March 20, 2014
Keynote presentation: What Do You Say When Data are Sparse?
George Gray, Ph.D.
Professor of Environmental and Occupational Health, and Director of the Center for Risk Science and Public Health,
George Washington University
Predictions about potential harm to humans or the environment from chemical exposures are subject to considerable qualitative and quantitative uncertainty. The degree of uncertainty depends, to some extent, on the amount of information available to assess risk. Tools exist to make statements about the potential risk of even poorly tested chemicals. Yet in high profile situations, like the recent spill of 4-methylcyclohexanemethalnol into the Elk River in West Virginia, the message that is often communicated is a lack of knowledge when data are sparse. I will explore, with this real life example, the possibilities and pitfalls of making and communicating risk statements while acknowledging uncertainty.
Environmental Risk Assessment
Chair: Peter Jutro, Ph.D., Associated Administrator for Homeland Security (Acting), U.S. Environmental Protection Agency
Stan Blacker, Former EPA Director of Quality Assurance Management and Goodman friend and colleague
Adam Finkel, Ph.D., Executive Director, Program on Regulation, University of Pennsylvania Law School
Dr. Daniel Goodman’s field was academically defined as ecology and population biology, but in reality became decision-making under uncertainty, which he approached from both a scholarly socio-historical and a pragmatic statistical point of view. His specialty was examining a complicated situation, breaking it down into appropriate analytical units, creating an appropriate analytical technique, undertaking the analysis, as then explaining to non-mathematically inclined people the essence of his approach, the reason for it, what it produced, and how to interpret and properly use the results. He accomplished what few others had: educating clients on how to properly interpret the information they had to deal with, how to recognize limits, and then teaching them how to do it themselves in the future.
In this session, we will address three environmentally based issues related to Dr. Goodman’s long involvement with the U.S. Environmental Protection Agency (EPA). First, Stanley Blacker will address his efforts to help develop a sound, defensible statistical framework for the arcane area of Quality Assurance (QA). This work, which he began for EPA, was the basis for subsequent work helping design cleanup of the Hanford Reservation for the Department of Energy (DOE). Second, in the context of EPA’s mission related to decontamination following almost any incident involving chemical, biological, or radiological contamination, Peter Jutro will discuss Dr. Goodman’s contributions to EPA and to the Department of Homeland Security’s (DHS) Advisory committee for the design of Biowatch, in which Dr. Goodman helped DHS and EPA both understand and address the Government Accountability Office’s (GAO) unease with the government’s approach to directed versus statistical sampling design for understanding the distribution of contamination. Finally, Adam Finkel will discuss three themes that permeated Dr. Goodman’s work in the area of risk assessment: the importance of quantifying uncertainty in order to make sense of risky choices; the folly of single-disciplinary approaches to complicated technical and societal problems; and the potential for analysis to become a much more useful force, if it is grounded in the comparative weighing of solutions to problems.
Jay Barlow, Ph.D., Leader of the EEZ Mammals and Acoustics Program, Southwest Fisheries Science Center, NOAA Fisheries
Eric Ward, Ph.D., Statistician, Northwest Fisheries Science Center, NOAA Fisheries
Rebecca Taylor, Ph.D., Research Statistician, Alaska Science Center, USGS
Lisa Schwarz, Ph.D., Assistant Research Biologist, EEB/Long Marine Lab, University of California, Santa Cruz
Dr. Daniel Goodman made great contributions to marine mammal conservation by revealing how demography accounts for much of what is needed to make sound management decisions when faced with uncertainty. We illustrate Dr. Goodman’s influence on marine mammal conservation with a series of case studies by his former students that use Dr. Goodman’s own research and build upon it. Barlow and Boveng developed a method to model age-specific mortality based upon general mammalian mortality patterns but applicable to the types and quantities of data that are typically available for marine mammal populations. B. Taylor summarizes similar work to develop a simple model allowing estimation of generation length for most cetaceans (whales, dolphins and porpoises) for use in the IUCN redlist categories. Ward demonstrates the importance of demographic variation alongside model uncertainty and environmental stochasticity for a small population of killer whales. R. Taylor and Udevitz combined a demographic model with multiple data types to estimate walrus population trend when population size data alone were too uncertain to be reliable. Finally, Schwarz shows how starting age structure and uncertainty in juvenile survival rate affect population projections of Antarctic fur seals, showing leopard seal predation is a recent but important driver in fur seal population dynamics. Although each approach differs, with some using likelihood, others a set of rules allowing use of parameters from similar species and others a Bayesian framework, they all contribute to marine mammal conservation using the basic principles of demography. The set of case studies also demonstrate the breadth of conservation applications from the IUCN redlist, to very species-specific problems and finally to ecological questions.
