**ENVR
765**

**Space/Time exposure mapping and risk assessment**

Spring of even years, 3 semester hours

Instructor : Marc Serre

**Course description: **

*Using environmental monitoring data and health surveillance
observations collected at sparse locations over space and time, how can we
assess the exposure at unmonitored points, assess the associated human health
risk across space and time, or directly construct disease space/time maps?*

These questions arise in many environmental and health
fields, such as **environmental risk
analysis**, **environmental epidemiology**,
and **medical geography**. These are exactly the questions that we will
visit in this course using a space/time exposure mapping and risk assessment
framework, and it’s implementation in the MATLAB^{©}
programming language. This framework
will consider the following issues

- How do environmental and health variables vary across space time, or more specifically, how can we quantify their space/time variability?

- What are sources of uncertainty associated with environmental and health data, such as analytical measurement error, secondary (satellite) variables, observation scale, sampling variability of health surveillance data, etc.?

- How can we combine data from multiple sources in a way that rigorously accounts for their uncertainty and space/time variability, and lead to the most accurate estimation at un-sampled space/time location?

- Using exposure maps and exposure/effect slope factors from the EPA integrated risk information system (IRIS), how can we assess risks on human health and characterize the associated uncertainty?

- Alternatively, how do we directly construct disease maps from space/time health surveillance count data?

Each of these issues will be addressed using space/time
statistics as well as computational lab applications (using MATLAB^{©}
programming) of space/time exposure mapping and risk assessment. Environmental
monitoring data arises when there is a need to evaluate and control the source
of exposure to some environmental contaminant, or when there is an interest to
model the relationship between exposure and a suspected health effect. The introduction to space/time statistics
covered in this course will allow students to create exposure maps from
space/time monitoring data. In an
epidemiologic context of exposure/health association analysis, we will see how
these maps are useful to provide exposure estimates at the unsampled location
and times of measured health outcomes.
When the exposure/health response curve can be constructed using published
slope factors (such as those of EPA IRIS), we will investigate how exposure
maps and the exposure/health response curve can be used to assess risks on
human health and characterize the associated uncertainty, which has interesting
policy implications. On the other hand, if health surveillance count data is
available, we will look at how to map disease rates using an estimation
framework that addresses the small count number problem.

The course is organized around 3 main themes: spatial statistics and risk assessment theory, computer programming, and a real-world application. The theory will include space/time random fields, linear and Bayesian geostatistics, and stochastic risk assessment. The computer programming will be done using the MATLAB language and spatiotemporal geostatistics BMElib package written in MATLAB. Writing their own code in MATLAB, the students will learn how to use the BMElib package for the exploratory visualization of space/time data, for the modeling of spatiotemporal variability, and to implement geostatistical estimation so as to create exposure, risk and disease maps. The aim of this course is to prepare its participants to use this toolbox and its concepts to analyze their own dataset in the context of exposure mapping and risk assessment, or of disease mapping.

As an example of application problem, we will consider throughout
the course the exposure mapping and risk assessment analysis of air pollutants
(particulate matter, ozone, lead, etc.) across a geographical region of
interest (such as the United States).
This problem will illustrate all the steps of the exposure mapping and
risk assessment framework, from analysis of space/time variability, modeling of
data uncertainty, exposure mapping, integration of
dose/response curves to estimate human health risk across space and time, and
finally incorporation of population density to arrive at a population health
impact assessment. Other environmental
and health variables will also be considered, such as environmental
carcinogens, water quality parameters, and health outcome variables including
asthma and HIV prevalence in

**Possible class topics**

*Space/time random fields: *

· Random variables,

· Random fields,

· Space/time variability,

· Covariance function

*Linear geostatistics: *

· The Best Linear Unbiased Estimator (BLUE),

· Simple kriging,

· Ordinary kriging,

· Universal kriging

*Bayesian geostatistics: *

· Bayesian Maximum Entropy (BME) theory,

· Bayesian Hierarchical modeling,

· Bayesian Melding

*Stochastic risk assessment: *

· The space/time risk assessment framework

· The EPA integrated risk information system (IRIS)

· Combining random variables (e.g. exposure and slope factors)

*Disease mapping: *

· The small count number problem

· Poisson kriging

· BME approaches to disease mapping

**Textbooks: **

Recommended:

George Christakos, Patrick
Bogaert, and Marc Serre
(2001) Temporal GIS: Advanced Functions for Field-Based Applications,

Peter Diggle and Paulo Ribeiro Jr (2007), Model-Based Geostatistics, Springer Series in Statistics, 230 p.

Also very useful:

George Christakos, Ricardo Olea, Marc Serre, Hwa-Lung Yu and Linlin Wang (2005) Interdisciplinary Public
Health Reasoning and Epidemic Modelling : The Case of Black Death.

Sudipto Banerjee, Bradley Carlin, and Alan Gelfand (2004) Hierarchical Modeling and Analysis for Spatial Data, Chapman & Hall/CRC, 452 p.

Lance Waller and Carol Gotway (2004) Applied Spatial Statistics for Public Health Data, Wiley, 494 p.

**Prerequisite: **

Multivariate
Calculus (MATH 233 or equivalent)

Linear
Algebra (MATH 547 or equivalent)

At least
one graduate level course in biostatistics, statistics or probability

Experience with computer
programming, or a strong appetite for using this course as a course on MATLAB
programming

**Philosophy and
grading: **

The students should learn the concepts of spatial statistics and risk assessment, they should learn how to implement these concepts in the MATLAB programming language, and they should learn how to apply them on a real world exposure mapping and risk assessment project. The students will do homework, an in-class exam, and a final project, which will approximately count for the final grade as follow:

Homework 60%

In-class exam and final project 40%