Social Justice in the Mississippi River Basin
Mandy Liesch 10/11/2021
There are many different ways to classify “risk” in rural areas. There are current environmental exposures, socio-economic variables, issues that result from culture shifts, population dynamic changes, changes in agricultural resilience, and future risks changes from shifting climates, crop yield changes, and environmental hazards.
This project focused on a 12 state region: The states along the main branch of the Mississippi River basin, the Ohio, and Tennessee River.
There are several different datasets that were utilized to build the One River Dataset:
Social Determinants of Health: Looking at the documentation, it aggregates several sources of health data. Data is availabe at both county and zip code levels. For this project, we are using the County data.
Including:
Along with the social determinents of health database, the Social Capital Index (SoCI) uses 19 indicators of public data to determine which communities are more capable of disaster resilience, based on how communities are organized and researched. They use three categories (bonding, bridging, and linking) to design an overall index by county. The base indicators that are not well represented with the above dataset were added into the Social Determinants of Health Dataset.
In addition, the ESRI County Health Rankings that were not in the other databases were added to the csv datafile.
There are still several variables that we want for analysis missing from the dataset. The USDA Economic Research Services (ERS), publishes a Rural Atlas of Rural and Small Town America. this site looks at variables of people, economic, and jobs data. The website above documents the maps and processes used to create this atlas.
This data spanned a 5 year range from 2015-2019. Defined as Drug Overdoses, Alcohol Related Deaths, Suicides, and Homicides, getting reliable county level estimates can be difficult, especially in rural areas. The CDC suppresses any county information where there are not at least 10 of each individual type of deaths. In areas with low population density, it can take 20+ years to get to 10 of these deaths. So, to fill in these blank data rates, the 2013 Urban and Rural classification average values were used by county type (Metropolitan Core, Micropolitan, non-core, etc.) as a substitution. This dataset deals with Non-Ignorable Missingness, meaning that the missing data trends toward rural areas.
There are 180 counties that are suppressed with Suicide and Homicide Data:
There are 222 counties that are suppressed for Alcohol and Drug Overdoses:
Deaths from Diabetes were calculated over the five year span of 2015-2019 from the CDC Wonder Data. Like the Deaths from Despair Data, there are several counties with less than 10 deaths.
There are 43 Counties with missing Diabetes data:
| State | Urbanization | Suicide | Homicide | DrugOD | Alcohol | Diabetes |
|---|---|---|---|---|---|---|
| Arkansas | Large Fringe Metro | 13.2 | 23.4 | 16.9 | 5.34 | 57.6 |
| Medium Metro | 17.8 | 7.2 | 14.5 | 9.40 | 31.9 | |
| Small Metro | 19.8 | 12.4 | 17.9 | 10.00 | 29.7 | |
| Micropolitan (Nonmetro) | 20.3 | 9.4 | 13.9 | 10.20 | 40.4 | |
| NonCore (Nonmetro) | 21.2 | 7.1 | 13.9 | 10.20 | 41.6 | |
| Illinois | Large Central Metro | 8.7 | 14.0 | 22.0 | 8.50 | 22.1 |
| Large Fringe Metro | 11.0 | 3.4 | 17.4 | 7.40 | 19.2 | |
| Medium Metro | 14.7 | 7.0 | 25.5 | 11.80 | 24.8 | |
| Small Metro | 13.9 | 5.2 | 19.7 | 9.60 | 21.7 | |
| Micropolitan (Nonmetro) | 16.6 | 2.6 | 17.8 | 10.20 | 28.2 | |
