Young-Seuk Park

Applications of artificial neural networks for patterning and predicting aquatic insect species richness in running waters

Abstract Two artificial neural networks (ANNs), unsupervised and supervised learning algorithms, were applied to suggest practical approaches for the analysis of ecological data. Four major aquatic insect orders (Ephemeroptera, Plecoptera, Trichoptera, and Coleoptera, i.e. EPTC), and four environmental variables (elevation, stream order, distance from the source, and water temperature) were used to implement the models. The data were collected and measured at 155 sampling sites on streams of the Adour–Garonne drainage basin (South-western France). The modelling procedure was carried out following two steps. First, a self-organizing map (SOM), an unsupervised ANN, was applied to classify sampling sites using EPTC richness. Second, a backpropagation algorithm (BP), a supervised ANN, was applied to predict EPTC richness using a set of four environmental variables. The trained SOM classified sampling sites according to a gradient of EPTC richness, and the groups obtained corresponded to geographic regions of the drainage basin and characteristics of their environmental variables. The SOM showed its convenience to analyze relationships among sampling sites, biological attributes, and environmental variables. After accounting for the relationships in data sets, the BP used to predict the EPTC richness with a set of four environmental variables showed a high accuracy (r=0.91 and r=0.61 for training and test data sets respectively). The prediction of EPTC richness is thus a valuable tool to assess disturbances in given areas: by knowing what the EPTC richness should be, we can determine the degree to which disturbances have altered it. The results suggested that methodologies successively using two different neural networks are helpful to understand ecological data through ordination first, and then to predict target variables.

Patternizing communities by using an artificial neural network

Abstract The Kohonen network, an unsupervised learning algorithm in artificial neural networks, performs self-organizing mapping and reduces dimensions of a complex data set. In this study, the network was applied to clustering and patternizing community data in ecology. The input data were benthic macroinvertebrates collected at study sites in the Suyong river in Korea. The grouping resulting from learning by the Kohonen network was comparable to the classification by conventional clustering methods. Through patternizing, the network showed a possibility of producing easily comprehensible low-dimensional maps under the total configuration of community groups in a target ecosystem. Changes in spatio-temporal community patterns may also be traced through the recognition process.

Self-Organizing Map

Ecological data are considered difficult to analyze because numerous biological and environmental factors are involved in ecological processes in a complex manner. The self-organizing map (SOM) has been an efficient alternative tool for analyzing ecological data without a priori knowledge. The unsupervised learning process was applied to provide a comprehensive view on ecological data through the use of ordination and classification. The SOM extracts information from multidimensional data and maps it onto two- or three-dimensional space. The network structure and learning algorithm are discussed to reveal the adaptive convergence of connection weights among computation nodes (i.e., neurons). Examples are provided to demonstrate the environmental impact gradient and sample unit clustering. SOM visualization is also presented to show profiles of the corresponding taxa and environmental variables.

Review of the Self-Organizing Map (SOM) approach in water resources: Commentary

We provide some additional input and perspectives on Kalteh et als review of the Self-Organizing Map (SOM) approach (Environ. Model. Softw. (2008), 23, 835-845). Map size selection is a key issue in SOM applications. Although there is no theoretical principle to determine the optimum map size, quantitative indicators such as quantization error, topographic error and eigenvalues have proven to be relevant tools to determine the optimal number of map units. Second, one of the most innovative applications of the SOM is the possibility of introducing a set of variables (e.g., biological) into a SOM previously trained with other variables (e.g. environmental). This can be achieved by calculating the mean value of each environmental variable in each output neuron of a SOM trained with biological variables, or by using a mask function to give a null weight to the biological variables, whereas environmental variables are given a weight of 1 so that the values for biological variables are visualized on a SOM previously trained with environmental variables only. We conclude that our different levels of expertise represent an opportunity for stimulating cross-fertilisation in the vast field of water research rather than simply yielding a collection of case studies to be re-examined.

Predicting the species richness of aquatic insects in streams using a limited number of environmental variables

Artificial neural networks were used to predict the species richness of 4 major orders of aquatic insects (Ephemeroptera, Plecoptera, Trichoptera, and Coleoptera, i.e., EPTC) at a site, using a limited number of environmental variables. EPTC richness was recorded in the Adour–Garonne stream system (France), at 155 unstressed sampling sites, which were characterized using 4 environmental variables: elevation, stream order, distance from the source, and maximum water temperature. The EPTC and environmental data were first computed with the Self-Organizing Map (SOM) algorithm. Then, using the k-means algorithm, clusters were detected on the map and the sampling sites were classified separately for each variable and for EPTC richness. Four clusters could be identified on the SOM map, according to the 4 environmental variables, and this classification was chiefly related to stream order and elevation (i.e., the longitudinal location of sampling sites within a stream system). Similarly, 4 subsets were derived from the SOM according to a gradient of EPTC richness. There was also a high coincidence between observed (field data) and calculated (predicted from the output neurons of the SOM) species richness in each taxonomic group. Species richness relationships between Ephemeroptera, Trichoptera, and Coleoptera for both observed and predicted data were highly significant. However, correlation coefficients among species richness of Plecoptera and the other groups were low. Last, a multilayer perceptron neural network, trained using the backpropagation algorithm, was used to predict EPTC richness (output) using the 4 above-mentioned environmental variables (input). The model showed high predictability (r = 0.91 and r = 0.61 for training and test data sets, respectively), and a sensitivity analysis revealed that elevation and stream order contributed the most among the 4 input variables. Prediction of species richness using a limited number of environmental variables is, thus, a valuable tool for the assessment of disturbance in a given area. The degree to which human activities have altered EPTC richness can be determined by knowing what the EPTC richness should be under undisturbed conditions in a given area.

