GeoLOGIC

GeoLOGIC: Visualizing Stratigraphic Knowledge and Reasoning

Overview

The ability of geologists to derive accurate conclusions from stratigraphic data is limited by the presence of systematic and conceptual uncertainty in that data. This project applies methods from logic and philosophy of science for classifying and understanding uncertainty, in order to improve the explanatory and predictive capabilities of geology. The benefits include a better understanding of stratigraphy, more accurate prediction of groundwater and hydrocarbon deposits, better understanding of long-term processes such as global climate change, and extending philosophers' analytical tools to an area of science which has previously received little attention. The techniques developed in this project for improving the explanatory and predictive power of scientific models in situations involving evidence that is inherently limited in scope and precision are also of use in other historical sciences such as evolutionary biology and astrophysics. This project lays the groundwork for the future development of collaborative research to be funded by national granting agencies.

Project Overview

To date very little has been done to try to introduce techniques for representing uncertainty in geology. Our aim is to fill this gap. We will first study and categorize the kinds of uncertainty in geology, and will then apply many-valued logic popularly known as "fuzzy logic" to represent uncertainty in computer models and visual representations of geological stratification.

The study of the relationships between rock layers, known as stratigraphy, is a foundational subdiscipline within geology. As a "historical" rather than an "experimental" science (to invoke a distinction employed by philosophers of science), stratigraphy often uses data that are conceptual and linguistic in nature, rather than numerical and exact. These data usually do not contain information regarding degrees of uncertainty. This is a major drawback: without a way to integrate and understand uncertainty in measurements and results, stratigraphic analyses are incomplete at best and can lead to inaccurate conclusions. A method is needed that makes scientists cognizant of the locations and degrees of the uncertainties in their data, so that they can make better-informed analyses and decisions. Such a method will be of practical as well as scientific importance, since it should enable better predictions about the sizes and locations of ground water and hydrocarbon resources.

Philosophers of science and logicians have devoted themselves to understanding reasoning under uncertainty, but they have done almost no work on these issues in the context of geology. This project will therefore also bring new and interesting material to the notice of philosophers.

The research team, composed of faculty from the Departments of Geology, Philosophy, and Computer Science, are surveying various types of uncertainty in stratigraphic data sets and developing new methods for representing information about uncertainty in computer models and visual media. Although it is not an immediate goal of the project, the techniques we develop should be applicable in other historical sciences such as evolutionary biology and astrophysics, where scientists try to infer the historical development of a system given only patchy data about current states. This project lays the groundwork for the future development of collaborative research to be funded by national granting agencies.

Parcell (Geology) contributes expertise in stratigraphy, and experience constructing computer models of geological formations and processes. He will also provides the field data (obtained under earlier grants) to be used to construct the computer models for this project. Vanderburgh (Philosophy) contributes expertise in logic and philosophy of science. In addition to the general philosopher's skill of conceptual analysis, Vanderburgh brings specific experience studying scientific methodology and evidential reasoning in science (especially astrophysics, another historical science). Radebaugh (Philosophy and Computer Science) contributes analytical and programming skills and experience.

Importance of Project

According to the traditional view (often attributed to Newton but in fact much older), uncertainty in science is radically undesirable. The Latin word "scientia" means "knowledge in the strictest sense." Ancient and Renaissance thinkers and even the Logical Empiricists in the early twentieth century interpreted this to mean that the ideal for scientific knowledge is the sort of perfect deductive-axiomatic system found in Euclid's geometry. On this view, if there is any room for doubt, there is no knowledge. Eliminating uncertainty from science is seen as a kind of epistemic progress (Demicco and Klir 2004).

This traditional view has been overthrown. Reducing uncertainty is still an important goal, but it is now recognized that uncertainty cannot be entirely eliminated from scientific reasoning. In disciplines as diverse as physics, medicine and sociology, statistical and other techniques have been developed to help deal with uncertainty due to measurement error, sampling problems, incomplete background knowledge, and other sources. These techniques do the important work of specifying the bounds of uncertainty in a given data set and in inferences drawn from it. In such cases the conclusions of the scientific studies are not perfectly certain, but these techniques for dealing with uncertainty nevertheless give us a clearer understanding of the degree to which those conclusions are uncertain, and in what ways. This both improves our knowledge and enables better-informed decision making on the basis of the scientific results.

