Abstract: Report on the ensemble model calibration and validation on the different sites
In response to the need to understand better the ecology of phytoplankton producers and predict the response of this key component of lacustrine food webs to changing environmental conditions, numerous computer models have been developed to simulate the seasonal development of lake phytoplankton. In this report we take two modelling approaches based on two different lakes, in quantifying the effects of future climate on the lake ecosystems. Most modelling studies that have attempted to predict future impacts on lake phytoplankton have utilised a single model and while
such studies have merit, applying multiple, independently developed models to a given lake system enable some of the inherent uncertainties in the model predictions to be quantified and reduced because the predictions of the individual models can be examined together. Hitherto, perhaps due to both the considerable resources required and the individual model expertise needed to apply a model to an aquatic ecosystem, no ensemble modelling studies have yet been carried out for prediction of lake phytoplankton dynamics.
The initial aim of this deliverable was to explore variations in model performance for a range of aquatic ecosystem models used within the REFRESH project and the intention was originally simply to compare performance of different models that had already been applied to different systems. However, given the expertise available from the international network of modelling experts involved in REFRESH, we were able, for the first time, to apply three autonomously developed models to the same lake, which comprise the first approach used in this report. We ran a series of potential future climate and nutrient load scenarios, and derived the predictions and uncertainties from the ensemble model run. Particular focus is given to changes in Cyanobacteria abundance. The lake ecosystem simulated by different models is Lake Engelsholm in Denmark, which, as is typical for lakes in many developed countries, is suffering from eutrophication as a result of decades of anthropogenic activities, with nutrient loading from both point sources and diffuse pollution.
Complex ecological models suffer from non-uniqueness, meaning that several different, but still realistic sets of model parameter values may yield similar simulated output. Utilising this knowledge we take a second approach in this report, in which we applied an ensemble set of model parameter combinations to a single ecosystem model. This approach was developed for a separate lake, the shallow Lake Arreskov in Denmark, for which the widely used PCLake model was adapted and run in ensemble. Utilizing the uncertainty range of simulated state variable such as cyanobacteria biomass and macrophyte coverage, as generated by simulations using the multiple parameter combinations, and by running a series of climate warming scenarios, we were able to elucidate the uncertainty of model predictions when evaluating the effects of future climate change on a lake ecosystem. The results presented here have in 2013 been submitted as two individual papers for peer-review.
Public Access: If you are unable to download this deliverable this means you are not logged in or are not a REFRESH participant. Most REFRESH deliverables are in the public domain. However, some are restricted to project participants. If you are interested in one of these deliverables please contact the co-ordinator (firstname.lastname@example.org) for further information