Research products from SSES projects are refereed papers and conference
presentations. Sometimes it is appropriate to present the research to
a more general audience using the web. We call these Web-Projects.
The Web-Project ENSO (El Nino Southern
Oscillation) shows long-lead forecasting of sea surface temperature in
the tropical Pacific Ocean. The Web-Projects TCO (Total Column Ozone) and CO2 (Carbon Dioxide) show daily maps, respectively,
of TCO and CO2 values, where each map comes with a second map that visualizes
its uncertainty. The
Web-Project Ice Streams shows the results of Bayesian
hierarchical modeling of ice streams' stress fields and velocity fields.
The Web-Project STB (Sources to Biomarkers) seeks to
characterize multi-pollutant human exposures by linking sources to biomarkers using
a hierarchical Bayesian statistical model.
Tropical Pacific sea surface temperatures (SST) and the accompanying El
Nino Southern Oscillation (ENSO) phenomenon are recognized as
significant components of climate behaviour. The atmospheric and
oceanic processes involved display highly complicated variability over
both space and time. Researchers have applied both physically derived
modeling and statistical approaches to develop long-lead
predictions of tropical Pacific SSTs. The comparative successes of
these two approaches are the subject of substantial inquiry and some
controversy. A new procedure for long-lead forecasting of Pacific SST
fields, that expresses qualitative aspects of scientific paradigms for
SST dynamics in a statistical manner, is presented. The investigators
(Berliner, Wikle, and Cressie) would like to acknowledge the help of
National Center for Atmospheric Research (NCAR) scientists in developing this
procedure. Through its
combining of substantial physical understanding and statistical
modeling and learning, the procedure acquires considerable predictive
skill. Specifically, a Markov model, applied to a low-order (EOF-based)
dynamical system of tropical Pacific SSTs, with a stochastic regime
transition is considered. The approach accounts explicitly for
uncertainty in the formulation of the model, which leads to realistic
error bounds on forecasts. The methodology that makes this possible is
Bayesian hierarchical dynamic (HiDyn) modeling.
This research is featured in an article on The Ohio State University
Research Communications webpage.
The Antarctic ozone-hole event has become a symbol of global ozone
depletion since its discovery in 1985. During the 1990s, the patterns
in the ozone hole during the Antarctic winter were similar from year to
year. Unfortunately, Total Ozone Mapping Spectrometer (TOMS)
ozone datasets are not complete because of
restrictions in sunlight availability, the coverage of satellite
orbits, and other engineering problems. In order to address this
deficiency in spatial coverage, Johannesson and Cressie (2004)
proposed the Multi-resolution Spatial Model (MRSM), which is an
effective statistical method for estimation of spatial processes based
on the change-of-resolution Kalman filter (Chou et al. 1994; Huang et
al. 2002) and variance-covariance likelihood inference.
In September 2002, the ozone hole split, unlike in any previous years
where data were available. Researchers have proposed several theories
to explain the 2002 ozone-hole splitting based on a diverse collection of ozone
datasets. TOMS is one of the most important resources of total column
ozone (TCO) data.
website shows complete TCO estimates based on the TOMS data and the
MRSM, along with a measure of each value's uncertainty. The website also contains brief background descriptions
of the MRSM, the atmospheric ozone distribution, and the 2002 ozone-hole splitting event.
This research is featured in an article on The Ohio
State University Research Communications webpage.
The mass balance and equilibrium state of the polar ice sheets are
complex functions of external climate forcings and internal dynamical
processes. To understand the behavior of vast ice sheets and to
assess future behavior, we seek to understand the dominant forces
controlling ice flow and how these forces have responded and will
respond to changes in climate and external forcings. We study
ice-stream dynamics via a fully Bayesian statistical analysis that
incorporates physical models that are not perfectly known, and using
data that are both incomplete and noisy. The physical-statistical
models we propose account for these uncertainties in a coherent,
hierarchical manner. Use of Bayes' Theorem allows us to make inference
on all unknowns given the data. The result of that inference is a
(posterior) distribution of possible values that can be summarized in
a number of possible ways. For example, the posterior mean of the
stress field gives average behavior at any location in the field, and
the posterior standard deviation associated with a posterior mean
value shows how variable the possible values are. There are no direct
measurements on stress; we infer it from basal-elevation data,
surface-elevation data, and velocity data. Forward smoothing methods
can be used, but their disadvantage is that they lack a coherent accounting of
uncertainties. This Ice Streams website analyzes data from the Northeast Ice Stream in
Greenland and indicates how scientific conclusions may be drawn from
Bayesian analyses. It also includes a Tutorial on Bayesian Statistics for Geophysicists.
