Premium
Ecoregionalization classification of wetlands based on a cluster analysis of environmental data
Author(s) -
Lechner Alex M.,
McCaffrey Nic,
McKenna Phill,
Venables William N.,
Hunter John T.
Publication year - 2016
Publication title -
applied vegetation science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.096
H-Index - 64
eISSN - 1654-109X
pISSN - 1402-2001
DOI - 10.1111/avsc.12248
Subject(s) - wetland , vegetation (pathology) , geography , range (aeronautics) , temperate climate , vegetation classification , cluster analysis , environmental science , cluster (spacecraft) , silhouette , ecology , physical geography , computer science , statistics , biology , mathematics , composite material , programming language , medicine , materials science , pathology , machine learning
Aim Effective vegetation conservation requires reasonable certainty regarding the distribution, extent and classification of plant communities and ecoregions for assessing rarity. In this paper we describe a multivariate clustering approach based on environmental data for objectively defining temperate treeless palustrine wetland communities. Location New South Wales ( NSW ), Australia. Methods In NSW no comprehensive state‐wide map of wetland vegetation exists, with more than 200 vegetation maps produced by local and state governments at a range of spatial resolutions and extents. Using the available vegetation spatial data, we produced a composite map which identified 6323 wetlands >1 ha. We then used the partitioning around medoids cluster analysis method for grouping wetlands based on 12 climate, topography, geology and soils spatial data layers and the wetland locations. We tested a range of cluster numbers from three to 20, and assessed the stability of the clustering by calculating mean silhouette widths. The derived classes were then characterized in terms of number of individual wetlands and their area, and also the number and area of individual wetlands found within protected areas such as national parks. Results We found a peak in the mean silhouette width at 11 clusters, indicating that this was the optimal number of clusters for classifying the wetland data. We produced maps of wetland density for each of the 11 clusters and described the mean and mode environmental characteristics of each cluster. Each cluster represented a unique combination of environmental variables. For example, wetlands in cluster 2 are typically in the south, in areas of low evaporation and low average temperatures. An assessment of rarity found that wetlands in the largest cluster class had an areal extent of 14 644 ha, compared to 1414 ha for the smallest cluster. All but one of the clusters had part of their range within protected areas. Conclusions Clustering environmental variables is an important but underutilized method for characterizing vegetation communities/ecoregions such as wetlands spatially. This approach can be used to produce objective, repeatable and defensible wetland community maps for assessing rarity.