Grantee Research Project Results
2000 Progress Report: Hierarchical Statistical Analysis of Global and Regional Environmental Data
EPA Grant Number: R827257Title: Hierarchical Statistical Analysis of Global and Regional Environmental Data
Investigators: Cressie, Noel A.C. , Berliner, Mark , Wikle, Christopher K.
Current Investigators: Cressie, Noel A.C. , Berliner, Mark , Wilke, Christopher K.
Institution: The Ohio State University , University of Missouri - Columbia
Current Institution: The Ohio State University , University of Missouri - Kansas City
EPA Project Officer: Hahn, Intaek
Project Period: September 1, 1998 through August 31, 2001 (Extended to March 31, 2003)
Project Period Covered by this Report: September 1, 1999 through August 31, 2000
Project Amount: $325,000
RFA: Environmental Statistics (1998) RFA Text | Recipients Lists
Research Category: Environmental Statistics , Human Health , Aquatic Ecosystems
Objective:
The first objective is to implement hierarchical spatio-temporal statistical modeling of environmental phenomena (e.g., sea-surface temperatures, disease mortality/morbidity rates, remote-sensing data). The second objective is to find computationally efficient ways to filter, statistically, massive spatio-temporal data sets, such as one obtains from polar-orbiting satellites. The third objective is to investigate ways to link two or more spatio-temporal environmental processes, one being explanatory for the other (e.g., meteorology might be explanatory for wetland ecology).
Progress Summary:
A Bayesian fingerprinting methodology for assessing anthropogenic impacts on climate is developed in the paper by Berliner, Levine, and Shea (2000). The authors develop a spatial CO2 fingerprint based on control and forced model output from the National Center for Atmospheric Research Climate System Model. The prior model for the amplitude of the fingerprint of this Bayesian approach is a mixture of two distributions: one reflects prior uncertainty in the anticipated value of the amplitude under the hypothesis of "no climate change," and the second reflects behavior assuming "climate change forced by CO2." A robust Bayesian analysis is used, which investigates the ranges of posterior inferences as prior inputs are varied.
In Berliner (2000), a collection of examples demonstrating the power of hierarchical modeling of environmental processes is presented. These include combining data sets and a variety of space-time modeling approaches. Notions and examples of how hierarchical Bayesian modeling provides a mechanism for developing large-scale analyses bridging different sciences are discussed.
The methodology of hierarchical Bayesian dynamical modeling is used in the paper by Berliner, Wikle, and Cressie (2000), in which a new procedure for long-lead forecasting tropical Pacific sea surface temperatures is presented. Through its combining of substantial physical understanding and statistical modeling and learning, the procedure acquires considerable predictive skills. The approach accounts explicitly for uncertainty in the formulation of the model, which leads to realistic error bounds on forecasts.
An infectious disease typically spreads via contact between infected and susceptible individuals. Since the small-scale movements and contacts between people are generally not recorded, available data regarding infectious disease are often aggregations in space and time. In the paper by Mugglin, Cressie, and Gemmell (2000), the authors developed a spatially descriptive, temporally dynamic hierarchical model to be fitted to such data. A Bayesian approach is taken, and using Markov chain Monte Carlo (MCMC) the posterior estimates of all parameters of interest are computed.
Another paper that deals with hierarchical modeling is Hrafnkelsson and Cressie (2000). The authors propose a method for a Bayesian hierarchical analysis of count data that are observed at irregular locations in a bounded spatial domain. The data are modeled on a fine regular lattice, although many sites do not have observations at them. The main goal is to predict the hidden Poisson-mean process at all sites on the lattice given the spatially irregular count data; to do this, an efficient MCMC is used.
In the paper by Zhu, Lahiri, and Cressie (2000), a spatio-temporal process is considered and the statistical methodology to analyze changes in the spatial cumulative distribution function over time is established. Hypothesis testing to detect a difference in the spatial random process at two time points is developed, as well as a prediction interval to quantify such discrepancy in the corresponding SCDFs. The methodology is applied to an ecological index for foliage condition of red maple trees in the state of Maine in the early 1990s.
