Tae-Soo Chon

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.

Self-Organizing Maps applied to ecological sciences

Abstract Ecological data are considered to be difficult to analyze because numerous biological and environmental factors are involved in a complex manner in environment–organism relationships. The Self-Organizing Map (SOM) has advantages for information extraction (i.e., without prior knowledge) and the efficiency of presentation (i.e., visualization). It has been implemented broadly in ecological sciences across different hierarchical levels of life. Recent applications of the SOM, which are reviewed here, include the molecular, organism, population, community, and ecosystem scales. Further development of the SOM is discussed regarding network architecture, spatio-temporal patterning, and the presentation of model results in ecological sciences.

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.

Stream biomonitoring using macroinvertebrates around the globe: a comparison of large-scale programs

Water quality agencies and scientists are increasingly adopting standardized sampling methodologies because of the challenges associated with interpreting data derived from dissimilar protocols. Here, we compare 13 protocols for monitoring streams from different regions and countries around the globe. Despite the spatially diverse range of countries assessed, many aspects of bioassessment structure and protocols were similar, thereby providing evidence of key characteristics that might be incorporated in a global sampling methodology. Similarities were found regarding sampler type, mesh size, sampling period, subsampling methods, and taxonomic resolution. Consistent field and laboratory methods are essential for merging data sets collected by multiple institutions to enable large-scale comparisons. We discuss the similarities and differences among protocols and present current trends and future recommendations for monitoring programs, especially for regions where large-scale protocols do not yet exist. We summarize the current state in one of these regions, Latin America, and comment on the possible development path for these techniques in this region. We conclude that several aspects of stream biomonitoring need additional performance evaluation (accuracy, precision, discriminatory power, relative costs), particularly when comparing targeted habitat (only the commonest habitat type) versus site-wide sampling (multiple habitat types), appropriate levels of sampling and processing effort, and standardized indicators to resolve dissimilarities among biomonitoring methods. Global issues such as climate change are creating an environment where there is an increasing need to have universally consistent data collection, processing and storage to enable large-scale trend analysis. Biomonitoring programs following standardized methods could aid international data sharing and interpretation.

Impact of agricultural land use on aquatic insect assemblages in the Garonne river catchment (SW France)

The impact of agricultural land use on the composition and structure of aquatic insect assemblages (i.e., taxa of Ephemeroptera, Plecoptera, Trichoptera, and Coleoptera (EPTC)) was investigated in tributary streams of the Garonne river basin, southern France. The self-organizing map (SOM) method was applied to compare both instream environmental conditions and EPTC assemblages between forest and agricultural streams. According to the SOM model, the study sites were classified into three main clusters corresponding to distinct EPTC assemblages. The SOM cluster associated with most of the agricultural sites had lower EPTC species richness and diversity. This cluster was also characterized by high levels of total dissolved solids, nitrate (NO3), and chemical oxygen demand. Overall, our study shows that agricultural streams when compared with forest streams had lower biological integrity. In accordance with the European Water Framework Directive, our results indicate that the sites most impacted by agricultural land use should be restored and that the least-impacted forest sites could serve as reference conditions.

Patterning and short-term predictions of benthic macroinvertebrate community dynamics by using a recurrent artificial neural network

Dynamic features of community data were extracted by training with a recurrent artificial neural network. Field data collected monthly from an urbanized stream consisted of densities of selected taxa in benthic macroinvertebrate communities. Sets of time-sequence data for communities were provided as the input for the network. The connectivity of computation nodes was arranged in such a way that the previous community data have recurrent feedback. In concurrence with the input of biological data, corresponding sets of environmental data such as water velocity and depth, sedimented organic matter, and volume of small substrates were also provided for the network. Through the connectivity of the network, environmental data were used as input to produce continuous, independent effects on determining community abundance. A trained pattern effectively represented the effects of habitat types and environmental impact on determining community dynamics. Short-term predictions of changes in the densities of selected taxa were made possible by a trained network after new sets of data were provided to the network.

Evidence for the Stepwise Behavioral Response Model (SBRM): The effects of Carbamate Pesticides on medaka (Oryzias latipes) in an online monitoring system

The Stepwise Behavioral Response Model (SBRM), which is a conceptual model, postulated that an organism displays a time-dependent sequence of compensatory Stepwise Behavioral Response (SBR) during exposure to pollutants above their respective thresholds of resistance. In order to prove the model, in this study, the behavioral responses (BRs) of medaka (Oryzias latipes) in the exposure of Arprocarb (A), Carbofuran (C) and Methomyl (M) were analyzed in an online monitoring system (OMS). The Self-Organizing Map (SOM) was utilized for patterning the obtained behavioral data in 0.1 TU (Toxic Unit), 1 TU, 2 TU, 5 TU, 10 TU and 20 TU treatments with control. Some differences among different Carbamate Pesticides (CPs) were observed in different concentrations and the profiles of behavior strength (BS) on SOM were variable depending upon levels of concentration. The time of the first significant decrease of BS (SD-BS) was in inverse ratio to the CP concentrations. Movement behavior showed by medaka mainly included No effect, Stimulation, Acclimation, Adjustment (Readjustment) and Toxic effect, which proved SBRM as a time-dependence model based on the time series BS data. Meanwhile, it was found that SBRM showed evident stress-dependence. Therefore, it was concluded that medaka SBR was both stress-dependent and time-dependent, which supported and developed SBRM, and data mining by SOM could be efficiently used to illustrate the behavioral processes and to monitor toxic chemicals in the environment.

Characterization of benthic macroinvertebrate communities in a restored stream by using self-organizing map

Abstract The Self-Organizing Map (SOM) was used for revealing the ecological states of streams in recovery through patterning of benthic macroinvertebrate communities. SOM was capable of showing different clusters of the sample sites in a small scale according to changes in environmental variables such as water velocity, depth, substrate roughness and the amount of silt. Community abundance correspondingly varied in different clusters of the sample sites. Within each cluster, data for community abundance were further grouped according to temporal changes in water quality. The patterns of benthic macroinvertebrate communities in the trained SOM were efficient in assessing recovery processes in the polluted sample sites, revealing the effects of river restoration projects in stream ecosystems. The study showed that spatial heterogeneity at the local level plays an important role in characterizing community patterns and consequently biological water quality assessment.

Use of an Artificial Neural Network to Predict Population Dynamics of the Forest–Pest Pine Needle Gall Midge (Diptera: Cecidomyiida)

Abstract The backpropagation algorithm in artificial neural networks was used to forecast dynamic data of a forest pest population of the pine needle gall midge, Thecodiplosis japonensis Uchida et Inouye, a serious pest in pine trees in northeast Asia. Data for changes in population density were sequentially given as input, whereas densities of subsequent samplings were provided as matching target data for training of the network. Convergence was reached, generally after 20,000 iterations with learning coefficients of 0.5–0.8. When new input data were given to the trained network, recognition was possible and population density at the subsequent sampling time could be predicted.