Mapping for More Inclusive Evacuation During Climate Emergencies

Below, 2019-2020 thesis department award winner Megan Morrow explains her research and why it matters more than ever in the current moment. For more thesis posts, click here.


Introduction

While much of the world has seemingly ground to a halt in 2020, it is clear that the effects of climate change have been unencumbered by stay-at-home orders and curfews. 2020 has brought yet another devastating wildfire season to California and several strong hurricanes and storm surges to the southeast coast of the United States. Many of these extreme weather events have required large-scale evacuations. Even without the challenges presented by a global pandemic, evacuations and evacuation planning can be extremely complicated, requiring the development of official routes, a network of shelters, emergency personnel, equipment, and, of course, emergency transportation. For those with access to personal vehicles, the evacuation process may be rather simple. It can certainly be frightening or stressful, but access to a car allows many to safely transport themselves out of harm’s way. However, for those who do not own cars, or who are otherwise transportation disadvantaged, evacuation planning and procedures can be quite complicated.

Transportation disadvantaged (TD) populations are sometimes referred to as low-mobility, vulnerable, or carless populations. TD is a metric that combines socioeconomic, environmental, and behavioral factors, including populations who are carless (whether by choice or not), minority, low income, elderly, disabled, those with limited mobility or health problems, homeless, children without adults present, or those with limited English proficiency.1,2 At its core, TD is a mobility and climate justice issue, and often the result of larger forces of long-term state domination or discrimination. Several of these factors are correlated, and therefore can exacerbate the level of TD for populations exhibiting more than one factor.

Transportation disadvantage is a daily disadvantage experienced by many, but it poses particular challenges when it comes to planning for the evacuation of these populations.

Among the many challenges, the primary issue is simply identifying where these populations are located. The multitude of factors that contribute to TD, coupled with the difficulty of pinpointing populations at scales more granular than typical census geographies, make it very difficult to know exactly where these populations may be living.

A look at the recent history of disasters demonstrates that a renewed attention to TD populations occurred following Hurricane Katrina, which severely impacted Louisiana and Mississippi in August 2005. Several government reports released following this disaster outlined the need for better planning for TD populations and faster response and financial assistance from the federal government when disasters occur. In the state of Massachusetts, evacuation plans are focused on hurricanes, as they are the most likely disasters to affect New England. The Massachusetts Emergency Management Agency (MEMA) is responsible for evacuation planning and prepares a specific plan for populations with Critical Transportation Need. The statewide plan does recognize the difficulty in identifying these populations, but only offers a few brief suggestions for how a municipality should attempt to do so.3

Several mapping techniques have been used to estimate TD and other vulnerable populations, yet there is no standardized method. Some researchers have utilized a simple choropleth technique, in which census geographies, such as block groups, are shaded by level of transportation disadvantage.4 Others have compared census geography demographics with elevation, flood hazard, storm surge, and transit layers to assess flood vulnerability.5

While these types of thematic mapping methods are common, they can misrepresent where actual populations are living. While a whole census block may be shaded the same color on a choropleth map, the actual people those statistics represent are not evenly distributed across that block. These maps suffer from what is commonly known to GIS and statistical scholars as the Modifiable Areal Unit Problem (MAUP), which is particularly of issue when trying to estimate population distribution based on census geographies relative to other overlaying data at different scales.6 

Research Question & Methods

Recognizing the deficiencies of standard choropleth mapping techniques and seeing the potential for improved population estimations, I decided to compare the standard choropleth technique with two dasymetric mapping techniques. Dasymetric methods utilize additional layers, such as land use raster data sets or parcel data, to disaggregate census-level data to smaller, more accurate areal units. The use of the ancillary layer helps to appropriately account for open space, industrial sites, or other areas that are typically uninhabited. In essence, these methods adjust the “denominator” to better represent rates of specific attributes.7 It was my goal to improve these population estimates, better understand where TD populations are located in the Boston area, and to assess their relative risk when it comes to extreme weather events.

