The primary focus of environmental risk assessment (ERA) for non-target organisms is on the direct effect of the toxicant. In some cases what are often termed large-scale sources of variation are also considered (e.g. expected crop distribution). However, these describe only part of the variation that occurs in the real world. Landscapes vary in structure, meaning that the field size and proximity to primary NTO habitats will vary. There is climatic variation driving changes in phenology and behaviour, and management changes in the proportion of crops grown, but also changes in how they are cultivated in time and space. All of these factors can affect the risk assessment. There is also another, difficult to observe, property of real systems, and that is the spatio-temporal dynamics associated with populations, climate, management, and ecology and behaviour, and the potential for feedback loops. These interactions can exacerbate or ameliorate impacts either via local feedback mechanisms e.g. multiple stressors, or by virtue of the spatial population dynamics. This uncertainty is normally considered as stochasticity in ERAs and the factors are often incorporated into a single general term, utilizing a safety factor to account for uncertainty. This, however, robs us of both understanding and predictive power, since probability distribution can only be based on statistical expectations of past events, which do not necessarily account for interactions in the future. An alternative approach capable of dealing with these system properties is agentbased modelling (ABM).ABMs are capable of integrating a range of drivers and actors in space and time and can represent detailed farming operations on a large scale, integrating these with realistic models of animal populations, and expressing each animal as an individual agent. As such they are capable of representing very complex dynamics in time and space. Another key model attribute is the ability of agents to respond to the information they gather from their local environment. This automatically integrates many of the dynamics difficult to capture in traditional models, e.g. source-sink dynamics are emergent properties and do not need to be imposed.ALMaSS is one such model system and is used here to exemplify aspects of complexity in population-level risk assessment for terrestrial mammals, birds, and arthropods exposed to pesticides. Results are used to argue for greater realism and population-level ERA.
Main Research Area:
6th SETAC World Congress / SETAC Europe 22nd Annual Meeting, 2012