[CRC Press] How are spatio-temporal modelling and environmental epidemiology connected?
[Gavin Shaddick] The London fog of 1952 was a landmark in the history of environmental epidemiology—an acute increase in air pollution levels inarguably led to thousands of deaths in the days following that episode. From this incident flowed the widespread recognition that air pollution was an environmental health risk and, with that, regulations intended to control it. This led to the time series approach to detecting acute health effects driven by hazards such as air pollution—where sharp increases in the air pollution are followed by increases in deaths or hospital admissions for things like asthma. This can provide the basis for possible causal connections. At this point, a relationship between statistical time series modelling and environmental epidemiology was established.
[Jim Zidek] It did not take long to figure out that there were also blips across space in the levels of these hazards to augment those in time. Adding these contrasts in the levels would add power to the time series approach. That led to the need to model these environmental processes over both space and time. Numerous studies have now been done based on such models.
[GS] These methods were the basis for accurately assessing the adverse health effects of environmental hazards due to such things as the high levels of atmospheric air pollution due to an increasing world population. So great is the concern about this phenomenon that when the results from a task force I led for the World Health Organization (in which Jim was involved) were released, showing that 92% of the world’s population lives in places where air quality levels exceed WHO limits, the findings were picked up and reported by more than 500 news outlets within 48 hours of the WHO news release.
[CRC] Why is your book necessary?
[GS] Few books are available for students and researchers in the rapidly evolving field of spatio-temporal modelling for environmental health risk analysis. There are a lot of key topics that we feel are of fundamental importance but which have not been included in other books.
[JZ] We also do not see other books that provide a bridge between spatio-temporal theory and environmental epidemiology that is so important in allowing statisticians, epidemiologists, public health researchers and policy makers to access state of the art methodology. Our motivation was inspired in part when we learned that statistics students were finding employment in schools of public health, even though they had no formal training in environmental epidemiology. We also found non-statisticians enrolling in short courses along with the one full course we gave on the subjects of the book, due to their need to understand and use the methods we presented. One of these people was a government employee in Mexico who needed to map arsenic levels in ground water, which, when consumed, led to illnesses and hospitalization.
[GS] We believe our book to be a unique contribution to students and researchers in environmental epidemiology, statistics, public policy and data science, in that it is up to date and includes key features not seen in any other books that are available in this field.
[CRC] What are some of those features?
[JZ] The book includes information on visualization, for example. We show how to draw Google Earth images with tacks that mark the spots in Holland where soil samples were taken. In the familiar “street view,” those tacks can be seen at a distance, and clicking on them shows the concentrations of various metals that were found in the sample taken at that site. We talk about the new world of post-normal science where you see great risks as well as great uncertainty. There we see how to handle “qualitative uncertainty.” For example, in an assessment carried out concerning the concentrations of lead coming out of paint used in construction, the quality of various different sorts of data has to be included.
[GS] The examples from that case study, and others seen within the book, are given in the online resources that we developed for the book. A webpage offers readers many other things, including the code and data needed to work the numerous examples given in the book. For instructors, it also provides suggested course structures and notes for delivery of the material to different audiences, such as example epidemiologists, statisticians, mathematicians and data scientists.
[JZ] By way the book introduces students to not only to R, but also to WinBUGS and INLA for performing Bayesian analyses, and an R package called EnviroStat. The latter provides a module for designing networks for monitoring environmental hazards and complements the book’s Chapter 13. EviroStat also includes a module for the method of Sampson and Guttorp for warping geographical to circumvent the problem often encountered in environmental statistics that the random field of the hazard is non-stationary.
[GS] We also include a chapter on emerging frontiers in spatio-temporal modelling and how to incorporate deterministic models within stochastic models of spatio-temporal phenomena.
[CRC] What background does one need to be successful in such an interdisciplinary ﬁeld?
