                                       
                                       
                         Risk and Technology Review - 
                                       
                       Analysis of Demographic Factors 
                         For Populations Living Near 
          Solvent Extraction for Vegetable Oil Production Facilities
                                       
                                 Prepared by:

                             SC&A Incorporated
                                       
                         501 Eastowne Drive, Suite 250
                            Chapel Hill, NC  27514
                                       
                                       
                                       
                         EPA Contract No. EP-W-12-011
                           Work Assignment No. 4-09
                                       
                                       
                                       
                                 Prepared for:
                                       
                 Terri Hollingsworth, Work Assignment Manager
                 Ted Palma, Alternate Work Assignment Manager
                          Air Toxics Assessment Group
                   Health and Environmental Impacts Division
                 Office of Air Quality Planning and Standards
                     U.S. Environmental Protection Agency
                 Research Triangle Park, North Carolina 27711
                                       
                                       
                                       

                                       
                                       
                                August 18, 2017
	Disclaimer

            Although the analysis described in this document has been funded wholly or in part by the United States Environmental Protection Agency under contract EP-W-12-011 to SC&A Incorporated (previously EC/R), it has not been subject to the Agency's review and therefore does not necessarily reflect the views of the Agency, and no official endorsement should be inferred.
                                   Contents



1.	Introduction	1

2.	Census Data	2

3.	Calculation Methods	3
3.1	Total Population Risks	4
3.2	Race/Ethnicity and Age Categories	5
3.3	Level of Education	5
3.4	Poverty Level	6
3.5	Linguistic Isolation	6

4.	Results	7
4.1	Proximity Demographics	7
4.2	Cancer Risk Demographics	8
4.3	Noncancer Risk Demographics	8

5.	Uncertainty Discussion	10

Appendix A	A-1
Appendix B	B-1
Appendix C	C-1
Appendix D	D-1



 Introduction

	This document provides summary results and describes the approach used to evaluate the potential cancer and noncancer risks associated with inhalation of hazardous air pollutants (HAP) by different socio-economic demographic groups within the population living near facilities that us Solvent Extraction for Vegetable Oil Production in the United States (U.S.). This work was done in support of the U.S. Environmental Protection Agency's (EPA) Residual Risk and Technology Review (RTR) for the Solvent Extraction for Vegetable Oil Production (VOP) source category emissions, subject to the rules and guidelines contained in 40 CFR Part 63 Subpart GGGG.

	In the RTR analysis, the Human Exposure Model, Version 3 (HEM-3)[,][,] was used to estimate cancer and noncancer risks due to the inhalation of HAP for populations residing within 50 kilometers (km) of VOP operations (the HEM-3 model domain). HEM-3 estimates cancer and noncancer risks at the level of census blocks using the AERMOD state-of-the-art air dispersion model developed under the direction of the American Meteorological Society (AMS) / EPA Regulatory Model Improvement Committee (AERMIC). Each census block typically includes approximately 50 people. Additional information is available in the dockets for the rules and guidelines contained in 40 CFR Part 63 Subpart GGGG. 

	The current analysis used EPA's Environmental Justice (EJ) Risk and Proximity Analysis Tool ("EJ Tool") to link HEM-3/AERMOD modeling results for the VOP source category with detailed census data, in order to evaluate the distribution of cancer and noncancer risks for different demographic factors (including racial, ethnic, age, economic, educational, and linguistically isolated population categories). In addition to evaluating risk distribution, this analysis also presents the demographic composition of the population located within close proximity (5 km) and within the overall HEM-3/AERMOD model domain (50 km) of the source category emissions (irrespective of risk). The following demographic groups were included in this risk and proximity analysis:
 Total population;
 White;
 Minority;
 African American (or Black);
 Native Americans;
 Other races and multiracial;
 Hispanic or Latino;
 Children 17 years of age and under;
 Adults 18 to 64 years of age;
 Adults 65 years of age and over;
 Adults without a high school diploma;
 People living below the poverty level;
 People living below two times the poverty level; and
 Linguistically isolated people.
      
	The HEM-3/AERMOD results, for a particular census block, reflect the estimated level of cancer or noncancer risk that would be experienced by an individual residing within the block boundaries for 70 years. In this analysis, the demographic composition of the population estimated to experience a cancer risk greater than or equal to 1 in 1 million and a noncancer hazard index greater than 1 from inhalation of emissions from VOP operations, is compared to the demographic composition of the overall nationwide population.