| NonCore (Nonmetro) | 17.5 | 2.3 | 15.4 | 6.90 | 33.6 | |
| Indiana | Large Central Metro | 14.6 | 17.4 | 35.6 | 14.40 | 26.9 |
| Large Fringe Metro | 15.3 | 5.8 | 24.9 | 9.90 | 27.2 | |
| Medium Metro | 14.9 | 7.5 | 21.7 | 12.90 | 32.1 | |
| Small Metro | 15.4 | 4.0 | 21.1 | 11.80 | 28.3 | |
| Micropolitan (Nonmetro) | 16.1 | 3.2 | 22.9 | 9.30 | 39.6 | |
| NonCore (Nonmetro) | 17.3 | 2.7 | 20.9 | 8.30 | 39.3 | |
| Iowa | Medium Metro | 15.7 | 3.5 | 14.5 | 12.80 | 22.8 |
| Small Metro | 12.6 | 2.3 | 8.1 | 10.90 | 22.6 | |
| Micropolitan (Nonmetro) | 17.7 | 3.0 | 10.9 | 14.40 | 36.6 | |
| NonCore (Nonmetro) | 14.8 | 1.5 | 7.5 | 11.20 | 40.5 | |
| Kentucky | Large Central Metro | 17.0 | 13.3 | 41.7 | 13.00 | 27.5 |
| Large Fringe Metro | 16.2 | 2.6 | 45.2 | 12.30 | 29.7 | |
| Medium Metro | 15.2 | 5.5 | 35.1 | 12.30 | 29.1 | |
| Small Metro | 17.1 | 4.9 | 17.3 | 10.20 | 31.0 | |
| Micropolitan (Nonmetro) | 18.7 | 5.0 | 28.6 | 9.70 | 35.7 | |
| NonCore (Nonmetro) | 18.9 | 5.1 | 28.5 | 9.80 | 44.8 | |
| Louisiana | Large Central Metro | 11.8 | 34.8 | 39.0 | 6.60 | 25.7 |
| Large Fringe Metro | 15.4 | 10.4 | 33.2 | 8.30 | 20.5 | |
| Medium Metro | 14.3 | 13.5 | 19.3 | 7.60 | 26.2 | |
| Small Metro | 16.7 | 9.9 | 21.9 | 7.40 | 36.5 | |
| Micropolitan (Nonmetro) | 16.8 | 10.3 | 20.6 | 8.40 | 38.0 | |
| NonCore (Nonmetro) | 16.8 | 8.2 | 15.7 | 6.20 | 35.6 | |
| Minnesota | Large Central Metro | 12.2 | 4.0 | 17.4 | 13.70 | 20.5 |
| Large Fringe Metro | 12.9 | 1.4 | 11.3 | 10.00 | 18.6 | |
| Medium Metro | 20.5 | 2.0 | 20.0 | 25.10 | 35.2 | |
| Small Metro | 13.0 | 1.4 | 10.7 | 11.20 | 19.0 | |
| Micropolitan (Nonmetro) | 15.3 | 1.9 | 12.1 | 11.90 | 28.2 | |
| NonCore (Nonmetro) | 17.5 | 2.3 | 12.9 | 13.70 | 40.5 | |
| Mississippi | Large Fringe Metro | 14.1 | 9.6 | 18.1 | 5.90 | 44.3 |
| Medium Metro | 15.5 | 12.4 | 14.2 | 7.40 | 28.2 | |
| Small Metro | 14.5 | 8.4 | 12.6 | 3.20 | 26.4 | |
| Micropolitan (Nonmetro) | 12.7 | 14.1 | 10.4 | 7.10 | 44.9 | |
| NonCore (Nonmetro) | 14.3 | 11.8 | 11.6 | 6.60 | 39.8 | |
| Missouri | Large Central Metro | 17.8 | 28.5 | 33.7 | 13.80 | 22.8 |
| Large Fringe Metro | 17.4 | 8.2 | 26.5 | 7.80 | 20.5 | |
| Medium Metro | 21.6 | 4.5 | 24.2 | 12.70 | 24.2 | |
| Small Metro | 17.7 | 5.4 | 13.6 | 7.40 | 28.7 | |
| Micropolitan (Nonmetro) | 20.8 | 4.5 | 18.5 | 7.30 | 33.9 | |
| NonCore (Nonmetro) | 20.7 | 4.9 | 16.7 | 8.70 | 35.1 | |
| Ohio | Large Central Metro | 13.0 | 11.2 | 40.1 | 10.10 | 27.6 |
| Large Fringe Metro | 14.3 | 2.7 | 33.3 | 8.90 | 24.8 | |
| Medium Metro | 16.5 | 6.9 | 41.9 | 11.80 | 34.2 | |
| Small Metro | 16.5 | 5.3 | 41.9 | 9.80 | 44.6 | |
| Micropolitan (Nonmetro) | 16.5 | 2.8 | 32.0 | 9.20 | 40.5 | |
| NonCore (Nonmetro) | 15.9 | 3.4 | 27.1 | 8.30 | 40.1 | |
| Tennessee | Large Central Metro | 11.8 | 18.1 | 28.1 | 10.20 | 25.9 |
| Large Fringe Metro | 16.8 | 4.6 | 23.9 | 9.40 | 20.1 | |
| Medium Metro | 17.9 | 6.1 | 34.6 | 14.70 | 29.8 | |
| Small Metro | 18.5 | 5.2 | 23.9 | 13.40 | 31.8 | |
| Micropolitan (Nonmetro) | 21.3 | 3.9 | 24.7 | 13.20 | 37.4 | |
| NonCore (Nonmetro) | 21.7 | 5.5 | 25.3 | 11.50 | 40.4 | |
| Wisconsin | Large Central Metro | 12.4 | 14.7 | 36.0 | 15.50 | 25.6 |
| Large Fringe Metro | 13.8 | 1.6 | 15.9 | 9.30 | 22.8 | |
| Medium Metro | 14.4 | 1.8 | 17.1 | 11.00 | 18.0 | |
| Small Metro | 16.5 | 2.0 | 15.8 | 13.30 | 24.3 | |
| Micropolitan (Nonmetro) | 16.2 | 1.5 | 16.3 | 14.10 | 29.0 | |
| NonCore (Nonmetro) | 18.0 | 1.8 | 13.2 | 16.80 | 35.5 |
The same five year span was used to calculate all firearm fatalities from the CDC Wonder Data.
