Author(s): Matthew R. Hipsey; Jorg Imberger; Andrea Paparini; Jason P. Antenucci; Rodolfo Soncini-Sessa; Vincenzo Romano-Spica; Chris J. Dallimore
Linked Author(s): Jörg Imberger
Keywords: Daptive management; Real-time decision support; Self-learning; Coupled model; Hydrodynamics; Water quality; Sustainability; Optimisation
Abstract: Across the globe, surface (e. g. wetlands, lakes, reservoirs, rivers and estuaries) and coastal waters face increasing pressures from development such as eutrophication and pollution from contaminants that are potentially deleterious to human and ecosystem health (e. g. pathogens, heavy metals, organic compounds). The sustainable management of such impacted systems requires a quantitative assessment of ecosystem dynamics and services to guide decision support activities. Models are used to support decision makers as they serve as virtual environmental laboratories where the functioning and sensitivities of systems can be explored, either in a natural or perturbed condition. They are also important since they support our capacity to reconcile theory with observation. Advances in cyberinfrastructure, sensors and observation networks have opened up new challenges for the development and application of models. Real-time data streams can now be utilised not only for model validation and testing, but they can also be dynamically integrated within a modelling system to enable adaptation of the model parameters as the simulation evolves. Although, such self-learning behaviour has been demonstrated for numerous physical systems (e. g. hydraulics), highly non-linear systems (e. g. aquatic ecology) paradoxically suffers from insufficient data which is essential for validation and further model development. This paper highlights the mismatch in water quality modelling between system and model complexity, and the available data-streams. An overview of new technologies such as flow-cytometry and real-time PCR that have potential to be applied in situ and operate in near real-time is presented. The integration of such advanced data-streams opens up a new paradigm for the way ecological models may adapt their operation and configuration in response to observed phenomena.