Ecological paradigms, change detection and prediction G. Harris CSIRO, Hobart, Tasmania Abstract: In an environment of massive environmental degradation it is very important that we get our basic assumptions and tools correct. Whether or not we are able to perceive the environmental problems being created, and whether we are able to respond effectively will depend on what evidence we can muster for environmental degradation, how we explain the phenomena we observe and how we plan, predict the outcomes of our actions and set policy. Hitherto we have tended to let theory drive observation rather than the reverse. There has been much emphasis on Gaussian statistics, sampling power, analysis of variance and various forms of dynamical simulation models. Effort has been placed on controlling “noise” in data rather than trying to understand and model it. Recent analyses of water quality data have revealed that what was hitherto thought to be noise in the data is actually small-scale information. The data reveal multi-fractal behaviour, and provide evidence for self- organised criticality and strong non-linear coupling at small scales. Different nutrient pools and biological components exhibit differing turnover times and contingent histories. Ignoring this small scale information means that many (if not most) ecological data are probably collected at inappropriate scales and are seriously aliased. Small scale interactions can have far reaching consequences in non equilibrium systems. The realisation that all water bodies contain much contingent small-scale information raises a serious question of indeterminism and questions the ability of widely accepted models to predict the outcomes of land use change on receiving waters. Models and predictions have been based around the properties of means and central tendencies only at scales in the region from hours to weeks – these are now seen to be an insecure basis for prediction and management. Techniques of data based modelling use the data itself to allow for the inclusion of prior experience and to define more parsimonious predictive models. Use of such models recognises the partial nature of our knowledge and requires adequate monitoring and adaptive management programs. Agile institutions and adequate data collection programs are the only solution to environmental management in this environment. Keywords: modeling, statistics, non-equilibrium systems, self organized criticality, complex adaptive systems 1. INTRODUCTION Just as in other areas of human endeavour the basic arguments and philosophical underpinnings of ecology (the fundamental science of environmental management) are changing. The basis of ecology lies in theories of the dynamics of populations and communities of organisms, based largely on nineteenth century ideas of plenitude and equilibrium (Kingsland 1985, McIntosh 1985). Much of ecological theory has been based around the role of competition in equilibrium communities (e.g. May 1973). More recently the sheer complexity of ecosystems and landscapes is being addressed with a range of “neutral” models which, rather than competition, stress the role of dispersal, chance and regional evolutionary histories (e.g. Hubbell 2001). Ecosystems are complex entities which show dynamic behaviour and spatial and temporal heterogeneity (Wu and Loucks 1995), discontinuities and multiple equilibria at a range of scales (O’Neill 1999). They display many of the properties attributed to Complex Adaptive Systems (CAS, Harris 1998) and show self- organised criticality (Sprott et al. 2002). Many properties of landscapes show fractal-like variability across a wide range of temporal and spatial scales (Lohle and Li 1996, Sole et al. 1999, Li 2000, Brown et al. 2002). With variability showing self similar properties across many scales, prediction, particularly for things like community composition and dynamics, is a matter of some difficulty (Lawton 1999). Fox Keller (2002) has recently noted that there is a strong distinction between the philosophy and practice – between the status of evidence and explanation – in physics and biology. In physics, theory has precedence; evidence is collected to confirm or refute the fundamental theoretical basis of the science. In biology on the other hand, the evidence has precedence over theory – so that biological explanations are contingent and may take the form of “just so stories”, contingent explanations or descriptions of natural history