Milo Adkison, Ph.D., Professor, University of Alaska Fairbanks
Randall Peterman, Ph.D., Professor Emeritus, Simon Fraser University
In this session we will examine some of Dr. Daniel Goodman’s work to show how science, management and conservation are linked, and how explicit recognition of this linkage can improve management and conservation, and how ignoring it can lead to poor results (or, as Dr. Goodman would say, “these activities usually work about as well as a train wreck”). Stanford will start with a short perspective on his years of work with Dr. Goodman on scientific review panels. Stanford also will show how Dr. Goodman helped demonstrate the trophic cascade that occurred in Flathead Lake as a consequence of the type 1 error by management when they introduced a strong interactor into the lake food web and how this big mistake has caused the management and policy train wreck that is ongoing. Milo Adkison will describe work that Dr. Goodman did in Alaska adding great insight into the pitfalls of stock-recruitment management of salmon. Randall Peterman will follow Milo with a very informed presentation how lack of quantitative attention to uncertainties in fisheries data and models can vastly complicate and often prevent successful conservation. Throughout the session the need for scientific rigor that Dr. Goodman always
Integrated Ecosystem Modeling and Fisheries Management Strategy Evaluation: Where are We and What is the Future?
Chair: Andre Punt, Ph.D., Associate Director of the School of Aquatic and Fishery Sciences, University of Washington
Francis Wiese, Ph.D., Former Science Director, North Pacific Research Board
Ivonne Ortiz, Ph.D., Research Scientist, University of Washington/Alaska Fisheries Science Center, NOAA Fisheries
George Hunt, Ph.D., Research Professor, School of Aquatic and Fishery Science, University of Washington
Traditionally, the advice provided to fishery managers has focused on the trade-offs between short- and long-term yields, and between future resource size and expected future catches. The control rules that are used to provide management advice consequently relate catches to stock biomass levels expressed relative to reference biomass levels. There are, however, additional trade-offs to consider when selecting fishery catch and effort controls. Ecosystem Based Fisheries Management (EBFM) aims to consider fish and fisheries in their ecological context, taking into account physical, biological, economic, and social factors. However, making EBFM operational remains a challenge. It is generally recognized that Integrated Ecosystem Modeling (IEM) should be key part of implementing EBFM, along with control rules which utilize information in addition to estimates of stock biomass. Dr. Daniel Goodman was central to identifying how IEM can be used to develop such control rules. Here we outline the process for selecting among alternative control rules in an ecosystem context and summarize a modeling effort funded by the National Science Foundation and the North Pacific Research Board which could implement this process. A key component of the Bering Sea Project is a vertically integrated group of models which include a management strategy evaluation component to compare alternative harvest control rules. Effective use of IEM requires that the models developed for a system are indeed integrated. We summarize the steps taken by the program managers to ensure integration of modeling efforts by multiple investigators. We also show how the results of ongoing monitoring efforts can be integrated into models and illustrate how the modeling efforts were designed to be able to address management needs. Finally, we highlight the lessons learned during the Bering Sea Project that can be used to guide future use of IEM.
March 21, 2014
Chairs: Jim Berkson, Ph.D., Unit Leader and Associate Professor, NOAA Fisheries RTR Unit at the University of Florida
Gary Lopez, Ph.D., Executive Director, Monterey Institute for Technology and Education
While mentoring benefits both the mentor and the mentee through enhanced personal and professional development, many professors merely advise their graduate students. Mentors develop a deeper, more personal relationship than advisors, advancing both the educational and personal growth of their students. Hallmarks of an effective mentoring relationship involve mutual trust and strong communication. In science, as in other disciplines, a mentor often guides a student to a way of problem solving, a way of thinking. As scientists, how we we’re taught to think was molded by the scientific process. We strive to understand the natural world by observation, experimentation, data gathering, and analysis. Does this way of thinking have value outside of scientific inquiry? This session will discuss two topics relating to mentoring: (1) a brief review of the definition of mentoring, its benefits, and how it’s accomplished, and (2) the value of thinking like a scientist for endeavors such as business, design, and development. Specific tenants and approaches that were emphasized by Dr. Daniel Goodman will be highlighted with real world examples.
