1 Modelling and Model Building MARK MULLIGAN AND JOHN WAINWRIGHT Modelling is like sin. Once you begin with one form improvement or optimization of these systems through of it you are pushed to others. In fact, as with sin, the development of new techniques, processes, materials once you begin with one form you ought to consider and protocols. other forms. ... But unlike sin – or at any rate unlike Research is traditionally carried out through the sin as a moral purist conceives of it – modelling is accumulation of observations of systems and system the best reaction to the situation in which we find behaviour under ‘natural’ circumstances and during ourselves. Given the meagreness of our intelligence experimental manipulation. These observations provide in comparison with the complexity and subtlety of the evidence upon which hypotheses can be generated nature, if we want to say things which are true, as about the structure and operation (function) of the well as things which are useful and things which are systems. These hypotheses can be tested against new testable, then we had better relate our bids for truth, observations and, where they prove to be reliable application and testability in some fairly sophisticated descriptors of the system or system behaviour, then ways. This is what modelling does. they may eventually gain recognition as tested theory (Morton and Suarez,´ 2001: 14) or general law. The conditions, which are required to facilitate research, include: 1.1 THE ROLE OF MODELLING IN ENVIRONMENTAL RESEARCH 1. a means of observation and comparative observation (measurement); 1.1.1 The nature of research 2. a means of controlling or forcing aspects of the system (experimentation); Research is a means of improvement through under- 3. an understanding of previous research and the state standing. This improvement may be personal, but it may of knowledge (context); also be tied to broader human development. We may 4. a means of cross-referencing and connecting threads hope to improve human health and well-being through of 1, 2 and 3 (imagination). research into diseases such as cancer and heart disease. We may wish to improve the design of bridges or aircraft through research in materials science, which provides 1.1.2 A model for environmental research lighter, stronger, longer-lasting or cheaper bridge struc- tures (in terms of building and of maintenance). We What do we mean by the term model?Amodelis may wish to produce more or better crops with fewer an abstraction of reality. This abstraction represents a adverse impacts on the environment through research in complex reality in the simplest way that is adequate biotechnology. In all of these cases, research provides for the purpose of the modelling. The best model is in the first instance better understanding of how things always that which achieves the greatest realism (mea- are and how they work, which can then contribute to the sured objectively as agreement between model outputs Environmental Modelling: Finding Simplicity in Complexity.EditedbyJ.WainwrightandM.Mulligan © 2004 John Wiley & Sons, Ltd ISBNs: 0-471-49617-0 (HB); 0-471-49618-9 (PB) 8 Mark Mulligan and John Wainwright and real-world observations, or less objectively as the more data, which suits managers and politicians as process insight gained from the model) with the least well as staff scientists and professors to whom it means parameter complexity and the least model complexity. more publications per unit time and thus an easier Parsimony (using no more complex a model or passage of the hurdles of annual evaluations and representation of reality than is absolutely necessary) has other paper-counting rituals. And it is more glamorous been a guiding principle in scientific investigations since to polish mathematical equations (even bad ones) in Aristotle who claimed: ‘It is the mark of an instructed the office than muddied boots (even good ones) in mind to rest satisfied with the degree of precision which the field. the nature of the subject permits and not to seek an (Klemes,ˇ 1997: 48) exactness where only an approximation of the truth is possible’ though it was particularly strong in medieval Amodelisanabstractionofarealsystem,itisa times and was enunciated then by William of Ockham, simplification in which only those components which in his famous ‘razor’ (Lark, 2001). Newton stated it as are seen to be significant to the problem at hand the first of his principles for fruitful scientific research are represented in the model. In this, a model takes in Principia as: ‘We are to admit no more causes of influence from aspects of the real system and aspects natural things than such as are both true and sufficient of the modeller’s perception of the system and its to explain their appearances.’ importance to the problem at hand. Modelling supports Parsimony is a prerequisite for scientific explanation, in the conceptualization and exploration of the behaviour not an indication that nature operates on the basis of of objects or processes and their interaction as a parsimonious principles. It is an important principle in means of better understanding these and generating fields as far apart as taxonomy and biochemistry and hypotheses concerning them. Modelling also supports is fundamental to likelihood and Bayesian approaches the development of (numerical) experiments in which of statistical inference. In a modelling context, a par- hypotheses can be tested and outcomes predicted. In simonious model is usually the one with the greatest science, understanding is the goal and models serve as explanation or predictive power and the least parameters tools towards that end (Baker, 1998). or process complexity. It is a particularly important prin- Cross and Moscardini (1985: 22) describe modelling ciple in modelling since our ability to model complexity as ‘an art with a rational basis which requires the use of is much greater than our ability to provide the data to common sense at least as much as mathematical exper- parameterize, calibrate and validate those same mod- tise’. Modelling is described as an art because it involves els. Scientific explanations must be both relevant and experience and intuition as well as the development of a testable. Unvalidated models are no better than untested set of (mathematical) skills. Cross and Moscardini argue hypotheses. If the application of the principle of parsi- that intuition and the resulting insight are the factors mony facilitates validation, then it also facilitates utility which distinguish good modellers from mediocre ones. of models. Intuition (or imagination) cannot be taught and comes from the experience of designing, building and using models. Tackling some of the modelling problems pre- 1.1.3 The nature of modelling sented on the website which complements this book will Modelling is not an alternative to observation but, help in this. under certain circumstances, can be a powerful tool in understanding observations and in developing and 1.1.4 Researching environmental systems testing theory. Observation will always be closer to truth and must remain the most important component Modelling has grown significantly as a research activity of scientific investigation. Klemes(1997:48)describesˇ since the 1950s, reflecting conceptual developments the forces at work in putting the modelling ‘cart’ before in the modelling techniques themselves, technological the observation ‘horse’ as is sometimes apparent in developments in computation, scientific developments modelling studies: in response to the increased need to study systems (especially environmental ones) in an integrated manner, It is easier and more fun to play with a computer than and an increased demand for extrapolation (especially to face the rigors of fieldwork especially hydrologic prediction) in space and time. fieldwork, which is usually most intensive during the Modelling has become one of the most power- most adverse conditions. It is faster to get a result ful tools in the workshop of environmental scien- by modeling than through acquisition and analysis of tists who are charged with better understanding the Modelling and Model Building 9 interactions between the environment, ecosystems and and nonliving (abiotic) systems and their interaction. the populations of humans and other animals. This Complexity increases greatly as the number of com- understanding is increasingly important in environ- ponents increases, where their interactions are also mental stewardship (monitoring and management) and taken into account. Since the human mind has some the development of increasingly sustainable means of considerable difficulty in dealing with chains of human dependency on environmental systems. causality with more than a few links, to an environ- Environmental systems are, of course, the same sys- mental scientist models are an important means of tems as those studied by physicists, chemists and biolo- breaking systems into intellectually manageable com- gists but the level of abstraction of the environmental ponents and combining them and making explicit the scientist is very different from many of these scien- interactions between them. tists. Whereas a physicist might study the behaviour of Nonlaboratory controllable.Theluxuryofcontrolled gases, liquids or solids under controlled conditions of • conditions under which to test the impact of individual temperature or pressure and a chemist might study the forcing factors on the behaviour of the study system interaction of molecules in aqueous solution, a