Abstract
Particulate Matter (PM) is a major cause of morbidity and mortality worldwide. This study examines the use of five models using a Bayesian Hierarchical Downscaling model structure to predict PM2.5 ( PM < 2.5 μm) across a region in central Italy in 2015. We build upon previous modeling work done in this region of Italy and provide an alternative way to create models to predict PM2.5 using fewer spatiotemporal and spatial predictors, smaller training data sets as well as the ability to calculate uncertainty measurements. The Bayesian models used in this paper predicted PM2.5 concentrations with a mean overall cross validation R2 of .72. Using extinction as our main predictor (aerosol optic density (AOD) divided by planetary boundary layer (PBL)) and data from NASA’s Multiple Angle Imager for Aerosol Ancillary Geographic Product (MAIA AGP) and Italian collaborators, we demonstrated that the MAIA AGP variables can be used to reliably predict PM2.5 and generate R2 values equivalent to those generated from models run with parameters processed by our Italian collaborators. The ability of our Bayesian model to integrate MAIA AGP variables and predict annual and daily PM2.5 concentrations with reasonable accuracy and uncertainty measurements provides future exposure studies with important data about model uncertainty, and the ability to predict PM2.5 across resource limited domains.
Table of Contents
This table of contents is under embargo until 18 May 2025
About this Master's Thesis
Rights statement
- Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School |
|
Department |
|
Degree |
|
Submission |
|
Language |
|
Research Field |
|
关键词 |
|
Committee Chair / Thesis Advisor |
|
Committee Members |
|