Eric Ward, Ph.D., Statistician, Northwest Fisheries Science Center, NOAA Fisheries
Rebecca Taylor, Ph.D., Research Statistician, Alaska Science Center, USGS
Lisa Schwarz, Ph.D., Post-doctoral researcher, EEB/Long Marine Lab, University of California, Santa Cruz
In Bayesian statistics, in contrast to classical frequentist statistics, probability is used to represent uncertainty about the state of knowledge. Bayesian statistics has often been associated with the idea that knowledge is subjective and that a probability distribution represents a personal degree of belief. Dr. Daniel Goodman considered this viewpoint problematic for issues of public policy. He sought to ground his Bayesian approach in data, and advocated the construction of a prior as an empirical histogram of “similar” cases. In this way, the posterior combined comparable previous data with case-specific current data, using Bayes’ formula. Dr. Goodman championed such a data-based approach, but he acknowledged that it was difficult in practice. If based on a true representation of our knowledge and uncertainty, Dr. Goodman argued that the decision process could be an exact science, despite the uncertainties. Bayesian statistics is a critical component of this science because a Bayesian analysis produces the probabilities of future outcomes. Indeed, Dr. Goodman maintained that the Bayesian machinery, following the rules of conditional probability, offered the best legitimate inference from available data. We give an example of a data-driven prior in a recent study of patterns of spatial use by Steller sea lions in Alaska.
Holly Doremus, J.D., Ph.D., James H. House and Hiram H. Hurd Professor of Environmental Regulation, Co-Director of the Center for Law, Energy & the Environment, Director, Environmental Law Program, University of California, Berkley
David Policansky, Ph.D., Research Scholar, Board on Environmental Studies and Toxicology, National Research Council
Dr. Daniel Goodman devoted much of his career to the enhancement of science, risk assessment, and decision-making under the Endangered Species Act. This section of the symposium will focus on three related topics of particular interest to Dr. Goodman. First, Professor Holly Doremus (University of California, Berkley) will describe the need for more structured decision-making under the Act. Dr. Goodman had long emphasized the need to integrate science and management decisions into decision frameworks that focused the science on critical issues and reduced – to the extent possible – the subjectivity in risk assessment and decision-making. Second, Dr. David Policansky (National Research Council) will describe Dr. Goodman’s involvement in a recent review of ecological risk assessment under the Endangered Species Act and the Federal Insecticide, Fungicide, and Rodenticide Act. Although Dr. Goodman died before the review report was finalized, he had disagreed with the review committee on several points and prepared a dissenting statement. Dr. Policansky will describe some of the issues of contention, as well as Dr. Goodman’s clarity of communication and purpose, and the importance of separating policy judgments from scientific ones. Third, Dr. Tim Ragen (Marine Mammal Commission, retired) will discuss Dr. Goodman’s contributions to Endangered Species Act implementation including listing decisions, recovery teams, recovery plans, and Section 7 consultations.
Chairs: Gina Himes Boor, Ph.D., Assistant Research Professor, Department of Ecology, Montana State University
Paul Wade, Ph.D., Research Biologist, National Marine Mammal Lab, Alaska Fisheries Science Center, NOAA Fisheries
Grant Thompson, Ph.D., Research Fisheries Biologist, Alaska Fisheries Science Center, NOAA Fisheries
The conservation and management of wildlife and other natural resources, at its fundamental level, involves making decisions. The application of science to the ecological decision-making process was something that Dr. Daniel Goodman thought deeply about. Over his career he worked in many areas that can be thought of as ecological decision-making, including the management of fish and wildlife harvests, the remediation of toxic waste sites, and dam removal. However, Dr. Goodman was particularly active in the science of threatened and endangered species management, including listing decisions, recovery planning, and delisting decisions. Our paper will focus on this latter topic, though much of the thinking is relevant to ecological decision-making in other contexts. We discuss the three main principles that Dr. Goodman espoused for good practice in ecological decision-making: 1) the results should be conditioned on all the data, 2) there must be a full characterization of all uncertainty, and it should be fully propagated into the result, and 3) doing so in the correct way results in a calculation of an accurate probability distribution for the state of nature (in this case, a probability of extinction). It then follows that the calculated probability distribution should be directly used for ecological decision-making. We discuss the practical implications of these principles in terms of how extinction risk should be defined and how Population Viability Analysis and Bayesian statistics are the most logical tools to make management and listing decisions about threatened and endangered species. To illustrate, we highlight two examples of the application of these tools and principles, a Bayesian PVA that Dr. Goodman produced for the Steller sea lion recovery plan and a Bayesian PVA of the Cook Inlet beluga whale population. Finally, we describe a specific methodology for incorporating such PVAs into a risk assessment framework.
Professor Anil Nerode, the Goldwin Smith Professor of Mathematics at Cornell University and a long-term friend and colleague of Dr. Daniel Goodman, will help integrate the topics discussed at the symposium. He will provide a global view of Dr. Goodman’s contributions to environmental science and policy and will describe Dr. Goodman’s special scientific talent for constructing testable statistical models to assess the consequences of environmental policies, independent of any ideology. Professor Nerode’s synthesis will provide an important basis for the panel discussion following his presentation.