	The total population statistics near facilities in the source category, irrespective of risk (i.e., at all risk levels) are also provided. In this analysis, the demographic composition of the population in close proximity to VOP facility operations (within 5 km) and within the model domain (50 km) is compared to the demographic composition of the nationwide population.

	The census data used in this analysis is described in Section 2. The algorithms used to compute the risk distributions for the total population and for each demographic category analyzed are presented in Section 3. The summary results of this analysis are presented in Section 4. The Appendices contain the detailed demographic results on which the summary results are based. 
 Census Data

	For cancer and noncancer risk modeling, HEM-3/AERMOD uses the decennial census data from tables in the 2010 Census Summary File 1 (SF1). The decennial census population data are provided at the census block level. 

	For the demographic analysis, statistics on total population, race, ethnicity, age, education level, low household income, poverty status and linguistic isolation are obtained from working files compiled in EPA's Environmental Justice Screening and Mapping Tool, EJSCREEN. The data compiled in EJSCREEN are originally derived from the Census' American Community Survey (ACS) 5-year averages for 2010-2014. These data are provided at the block group level. A census block contains about 50 people on average; and a block group contains about 28 blocks on average, or about 1,400 people.

	Table 1 summarizes the census data used in the analysis, showing the source of each dataset and the level of geographic resolution. 

     Table 1.  Summary of Census Data used to Analyze Risks for Different 
                              Demographic Groups
Type of population category
Source of data
Level of geographic resolution
Total population (HEM-3 risk analysis)
2010 Census SF1
Census block
Total population (Demographic analysis)
ACS Table B03002 
Block group
Race/ethnicity categories
ACS Table B03002
Block group
Age groups
ACS Table B01001
Block group
Level of education - adults 25 years and older without a high school diploma
EJSCREEN_Full_V3_USPR 
Block group
Individuals living in households earning 	below the poverty level
ACS Table B17002
Block group
Individuals living in households earning 	below two times the poverty level
ACS Table C17002
Block group
Linguistically isolated people
EJSCREEN_Full_V3_USPR 
Block group
 Calculation Methods

	HEM-3/AERMOD models the cancer risk and noncancer hazard index ("risk") at a point near the geographic center of each census block and optionally at user-defined receptors representing populations located within each census block. For the current analysis, these risk estimates are assumed to apply to all individuals residing in the census block. The EJ Tool uses the modeled census block and census block group identification codes to link the HEM-3/ AERMOD modeling results for each census block to the appropriate ACS and EJSCREEN census block group demographic statistics. This allows us to estimate the number of people in different demographic categories for each census block modeled by HEM-3/AERMOD.

	For the demographic analysis, the total nationwide population is determined by summing the total population of all census block groups in the Census' ACS 5-year average for 2010-2014 ("2014 ACS"). The demographic characteristics of the population potentially impacted by VOP operations is determined by applying the characteristics of a census block group to the HEM-3 modeled census block populations located within that block group. For the purposes of this analysis, the demographic characteristics of a given block group  -  that is, the percentage of people in different races/ethnicities, the percentage in different age groups, the percentage without a high school diploma, the percentage at certain low income levels and the percentage that are linguistically isolated  -  are presumed to also describe each block located within that block group. 

	Section 3.1 describes the calculation method used to estimate the total population exposed to different risk levels. Sections 3.2 through 3.5 describe calculation methods used to compute risks for racial, ethnic, age, education status, low household income, poverty status, and linguistic isolation demographic categories. The detailed results from the application of these calculation methods are provided in the tables included in the Appendices. 
    Total Population Risks

	The calculation of risk distributions for the total population in this analysis involved a block-by-block accumulation of people at various risk levels. We began by identifying a set of bins reflecting the level of risk. The population of each block was then assigned to the appropriate risk bin based on the HEM-3/AERMOD modeled risk level in the block. The numbers of people in each risk bin were then added together for all of the blocks modeled for the VOP source category. 