[GS] Technological change has led to the production of very large and often complex databases; the required analysis can be quite challenging. These days, to be successful environmental epidemiologists will need a wide variety of skills, including statistics, data analytics, computation, and data governance, in addition to core epidemiological skills. In the WHO study I mention above, the data included measurements from monitors on the ground, estimates from remote sensing satellites, and simulations from numerical computer models that embraced the transport and creation of atmospheric pollutants. These disparate sources of information needed to be integrated in a coherent fashion and by combining them all, we were even able to infer levels of particulate air pollution in countries where there were no monitors on the ground!
[JZ] At the same time, statisticians should ideally know something about the nature of the population health effects they are studying. Some of the relevant material would be found in biostatistics courses. For example, the relative risk of a one-unit increase in the level of a pollutant sometimes needs to be estimated and traditionally that topic might have been encountered in such a course. However, we have made the book self-contained and the background needed to understand the more complex material is included in the book.
[CRC] What made you become interested in the intersection of environmental epidemiology and spatio-temporal statistics?
[GS] My first job was in an epidemiology department looking at the health risks associated with industrial point sources of pollution. We looked at patterns of risk in relation to distance from things like incinerators and powerlines. Traditionally, epidemiological studies were performed either over time, or over space, often due to the way that data was collected; however, with technological advances and the reduction in the cost of storage, increasingly data became available on a massive scale, recorded over both space and time, often at very high resolutions. In order to fully realize the potential of such data in epidemiological analyses, new methods and computational techniques were required.
[JZ] A long while ago, I was contracted by the Government of Canada to develop a methodology for estimating the population levels of exposure to air pollutants, notably carbon monoxide and ozone. This naturally led me into the world of environmental epidemiology. It also acquainted me with the other side of the same coin, setting regulatory standards for air quality, which is a topic in which I remain very much interested.
[CRC] So Jim, I expect that experience in Canada must have had some role in your appointment to the Environmental Protection Agency’s scientific advisor committee for setting ozone standards, correct? What aspects of this job did you ﬁnd interesting and/or frustrating?
[JZ] You may be right, although I don’t actually know why I was invited to join the panel of 24 scientists from all sorts of different fields, such as toxicology. Anyway, my experience, I think, proved of value to the committee, especially since I had focused earlier in the Canadian study on estimating the levels of personal exposure sustained by a population given the levels of ozone being measured at the maybe ten or twenty ambient monitors in an urban area. Gavin joined me in that type of analysis and we published a couple of papers on it. The committee’s work proved a great learning experience about the relationship of modelling and the setting of regulatory standards.
[GS] The point of the work Jim describes above is that not all potential sources of things like carbon monoxide or radon are outside—some are right in your own home. It is important that environmental health risk analysis is able to incorporate all potential sources of exposures.
[CRC] What types of real-world problems does the book address?
[GS] Jim and I have had a lot of experience in applying the methods described in the book to actual studies of the health effects of environmental pollution. We have noted some examples earlier in this interview of the type of applications to which the methods in the book can be applied. These include a look at levels of the concentrations of nitrogen dioxide over Europe, the relationship between outdoor air pollution and mortality in Great Britain and the relationship between asthma in children and their proximity to the roads. The book also contains other examples, including the infamous London fog of 1952, monitoring ozone in New York state, hospital admissions in the UK, and temperatures in California, amongst many others. In each application addressed in the book, we provide the necessary code and data sources to enable readers to get some hands-on experience. We used a lot of these examples in exercises and courses we gave in the past.
[CRC] How does statistical computation factor into the book?
[GS] Extensively. We see computation as essential to the modern practice of data science that is reflected throughout the book. Code is interwoven into the text and complete code (with data) is given in the online resources. We see code as expressive, in the same way as written language, so we strive to have meaningful interplay between the code, the discussion of how the code works, and the overall story that is being told.
[CRC] Why is this book signiﬁcant to today’s statisticians and environmental epidemiologists?
[JZ] Over the years, we have had a lot of interaction with folks (students, faculty, and professionals) who are interested in environmental health risk analysis. That led to the courses we have given, as well as a lot of the research we have done. This book is our attempt to give a readable, comprehensive account of the knowledge we have gained from our experience and interactions with others.
