Product Description

Validation is a central issue to future model design in environmental science. This book is the first to provide a critical appraisal of today′s validation needs, capabilities, and required changes in philosophy. It takes examples from four different scales: hillslope and river channel, catchment, regional, and global.

This timely book offers unique, multifaceted coverage of model validation in hydrological science today. Topics covered include calibration procedures, data assimilation, scaling, critical future need in validation, and evidence of field data.
∗ State–of–the–art research book on an important new topic
∗ End–of–section discussion chapters written by leading international researchers

Product Details

• Amazon Sales Rank: #1923401 in Books
• Published on: 2001-04-23
• Original language: English
• Number of items: 1
• Binding: Hardcover
• 512 pages

From the Back Cover
Model Validation is a fundamental issue in modern hydrological science where increased demands for prediction and process understanding has been driven by advances in numerical modelling and environmental legislation. Model Validation: Perspectives in Hydrological Science is the first book to deal with this subject in hydrology and environmental science, as well as in other fields.

Model Validation brings together philosophers, modellers and legal experts to comment on model validation issues and gives an evaluation of how we interpret scientific evidence drived from numerical models. It shows how much issues underpin research across the discipline of hydrological science, and also in legal and philosophical frameworks, by addressing major questions concerning acceptable levels of proof in the area. This book will contribute to research within the discipline, in addition to interdisciplinary studies in environmental science and to the wider debate over the nature of scientific prediction.

This is essential reading for academic researchers, policy makers, and legal experts across the spectrum of environmental science, and of great use for 3rd year undergraduate and postgraduate courses.

Excerpted from Model Validation by M.G Anderson, Paul Bates. Copyright © 2001. Reprinted by permission. All rights reserved.
CHAPTER 2 – KINDS OF MODELS

Adam Morton and Mauricio Suárez
University of Bristol, UK

2.1 The M Word

‘Model’ is a term of the working scientist’s self-explanatory and self-justifying vocabulary. ‘Here is my model of the phenomenon’, it follows from our model that…’, ‘our model does not capture the following aspects of the data, but this is no problem since it is just a model…’, ‘to model the data we have made the following assumptions…’. (We are modelling the river as a drunken snake’, ‘the 300% discrepancy between predicted sedimentation and observation is very satisfactory for a purely theoretical model.’) In some such assertions the word ‘model’ could be replaced with ‘theory’ or ‘hypothesis’ with no big loss of meaning. But in many it could not. There are examples of both in the chapters in this volume, as we argue in section 2.3. When scientists describe their creations as models they often intend to take advantage of some of the following features of models, as opposed to theories:

- Two models can be inconsistent with each other, and both can be good models.
- A model can contradict some aspects of the observed phenomena and not be refuted.
- A model can contain assumptions which there are theoretical reasons to believe to be false.
- A model can contain assumptions which observation shows to be false.

How can we have the features on the second list and also those on the first? How do we get our cake and eat it? It may not be as difficult as it seems, depending on how we interpret the ideas of evaluation (and its variants confirmation and validation or vindication – see Section 2.2) and explanation (and its cousins prediction and derivation). One aspect of modelling builds on the familiar idea of a harmless idealisation, with objects as point masses in Newtonian mechanics, gases as homogeneous fluids or as collections of randomly distributed point masses. It is clear that we can use such idealisations in formulating hypotheses which have explanatory value and can be tested. But some of the false consequences of such hypotheses are not to count as refuting them: those that are the direct results of the idealisation rather than of those aspects of the hypothesis that is intended as a description of the subject matter.

The danger now is that all theories will be models, that the theory-model distinction will collapse. Instead of an interesting category of intellectual constructs called ‘models’ we will have an account of the conditions under which a false consequence does not refute a hypothesis. (This would be worth having, but it could be had without any use of the M word.)

We strongly agree with Bates and Anderson’s arguments in this volume (chapter 13) that models are typically propositional and truth-valued – and thus that they can and must be empirically tested and verified. It is very important to remind ourselves of that. An old philosophical tradition used to take it that models were merely interpretations of theories, and therefore could neither be verified nor falsified empirically, Models could not be tested only theories could. This is a tradition that we oppose, and that we have independently criticised elsewhere. Typically, in putting forward a model, we claim, a scientist (among other things) puts forward a hypothesis – i.e. a set of claims that are subject to empirical scrutiny. But we do not believe that ‘model’ just means ‘theory’ (or ‘hypothesis’ or ‘assumption’). Modelling is a distinct activity from mere theorising: in addition to putting forward hypotheses, in talking of models, scientists usually are signalling that they are!
making some very specific uses of idealising assumptions. The problem is that there are many different such signals they can be sending. Here are some worthwhile non-redundant claims that can be intended when a scientist describes her creation as a model:

a) Models as tamed theories. Often there are theoretical reasons for thinking that the laws governing a certain domain should take a certain form, but the result is a theory that is very hard to work with. The difficulty can take the form of problems in deriving practical consequences or problems deriving the kinds of observable predictions which could be evidence for theory. Then simplifications or idealisations are needed. Fluid mechanics is the obvious example. Newtonian physics conceives of physical systems as complexes of discrete particles, so that when we apply this way of thinking to continua it is not surprising that we get such intractable monsters as the Navier-Stokes equations. (So modelling in hydrology inevitably runs into the hard questions.) A model in this case will be obtained by taking the background theory and changing some of the assumptions that make it badly behaved or hard to understand, hoping that the particular changes will not produce inaccurate results in the particular application that one has in mind.

In fact there are two distinct directions in which one can tame a theory. One direction is towards intelligibility or intuitive understanding. For example, the equations of general relativity permit mind-bogglingly many solutions. So expositions of the theory usually subtly restrict the range of solutions so that only those that are cosmologically plausible or geometrically manageable are considered. The other direction is towards the deduction of observable consequences. This can be motivated by the need to test the background theory or by practical applications. Most of the models in this book that are tamed models are of this second kind, which we shall refer to as theory-based models.

b) Models as analogies with other (real) systems. We model the atom as a planetary system (warily); we model gases as fluids; we model an economy as a collection of independent self-interested perfectly rational agents. In each of these cases we know perfectly well that our assumptions are false. Electrons are much more unlike planets than they are like them, and electrostatic force and gravitation are fundamentally different. Gases are composed of molecules with empty space between them. Economies consist of people of unlimited…