Stratigraphy is fundamental to modern studies in the earth sciences because it provides a chronological and environmental framework for studying global climate change and evolutionary trends through time. It also provides a tool for predicting groundwater resources and hydrocarbon deposits. However, no formal system has yet been developed to represent or visualize uncertainty in the stratigraphic sciences. This project will fill this important gap, thus providing resources for improving the explanatory and predictive capabilities of geology.

Methodology/Approach

The research team draws on the resources of several related fields to develop techniques for representing uncertainty in geology. In recent years information science has worked out a theoretical classification that incorporates uncertainty due to nonspecificity, vagueness, dissonance, and confusion (McNeill and Freiberger 1993). Mathematical and logical analyses of uncertainty are well established. These include evidence theory, possibility theory, classical and fuzzy set theory, and probability theory (Klir 1999). Philosophers of science have also devoted considerable attention to evidence theory, set theory, probability theory and multi-valued logics. A major focus of attention in philosophy of science is on scientific method, evidential reasoning and the problem of choosing between competing theories all of which hare examples of the general issue of how best to ensure that inferences from incomplete and/or error-laden evidence are reliable. This project examines these various classifications and measures in order to adapt them to the data types and conceptual methods found in stratigraphy.

This project is led by two WSU investigators: Dr. William Parcell (stratigraphy) from the Department of Geology and Dr. William Vanderburgh (philosophy of science and logic) from the Department of Philosophy. Additionally, Dr. Day Radebaugh (cognitive science, philosophy, and computer science), a Visiting Assistant Professor jointly appointed between the Departments of Philosophy and Computer Science, acts as technical assistant to the project. Expertise in logic and philosophy of science is critical in order to understand the origins of empirical uncertainty and conceptual uncertainty in stratigraphy. Additionally, many stratigraphic measurements are based on "linguistic" (rather than "numerical") representations of geological information. Therefore, the analysis of stratigraphic uncertainty also requires expertise in cognitive science.

Methods Description

This project has two main components. First, we are developing a classification of the various types of uncertainty found in the stratigraphic sciences and analyze the best methods for representing them. Second, we are beginning the development of both computer models and visual (2D and 3D) representations of uncertainty in stratigraphic data sets.

There are various kinds of uncertainty in science. Uncertainty emerges at the empirical level (e.g., measurement error, sampling error), the cognitive level (e.g., vagueness and ambiguity in natural language), and the conceptual level (e.g., uncertainty regarding relationships between data). Stratigraphy is laden with all these various types of uncertainty. For example, studying preserved 'fossilized' records of what occurred millions or even billions of years ago requires an interpretation that is fraught with uncertainty. An inherent problem is that the event that formed a set of deposits cannot be directly observed. Other varieties of uncertainty arise in geology due to human classification of naturally gradational systems. (It is worth noting that some of the misunderstanding amongst the general public regarding the evidence for biological evolution may have its basis in the poor representation of uncertainty in stratigraphic measurements and concepts. For when doubt is cast on a stratigraphic model, the lay person may take this to undermine the whole science, when in fact the actual uncertainty is small and is in the details, not the broad strokes.)

Stage 1 requires a representation of the various data types used in stratigraphy. Parcell is providing physical (lithostratigraphic), biological (biostratigraphic), and chemical (chemostratigraphic) data from outcrop and the subsurface control points derived from his research in ancient rocks of the Western U.S. and the Gulf of Mexico. Also in Stage 1, the analytical and philosophical skills of Vanderburgh and Radebaugh are used to examine the kinds and sources of uncertainty associated with various data types and conceptual methods. Stage 2 requires development of methods to graphically represent uncertainty. Parcell and Radebaugh will combine their expertise in stratigraphy and computer science to develop these methods.

Significance

To date, no formal system has been developed to represent or visualize uncertainty in the stratigraphic sciences. Without the integration of uncertainty in measurements and results, the analysis of these data is incomplete at best and often leads to inaccurate or incorrect conclusions. A method is needed to represent uncertainty so that scientists are made aware of the locations and degrees of uncertainties in their data. This will also have practical benefits in that it should facilitate better-informed analyses and decisions that depend on geological data. Improving the interpretation of geological data in this way should, for example, facilitate more reliable and more efficient discovery of ground water and hydrocarbon deposits, something that would clearly be of major benefit to the national and global economy (and even to national security in that it would reduce reliance on foreign sources). This project will develop the framework for future research concerning computer models and visual representation of uncertainty in stratigraphic data sets. Philosophers of science and logicians have devoted themselves to understanding reasoning under uncertainty, but they have done almost no work on these issues in the context of geology. This project will therefore also bring new and interesting material to the notice of philosophers.