There are several different geospatial analysis that need to be done to get the spatial and environmental data into this analysis.
The Census tract shapefiles were downloaded from the TIGER Shapefile website. The individual state census tracts were then merged in ArcMap 10.8.1. All shapefiles with 0 acres of land area were removed from analysis (located mostly in coastal counties and in the Great Lakes Region).
The overall estimatd census block population was pulled from the 2019 American Community Data Survey on the census bureau website. Population density (per/km) was calculated from the total land area of the census block. This shapefile for total population density was converted into a raster file.
Energy infrastructure and pipeline shapefiles from the US Energy Administration were accessed and downloaded into a map. All pipeline shapefiles were buffered to 5 km, and then the raster was clipped by the buffer, and the values for each county was extracted using zonal statistics at the county level.
Every year, the EPA puts out the EJScreen environmental justice index is a geodatabase that is at the US Census Tract or Block level. Socio-economics and demographic data are included (for this study, they were not used, as they were present in the Social Determinents of Health spreadsheet). Present data risk includes:
The EPA Safe Water Dashboard has a quarterly lookup of the total number of drinking water violations in a specified state, as well as the total amount of nitrate violations, with the average nitrate value. The total amount of quarterly violations were averaged over a one year period, along with the nitrate data values. The data were aggregated by zip code, and then by county.
As seen above, the EJ Screen contains mostly air quality parameters because of many different data overlap concerns and issues. Prior to the EJScreen, the EPA had an Environmental Quality Index. This index quantified environmental risks into five major categories: air, water, land, built, and sociodemographic environments, at a county level, calculating and overall principal components value.
The US EPA has a Social Vulnerability Report from their Climate Change Impacts and Research Analysis group. This report breaks down future climate risk by census district from:
These CSV files were cleaned and merged together in R to get average data per county for present and future risks.
The Bureau of Economic Analysis has several different values measuring income, jobs, and personal finances at a county level data from 2001-2019.
Traditionally speaking, rural America relied heavily on manufacturing in proportion to Urban America. The CAEMP25N data table has all of the jobs in each county, from the total jobs (Line Code 10), percentages of each industry were divided by a total, and the linear space time trend (slope), and initial percentages (intercept) were used for economic output in these categories:
The World Resources Institute has an Aqueduct Water Resources Risk atlas that looks at all of the potential threats that impact water use both currently and in the future. For 2030 and 2040 data, the RMP 8.5 scenario (worst case projections) for the major variables were calculated with zonal statistics in ArcMap 10.8.
Future scenarios from the RMP 8.5, were used from the “Estimating Economic Damage from Climate Change in the United States” by Hsiang, Kopp, Rising, Jina et al. (2017). There is a Zenodo repository that has the downloadable percentiles (q50 is used in this analysis, as the likely middle ground) from multiple Monte Carlo simulations at the county level for three different time periods (2020-2039), (2040-2059), (2080-2099) for:
The Resource Watch Land Cover Dataset has 30 m satellite resolution data that looks at cropland gained (from 2000-2003 and 2016-2019) over a 19 year period. Both of these rasters are averaged and clipped to the county level.
These mapping products were derived through terrain analysis and a technique of pattern classification performed on hydrologically conditioned DEMs by Manfreda, S., and Troy, T.J., in 2017.