affects exposure. She has also mined the GPS data to derive information on time-activity patterns and combined those results with land use and building footprint data to determine whether activities occurred indoors or outdoors and in which context.

Working with researchers at the Los Angeles Pediatric Research using Integrated Sensor Monitoring Systems (LA PRISMS) Center, Habre has developed an informatics platform to communicate wirelessly and in real time with a variety of personal exposure monitors, health monitors, and other sensors. In a study of children with asthma, this platform detected peaks in air pollution generated by primary combustion sources, triggering a smartphone app-based survey to ask the individual to identify the pollution source that they were near just before the text alert arrived.⁸ Taking that a step further, current methods and technologies can be leveraged to design just-in-time adaptive interventions to change behaviors and reduce personal exposures in real time.

Embedding Mobile Health and Deep Learning into Geospatial Epidemiology

There is growing research, said Peter James (Harvard Medical School), that suggests living in greener areas may produce better health behaviors and outcomes, including higher levels of physical activity, better mental health, and lower mortality rates. The challenge, he added, is to study this idea in large cohorts. To do this, James takes a geocoded address of a study participant and overlays spatial datasets such as satellite-based vegetation indices. He found that participants in the Nurses’ Health Study who lived in the highest quintile of green space had a 12 percent lower mortality rate than participants living in the lowest quintile of green space.⁹

Since people spend more than half of their time away from home, measurement error is embedded in these results based on residential estimates of green space, said James. To correct for that error, James outfitted a subset of the newest cohort of the Nurses’ Health Study with a Fitbit device that measures physical activity, heart rate, and sleep every minute and a custom smartphone app that measures location every 10 minutes. The data provided a personalized measure of all exposures compared to residence-based exposures and revealed the types of environments in which people are more or less likely to be active. He plans to collect these data prospectively along with chronic disease outcomes and is looking to couple these metrics with other measures such as cognition in real time that may reveal whether certain environmental exposures affect cognitive function.

One limitation of all spatially oriented research is the lack of fine-grained spatial data on the built and natural environment. “Thankfully, advances in deep learning have enabled us to make sense of what is in any image at any location and at rapid scale,” said James. By analyzing satellite imagery and Google Street View imagery using a deep learning algorithm, he has been able to determine with high accuracy what a participant would be looking at as they walk out their front door and is testing whether the results are associated with activity, cognition, or health outcomes.

Leveraging Earth Observations to Estimate Climate Exposures Across Scales

Benjamin Zaitchik (Johns Hopkins University) discussed approaches to compensate for the lack of concurrently collected or temporally relevant geospatial environmental data that can be linked to health data. For example, in a study of childhood enteric infectious disease, researchers collected active surveillance, pathogen-specific data for children with and without diarrhea to understand which pathogens are causing these infections. However, none of the teams involved collected environmental or geospatial data. Fortunately, said Zaitchik, the world is rich in geospatial data at fairly high resolution that can be applied to obtain the probability of infection for children in different age groups at the community level, which, he added, has direct relevance to an individual’s risk profile.¹⁰ The probability of infection, he said, is also pertinent for questions of environmental justice associated with exposure to these infections across communities as a function of built environment, communities, and the physical environment. The lesson from this type of analysis, he said, is that there is power in aligning detailed health studies with geospatial databases. Ideally, though, researchers could collaborate at the outset of a study to optimize collection of geospatially relevant data.

As the COVID-19 pandemic unfolded, researchers wanted to determine if meteorological and air quality factors were linked to disease transmission or exacerbated symptoms. However, geospatial infrastructure and reliable, rapid analytical techniques were unavailable to identify precision environmental risk factors for COVID-19. Zaitchik said that public, near-real-time geospatial databases are available, but they are unwieldy, sometimes error prone, and not ready to be integrated consistently across the globe. He then spoke about using computationally derived artificial populations constructed from Census-generated demographic data, cohort study data, existing databases on activities associated with different demographic groups, and geospatial data to model populations and their mobility patterns in space and time. Such models can link environmental conditions to rates of exposure, but they are contextual and can be question specific.