	H(Rab) = ∑i[(Ra<=Ri<Rb)] [N(i)]	(1)
where: 

	H(Rab) =	the population count for risk bin Rab, which is between Ra and Rb 
	Ri =	the modeled risk level in block "í" (estimated lifetime cases of cancer per million population)
	∑i[(Ra<=Ri<Rb)]	refers to the summation over all blocks i where Ri falls in bin Rab, between Ra and Rb 
	N(i) =	the number of people in block i

The average risk for the total population was then calculated using the following equation:

		A(s) = ∑i [N(i) x Ri] ∕ ∑i [N(i)]	(2)
where:
	A(s) =	the average risk for the population in the modeling domain (estimated lifetime cases of cancer per million population)
	∑i	refers to the summation over all blocks "í" modeled for the emission source category
	N(i) and Ri 	were defined above 

These risk bins, populations within each risk bin and average risk results are shown in the tables presented in the Appendices.
    Race/Ethnicity and Age Categories

	Table B03002 of the 2010-2014 ACS data provides race/ethnicity statistics for each census block group nationwide. Table B01001 provides age statistics for the population by ranges (in years) for each census block group nationwide. For each modeled census block in this analysis, the race/ethnicity (White, African American, Native American, Multiracial/Other and Hispanic or Latino) and age range (0-17, 18-64 and >=65 years) for that block is estimated based on the demographic information provided at the block group level, as follows:

		N(s,b/bg) =  N(t,b/bg) x N(s,bg) ∕ N(t,bg)	
where:

	N(s,b/bg) =	number of people in racial/ethnic or age subgroup "s", in block "b" of block group "bg"
	N(t,b/bg) =	total number of people in block "b" of block group "bg"
	N(s,bg) =	number of people in racial/ethnic or age subgroup "s", in a block group "bg"
	N(t,bg) =	total number of people in block group "bg"

Equation (1) was then used to generate risk distributions based on the block-level results, and Equation (2) was used to compute the average risk for people in each racial/ethnic or age subgroup. 

The detailed risk results based on race/ethnicity and age demographic categories are presented in the Appendices in the Racial and Ethnic Groups and in the Different Age Groups tables.
    Level of Education

	Table EJSCREEN_Full_V3_USPR, which was derived from the 2010-2014 ACS data, provides education attainment statistics for each census block group nationwide. For each modeled census block in this analysis, the number of people 25-years and older without a high school diploma is estimated based on the demographic information provided at the block group level, as follows:

		N(nhs,b/bg) =  N(t,b/bg) x N(nhs,bg) ∕ N(t,bg)	
where:

	N(nhs,b/bg) =	number of people 25-years and older without a high school diploma "nhs", in block "b" of block group "bg"
	N(t,b/bg) =	total number of people in block "b" of block group "bg"
	N(nhs,bg) =	number of people 25-years and older without a high school diploma "nhs", in a block group "bg"
	N(t,bg) =	total number of people in block group "bg"
Equation (1) was then used to generate risk distributions based on the block-level results, and Equation (2) was used to compute the average risk for adults without a high school diploma. 

The detailed risk results based on the attained level of education demographic category are presented in the Appendices in the Adults without a High School Diploma tables.
       Poverty Level

	Tables B17002 and C17002 of the 2010-2014 ACS data provide the following two income statistics related to the poverty level for each census block group nationwide: individuals living in households earning incomes below the poverty level and individuals living in households earning incomes below two times the poverty level, respectively. For each modeled census block in this analysis, the block's household income level is estimated based on the demographic information provided at the block group level, as follows:

	N(hi,b/bg) =  N(t,b/bg) x N(hi,bg) ∕ N(t,bg)

where "hi" indicates household income, whether below the poverty level or below two times the poverty level, depending on the statistic, and:

	N(hi,b/bg) = 	number of people living in low income households "hi" relative to the poverty level, in block "b" of block group "bg"
	N(t,b/bg) =	total number of people in block "b" of block group "bg"
	N(hi,bg) =	number of people living in low income households "hi" relative to the poverty level, in a block group "bg"
	N(t,bg) =	total number of people in block group "bg"

Equation (1) was then used to generate risk distributions based on the block-level results, and Equation (2) was used to compute the average risk for people living in low income households. 

