Making eco logic and models work : an integrative approach to lake ecosystem modelling

Abstract

Dynamical ecosystem models are important tools that can help ecologists understand complex systems, and turn understanding into predictions of how these systems respond to external changes. This thesis revolves around PCLake, an integrated ecosystem model of shallow lakes that is used by both scientists and water quality managers to understand and predict eutrophication effects in shallow lake ecosystems. Shallow lakes provide some of the clearest examples of alternative stable states in natural systems. PCLake can be used to calculate the critical nutrient loading, that is, the nutrient loading where an abrupt regime shift occurs from a clear aquatic plant dominated state to a turbid phytoplankton dominated state, or vice versa. Four different aspects of modelling with PCLake are addressed in this thesis: (1) making the model better accessible for the modelling community, (2) improving the model, (3) developing scientific theory, and (4) exploring new applications for water quality management. Following a general introduction to the thesis in chapter 1, the Database Approach To Modelling (DATM) is introduced in chapter 2. DATM is invented to make dynamic models more accessible. The idea of DATM is that the mathematical equations of a model are stored in a database independently of program language and software specific information. From the database, the information can be automatically translated, augmented and compiled into working model code of various different modelling frameworks (software programs). In chapter 3 the weak link between ecosystem models and real ecosystems is discussed in relation to model calibration and improvement. In a previous stage PCLake has been calibrated using data of more than 40 lakes to obtain a best overall fit, which has greatly increased the scope of the model by making it suitable for more generalized studies on temperate shallow lakes. However, because of this calibration, adding missing functional components to the model at a later stage does not automatically increase the validity of the model, as it may bring the model ‘out of balance’. This is exemplified by adding filter feeding zoobenthos to PCLake, which were previously ignored. In chapter 4, the relation between food-web theory and alternative stable states theory is scrutinized. Both theoretical paradigms are highly influential in modern ecology as they help scientists understand how stability emerges in complex natural ecosystems. Unfortunately, they developed independently and it is largely unclear how the resilience of a food web relates to the stability of the complete ecosystem. For this study PCLake was used as a virtual reality from which ‘empirical’ information is sampled to parameterize a food web model, following traditional food web methods. This allowed calculating the stability of the food web along a gradient of environmental change, knowing that the complete ecosystem shows a regime shift once the critical nutrient loading is exceeded. In chapter 5 the question is asked to what extent models of a different form can be used to describe the same natural phenomenon, and hence, how these models can be used for a better understanding of such natural phenomena. Using three classical extensions of the famous Lotka-Volterra equations, which unlike PCLake can be fully mathematically understood, we analyze the consequence of changing a system with a sophisticated functional response term (e.g. Holling type II or III) into a system with a simpler functional response term while maintaining equilibrium densities and material fluxes. These results give new insight into when empirical data can be linked to mathematical models to estimate the stability properties of real ecosystems. Although PCLake is predominantly applied in the context of ecosystem restoration of turbid phytoplankton dominated lakes, chapter 6 focusses on the clear water state after the reestablishment of aquatic plant dominance as occured. Dense stands of aquatic plants easily cause nuisance, and hence the removal of aquatic plants is an emerging management issue. Yet, because aquatic plants play an important role in stabilizing the clear water state, the removal of plant biomass can potentially trigger a critical transition back to the turbid water state. Currently there is only limited empirical and theoretical understanding of how harvesting of aquatic plants affects ecosystem functioning, which frustrates effective and efficient ecosystem management. With PCLake the impact of harvesting is evaluated, in terms of reducing nuisance and ecosystem stability, for a wide range of external nutrient loadings, mowing intensities and timings. Additionally, the model is used to estimate how much phosphorus is removed from the system during harvesting. In chapter 7 I discuss the added value of taking an integrative approach to modelling, and discuss the integrated nature of the studies presented in this thesis. It’s also important to note that these studies were part of a larger research project with the overall aim of increasing the usefulness and the validity of PCLake and its twin model PCDitch, and to enhance the confidence in the models among water quality managers. A synopsis of the overarching collaborative research project on PCLake and PCDitch is presented in chapter 8.