Discussion

Reflecting on the three presentations, Erik Svendsen (CDC) pointed out that governments regulate at the community level, not at the level of an individual’s exposure to an environmental burden. He noted that it is apparent, however, that community-level measurements miss important details about exposures, resulting in the environmental justice issues highlighted in the workshop’s first session. As a result, Svendsen said, scientists are trying to empower the public to obtain information about their personal and community-level exposures.

To Svendsen, the challenge is to take new scientific and technological developments and push the frontiers of knowledge in a way that will protect public health and empower best practices and sound decisions. In some ways, he said, the nation may be undergoing a paradigm shift in the way it regulates the public’s exposures, switching from a top-down approach to a bottom-up approach using technology to gather data on a local and even personal scale, and engaging the community to help answer the questions that are important to it.

In a discussion with all of the session participants, when asked to identify some of the opportunities for novel geospatial technologies in the environmental health field, Habre said this is the chance to learn what makes people different in their responses to environmental exposures by improving the inputs to the models that researchers have developed so as to generalize to larger populations. James offered that plenty of data are available; the challenge is to integrate them in a way that is useful to individuals. He noted the challenge of using data to create more green spaces in underserved communities without triggering gentrification, and the challenge of using geospatial technologies to identify extreme temperatures and tackle the planet’s climate emergency.

Zaitchik said that the new appreciation from the medical and public health research communities of the importance of environmental data creates opportunities regarding data collection, and he questioned why these data are not collected in electronic health records (EHRs). Privacy concerns may be an obstacle, he said, but the environmental justice community could develop and implement solutions to protect privacy. He noted that CDC is undertaking a data modernization initiative aimed at obtaining better data to make better and faster decisions. Going forward, Zaitchik felt it important to collect data to protect the public, which can also be used to address questions at the individual exposure level, including whether the current exposure limits and measurements are doing the job for which they are intended.

Both James and Svendsen said that it will be important to respect data privacy, particularly as an increasing number of smartphone apps gather location data without the permission of users. Svendsen suggested that partnerships at state, regional, and federal levels may be useful for looking at issues of privacy and data access in a more holistic and unified manner. All panelists also noted that citizen science is driving the need for a more spatially granular approach to data collection that can inform local scenarios and give communities the ability to carry out interventions at a local level.

Quantifying the Health Effects of Saharan Dust

Highly technical remote sensing information needs to be translated into a more user-friendly format that public health practitioners can use to improve decision-making, said Pablo Méndez-Lázaro (University of Puerto Rico, Medical Sciences Campus). In his case, he and his colleagues with expertise in epidemiology, environmental science, environmental health, remote sensing, chemistry, and other related disciplines are using data from Earth observation satellites and ground stations to quantify the effect of dust from the Sahara Desert on public health in Puerto Rico. An objective is to develop a public health early warning system. Some 19 organizations, he noted, are working to build this networked public health system.

Saharan dust, an environmental health hazard that includes metal particles, organic matter, and bacteria, arrives in the Caribbean mostly between May and September, with June through August being peak dust season in Puerto Rico, explained Méndez-Lázaro. Aerosol optical depth data from satellite imagery provide measurements of PM_2.5 levels, which correlate with respiratory problems, including emergency room visits and hospitalizations among children, as seen throughout the Caribbean. Using an online survey developed with help from physicians and health clinic staff, Méndez-Lázaro found that almost 90 percent of the more than 1,500 respondents, along with family members, experienced health consequences, particularly asthma flare-ups, coincident with the arrival of a large wave of Saharan dust in 2020.

Quantifying Health Effects of Wildland Fire Smoke: The Geospatial Toolbox

As a result of emergency response efforts, deaths from acute exposure to wildfires remain low, said Sheryl Magzamen (Colorado State University), but that does not mean that communities do not experience a high degree of morbidity and mortality from wildfire smoke exposure. She noted that the past four decades have seen wildfire duration, intensity, area burned, and length of wildfire season increase significantly across the western United States, with further increases expected due to climate change.

Through a series of case studies, Magzamen illustrated how remote sensing tools can quantify smoke in different settings and differentiate it from anthropogenic sources of PM_2.5 and other particulate matter. The first study aimed to quantify the smoke produced during Washington State’s 2012 wildfire season and its effects on human health.¹¹ This project used three inputs—daily PM_2.5 concentrations simulated using a chemical transport model, data from a ground-based monitoring network, and satellite remote-sensed aerosol optical depth data—with each approach producing different results. Linking the results of the different inputs to health data produced different conclusions and inferences related to specific adverse health effects.

