COAT is an ecosystem-based observation system aiming at detecting, documenting and understanding the impacts of climate change on arctic tundra in order to enable scientifically evaluated management actions. The achievement of these goals requires a monitoring approach which is question and hypothesis driven, builds on conceptual models for the system studied and is relevant for management. The framework chosen by COAT uses the paradigm of adaptive monitoring developed by Lindenmayer and Likens (2010). Here a conceptual model of the investigated system is at the core of the monitoring program (Figure 1). This model, based on the present state of knowledge, determines key monitoring targets, key stressors, such as climatic drivers, and possible management actions. It is used to formulate predictions for the impact of drivers of change on the state of target ecosystem components, and forms the basis for developing the monitoring design. Management actions can enter this design in an experimental fashion and thus be tested and adapted to become maximally rational and effective. In this sense, adaptive monitoring is closely related to the concept of adaptive management. Obtained data are analyzed to improve knowledge, and the new understanding is used to improve the conceptual model, which may result in new predictions and an adjusted design. An adaptive monitoring protocol allows updates in the design and analyses to use new technologies, when they become available. Finally, stakeholders are involved systematically to adapt the monitoring protocol to changing needs of knowledge in the society.
Figure 1. The protocol for adaptive monitoring of Lindenmayer & Likens (2010), here modified to be tailored to climate change impact monitoring.
A food web approach
COAT takes a food web approach to ecosystem-based monitoring. Some of the strongest climate change impacts on tundra ecosystems are indeed mediated by strong trophic interactions in the food web. A food web approach provides also a solid theoretical foundation for developing the conceptual models central to the adaptive monitoring protocol, as there is a long tradition for conceptualizing the functioning of food webs in terms of models in ecology. Moreover, humans often affect ecosystems by their involvement in the plant-based food webs through impacts on vertebrate populations. COAT builds on a set of conceptual models that outlines how the tundra food webs of low arctic Varanger and high arctic Svalbard are expected to be impacted by climate change. These models define the framework for what shall be monitored (the monitoring targets) and how (the monitoring design). The monitoring targets are species or species assemblages which are known or expected to have key functions in the food web, and which often are management relevant either as natural resources or as species of conservation concern. In addition, known or expected sensitivity to climate change was a key criteria for prioritizing monitoring targets. COAT consists thus of a series of food web modules built on specific climate impact models for compartments of the plant based food web in the arctic tundra.
Conceptual climate impact path models
Conceptual climate impact path models with a similar structure have been developed for each food web module (Figure 2). Climate or climate change and management actions (for instance harvesting) represent the two main external drivers to be addressed. The arrows define the predicted climate and management impact paths onto the monitoring targets (species or functional species groups) of the module. Statistical versions of the climate impact path models will be developed to analyze monitoring data and obtain quantitative estimates of effects. In the statistical path models each monitoring target, climate and management will be represented by quantitative state variables such as abundance of a species, number harvested or climatic variables. The monitoring targets in the modules are of two kinds: response targets that are focal to the present module, and predictor targets that are focal responses of other food web modules. This structure creates conceptual links between the modules. Likewise, impacts (broken arrows) pointing at predictor targets denote effects that are estimated in other models, while whole arrows pointing at response targets are focal impacts to be estimated as effects in the present model. Two-headed arrows denote interaction effects between response targets and loop-arrows denote feed-back effects within response targets (for instance density dependence).