The detailed risk results based on poverty level are presented in the Appendices in the People Living in Households Below Two Times the Poverty Level and Below the Poverty Level tables.
    Linguistic Isolation 

      Linguistic Isolation is defined by EJSCREEN as "a household in which all members age 14 years and over speak a non-English language and also speak English less than "very well" (have difficulty with English)." Table EJSCREEN_Full_V3_USPR, which was derived from the 2010-2014 ACS data, provides the number of households in linguistic isolation in each block group. For each modeled census block in this analysis, the number of people living in linguistic isolation is estimated based on the demographic information provided at the block group level, as follows: 

		N(li,b/bg) =  N(t,b/bg) x N(li,bg) ∕ N(t,bg)	

where:

	N(li,b/bg) =	number of people living in linguistic isolation "li", in block "b" of block group "bg"
	N(t,b/bg) =	total number of people in block "b" of block group "bg"
	N(li,bg) =	number of people living in linguistic isolation "li", in block group "bg"
	N(t,bg) =	total number of people in block group "bg"

Equation (1) was then used to generate risk distributions based on the block-level results, and Equation (2) was used to compute the average risk for people living in linguistically isolated households. 

The detailed risk results based on the linguistically isolated demographic category are presented in the Appendices in the People Living in Linguistic Isolation tables.

      Results
	The demographics of the population within close proximity to the modeled facilities is presented in Section 4.1. The demographics of the population surrounding the modeled facilities at various levels of cancer and noncancer risk from source category emissions are presented in Sections 4.2 and 4.3.
 Proximity Demographics

	The proximity results describe the demographics of the population surrounding the modeled facilities irrespective of risk. Table 2 presents the demographic composition of the population located within close proximity (5 km) and within the overall HEM-3/AERMOD model domain (50 km) of the source category emissions, irrespective of risk. For context, Table 2 also provides the nationwide percentages of these various demographic groups. The detailed demographic results used to create Table 2 can be found in the Appendices.

	The proximity results presented in Table 2 indicate that the population percentages for certain demographic groups within 5 km of the VOP manufacturing facility emissions are greater than the corresponding nationwide percentage for those same demographics. The following demographic percentages for populations residing within 5 km of VOP manufacturing facilities are higher than the corresponding nationwide percentages: the minority population (41% compared to 38% nationally), the African American population (19% compared to 12% nationally), ages 0 to 17 (24% compared to 23% nationally), ages 18 to 64 (64% compared to 63% nationally), the population over 25 without a high school diploma (18% compared to 14% nationally), and the population below the poverty level (21% compared to 14% nationally). The other demographic groups included in the assessment within 5 km of VOP emissions were the same or lower than the corresponding nationwide percentages.
 Cancer Risk Demographics

	The demographic composition of the population (if any) at a lifetime cancer risk greater than or equal to 1-in-1 million from inhalation of emissions associated with VOP operations is shown in Table 2. The cancer risk results shown are based on the standard HEM-3/AERMOD domain of a 50-km radius around each modeled facility. The detailed demographic risk results used to create Table 2 can be found in Appendix A (based on 50 km). Appendix B provides similar results based on a 5-km radius, although these closer proximity results were used primarily to inform the proximity analysis (irrespective of risk) described in Section 4.1. 

	The results of the risk analysis presented in Table 2 indicate that there are no people within a 50-km radius of modeled facilities exposed to a cancer risk greater than or equal to 1-in-1 million as a result of emissions from VOP operations.
 Noncancer Risk Demographics

	The demographic composition of the population (if any) at a noncancer neurological hazard index (HI) greater than 1 from inhalation of emissions associated with VOP operations is shown in Table 2. The noncancer risk (i.e., maximum HI) results shown are based on the standard HEM-3/AERMOD domain of a 50-km radius around each modeled facility. The detailed demographic HI results used to create Table 2 can be found in Appendix C (based on 50 km). Appendix D provides similar results based on a 5-km radius, although these closer proximity results were used primarily to inform the proximity analysis (irrespective of risk) described in Section 4.1. 

	The results of the risk analysis presented in Table 2 indicate that there are no people within a 50-km radius of modeled facilities exposed to a noncancer risk (in this analysis, neurological HI) greater than 1 as a result of emissions from VOP operations. 