The second study aimed to quantify smoke exposure over a diverse topographical region, which proved challenging when using chemical transport models. Instead, Magzamen estimated daily wildfire smoke PM_2.5 by subtracting seasonal median PM_2.5 concentrations on days with no smoke from daily concentrations measured by a ground-based monitoring network on days with wildfire smoke. One result from this study was that local smoke exposure was negatively associated with hospitalizations, which could be the result of successful efforts to evacuate residents from the affected areas.¹² Another finding was that wildfire smoke can interfere with urban measurements of ground-level black carbon, a component of PM_2.5 formed by the incomplete combustion of fossil fuels and an indicator of traffic patterns.

The third study, she explained, implemented lessons from the second study to identify 104 wildfire days for a study area in Alaska, which has limited ground-based monitoring, and matched those results with data on residents seeking health care.¹³ Community-based stratification of the data indicated that residents in more rural areas may delay seeking health care, particularly for cardiovascular disease.

Magzamen listed the five key lessons she wanted the participants to take from her presentation:

1.

Although smoke is a mixture of particles and gases, PM_2.5, a major combustion product and health hazard, is an indicator of wildfire smoke.

2.

Differentiating between background PM_2.5 and wildfire smoke PM_2.5 is a major challenge that is critical for disease etiology, research, regulatory intervention, and policy setting.

3.

Wildfire smoke exposure is hyperlocal, demonstrating both temporal and spatial shifts, while patterns of anthropogenic PM_2.5 sources from fixed and mobile sources show only temporal shifts. Thus, wildfire smoke exposure is harder to capture with extant on-ground monitoring systems.

4.

Current ground-based monitoring systems were built to detect anthropogenic pollution or to measure visibility and regional haze.

5.

Remote sensing provides a variety of tools to estimate wildfire smoke exposure, though the variability in approaches for estimating that exposure may limit the comparability of epidemiologic findings.

Looking to the future, Magzamen said that the field needs to harmonize exposure assessment approaches to understand potential differences in concentration response functions resulting from exposure to smoke and leverage new, low-cost technology to augment current exposure assessment methods. The combination of current monitoring, remote sensing, and low-cost sensors holds great promise, especially for rapid response, said Magzamen. She would also like to increase the use of remote sensing data to inform smoke warning systems. For example, her team has leveraged the U.S. Forest Service’s BlueSky Forecasting System to predict the burden of emergency department visits for respiratory problems associated with wildfire smoke for counties across the United States.

Planning for Environmental Justice, Hazard Vulnerability, and Critical Infrastructure in Communities of Color

Urban flooding is an environmental hazard that is becoming more frequent, with Hurricane Harvey’s inundation of the Houston area in 2017 being a prime example, said Marccus Hendricks (University of Maryland). Similar events, though not at that scale, are occurring across the country. Many of these events are overwhelming storm water and wastewater systems and triggering sewer system backups into homes, explained Hendricks, with direct implications for environmental health in affected communities.

Common features of urban flooding events include heavy rainfall with impacts beyond coastal and riverine flood zones, and increased runoff from that rainfall due to the increase in impervious cover, such as roads and building lots in high-density areas. Storm water infrastructure that is unable to cope with the amount of storm water runoff during these events further increases flooding, noted Hendricks. While the nation’s aging and decaying infrastructure has been much discussed, little to no attention has been paid to where and on whom the burdens of this decaying infrastructure fall heaviest, he added.

In fact, said Hendricks, some communities have not experienced the fair application of the country’s environmental laws and policies. He noted that social stratification based on race, income, disability, gender, age, and nationality are among the factors contributing to differential risks and adverse effects from disasters.¹⁴ He pointed out that Wilson raised this issue in 2008 when he called out municipalities for their failure to install up-to-code sewer and water infrastructure and the likely negative effect on the health and quality of life among poor people of color. Today, a huge data gap may impede elucidating these disparities in infrastructure, particularly storm water systems, and in maintaining and rehabilitating that infrastructure. Even when the data exist, said Hendricks, municipalities are reluctant to make it available to researchers or the public.

