# Evaluating a Land Use Regression Model for Estimating Metals in Fine Particulate Matter across the Denver Metro Area: The Healthy Start Study

## TL;DR

This study evaluates a model to estimate metal levels in fine particulate matter across Denver, finding that traffic-related metals like copper and iron are predicted well, but other metals are not.

## Contribution

The study introduces a land use regression model for estimating seven metals in PM2.5 in Denver, emphasizing traffic-related predictors.

## Key findings

- The model predicts copper and iron well during fall (R² = 0.56 and 0.63, respectively).
- Traffic-related predictors are the strongest and most consistent across models.
- The model fails to capture non-traffic-related metals year-round (R² < 0.40).

## Abstract

Few studies examine
health effects of metals in ambient fine particulate
matter (PM2.5), as measurements of elemental composition
are sparse. To facilitate intraurban studies in Denver, Colorado,
we developed land use regression models for seven speciescopper
(Cu), iron (Fe), titanium (Ti), zinc (Zn), potassium (K), calcium
(Ca), and magnesium (Mg). As part of the Healthy Start Cohort study,
we collected filter-based PM2.5 using personal air samplers
at 67 locations across Denver. Sample collection occurred from May
2018 through March 2019, accounting for all meteorological seasons.
Exposure models were informed by 83 geospatial covariates, with traffic-related
predictors as the strongest and most consistent across models. Model
performance was evaluated using 10-fold cross validation and overall,
varied by sampling campaign and season, with R
2 values ranging from 0 to 0.63. At best, our model predicts
Cu and Fe during fall (R
2 = 0.56 and 0.63,
respectively); whereas it fails to capture species unrelated to traffic
year-round (R
2 < 0.40). This highlights
the influence of missing predictors (e.g., wildfire smoke, atmospheric
transport and other meteorological factors) on PM2.5 concentrations
and spatial gradients. Despite limitations, resulting models enable
estimation of intraurban metal exposures and support future analyses
of long-term health impacts in Denver.