Limnologists have long recognized that one of the goals of their discipline is to increase its predictive capability (Peters 1986). In recent years, the role of prediction in applied ecology escalated, mainly due to man's increased ability to change the biosphere. Such alterations often came with unplanned and noticeably negative side effects mushrooming from lack of proper attention to longterm consequences. The current loss of biological diversity and global climate change are paradigmatic examples of such negative effects. Their mitigation requires actions based on efficient models for ecological forecasting (Clark et al. 2001). Past applications of predictive limnology proved fundamental, for example, to eutrophication control. Vollenweider (1968), searching for effective responses to the eutrophication problem, formulated successful deterministic models for lake management based upon quantitative relations between nutrients and production. Regression analysis of common limnological parameters have also been successfully applied to develop predictive models relating the variability of limnological parameters to specific key causes. These approaches, though, are biased by the requirement of a priori cause-relation assumption, oftentimes difficult to find in the complex, nonlinear relationships entangling ecological data. In addition, it is often difficult to satisfy the restrictive assumptions required by conventional parametric approaches. One promising set of quantitative tools that can help addressing current environmental challenges avoiding such restrictions is currently being researched and developed within the framework of ecological informatics. This is an interdisciplinary framework promoting the use of advanced computational technology to reveal ecological processes and patterns across levels of ecosystem complexity. Machine learning (ML) is a rapidly growing area of eco-informatics that is concerned with identifying structure in complex, often nonlinear data and generating accurate predictive models. One of these approaches attempting to model the relationship between a set of inputs and known outputs, is based on genetic algorithms and programming (GP). This stochastic optimization tool is based on the process of evolution in natural systems and was inspired by a direct analogy to sexual reproduction and Charles Darwin's principle of natural selection. GP works through genetic algorithms that use selection and recombination operators to generate a population of equations. These are competing solutions to a problem that evolve over time to converge to an optimal solution (Recknagel, 2001). The best solution, i.e. the best predicting equation, can be tested on a subset of data from the time series used to construct the model. Thanks to a 25-years long time-series of regular limnological data, the deep, oligotrophic and large Lake Maggiore (Northern Italy) is the ideal casestudy to test the predictive ability of GP. In this lake, the trend over time of physical, chemical and biological variables in the size range of the microbial food chain is well studied (Bertoni et al, 2010). Nevertheless, unsurprisingly, it appears that variables not included in deterministic models can influence the evolutionary trend of this oligotrophic ecosystem. The application of GP may lead to the discovery of non-obvious relationships between variables, and to the formulation of new stochastic models. Here we discuss the forecasting efficacy of some models emerging from GP application in which no deterministic assumption is made.
A non-deterministic approach to forecasting the trophic evolution of lakes
Roberto Bertoni;
2014
Abstract
Limnologists have long recognized that one of the goals of their discipline is to increase its predictive capability (Peters 1986). In recent years, the role of prediction in applied ecology escalated, mainly due to man's increased ability to change the biosphere. Such alterations often came with unplanned and noticeably negative side effects mushrooming from lack of proper attention to longterm consequences. The current loss of biological diversity and global climate change are paradigmatic examples of such negative effects. Their mitigation requires actions based on efficient models for ecological forecasting (Clark et al. 2001). Past applications of predictive limnology proved fundamental, for example, to eutrophication control. Vollenweider (1968), searching for effective responses to the eutrophication problem, formulated successful deterministic models for lake management based upon quantitative relations between nutrients and production. Regression analysis of common limnological parameters have also been successfully applied to develop predictive models relating the variability of limnological parameters to specific key causes. These approaches, though, are biased by the requirement of a priori cause-relation assumption, oftentimes difficult to find in the complex, nonlinear relationships entangling ecological data. In addition, it is often difficult to satisfy the restrictive assumptions required by conventional parametric approaches. One promising set of quantitative tools that can help addressing current environmental challenges avoiding such restrictions is currently being researched and developed within the framework of ecological informatics. This is an interdisciplinary framework promoting the use of advanced computational technology to reveal ecological processes and patterns across levels of ecosystem complexity. Machine learning (ML) is a rapidly growing area of eco-informatics that is concerned with identifying structure in complex, often nonlinear data and generating accurate predictive models. One of these approaches attempting to model the relationship between a set of inputs and known outputs, is based on genetic algorithms and programming (GP). This stochastic optimization tool is based on the process of evolution in natural systems and was inspired by a direct analogy to sexual reproduction and Charles Darwin's principle of natural selection. GP works through genetic algorithms that use selection and recombination operators to generate a population of equations. These are competing solutions to a problem that evolve over time to converge to an optimal solution (Recknagel, 2001). The best solution, i.e. the best predicting equation, can be tested on a subset of data from the time series used to construct the model. Thanks to a 25-years long time-series of regular limnological data, the deep, oligotrophic and large Lake Maggiore (Northern Italy) is the ideal casestudy to test the predictive ability of GP. In this lake, the trend over time of physical, chemical and biological variables in the size range of the microbial food chain is well studied (Bertoni et al, 2010). Nevertheless, unsurprisingly, it appears that variables not included in deterministic models can influence the evolutionary trend of this oligotrophic ecosystem. The application of GP may lead to the discovery of non-obvious relationships between variables, and to the formulation of new stochastic models. Here we discuss the forecasting efficacy of some models emerging from GP application in which no deterministic assumption is made.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