Hendricks has developed a framework demonstrating a relationship between social vulnerability and environmental justice; neighborhood factors and inequalities; physical vulnerability and critical infrastructure; and hazard risks, exposure, and recovery. This framework shows how potential differences in the quality, inventory, and condition of these infrastructure systems can have consequences for hazard risk exposure and disaster recovery.

In his view, there is an opportunity for environmental justice that works toward and fosters urban resilience. He feels that realizing this opportunity will occur only if researchers and advocates for environmental justice can access more and better data related to infrastructure, and only if communities are engaged to transform these systems using hybrid gray and green infrastructure approaches that will best manage storm water holistically.

As far as actually achieving resilience while also engaging communities in a meaningful fashion, Hendricks has developed a participatory assessment technique for infrastructure that adapted a visual inspection process employed by trained engineers for use by community members.¹⁵ He views involvement of community-based citizen scientists in gathering infrastructure monitoring data at the local level as a means of closing gaps in the data and getting communities to advocate on their own behalf for infrastructure investments. He addressed questions about the validity and reliability of data collected by citizen scientists by showing that the quality of the data that they produced was similar to the quality of data produced by trained engineers.

Hendricks concluded his remarks by noting that there is an opportunity to not only invest funds in a large infrastructure bill, but to do it by involving and engaging communities. Equity in infrastructure includes procedural, distributive, and restorative justice, said Hendricks. The built environment does not happen in a vacuum, and he noted that it must be recognized as a continuation of social circumstances. Involving citizen scientists in tracking the health of community infrastructure is not merely a gateway to more data, he added, but also to a more healthy, just, and resilient society.

Discussion

In a discussion with all of the session participants, Elliott returned to the subject of data access and noted that communities are looking for ways of accessing data for their own use. Magzamen pointed out that the National Aeronautics and Space Administration (NASA) has a significant initiative, one that this community could learn from, that puts its technical data into the hands of community members as a means of increasing community engagement and addressing inequities. Hendricks added that agencies need to make more funds available for engaging communities in research and data gathering. At the same time, it is important not just to engage a community to acquire data but also to report back to those communities and help them use those data.

Data Availability

Participants in this group discussed opportunities for involving community groups, citizen scientists, schools, and local businesses in data collection and use; making sure new satellite data programs account for who will use the data and for what purpose; and continuing the dialogue started by this workshop as a means of encouraging transdisciplinary research. Barriers discussed by this group included a lack of standardization and a centralized hub for accessing and using data, the difficulty of accessing and using ever-larger datasets, and the reluctance of the health community to use geospatial data.

Some members of the group suggested a range of best practices, including ideas to create data-sharing standards, link geospatial data to census data and large health datasets, commit to culturally aware community engagement and outreach, and establish training workshops. The group also discussed roles that federal agencies can play in creating platforms and standards for data access and promoting extended user availability of datasets.

Data Integration

The opportunities that some members of this group suggested include integrating microdata and other personalized data sources to provide more accurate and fine-grained exposure estimates that account for lived experiences. These data could potentially better inform policies to protect those who are most exposed, though a mechanism that protects individual privacy would be an important consideration. For barriers, the group noted the lack of sophisticated spatial statistical methods to use with hyperlocal datasets at large scales and with mismatched datasets. Data-sharing policies can also impede research and create challenges to communities wanting to integrate their data with other data sources.

Best practices may include ensuring that the data infrastructure maintains privacy but allows for fine-grained analysis. Federal agencies have produced guidance in this area. Many members of this group suggested that federal science agencies may foster transdisciplinary collaboration and could consider better data integration across agencies. Some participants also suggested that it can be useful for regulatory agencies to rethink historical approaches to policymaking to incorporate the richness of new integrated geospatial data sources.

Training

A potential barrier for training scientists and health policy experts is the siloed nature of universities and the separation between environmental health sciences and the geosciences. University culture can also act as an impediment for early career researchers that may be more solution-oriented. Some breakout participants noted that students are eager to work on solutions, which may create an opportunity for universities to change their focus and play a bigger role in fostering solution-oriented research. Universities, and particularly minority-serving institutions, could work with professional organizations to create environmental health training programs that include the use of geospatial technologies and geoscience programs that include environmental health. Discussion among members of this breakout pointed to the benefits of interdisciplinary training early in students’ careers, and the Department of Education could potentially take a more active role in encouraging this type of training. A few participants in the group also suggested that federal agencies could increase their focus on international collaborations and their work with the private sector, which is investing in this field.

