Elsevier

Ecological Modelling

Volume 203, Issues 1–2, 24 April 2007, Pages 147-156
Ecological Modelling

Predictive function and rules for population dynamics of Microcystis aeruginosa in the regulated Nakdong River (South Korea), discovered by evolutionary algorithms

https://doi.org/10.1016/j.ecolmodel.2006.03.040Get rights and content

Abstract

Two algorithms of evolutionary computation, an algebraic function model and a rule-based model, were applied for model development with respect to 8 years of limnological data from the lower Nakdong River. The aim of the modelling was to reproduce the abundances of the phytoplankton species, Microcystis aeruginosa, based on physical, chemical and meteorological parameters. The algebraic function model overestimated or underestimated abundance values, but correctly recognized the timing of high abundances. The rule-based model detected not only the timing of algal blooms well but also the magnitude of abundances. Sensitivity analysis indicates that high water temperature influences high abundances of M. aruginosa. In addition, dissolved oxygen, pH, nitrate and phosphate are shown to be explainable in relation to deoxygeneration, carbon dioxide transformation and nutrient limitations.

Introduction

Population and community dynamics in ecosystems are determined by ongoing complex interactions between living and non-living components. It is difficult to predict ecosystem phenomena, and many scientists are trying to clarify the interaction and relationship between ecological constituents. Numerous approaches have been attempted in order to explain and understand ecosystem dynamics, ecological modelling is one of these. The development of computational technologies has made ecosystem analysis more accurate and useful for prediction and elucidation of ecological phenomena.

The characteristic of ecological data is complex and dynamic. It can be very hard to understand and explain the causality of ecological interactions and events with just datasets. Machine learning techniques can support ecological scientists by enabling them to make sense of datasets. Recently, many machine learning techniques have been introduced and applied to ecological study [e.g. artificial neural networks, genetic programming, self organizing map, etc.] (Lek et al., 1996, Recknagel, 1997, Fielding, 1999, Whigham and Recknagel, 2001a).

Evolutionary algorithms (EAs) are a branch of machine learning techniques. These are recently developed methods inspired by principle of natural selection. Artificial neural networks (ANNs) have gained inspiration from signal transmission in neuron synapses, while EAs are based on the principle of biological evolution, such as crossover, mutation and chromosome's alteration. Previous applications of both ANNs and EAs suggest that machine learning techniques can be a good solution in ecological modelling so as to forecast and explain target systems (Recknagel, 2001). Many recent researches have reported successful results using machine learning such as ANNs and EAs (Ngan et al., 1999, McKay, 2001, Neely and Weller, 2001, Weekes and Fogel, 2003). In particular, ANNs showed successful cases with respect to classification or prediction of noisy and multivariate data (Reed and Mark, 1999). EAs also exhibited several promising results to predict and elucidate specific ecological phenomena (Whigham and Recknagel, 2001b, Jeong et al., 2003a).

The aim of this study is to evaluate and examine the application of EAs in order to explain relationships or interactions between algal bloom components in a eutrophic river ecosystem by developing EA models and using sensitivity analysis. Eutrophication in lake and river ecosystems has been recognized as a global environmental issue, as the demand for water resources increases with growing human population and developing industries. Especially, the proliferation of blue green algae has an effect on the management of water quality in the catchment area (NRA, 1990). Blue green algae generally contain harmful toxins, so its toxicity can influence structural change in aquatic ecosystems (Carr and Whitton, 1973, Fogg et al., 1973, Bold and Wynne, 1985).

Nakdong River, which is the longest river in South Korea, plays a key role in the supply of water resources for the south eastern area of S. Korea. This river is unique, as it is a regulated river system. In the upper part of the river, there are several multi-purpose dams, while in the lower part of the river there is an estuarine barrage. The flow of the river is regulated by both methods of regulation for stable water supply and storage in preparation for the next year's irrigation. Due to the construction of these artificial structures, the water velocity has decreased and the river has been seasonally turned into a lake-like system (Joo et al., 1997). The above characteristics of Nakdong River cause blue green algal blooms in the summer time, especially when there is insufficient summer rainfall (Ha et al., 1999, Ha et al., 2000).

