Information-based Network Environ Analysis: A system perspective for ecological risk assessment

Abstract

Ecological risk assessment, aiming at evaluating a wide range of undesirable consequences initiated by a possible eco-environmental hazard, has been the center of concern for ecosystem management in recent years. However, when it comes to disturbed natural ecosystems, most models developed for ecological risk assessment are restricted to instant cause–effect computation of single factors and often ignore the indirect effects, therefore fail to implement a holistic assessment at an ecosystem scale when interactions of different risk receptors are obvious. In this study, we developed a risk-based network model based on a new control analysis termed control allocation and a conceptual conversion of flow currency in NEA. By taking a river ecosystem intercepted by dam construction as an example, risk propagation between all functional guilds of the ecosystem concerning both direct risk and integral risk dynamic were quantified and illustrated in the network model. The results of this new risk assessment showed that there were significant differences between network integral risk and input risk, and although the phytoplankton received the instantaneous impact of chromium pollution among all the functional guilds, it was the piscivorous fish which obtain the greatest overall risk threat. On the basis of the model results, we proposed the network-based indicators for assessing the system-wide risk condition and component-specific risk scenarios of disturbed ecosystems exposed to single or multiple stressors. This study could provide a novel perspective and methodology for assessing ecological risk at the system scale, and concurrently, serve as an elicitation of how we can effectively evaluate ecosystems on the same analytic basis of information-based networks.

Introduction

Risk science is the science of loss. It was developed to calculate the value of possible costs or damages in the face of knowable profitable outcome (Kaplan and Garrick, 1981). The conception of probabilistic risk analysis is so deeply embedded into various disciplines that it plays a crucial role in almost all safety evaluations and managements of human society involving economics (Kahneman and Tversky, 1979), politics (Slovic, 1999), engineering (Kumamoto and Henley, 1996), environmental health (Hallenbeck, 1986), etc. But the application of risk assessment for ecological theory is, comparatively, quite a recent interest. By definition ecological risk assessment (ERA) is focused on the rational appraisal of the possible damages or potential diverse effects by computing the risk values associated with possible eco-environmental hazards under uncertainty (USEPA, 1992, Freedman, 1998, Suter II, 2007). The goal of ERA is to provide information about the statistical distribution of possible ecological effects arising from exposure to one or more stressors (USEPA, 1998, Findlay and Zheng, 1999). In order to achieve this goal, three elements complete the basic profile of ERA: (1) the scenario, (2) the likelihood and (3) the consequence, or rather—what can happen; how likely things are to happen; and what are the end points from sets of occurrences (i.e., “sets of triplets” in Helton, 1993). Clear as it seems, the random and nonlinear characteristics inherent in ecosystems often make it difficult to predict the precise ecological fate, furthermore, multi-process scenarios (multi-source, multi-factor and multi-receptor scenarios) are what environmental managers inevitably encounter when performing a risk analysis, all of which urge them to resort to powerful models, preferably mechanistic ones. So far, mathematical models employed for ERA includes holographic neural networks (Findlay and Zheng, 1999), Bayesian networks (Lee and Lee, 2006, Pollino et al., 2007), comprehensive aquatic systems models (DeAngelis et al., 1989, Bartell et al., 1999), and environmental contaminant dispersion models (Chen et al., 2010), etc. Helpful and instructive as they are for guiding risk-based decision making, most of them are either restricted to the evaluation of species biology and population on the microcosm scale under circumstances of single stressors, or developed on the profile capable of limited risk receptors, therefore they have relatively poor capacity for fitting into iterative and adaptive management (Yeardley and Roger, 2000). Furthermore, the instant cause–effect type of computation might neglect the information of the indirect effect carried by the interactive components within communities or ecosystems.

Alternatively, Network Environ Analysis (NEA), an important branch of network analysis first developed by Patten, 1978a, Patten, 1978b, Patten, 1982, is a system-oriented modelling technique for examining the structure and flow of materials in ecosystems (see also the pioneer works by Leontief, 1951, Leontief, 1966 and Hannon (1973)). NEA places great emphasis on the interactions between components rather than the characteristics of individuals, and the dynamic attributes within the system are identified and quantified via network structural and functional analytic methods (e.g., storage analysis, throughflow analysis, utility analysis, control analysis, etc.) (Fath, 1998, Fath, 2004a, Fath, 2004b, Fath and Patten, 1998, Fath and Patten, 1999, Fath and Borrett, 2006, Kazanci, 2007, Schramski et al., 2006, Schramski et al., 2011, Ulanowicz, 2004). Fundamentally, the underlying strength of these concepts and methods is the incorporation of direct and indirect effects which construct the whole regime of the interacted network, and that system wholeness is arguably more critical in determining the system's behaviour than the direct effects alone, or further, the holistic picture of the concerned system can only be delineated when interactions of all lengths are clarified.

In view of these important insights, one of the most promising applications of NEA is identified as a methodology platform for modelling the integrated eco-environmental impact of natural systems under human interference (Fath, 2004a). The implication is that NEA is conceivably promising for indexing the holistic ecological risk of perturbed ecosystems. In fact, ecological network analysis (a more general version of NEA) has been proved useful as a complementary tool for assessing disturbed ecosystems in the context of system-based management. The most recent cases concerned the determination of possible ecosystem impacts of fishing on estuarine ecosystem (Manickchand-Heileman et al., 2004), the evaluation of environmental stress due to soil contamination on terrestrial ecosystem (Tobor-Kapłon et al., 2007) and the functional assessment of an estuary ecosystem exposed to eutrophication (Christian et al., 2009). Unfortunately, the fact that the model operation may encounter flow incompatibility in a material- or energy-oriented NEA remains impeditive when evaluating the adverse impact or managing the transitive risk on a system scale. In other words, energy and material, the conventionally used mediates for network synthesis, are not essentially adaptable for a system-wide ERA. A fundamental conversion of the flow currency for NEA is needed.

Herein, we presented a novel approach for holistic ecological risk assessment based on NEA. The information-based network analysis was developed to address ecological risk in which both direct and indirect effects were together considered and various risk factors and receptors were technically compatible in the same model. The reminder of this paper was arranged as follows: The general framework of the model was described in Section 2. In Section 3, the development of risk-based flow was formulated. Following that in Section 4 we illustrated the operation of risk network through a case study of a river ecosystem intercepted by dam construction. Then in Section 5, we discussed the application of information-based NEA for ecosystem management, and finally, a range of conclusions were presented in Section 6.