Privacy and Ethics

Focusing on barriers, some members of this group highlighted the need to balance the risks and benefits of the level of detail that researchers desire against the risk of being too intrusive in people’s lives. Discussions also focused on the different conceptions of privacy that different stakeholders might have. Some breakout members noted that consent is a gold standard for research involving individuals, and they suggested that perhaps it should be the gold standard for community-oriented research, too.

Many breakout participants noted that it can be helpful for researchers to be aware that data sources could be combined to reveal individual or community identities, or violate privacy agreements. Rural versus urban data collection might raise or obscure different details and treat certain zip codes as homogenous when in fact they are not. In addition, stigmatization based on location data can create unanticipated issues. Data rights and licensing may benefit from additional consideration, particularly when communities are active participants in research.

When planning a study, some participants noted that it can be important to think deliberately about the type of data that one will collect, how individualized they can be, and what the privacy risks are and how to minimize them. Obfuscation, commonly used in medical research, is one potential privacy solution. Making data available in the cloud for computation but not for downloading could limit the unauthorized spread of data. Participatory research that involves citizen scientists and communities from conception to publication may allow researchers to consider privacy issues alongside those who are potentially the most affected by the data collected in a study. Some members of this group felt that this is particularly important when working with communities that have had negative experiences in the past. Finally, the group cautioned against setting more barriers and regulations in public repositories than there are in place for private repositories.

Several breakout participants felt that federal agencies could play a role in furthering the discussions on privacy and ethical issues, potentially through research on possible approaches to those issues. Agencies might also consider a data-sharing platform, similar to ClinicalTrials.gov, to avoid the time and cost of having each project address these issues.

Discussion

The final discussion period addressed issues related to funding transdisciplinary research and training and the ethical and privacy considerations that can accompany the application of geospatial technologies to inform environmental health decisions. Suggestions from some participants included adding community representatives to grant review boards, as is done with Superfund research grants and several NASA-funded programs, and to look at what the Silent Spring Institute has done regarding privacy and ethical issues as they pertain to community-based data collection.

Another topic of discussion was the challenge of establishing partnerships across disciplines, and some participants’ comments stressed the importance of creating networking platforms that go beyond one-off workshops and getting professional organizations involved in that effort. The American Geophysical Union’s Thriving Earth Exchange and Research Data Alliance are examples of such platforms, and the NASA Health and Air Quality Applied Sciences Team, the National Science Foundation’s sustainability research networks, and the One Health Initiative are other example approaches to fostering partnerships across disciplines. Messaging and communication were noted by many participants as being important aspects of transdisciplinary collaborations.

The challenge of getting physicians to use geospatial data in patient care was discussed, and it was noted that geospatial data are starting to get integrated into EHRs. A few participants suggested that data harmonization across medical health records is a problem, but physicians are interested in using those data if they are accessible in the EHR. The precision environmental health piece that connects specific exposures to individual health issues may be useful, and medical schools and residencies could consider including environmental health in their curricula.

Closing Remarks

In closing the workshop, Anenberg provided a summary of the key themes that emerged throughout the 2 days. She pointed out that the workshop emphasized how place matters for health, as highlighted by the many scenarios, analyses, and contexts that were discussed by workshop participants. Place includes the built environment and infrastructure, which Anenberg said must be considered and linked with historical reasons why differential health outcomes are experienced in different places and by different populations.

Next, Anenberg discussed how people are living at a confluence of several events. New government initiatives are putting equity and environmental justice front and center, and action is happening from the local, state, and federal levels, and even at the global scale. Many sources of data are now available in the “geospatial toolbox.” However, Anenberg noted, there are still gaps in linking the collection of geospatial data, understanding what the data mean, and operationalizing the data to meet end user needs.

Anenberg closed by reiterating the viewpoints heard in several of the workshop’s discussions. She noted that some workshop participants put forth a challenge to geospatial scientists and environmental health researchers to partner with end data users and communities directly, to engage with governmental environmental justice initiatives, and to continue fostering cross-disciplinary capacity building. Anenberg encouraged participants to think about the community and individual next steps that can be taken to meet that challenge.