Defining indirect uncertainty in system-based risk management

Abstract

This paper defined a new concept concerning information variability within the risk modeling procedure in order to illuminate the handling of uncertainty on the system scale. With respect to scientific uncertainty, most definitions proposed are largely focused on the knowledge insufficiency or data imperfection of a model that notoriously impairs the robustness of simulation output. However, the uncertainty inherent in the indirect effect of a multi-component system (e.g., an ecosystem), which serves as the major source of the information variability responsible for the ultimate fate of the system, has never been expressly unfolded. Herein, we developed a new concept called indirect uncertainty (IU) to represent the variability along with the indirect process of information propagation throughout the system. Particularly, we adapted IU to the system-based ecological risk assessment by using qualitative reasoning, whereby hopefully we can reveal the potential application of IU analysis. Based on this, IU analysis was recommended to be incorporated into the risk management framework as a necessary complement.

Introduction

Risk assessors are always concerned with uncertainty, as it might give occasion to significant deviation of simulation output from the real situation and thereby elicit a delusive picture in front of decision makers. In environmental assessment and management activities, uncertainty is more or less unavoidable and its characterization is essential if we are to understand the intrinsic functioning of perturbed ecosystems as well as our effects on these ecosystems (Ascough et al., 2008, Harwood and Stokes, 2003, USEPA, 1989). There are compelling reasons to recognize the existence of uncertainty and avoid the mistaken impression that ecological assessments are always precise and well-understood (Reckhow, 1992, Reckhow, 1994). One is that the modeling procedure of the ecological risks propagated throughout the perturbed environment is inevitably filled with imprecision due to a lack of sufficient knowledge and incomplete information. Furthermore, environmental managers need to know the expected uncertainty in each assessment system and choose the most dependable direction. For these and other reasons, uncertainty analysis has been recommended as a routine part of ecological risk assessment (ERA) (Morgan and Henrion, 1990, USEPA, 1992). Currently, it is frequently used in detailed pollutant transport and fate models to indicate the indeterminate factors or processes in risk evaluations (Reckhow, 1994).

In the modeling of ecological risk, scientific uncertainty is usually classified into two facets in terms of different sources, i.e., epistemic uncertainty and aleatory uncertainty (Chowdhury et al., 2009, Granger and Henrion, 1990). Basically, epistemic uncertainty results from a complete or partial lack of pivotal knowledge or information about a system, which is related to built-in framework, model parameters, boundary conditions and different choices of solution approaches, etc (Roger and Pielke, 2003, Zimmermann, 2001). To put it another way, epistemic uncertainty is a product of limited perception and deflective reasoning of modelers and can be possibly avoided once the required second remedies of the model are provided. However, aleatory uncertainty is quite different from the former in that it is generally induced by the information indeterminacy or variability inherent in the natural ecosystems rather than influenced by human initiative. It is closely associated with the intrinsic structure and random processes of the natural system we targeted and coupled with the intricate response when subjected to external interventions. Ecological systems, as is known, are usually complex, nonlinear and strongly influenced by stochasticity (Lek, 2007, Mangel et al., 1996). However, the available information about the way that these systems function is often paradoxical or equivocal. Therefore, it is almost impossible to fully unfold the dynamics of these systems in detail and little confidence is left with respect to the explicit apprehension of certain perturbation effects. As a result, almost all the developed methods concerning system uncertainty in risk assessment including probabilistic approaches (such as typically Monte Carlo simulation and asymptotic reliability analysis) and non-probabilistic approaches (e.g., interval analysis, fuzzy set theory and possibility theory) are actually the accounts of epistemic uncertainty in ecological models (Chowdhury et al., 2009, Mamdani, 1977, USEPA, 1996), whereas no formal methods have been proposed yet, with the problem of intrinsic uncertainty within systems per se lying on the table.

The structure and dynamics of a system, nonetheless, have been increasingly modeled based on the propagation of environmental information (typically material flow, energy flow and information flow) in an integrated way (Fath, 2004a, Fath, 2004b, Kazanci, 2009, Patten, 1978, Ulanowicz, 2004). These formulations of the system, by design, provide an important insight into how system components are tied to a larger web of interactions by their uncovering patterns and influence among all the objects in a system under uncertainty (Fath, 2004a). In fact, models focused on these interactions have been constructed for ERA on the system scale as well as other ecosystem-based management. In addition to the direct interactions among components, the indirect effects are expressly taken into consideration when implementing the environmental assessment, as it also plays an important part in the integral propagation of information (Chen et al., 2011, Christian et al., 2009 Fath, 2004b, Schramski et al., 2011, Tollner et al., 2009).

This network perspective serves as the first inspiration of the present study. From the system-oriented perspective, we explored the intrinsic uncertainty within the ecosystem modeling based on the development of new concept associated with indirect effect. The implication of this concept was, therefore, applied to address ERA for ecosystems in the face of modeling indeterminacy by using qualitative reasoning. Finally, we recommended the concept be incorporated into the risk management framework as a useful complement. With the proposal of a new concept, this study could further the apprehension of ecosystem's functioning under uncertain condition, thus aiding in eliciting a more robust version of ERA as well as other ecosystem-based assessments.