Biological early warning system based on the responses of aquatic organisms to disturbances: a review.

Aquatic ecosystems are subject to a number of anthropogenic disturbances, including environmental toxicants. The efficient monitoring of water resources is fundamental for effective management of water quality and aquatic ecosystems. Spot sampling and continuous water quality monitoring based on physicochemical factors are conducted to assess water quality. However, not all contaminants or synergistic and antagonistic toxic effects can be determined by solely analyzing the physicochemical factors. Thus, various biotests have been developed using long-term and automatic observation studies based on the ability of the aquatic organisms to continuously sense a wide range of pollutants. In addition, a biological early warning system (BEWS) has been developed based on the response behaviors of organisms to continuously detect a wide range of pollutants for effective water quality monitoring and management. However, large amounts of data exhibiting non-linearity and individual behavioral variation are continuously accumulated over long-term and continuous behavioral monitoring studies. Thus, appropriate mathematical and computational data analyses are necessary to manage and interpret such large datasets. Here, we review the development and application of BEWS by using various groups of organisms and the computational methods used to process the behavioral monitoring data.

Determining temporal pattern of community dynamics by using unsupervised learning algorithms

Analysis of patterns of temporal variation in community dynamics was conducted by combining two unsupervised artificial neural networks, the Adaptive Resonance Theory (ART) and the Kohonen network. The field data used as input for training represented monthly changes in density and species richness in selected taxa of benthic macroinvertebrates collected in the Suyong River in Korea from September 1993 to October 1994. The sampled data for each month was initially trained by ART, the weights of which preserved conformational characteristics among communities during the process of the training. Subsequently these weights were rearranged sequentially from 2 to 5 months, and were provided as input to the Kohonen network to reveal temporal variations in communities. The network was then able to extract the features of community dynamics in a reduced dimension covering the specified input period.

Modelling the factors that influence fish guilds composition using a back-propagation network: Assessment of metrics for indices of biotic integrity

Abstract Fish assemblages are reckoned as indicators of aquatic ecosystem health, which has become a key feature in water quality management. Under this context, guilds of fish are useful for both understanding aquatic community ecology and for giving sound advice to decision makers by means of metrics for indices of biotic integrity. Artificial neural networks have proved useful in modelling fish in rivers and lakes. Hence, this paper presents a back-propagation network (BPN) for modelling fish guilds composition, and to examine the contribution of five environmental descriptors in explaining this composition in the Garonne basin, south west France. We employed presence–absence data and five variables: altitude, distance from the river source, surface of catchment area, annual mean water temperature, and annual mean water flow. We found that BPN performed better for predicting species richness of guilds than multiple regression models. The standardised determination coefficient of observed values against estimated values was used to characterise model performance; it varied between 0.55 and 0.82. Some models showed high variability which was presumably due to spatial heterogeneity, temporal variability or sampling uncertainty. Surface of catchment area and annual mean water flow were the most important environmental descriptors of guilds composition. Both variables imply human influence (i.e. land-use and flow regulation) on certain species which are of interest to environmental managers. Thus, predicting guilds composition with a BPN from landscape variables may be a first step to assess metrics for water quality indices in the Garonne basin.

Characterizing effects of landscape and morphometric factors on water quality of reservoirs using a self-organizing map

Understanding the pattern of reservoir water quality in relation to morphometry and other landscape characteristics can provide insight into water quality management. We investigated the water quality of 302 reservoirs distributed nationwide in Korea by classifying them using a self-organizing map (SOM), examining how hydrogeomorphometry variables are related to reservoir water quality, and evaluating the effects of variables at different categories including geology, land cover, hydromorphology, and physicochemistry on reservoir water quality through a theoretical path model. The SOM classified the reservoirs into six clusters, from least to most polluted, with differences in physicochemical and hydrogeomorphometry variables between clusters. Water quality exhibits strong relationships with the proportions of urban, agricultural, and forest land cover types in the watersheds. Finally, our results revealed that hydrogeomorphometry of reservoirs and percentages of land cover types within watersheds have a considerable impact on the water quality of adjacent aquatic ecosystems. Water quality of agricultural reservoirs was characterized with variables in multiple spatial scales.SOM classified 302 agricultural reservoirs into six different clusters based on water quality.The water quality also has strong relations with the proportions of land cover types in watersheds.The hydrogeomorphometry of reservoirs have a considerable impact on the water quality of adjacent aquatic ecosystems.

Use of unsupervised neural networks for ecoregional zoning of hydrosystems through diatom communities: case study of Adour-Garonne watershed (France)

Knowing that diatoms are good indicators of stream ecological conditions, the aim of our research program was to test on a pilot data-set the interest and efficiency of using a Self-Organizing Map (SOM) as an ordination technique to determine and to classify types of river ecosystems. Such neural networks have already been successfully used for other aquatic communities patterning. Diatoms, waterchemistry and stream morpho-dynamical parameters were characterised for 49 non impacted sampling stations spread over the Adour-Garonne watershed (South-Western France). Combining the SOM to the Structuring Index we selected in a second step the most relevant species (called structuring species) influencing this typology. In this way, three main homogeneous regions were characterised, as regards to diatom communities and abiotic parameters, allowing us to meet the Water Framework Directive requirements concerning stream ecoregional classification.