[GS] Most of the resources that are out there tend to focus on speciﬁc areas. We wanted to provide a comprehensive account that covers the entire spectrum of methods and application in this area, focusing on meaningful applications. However, unless the book was to be thousands of pages, we realized early on that we would have to strike a balance between breadth and depth in order to achieve our aim. The preface describes the different emphases that may be placed on different chapters to reflect the difference in the training required for different audiences. Much of the discussion is tutorial in nature and a modern bibliography is included with reference to an extensive toolbox of methods. We believe the book and the online resources, which provide the computer code needed for the examples, will prove valuable to researchers and consultants. Our hope is that this book can serve as the primary text for the interaction between environmental epidemiology and statistics
Gavin Shaddick is a reader in statistics in the Department of Mathematical Sciences at the University of Bath. He received his master’s in applied stochastic systems from University College London and his PhD in statistics and epidemiology from Imperial College London. His research interests include the theory and application of Bayesian statistics to the areas of spatial epidemiology, environmental health risk, and the modeling of spatio-temporal fields of environmental hazards. Of particular interest are computational techniques that allow the implementation of complex statistical models to real-life applications where the scope over both space and time may be very large. Dr. Shaddick is actively involved in a number of substantive epidemiological projects related to the effects of air pollution to health. He has worked on many large-scale funded projects, including the high-resolution mapping of environmental pollutants, the utilization of information from multiple sources in estimating exposures to environmental hazards, and the characterization of uncertainty in scenario assessment and policy support. He is a co-author of the Oxford Handbook of Epidemiology for Clinicians, which was Highly Commended in the Basis of Medicine Category, BMA Book Awards 2013.
About Jim V. Zidek
James V. Zidek is a professor emeritus in the Department of Statistics at the University of British Columbia. Professor Zidek received his MSc and PhD in statistics from the University of Alberta and Stanford University, respectively. He began his research career working on Wald’s statistical decision theory. That interest shifted into Bayesian decision analysis. His interest in applications also emerged early in his career and as a consultant, published with engineering collaborators, the first design code for long-span bridges, such as the famous Golden Gate Bridge in San Francisco. The combination of theory and practice led him to an EPA project on acid rain where he, with a few of his collaborators, started to lay the foundations of environmetrics as it is now called, notably on the design of environmental monitoring networks and spatio-temporal modeling of environmental processes. That work led naturally into spatio-temporal epidemiology, which remains an area of interest. He has published about 100 refereed articles and a book on modeling environmental processes. His contributions to statistics have been recognized by a number of honors. He is a fellow of the ASA, IMS, and Royal Society of Canada; member of the ISI; and a recipient of the Gold Medal of the Statistical Society of Canada (its highest honor).
"The authors of this text, both accomplished researchers in the area, provide a much-needed consolidation of spatio-temporal modelling methods…The textbook condenses many complex topics into accessible and manageable chapters addressing key elements of modern spatio-temporal analyses of environmental epidemiologic data…The authors provide helpful R examples throughout…Analytic challenges such as missing data, measurement error, and preferential sampling often arise in environmental epidemiology and are each described in detail along with focused data examples and accompanying code…The text covers a remarkable number of topics in its 318 pages (including many full color graphics and examples of code and output). The structure outlined above provides excellent coverage of many areas of recent development, held together with compelling examples and illustrations…Overall, I found the book a comprehensive overview placing many different topics into a logical perspective with focused, helpful examples. I enjoyed reading the book, am already recommending it to colleagues, and anticipate referring to it often in my future work."
—Lance A.Waller, Emory University, The American Statistician, November 2016
Follow CRC Math & Stats on social media:
Join Our Community!
Sign up for email alerts and be the first to know about new books in your area of interest. You will also be eligible for:
Only available in our email and print promotions. Take full advantage of your insider savings and free shipping when you buy directly from routledge.com