Waterfowls habitat modeling: Simulation of nest site selection for the migratory Little Tern (Sterna albifrons) in the Nakdong estuary
Highlights
► This paper aims to find patterns in nest site selection by Little Terns Sterna albifrons, using an empirical modelling algorithm; genetic programming (GP). ► Nest distribution was surveyed for the species, and physical environment data were used as forcing functions. ► The developed model provided useful information for the explanation of Little Tern nest distribution patterns. ► Monospecific colony of Little Tern exhibited different patterns (sensitivity to mean elevation from sea level) from mix colonies of multi-species. ► The results of this study are challenging due to the importance of estuarine areas globally as refuge sites and because of the limited effort in modeling research based on animal species in these areas.
Introduction
Habitat degradation will lead to decreases in the number of species and/or individuals, which is very apparent for seabirds assemblages, including migratory birds (Davies, 1981). In particular, many migratory avian species are protected due to limited numbers and sensitivity to habitat change. Nevertheless it is important to note that more mathematically quantitative standards are used as analytical techniques except in some experimental approaches. In this regard, modeling bird habitat can; provide objective criteria for habitat protection and reinforce bio-conservation strategies in establishing a natural sanctuary. A purpose of habitat modeling is to directly simulate or indirectly infer behavioral patterns of ecological entities in response to changes in environmental factors, such as human disturbance, subsidence of estuarine area, or global warming. There have been numerous efforts at predicting or patterning habitat suitability for avifauna, such as using regression for prediction (Brotons et al., 2004), multi-variate statistics like ordination (Fielding and Haworth, 1995), or Lidar-derived habitat structure analysis (Broughton et al., 2006, Goetz et al., 2007). The first two approaches are based on a collected habitat characteristics database, such as number of nests, clutch size, or bird species individual, in order to adapt the initial model structure, while the last explores relationships between birds and habitat according to species territory and diversity.
Recent advances in computer science enable researchers to speed up the calculation process for predictive model development, which may provide opportunities for importing numerous forcing functions into model construction. Comparatively, using sophisticated computation algorithms such as genetic programming (GP) is thought of as an alternative pathway to discover underlying patterns in the collected dataset. Among the various machine learning algorithms, GP has two advantages in the so-called ‘global search for optimal solution’ and ‘evolutionary variable selection’. The former encourages its utility when a local minima problem is expected, which has to be considered in connectionist approaches (Gil-Tena et al., 2010, Lee et al., 2007, O’Hanley, 2009). Increases in the number of forcing function do not guarantee the increase of model performance (Jørgensen, 1997), and even using a small number of forcing functions, which are strongly related to the output variable, is more crucial for model predictability (Jeong et al., 2006a).
Therefore, in this study, we adopted GP, an efficient branch algorithm of evolutionary computation (EC) supported by previous research (Jeong et al., 2003, Jeong et al., 2010, Kim et al., 2007a), to develop a prediction model for the distribution of Little Tern Sterna albifrons nests on a barrier islet, in accordance with physical habitat characteristics. This species migrates between South Korea and Australia, and nests in the Nakdong estuary, located in southern Korea, from April to May. We surveyed the distribution of Little Tern nests on a barrier islet which supports this species, and developed an ecological model for the explanation of habitat preference using GP. Predictability and model structure are discussed in accordance with ecological characteristics of the species.
Section snippets
Study site and data collection
The Nakdong estuary is located in the southeastern part of South Korea (35° 05′N, 128°55′E) near the middle of the Australasian Flyway (Fig. 1A). The estuary supports more than 200 avifaunal species including migratory birds (Fig. 1B; Lee et al., 2010). Rapid industrialization and urbanization has affected the loss of estuarine wetlands and riparian zones along the lower Nakdong River and the estuary since the 1980s. We surveyed the Little Tern Sterna albifrons nests distribution on the islet
Results
From the survey results, an increasing tendency in the number of nests was observed in the quadrats where a relatively high mean elevation was measured (Fig. 4). The mean elevation increased by ca. 50 cm in the quadrats No. 55–80, and in those quadrats, the elevation variation was relatively larger (Fig. 4A and B). Quadrats with higher elevation were strongly vegetated compared to other areas (Fig. 4C and D). The number of Little Tern's nests tended to increase in those quadrats, however
Discussion
The application of GP to nest number prediction was successful in terms of predictability and variable input selectivity. Even though the prediction accuracy of GP was relatively moderate, selecting a small number of influencing input variables is a feasible in ecological model development (Jeong et al., 2006a). The application of GP to a long-term ecological database indirectly provides evidence that a small number of input variables, selected based on a global search rather than using all
Conclusion
In this study, a hundred quadrats were established to survey the distribution of Little Tern nests in the Nakdong estuary. A barrier islet where this species was abundant was selected for the survey. The number of nests used for monitoring was assessed along with other environmental variables, such as elevation, vegetation coverage and height. A non-linear modeling algorithm using genetic programming was applied to the survey results, producing a simple rule set model (consisting of
Acknowledgments
The authors appreciate Ms. Hee-Sun Park for her assistance of identifying the plant species during the survey. In addition, we thank Mr. Maurice Lineman for English proof reading in the process of manuscript preparation. This study was partly financially supported by the BK21 Research Group for Marine and Silver Biotechnology in the Pusan National University (post-doc research grant for Dr. K.-S. Jeong). Also this research was supported by the Korea National Long-Term Ecological Research
References (28)
- et al.
