Team: 1
School: Cleveland High
Area of Science: Environmental Sciences
Proposal: Fine particulate matter consists of particles < 2.5 micrometers in diameter and poses a severe threat to human health worldwide, associated with increased rates of lung cancer, asthma, heart disease, and more. While the US, East Asia, and Europe have high-resolution 1kmx1km continuous PM2.5 maps, Mexico and many developing countries lack air quality monitoring for hundreds or thousands of miles. Satellites, however, have global daily coverage and can be used to estimate ground particulate matter concentrations. Aerosol optical depth (AOD) is a measure of the extinction of light (such as from aerosols), but its relationship with PM2.5 is complicated by clouds, altitude, terrain, and other uncertainties. In situations like these, machine learning algorithms are a powerful tool to find these complicated nonlinear relationships.
The goal of this project is to create a machine learning algorithm accounting for land use (INEGI), meteorological (ECMWF ERA-5), historical ground measurements (SINAICA), and remote sensing (MAIAC MCD19A2) variables to predict particulate matter at the 1km resolution across the entire country.
Because of Mexico's diverse geography and climate, an ensemble-based approach and spatiotemporal analysis will be considered to capture seasonal or spatial changes in the relationships between AOD and particulate matter.
The data collection and models will be created in R and Python, and the accuracy of the model will be evaluated with cross-validation. Its performance will be compared to NASA's Merra-2 surface dust reanalysis product (M2T1NXAER) over the region, and a bias correction algorithm will be created to correct its historical estimates of ground particulate matter.
Team Members:
Sponsoring Teacher: Ashli Knoell