Spatiotemporal modeling of irregularly spaced aerosol optical depth data

Jacob J. Oleson, Naresh Kumar, Brian J. Smith

Research output: Contribution to journalArticlepeer-review

4 Scopus citations


Many advancements have been introduced to tackle spatial and temporal structures in data. When the spatial and/or temporal domains are relatively large, assumptions must be made to account for the sheer size of the data. The large data size, coupled with realities that come with observational data, make it difficult for all of these assumptions to be met. In particular, air quality data are very sparse across geographic space and time, due to a limited air pollution monitoring network. These "missing" values make it difficult to incorporate most dimension reduction techniques developed for high-dimensional spatiotemporal data. This article examines aerosol optical depth (AOD), an indirect measure of radiative forcing, and air quality. The spatiotemporal distribution of AOD can be influenced by both natural (e.g., meteorological conditions) and anthropogenic factors (e.g., emission from industries and transport). After accounting for natural factors influencing AOD, we examine the spatiotemporal relationship in the remaining human influenced portion of AOD. The presented data cover a portion of India surrounding New Delhi from 2000-2006. The proposed method is demonstrated showing how it can handle the large spatiotemporal structure containing so much missing data for both meteorologic conditions and AOD over time and space.

Original languageEnglish (US)
Pages (from-to)297-314
Number of pages18
JournalEnvironmental and Ecological Statistics
Issue number2
StatePublished - Jun 2013


  • Air quality
  • AOD
  • Autoregressive
  • Bayesian
  • Spatial correlation
  • Temporal correlation

ASJC Scopus subject areas

  • Environmental Science(all)
  • Statistics, Probability and Uncertainty
  • Statistics and Probability


Dive into the research topics of 'Spatiotemporal modeling of irregularly spaced aerosol optical depth data'. Together they form a unique fingerprint.

Cite this