Polar orbiting satellites remotely sense the earth and its atmosphere, producing data sets that give daily global coverage. For any given day, the data are many and spatially irregular. The goal of the article by Huang, Cressie, and Gabrosek (2000) is to predict values that are spatially regular at different resolutions, and also to process the data very rapidly. The authors present a new statistical prediction methodology that preserves "mass balance" across resolutions and computes spatial predictions and prediction (co)variances extremely fast.
In the paper by Aldworth and Cressie (2000), a spatial empirical Bayes model for environmental data collected over a region is proposed. The authors also propose a predictor, based on the kriging methodology with extra constraints, that has useful unbiasedness properties in predicting nonlinear spatial functionals. This predictor, called covariance-matching constrained kriging, is an optimal linear predictor that matches not only first moments but second moments as well. Cressie and Johannesson (2000) give a practical algorithm for computing the predictor.
The paper by Cressie and Richardson (2000) investigates how regression coefficients measuring covariate effects at the point level are modified under aggregation. Changing the level of aggregation leads to completely different conclusions about exposure-effect relationships, a phenomenon often referred to as ecological bias. The notion of maximum entropy is used to approximate that part of the within-area distribution of the exposure variable that is unknown. The approximation is then used to obtain an expression for the ecological bias. The methodology is applied to a radon-exposure study conducted in France.
Future Activities:
In the final year of the grant, we shall consider three outstanding problems. The first, prompted by the results of Aldworth and Cressie (2000), involves covariance matching in spatio-temporal settings; this is natural, for example, when compliance history is as important as current compliance in a monitoring network. The second problem involves adapting the multi-resolution Kalman filter to include a dynamic component. The third, and most open-ended, is using Bayesian thinking to model explanatory links between space-time environmental processes. We are starting to look at precipitation and wetland habitat in the Prairies region of the USA.
Journal Articles on this Report : 3 Displayed | Download in RIS Format
Other project views: | All 103 publications | 29 publications in selected types | All 19 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Berliner LM, Levine RA, Shea DJ. Bayesian climate change assessment. Journal of Climate 2000;13(21):3805-3820. |
R827257 (2000) R827257 (Final) |
Exit Exit Exit |
|
Berliner LM, Wikle CK, Cressie N. Long-lead prediction of Pacific SSTs via Bayesian dynamic modeling. Journal of Climate 2000;13(22):3953-3968. |
R827257 (2000) R827257 (Final) |
Exit Exit Exit |
|
Berliner M. Hierarchical Bayesian modeling in the environmental sciences. Allgemeines Statistisches Archiv (AStA Advances in Statistical Analysis) 2000;84(2):141-153. |
R827257 (2000) R827257 (Final) |
Exit Exit |
Supplemental Keywords:
ambient air, atmosphere, global climate, stratospheric ozone, precipitation, health effects, ecological effects, human health, Bayesian, biology, ecology, epidemiology, mathematics, modeling, climate models, satellite, remote sensing, agriculture., RFA, Economic, Social, & Behavioral Science Research Program, Scientific Discipline, Air, Ecosystem Protection/Environmental Exposure & Risk, Ecology, air toxics, Ecosystem/Assessment/Indicators, Mathematics, climate change, Ecological Effects - Environmental Exposure & Risk, Environmental Statistics, atmospheric, risk assessment, regional environmental data, remote sensing, EMAP, environmental monitoring, stratospheric ozone, Bayesian space-time model, global environmental data, mortality rates, satellite data, statistical models, climate models, global warming, hierarchical statistical analysis, statistical methods, EOSRelevant Websites:
http://www.stat.ohio-state.edu/~sses
(Program in Spatial Statistics
and Environmental Sciences)
http://www.stat.ohio-state.edu/~sses/research
epa.html
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.