An important problem in human-exposure assessment is to characterize links from sources to biomarkers (STB).
In this Web-Project, we use a multi-scale (areal, residential, and personal)
Bayesian hierarchical model (BHM), which describes how multi-media pathways contribute to direct routes
of exposure (inhalation, ingestion, dermal). The statistical-modeling framework coherently
manages and accounts for variability and uncertainty and has explicit stages for sources, areal
environmental levels, indoor (residential) environmental levels, personal exposures, and biomarkers.
The primary data sources are the National Human Exposure Assessment Survey (NHEXAS) Phase I data from EPA
Region 5 (the six Midwest states of Illinois, Indiana, Michigan, Minnesota, Ohio, and Wisconsin) and Arizona, supplemented by census data, ambient-air monitoring data, and emissions data.
NHEXAS data provide information to stages of the model addressing areal and indoor environmental conditions,
personal exposures, and biomarkers. These stages combine this information in a manner akin to structural-equation
modeling to discern pathways and routes of exposure. The results include characterizations of the distribution of
biomarkers across the population as a whole and within subpopulations, as well as the relative contribution to biomarkers from various
pathways and routes of exposure.
There is increasing concern about climate change; monitoring
CO2 (a leading greenhouse gas) has become a crucial scientific
endeavor. Global monitoring relies in large part on satellite instruments. For example, the Orbiting Carbon Observatory-2 (OCO-2) instrument to determine CO2 sources and sinks, will be launched in 2014; and the Atmospheric InfraRed Sounder (AIRS)
on NASA's Aqua satellite is capable of measuring radiances from
which derivation of mid-tropospheric CO2 concentrations can be
obtained. Daily retrievals of CO2 are sparse and incomplete
with respect to the globe as a whole. Hence, statistical methods
are needed that are able to exploit spatial and temporal dependencies
in the data to produce complete global maps of CO2, together
with the associated uncertainty estimates. Such methods need to
be flexible to accommodate differing spatial variability of CO2
around the globe. In addition, they need to allow for computationally
feasible statistical inference, for very-large-to-massive datasets.
In this Web-Project, we describe the Spatial Random Effects (SRE)
model and a related optimal spatial prediction method called Fixed
Rank Kriging (FRK). We discuss parameter estimation and spatio-temporal
extensions. The website gives an analysis of spatial data derived
from a CO2 transport model, and it shows results of a spatio-temporal
analysis of mid-tropospheric CO2 from the AIRS instrument and data and associated Matlab code are provided. It also includes a
Tutorial on Fixed Rank Kriging (FRK) of CO2 data.
This research is featured on a NASA webpage devoted to the AIRS instrument.
This Web-Project represents an accounting of temperature change that is projected for North America in 2041-2070. The preponderance of our results throughout all regions of North America is one of warming, usually more than 2°C (3.6°F). Climate models have become the primary tools for scientists to project future climate change and to understand its potential impact. General Circulation Models (GCMs) usually oversimplify the regional climate processes and geophysical features, such as topography and land cover. Since local/regional climate effects are more relevant to natural-resource management and environmental-policy decisions, Regional Climate Models (RCMs) have been developed to produce high-resolution outputs on scales of 20 to 50 km. RCMs can simulate 3-hourly "weather" over long time periods and generate a vast array of outputs, from which long-run averages are commonly used as a summary of how a climate model approximates the Earth's climate. With anthropogenic forcings incorporated, they provide a means to assess natural and anthropogenic influences on climate variability.
In this Web-Project, we consider a subset of the climate-model experiment associated with the North American Regional Climate Change Assessment Program (NARCCAP). Regional Climate Models (RCMs) are run 60 years into the future for small, 50 km x 50 km regions in North America, from which we obtain temperature-change projections for all regions and all four seasons. The statistical framework to analyze the approximately 100,000 data is based on a Bayesian hierarchical spatial analysis of variance (ANOVA) model that incorporates dimension reduction.