Clustering of ant communities and indicator species analysis using self-organizing maps

Abstract

To understand the complex relationships that exist between ant assemblages and their habitats, we performed a self-organizing map (SOM) analysis to clarify the interactions among ant diversity, spatial distribution, and land use types in Fukuoka City, Japan. A total of 52 species from 12 study sites with nine land use types were collected from 1998 to 2012. A SOM was used to classify the collected data into three clusters based on the similarities between the ant communities. Consequently, each cluster reflected both the species composition and habitat characteristics in the study area. A detrended correspondence analysis (DCA) corroborated these findings, but removal of unique and duplicate species from the dataset in order to avoid sampling errors had a marked effect on the results; specifically, the clusters produced by DCA before and after the exclusion of specific data points were very different, while the clusters produced by the SOM were consistent. In addition, while the indicator value associated with SOMs clearly illustrated the importance of individual species in each cluster, the DCA scatterplot generated for species was not clear. The results suggested that SOM analysis was better suited for understanding the relationships between ant communities and species and habitat characteristics.

The dynamics of ant mosaics in tropical rainforests characterized using the Self‐Organizing Map algorithm

“…The SOM (Kohonen, ), a heuristic model used for exploring and representing the linear and nonlinear relationships in high‐dimensional data sets, has also been used to characterize different patterns of communities (Céréghino et al ., ; Chon, and papers cited therein), but rarely in studies on terrestrial ecosystems (Groc et al ., ; Delabie et al ., ; Park et al ., ). This technique uses an array of prototype vectors that analyze the variance structure of the data set and converts relationships between the data into graphical representations in two dimensions.…”

“…Because most ant species have sedentary colonies, they are sensitive to changes in the environment and so have been used as indicators of these changes (Park et al ., ). For arboreal ants, the changes correspond to different tree sizes (ontogenic succession of TDAs as the trees grow; Dejean et al ., ) and to the different tree species related to the succession process leading to climactic forests (for the succession in Africa, see White, ).…”

“…However, the data treated are usually bulky, nonlinear, and complex, showing noise, redundancy, outliers, and complex interactions between explanatory and response variables, rendering conventional multivariate analyses ineffective (but see Taylor & Adedoyin, who used a principal component analysis). These difficulties can be overcome by using the Self‐Organizing Map (SOM, neural network; see Park et al ., who compared three ordination analyses). We therefore used the SOM to illustrate the succession of ant species as the trees grow and their selection between 3 tree species.…”

How territoriality and host-tree taxa determine the structure of ant mosaics

“…The occurrence probability for each ant species in a given area is represented on the SOM map in gray scale, allowing us to analyze the effect of each species on the patterning input dataset (trees). A detailed description of the SOM is provided in Appendix S2 (BElectronic Supplementary Material^; for the ant distribution, see Park et al 2003Park et al , 2014Dejean et al 2015).…”

Winter Activity of Ants in an Urban Area of Western Japan

“…The five most abundant species in this study are also the most common ant species observed in the study area during spring to autumn (Ogata et al, 1998;Park et al, 2014aPark et al, , 2014b. The number of ant species observed at each study site represented ca.…”