In this study, EAs are applied in order to represent population dynamics of Microcystis aeruginosa dependent on physico-chemical variables of the Nakdong River by means of predictive algebraic function and rule sets. Each model examines the applicability of EAs for the elucidation of ecosystem dynamics. There has been some research to introduce and apply predictive models to the Nakdong River during the past few years (Park and Lee, 2002, Jeong et al., 2003a). Park and Lee (2002) demonstrated a water quality model using physical and chemical data, and Jeong et al. (2003a) described M. aeruinosa bloom dynamics using evolutionary computation with 25 limnological data. The two above modelling structures have a difference from a standpoint of deductive or inductive approach, and so far, it is difficult to judge which approach is more accurate and feasible for ecological modelling. Our research explains specific phytoplankton bloom dynamics, M. aeruginosa, in the Nakdong River in order to apply EAs.

Section snippets

Sampling and data selection

Water samplings were conducted weekly at Mulgum water intake station (27 km above the estuarine barrage), which is located in the lower Nakdong River. Water samples were taken from the river surface at approximately 50 cm depth. The parameters for sample were water temperature, dissolved oxygen (DO), turbidity, conductivity, Secchi disc transparency, pH, alkalinity, chlorophyll a, phytoplankton biovolume, zooplankton abundance and nutrient concentrations such as nitrate (NO3-N), ammonia (NH4+

Limnological aspects

All parameters are exhibited in forms of averages and standard deviation, and several inter-annual variations over an 8 years period were observed (Table 1). The concentration of chlorophyll a was 43 ± 62 μm/L (mean ± standard deviation), which is at a eutrophic level. Chlorophyll a concentration was high in 1994 (ca. 70 μm/L) and 1995 (ca. 66 μm/L), and the abundance of M. aeruginosa was also relatively high. Many physico-chemical parameters are related to rainfall in a certain year. For example, in

Discussion and conclusion

Blue-green algae are some of the predominant species in eutrophicated lakes and rivers. Previous researches exhibited several parameters for dominance of blue-green algae from light, temperature and concentrations of nitrate, phosphate and oxygen (Fogg et al., 1973). Shapiro (1990) presented various factors for the growth of blue-greens such as high water temperature and low light intensity. It was shown that high water temperature influenced the growth of blue-green algae, which describes a

Acknowledgements

This work was supported by the Long-Term Ecological Research (LTER) Project from the Ministry of Environment in South Korea. This study was financially supported for Dr. Kwang-Seuk Jeong by Pusan National University in program, Post-Doc. 2006. The authors are very grateful to Dr. Kyung. Ha of Durham University in U.K. for identifying phytoplankton. We also thank Dr. Hugh Wilson for discussing and commenting about algorithms in University of Adelaide. This paper is contribution No. 47 of the

References (45)

  • P.A. Whigham et al.

    Predicting chlorophyll-a in freshwater lakes by hybridising process-based models and genetic algorithms

    Ecol. Model.

    (2001)
  • T. Bäck et al.

    Evolutionary computation: comments on the history and current state

    IEEE Trans. Evolutionary Computation

    (1997)
  • W. Banzhaf et al.

    Genetic Programming: An Introduction On the Automatic Evolution of Computer Programs and Its Applications

    (1998)
  • R.C. Berkan et al.

    Fuzzy Systems Design Principles: Building Fuzzy IF-THEN Rule Bases

    (1997)
  • H.C. Bold et al.

    Introduction to the Algae

    (1985)
  • A.J. Boulton et al.

    Australian Freshwater Ecology Processes and Management

    (1999)
  • N.G. Carr et al.

    The Biology of Blue-Green Algae

    (1973)
  • V. Cassie

    A Contribution to the Study of New Zealand Diatoms

    (1989)
  • R.L. Dennis

    Dispersion models for management of urban air quality

  • U. Einsle
    (1993)
  • A.H. Fielding

    An introduction to machine learning methods

  • E. Foged

    Diatoms in Eastern Australia

    (1978)
  • Cited by (0)

    View full text