The three mapping methods were as follows:

  1. A standard choropleth method, which utilized a binary system to determine levels of TD and a thematic shading symbology;
  2. a dasymetric mapping method which used the Environmental Protection Agency’s Intelligent Dasymetric Mapping (IDM) Toolbox to redistribute population based on a land use raster dataset;8 and
  3. the Cadastral-Based Expert Dasymetric System (CEDS) which utilized parcel data to redistribute population along with a principal component analysis of vulnerability attributes.9

The study area included 97 cities and towns in the Boston Regional Metropolitan Planning Organization and seven vulnerability attributes were used to represent levels of TD:

  • Children (the percent of population 18 and below)
  • Elderly (the percent of population 65 and above)
  • Poverty (percent of population with income/poverty ratio below 1.49)
  • Vehicle Access (percent of households with no vehicle available)
  • Disability (percent of population with disability)
  • Language (percent of population that only speak English), and
  • Race (percent of population white)

To assess the relative risk of TD populations, I compared their population distribution with the official hurricane evacuation zones and “walksheds” representing walking access to emergency shelters. In addition to performing these mapping methodologies, I conducted semi-structured interviews with representatives from community organizations in particularly vulnerable neighborhoods and from state agencies involved in emergency planning. The purpose of the interviews was to ground-truth the mapping results, as well as to understand the evacuation resources for these populations at a local and state level.

Results & Discussion

As shown in the example of Chelsea, MA below, each consecutive mapping methodology improved in accuracy and detail as the methods utilized increasingly finer-grained layers and processes to represent population distribution and levels of TD. The underlying population estimates also improved in accuracy. In total, the CEDS results suggested 14% of the population in the study area (487,389 people) had a high propensity to be TD, compared to just over 2% predicted by the Choropleth and IDM methods. 

Figure 1. These maps display levels of transportation disadvantage in Chelsea, MA (Left: Choropleth Results; Middle: IDM Results; Right: CEDS Results).

The graph below shows the population estimates for those living in the highest risk locations: both in an evacuation zone and greater than a 20-minute walk from a shelter. The CEDS method estimated a much higher proportion of these populations as TD: 15.3% compared to a more modest 4.4%. And while this represented a smaller proportion of the total TD population in the study area (14.5% compared to 39.2%), this amounted to more than double the number of people estimated by the Choropleth and IDM methods (70,836 compared to 32,409). Compared with estimates from MEMA, which assumes only 6% of the population in evacuation zones A and B could require assistance, these numbers were significantly higher and could represent a severe underestimation in the amount of people that could require evacuation assistance.10 

Figure 2. Populations within evacuation zones and >20-minute walk from shelter.

There are several reasons why the CEDS method may have resulted in large increases in population estimates for both total and TD population. While the IDM system redistributed population density based on intensity of land use, it still assumed uniform density across similar land use designations. In contrast, the CEDS system estimated population on a per-parcel level, based on either the number of residential units or the residential area of that parcel. Therefore, the CEDS method was better suited to estimate population in multi-family buildings, such as apartment buildings, in which lower-income populations often live. Additionally, the CEDS system only analyzed parcels with residential classification codes, reducing the possibility that high-intensity commercial areas were included in the analysis.

Despite these differences in population distribution, each method overwhelmingly identified the same cities and neighborhoods within the study areas as having high levels of TD. These areas included Chelsea, Dorchester, East Boston, Framingham, Lynn, Mattapan, Revere, Randolph, Roxbury, and the South End of Boston. However, the CEDS method revealed the inconsistencies in residential population distribution within each of these high TD areas and the importance of accurately identifying where these vulnerable populations were located precisely within these neighborhoods.

Figure 3. Areas of consistently high TD among the three mapping methodologies

The interviews with representatives from community and state-level organizations revealed a real concern for extreme weather and climate resiliency. All organizations expressed concerns for the effects extreme weather could have on energy and transportation systems and how potential outages or lapses in service could affect the delivery of social or medical services. Increased community involvement, public education, training, government agility, and IT resiliency strategies were proposed as ways to increase the resiliency of communities during extreme weather events.

Conclusion

According to this analysis, hundreds of thousands of residents in the Boston metropolitan area may be transportation disadvantaged. This vulnerability affects these residents’ lives on a daily basis, as they rely on transit, paratransit, walking, or biking to travel to work, run their errands, or visit friends and family. For some, their level of TD may be so high that they are unable to leave their homes at all without a great deal of assistance. In the event of an emergency, this disadvantage is increased tremendously, as residents may be required to evacuate their homes and make their way to a shelter. Alternatively, in a different kind of emergency scenario, such as the pandemic we are experiencing today, those who are TD may not be able to access essential services, food retailers, or drive-through testing sites. Ironically, both stay-at-home and evacuation orders are likely difficult for TD populations to adhere to due to their reliance on transportation assistance. 

