Reduced-Rank Spatio-Temporal Modeling of Air Pollution Concentrations in the Multi-Ethnic Study of Atherosclerosis and Air Pollution

• Activity Number: 86
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Casey Olives*+ and Lianne Sheppard and Johan Lindstrom and Paul D. Sampson and Joel D. Kaufman and Adam Szpiro
• Companies: and University of Washington and Lund University and University of Washington and University of Washington and University of Washington
• Keywords: spatio-temporal ; kriging ; splines ; air pollution ; prediction ; epidemiology

Abstract

Prediction of individual air pollution exposure in the Multi-Ethnic Study of Atheroscelerosis and Air Pollution (MESA Air) study relies on a flexible spatio-temporal prediction model that integrates land-use regression with kriging to account for spatial dependence in pollutant concentrations. Temporal trends are estimated via modified singular value decomposition and temporally varying spatial residuals.

Spatio-temporal models are limited in their efficacy for large datasets due to computational burden. We develop reduced-rank versions of the MESA Air spatio-temporal model that employ low-rank kriging (LRK) and thin plate regression splines (TPRS) to account for spatial variation. We compare the performance of these models using regulatory and supplemental MESA Air monitoring data for predicting concentrations of NOx in Los Angeles via cross-validation.

We show that reduced-rank models are competitive with their full-rank counterparts and can improve computational efficiency in certain cases. LRK and TPRS were competitive across the formulations considered. We conclude that the use of either by LRK or TPRS is an attractive option for spatio-temporal prediction in MESA Air.