How predicted microplastic exposure connects with health, geography, and inequality across the U.S.
This tool visualizes how wastewater, geography, and chronic disease data come together to show where microplastics may be more common – and what those patterns really mean. It’s not a toxicology study or risk calculator. Instead, it helps explore how infrastructure and inequality shape both exposure and health.
Wastewater + Geography + Population
↓
Predicted Microplastics (µg/L)
↓
Compare with Chronic Disease Rates
Microplastics are in headlines, bottles, and panic threads. National measurement is not. This project uses a transparent exposure proxy and real health data so you can explore the big picture yourself.
Why a proxy - there is no national grid of measured microplastic concentrations. We predict county exposure using variables that move microplastics through systems: wastewater design flow, treated population, treatment level, impervious cover, and distance to coast.
What the value means - Predicted microplastics is in µg/L equivalent. It is a model estimate trained where raster measurements existed, then applied to all counties.
Uncertainty - each county has a 95 percent prediction interval. We mark wide when the interval is large and OOR when inputs fall outside the training range. Treat those as provisional.
The color ramp shows the selected layer. Start with predicted microplastics. Click a county to open its profile with exposure, confidence, SES context, and disease metrics.
Not proof that microplastics are safe. Not individual risk. Not a reason to ignore plastic pollution.
It means that in national county data, urbanization beats exposure as an explanation for differences we see.
Why geospatial nuance matters - counties bundle people together. States differ in reporting, demographics, climate, and care systems. Spatial autocorrelation means neighbors look alike. If you ignore those facts, you can read patterns as exposure effects when they are really geography.
What a better test would look like - individual level design with measured microplastic biomarkers, age-adjusted outcomes, smoking and PM2.5 controls, and spatial or multilevel modeling. Until then, treat predictive exposure surfaces as screening tools that guide where measurement would be most useful.
Standalone viewer - double-click to open. No server required (internet needed for the basemap tiles).
Built for DATA 467: Data Science Applications, University of Arizona. Instructor: Dr. Haverland. AI assistance by ChatGPT for integration and documentation. Citations and data provenance verified manually.