In Massachusetts, a municipality is technically responsible for providing transportation assistance to these residents of “critical transportation need” so they can safely reach a local shelter; however, this analysis revealed that many more people may require assistance than MEMA estimates and the current resources offered to municipalities to identify TD populations are limited.11 Therefore, the pursuit of mapping methods beyond standard thematic analyses, such as the CEDS method, could help planners more accurately predict areas with high TD and to appropriately tailor evacuation plans and target resources to ensure that help is provided to those who require assistance. These methods could be performed using publicly available census and parcel data and relatively simple spatial analysis tools. In addition to conducting mapping analyses, evacuation planning for TD populations should be combined with other identification methods, such as in-person outreach, registries, or surveys, that will allow identification on a household level.

My hope is that the results of this research can help emergency planning professionals proactively prepare for evacuation scenarios in order to avoid a disaster response like that of Hurricane Katrina. Massachusetts has been lucky to avoid the brunt of major storms in the past few decades, but the state is bound to be faced with a powerful storm at some point in the near future. Planners must proactively prepare for this event, so when the time comes, everyone—including the most vulnerable and transportation disadvantaged—are able to get out of harm’s way.

Sources

  1. United States Government Accountability Office. 2006. “Transportation-Disadvantaged Populations: Actions Needed to Clarify Responsibilities and Increase Preparedness for Evacuations.” Report to Congressional Committees, December 2006.
  2. Renne, John L., Thomas W. Sanchez, and Todd Litman. 2011. “Carless and Special Needs Evacuation Planning: A Literature Review.” Journal of Planning Literature 26, no. 4 (November 2011): 420–31. https://doi.org/10.1177/0885412211412315.
  3. Mass.gov. 2019b. “Massachusetts Emergency Management Agency.” Mass.gov, 2019. Accessed March 30, 2020. https://www.mass.gov/orgs/massachusetts-emergency-management-agency.
  4. Shay, Elizabeth, Tabitha S. Combs, Daniel Findley, Carl Kolosna, Michelle Madeley, and David Salvesen. 2016. “Identifying Transportation Disadvantage: Mixed-Methods Analysis Combining GIS Mapping with Qualitative Data.” Transport Policy 48 (May 2016): 129–38. https://doi.org/10.1016/j.tranpol.2016.03.002.
  5. Pulcinella, Joshua A., Arne M. E. Winguth, Diane Jones Allen, and Niveditha Dasa Gangadhar. 2019. “Analysis of Flood Vulnerability and Transit Availability with a Changing Climate in Harris County, Texas.” Transportation Research Record: Journal of the Transportation Research Board 2673, no. 6 (June 2019): 258–66. https://doi.org/10.1177/0361198119839346.
  6. Fotheringham, A S, and D W S Wong. 1991. “The Modifiable Areal Unit Problem in Multivariate Statistical Analysis.” Environment and Planning A: Economy and Space 23 (1991): 1025–44. https://doi.org/10.1068/a231025.
  7. Maantay, Juliana Astrud, Andrew R. Maroko, and Christopher Herrmann. 2007. “Mapping Population Distribution in the Urban Environments: The Cadastral-Based Expert Dasymetric System (CEDS).” Cartography and Geographic Information Science 34, no. 2 (2007): 77–102.
  8. United States Environmental Protection Agency (EPA). 2017. “Dasymetric Toolbox.” Data and Tools, July 5, 2017. Accessed November 15, 2019. https://www.epa.gov/enviroatlas/dasymetric-toolbox.
  9. Maantay, Juliana Astrud, Andrew R. Maroko, and Christopher Herrmann. 2007. “Mapping Population Distribution in the Urban Environments: The Cadastral-Based Expert Dasymetric System (CEDS).” Cartography and Geographic Information Science 34, no. 2 (2007): 77–102.
  10. Massachusetts Emergency Management Agency (MEMA). 2019. “Commonwealth of Massachusetts Critical Transportation Need (CTN) Evacuation Operations Annex,” January 2019. Accessed November 5, 2019. 
  11. Massachusetts Emergency Management Agency (MEMA). 2019. “Commonwealth of Massachusetts Statewide Evacuation Coordination Plan,” January 2019. Accessed November 4, 2019.  https://www.mass.gov/files/documents/2019/07/02/Statewide%20Evacuation%20Coordination%20Plan%20January%202019%20FINAL.pdf.

Cover image by paulbr75/Pixabay