Table 2.  Summary of Demographic Assessment for the Solvent Extraction for Vegetable Oil Production Source Category
                               Population Basis
                                   Parameter
                               Demographic Group

                                       
                                     Total
                                  Minority[1]
                               African 
American
                               Native 
American
                             Other and
Multiracial
                              Hispanic
or Latino
                                 Ages 0 
to 17
                                Ages 18 
to 64
                                Ages 65 
and up
                         Over 25 Without a HS Diploma
                            Below the Poverty Level
                                       
                                       
                             Linguistic Isolation
Nationwide 
                                      n/a
                                  317,746,049
                                      38%
                                      12%
                                     0.8%
                                      7%
                                      18%
                                      23%
                                      63%
                                      14%
                                      14%
                                      14%
                                      6%

                                   Proximity
            Population Surrounding the Source Category Emissions[2]
                                Source Category
                    Within HEM-3 Model Domain       (50 km)
                                  3,9171,369
                                      38%
                                      16%
                                     0.4%
                                      6%
                                      15%
                                      24%
                                      63%
                                      13%
                                      14%
                                      14%
                                      5%
                                Source Category
                     Within close proximity        (5 km)
                                   2,105,855
                                      41%
                                      19%
                                     0.5%
                                      4%
                                      17%
                                      24%
                                      64%
                                      12%
                                      18%
                                      21%
                                      5%

                         Maximum 
Risk 
(in 1 million)
    Population With Cancer Risk Greater Than or Equal to 1 in 1 million[3]
                                Source Category
                                      0.2
                                       0
                                      0%
                                      0%
                                      0%
                                      0%
                                     0.0%
                                      0%
                                      0%
                                      0%
                                      0%
                                      0%
                                      0%
                                       
                             Maximum 
Hazard Index
                                (Neurological) 
               Population With Noncancer Risk Greater Than 1[4]
                                Source Category
                                      0.7
                                       0
                                      0%
                                      0%
                                      0%
                                      0%
                                     0.0%
                                      0%
                                      0%
                                      0%
                                      0%
                                      0%
                                      0%
Notes:

1 Minority population is the total population minus the white population.

[2] Proximity population statistics are provided irrespective of cancer and noncancer risk.

3 Population statistics based on cancer risk are for the population residing within the HEM-3/AERMOD model domain (50-km radius from modeled facility centers) with cancer risk estimated to be greater than or equal to 1 in 1 million.

[4] Population statistics based on noncancer risk (maximum hazard index) are for the population residing within the HEM-3/AERMOD model domain (50-km radius from modeled facility centers) with noncancer risk estimated to be greater than 1.


 Uncertainty Discussion
      
	Our analysis of the distribution of risks across various demographic groups is subject to the typical uncertainties associated with census data (e.g., errors in filling out and transcribing census forms), which are generally thought to be small, as well as the additional uncertainties associated with the extrapolation of census block group data down to the census block level.  

	The uncertainties in these risk estimates include the same uncertainties in emissions data sets, in air dispersion modeling, in inhalation exposure and in dose response relationships that are associated with our source category risk estimates.

	The methodology for our demographic analyses combines EPA's RTR methodology using HEM-3/AERMOD risk results with EJSCREEN demographic data. While this is our best attempt to provide useful information now, our thinking is continuously advancing. The EPA has developed technical guidance for environmental justice analyses. We present these analyses, with their associated uncertainties, to EPA decision makers and the public as additional analyses to inform RTR decisions. 







                                       
                                  Appendix A

        Demographic Cancer Risk Results within the HEM-3 Model Domain 
                 (50 kilometers) of Source Category Emissions

                                       
                                        
                                         
                                       


                                  Appendix B

    Demographic Cancer Risk Results within Close Proximity (5 kilometers) 
                         of Source Category Emissions

                                       
                                       
                                       

                                       

                                       
                                       

                                       
                                       
                                       

                                       

                                       
                                       

                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                  Appendix C

       Demographic Noncancer Risk Results within the HEM-3 Model Domain 
                 (50 kilometers) of Source Category Emissions
                                       


                                       
                                       
                                       
                                       
 


                                       
                                       

                                       

                                       

                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                  Appendix D
Demographic Noncancer Risk Results within Close Proximity (5 kilometers) of Source Category Emissions



                                       


                                       
                                       
                                        
                                       
                                       

                                       
                                       

                                       
                                       

                                       
