Combining Environmental Data Using Hierarchical Bayesian Space-Time Models

EPA Grant Number: R826887
Title: Combining Environmental Data Using Hierarchical Bayesian Space-Time Models
Investigators: Zimmerman, Dale , Cowles, Mary Kathryn
Institution: University of Iowa
EPA Project Officer: Louie, Nica
Project Period: January 1, 1999 through December 31, 2001
Project Amount: $199,904
RFA: Environmental Statistics (1998) RFA Text |  Recipients Lists
Research Category: Environmental Statistics , Health , Ecosystems


Environmental data often come from multiple, disparate sources or consist of subsets which are collected under markedly different conditions in space or time. To obtain the most informative conclusions from the data, it is preferable (and sometimes necessary) to base a statistical analysis upon the combined data from all these sources or subsets. Some specific examples include: grafting environmental time series from different measurement systems to better detect long-term trends; streamflow record reconstruction at a previously ungauged station; and combining data from monitoring networks operated by two different organizations for the purpose of improved prediction of the underlying processes) at unsampled sites and times. The primary objectives of this research are to build useful models for combined environmental space-time data and to develop sound, easy-to-use methods for fitting those models and using them for estimation, prediction, and other inferences regarding the underlying space-time processes).


We shall develop general, flexible Bayesian hierarchical models suitable for combined space-time data. The models will accommodate spatial and temporal dependence at several scales and non-trivial spatio-temporal interaction. Furthermore, they will be able to deal with the lack of rectangularity in the combined data. We shall fit these models using Markov Chain Monte Carlo (MCMC) methods, treating missing values which arise from 19 combining the data as additional quantities to be estimated. High dimensionality, high correlation among model parameters, and increased uncertainty introduced by missing data all contribute to slow convergence of the MCMC sampler. Consequently, we will investigate several strategies for assessing and accelerating convergence. Model performance and the merits of combining the data will be investigated by simulation. We will make computer code for fitting our models publicly available.

Expected Results:

The main result of the research will be the provision of a convenient method for users of environmental data drawn from disparate sources to carry out more efficient estimation and prediction than is possible from separate analyses of the data components. Clearly, those risk assessment procedures which rely on accurate characterizations and projections of an environmental variable across space and time stand to benefit. For example, our approach should make possible more rapid detection of global or local climate change or other kinds of environmental change and may also serve to reduce the cost of present and future environmental monitoring programs.

Publications and Presentations:

Publications have been submitted on this project: View all 16 publications for this project

Journal Articles:

Journal Articles have been submitted on this project: View all 5 journal articles for this project

Supplemental Keywords:

Monte Carlo methods, trends, streamflow, prediction, RFA, Economic, Social, & Behavioral Science Research Program, Ecosystem Protection/Environmental Exposure & Risk, Ecosystem/Assessment/Indicators, Ecological Effects - Environmental Exposure & Risk, Environmental Statistics, risk assessment, hierarchical Bayesian space-time models, Bayesian space-time model, co-pollutant effects, global environmental data, prediction, simulation studies, streamflow, statistical models, combined data, data analysis, spatial-temporal methods, data models, global warming, innovative statistical models

Relevant Websites:

Progress and Final Reports:

  • 1999 Progress Report
  • 2000 Progress Report
  • Final