Can genetic programming improve software effort estimation? A comparative evaluation
Inform. Software Technol.
(2001) - et al.
Laser remote sensing of canopy habitat heterogeneity as a predictor of bird species richness in an eastern temperate forest, USA
Remote Sens. Environ.
(2007) - et al.
Prediction and elucidation of phytoplankton dynamics in the Nakdong River (Korea) by means of a recurrent artificial neural network
Ecol. Model.
(2001) - et al.
River phytoplankton prediction model by Artificial Neural Network: model performance and selection of input variables to predict time-series phytoplankton proliferations in a regulated river system
Ecol. Inform.
(2006) - et al.
Delayed influence of dam storage and discharge on the determination of seasonal proliferations of Microcystis aeruginosa and Stephanodiscus hantzschii in a regulated river system of the lower Nakdong River (South Korea)
Water Res.
(2007) - et al.
Modelling Microcystis aeruginosa bloom dynamics in the Nakdong River by means of evolutionary computation and statistical approach
Ecol. Model.
(2003) - et al.
Predictive function and rules for population dynamics of Microcystis aeruginosa in the regulated Nakdong River (South Korea), discovered by evolutionary algorithms
Ecol. Model.
(2007) - et al.
Patterning habitat preference of avifaunal assemblage on the Nakdong River estuary (South Korea) using self-organizing map
Ecol. Inform.
(2010) - et al.
Classification of breeding bird communities along an urbanization gradient using an unsupervised artificial neural network
Ecol. Model.
(2007) - et al.
Presence–absence versus presence-only modelling methods for predicting bird habitat suitability
Ecography
(2004)
Marsh Tit Poecile palustris territories in a British broad-leaved wood
IBIS
Are salinas a suitable alternative breeding habitat for Little Terns Sterna albifrons?
IBIS
Development and behaviour of Little Tern chicks
Br. Birds
Nest habitat selection by eight syntopic species of Mediterranean gulls and terns
Colon. Waterbirds
Cited by (8)
Evolutionary algorithms for species distribution modelling: A review in the context of machine learning
2019, Ecological ModellingCitation Excerpt :For the first, hyperparameter optimization methods are already available giving satisfying results for multiple problems (Gibbs et al., 2008; Gobeyn et al., 2017). The second, the need for modelling know-how will require standardization, documentation and refinement of the algorithm development and application process (Jakeman et al., 2006; Grimm et al., 2010) going hand in hand with the philosophy of open science. In this review paper, a number of suggestions with respect to the definition of the model, algorithm and implementation are given.
Search of optimal locations for species- or group-specific primer design in DNA sequences: Non-dominated Sorting Genetic Algorithm II (NSGA-II)
2015, Ecological InformaticsCitation Excerpt :Evolutionary algorithm has been highlighted by favor of its global search ability, which is known to be able to avoid over-fitting in the training process (Muttil and Lee, 2005; Whigham, 2000). This algorithm often is used in parameter optimization (e.g. Kim et al., 2014; Moradi-Jalal et al., 2004) or equation discovery (e.g. Jeong et al., 2011; Kim et al., 2007), which suggests successful solutions for given problems by means of heuristic approach in ecological research field. Multi-Objective Optimization (MOO) has become a useful strategy in the solution of many modern engineering problems, and D'Souza et al. (2010) emphasized the utility of evolutionary algorithms as having been very successful in carrying out the optimization of multiple objectives.
Long-term adaptations of a migratory bird (Little Tern Sternula albifrons) to quasi-natural flooding disturbance
2015, Ecological InformaticsCitation Excerpt :This positive pattern indicates that flooding events are a disturbance factor for this species in their habitat utilization. Little Terns prefer breeding sites that are elevated, dry, and near water for foraging (Fasola and Canova, 1992; Goutner, 1990; Jeong et al., 2011a). Flooding in the estuary caused by KM events inundated mostly the islets (i.e., breeding sites).
Prey Identification of the Little Tern, Sternula albifrons (Pallas, 1764), by Applying DNA Barcoding to Fecal Materials
2022, Sustainability (Switzerland)Long-term ecological research in the Nakdong River: Application of ecological informatics to harmful algal blooms
2017, Ecological Informatics: Data Management and Knowledge Discovery: Third EditionChanges in Suitable Waterbirds’ Habitats from 1990 to 2010 in the Bielahong Watershed of Northeast China
2015, Journal of the Indian Society of Remote Sensing