The management of vegetation classifications with fuzzy clustering

Abstract

Questions: Does fuzzy clustering provide an appropriate numerical framework to manage vegetation classifications? What is the best fuzzy clustering method to achieve this? Material: We used 531 releve´s from Catalonia (Spain), belonging to two syntaxonomic alliances of mesophytic and xerophytic montane pastures, and originally classified by experts into nine and 13 associations, respectively. Methods: We compared the performance of fuzzy C-means (FCM), noise clustering (NC) and possibilistic C-means (PCM) on four different management tasks: (1) assigning new releve´data to existing types;(2) updating types incorporating new data; (3) defining new types with unclassified releve´s; and (4) reviewing traditional vegetation classifications. Results: As fuzzy classifiers, FCM fails to indicate when a given releve´does not belong to any of the existing types; NC might leave too many releve´s unclassified; and PCM membership values cannot be compared. As unsupervised clustering methods, FCM is more sensitive than NC to transitional releve´s and therefore produces fuzzier classifications. PCM looks for dense regions in the space of species composition, but these are scarce when vegetation data contain many transitional releve´s. Conclusions: All three models have advantages and disadvantages, although the NC model may be a good compromise between the restricted FCM model and the robust but impractical PCM model. In our opinion, fuzzy clustering might provide a suitable framework to manage vegetation classifications using a consistent operational definition of vegetation type. Regardless of the framework chosen, national/regional vegetation classification panels should promote methodological standards for classification practices with numerical tools.

Expanding an existing classification of New Zealand vegetation to include non-forested vegetation

“…We ran Noise Clustering iteratively, with each iteration having the 29 original woody alliances plus n new alliances, where n = 1 to 50 clusters using δ = 0.84 and m = 1.1. In all cluster analyses, the 29 centroids of the previously defined woody alliances were used as fixed elements so that the newly defined alliances would be as distinct from them as possible (De Cáceres et al 2010). We repeated the same analyses at the association level using δ = 0.76, m = 1.1 and the 79 centroids of the previously defined woody associations as fixed elements.…”

“…Our approach to semi-supervised clustering was based upon a fuzzy classification algorithm called Noise Clustering (Dave 1991;De Cáceres et al 2010;Wiser & De Cáceres 2013). The use of fuzzy classification explicitly addresses the understanding that vegetation composition varies along a continuum.…”

“…As such, one would be reluctant to develop a quantitatively based national classification until there was comprehensive coverage of vegetation plot data (the term 'plot' is used here in a broad sense to incorporate both quadrat and transect data) collected in a consistent way. A framework has recently been developed, termed 'semisupervised clustering' (Tichý et al 2014), that allows new data to be incorporated into a pre-existing classification, while retaining types defined in the original classification (De Cáceres et al 2010). For New Zealand's woody vegetation, adopting this framework allowed the 17 vegetation alliances described by Wiser et al (2011) to be retained and also related to a finer thematic level when new data were analysed, and 12 new alliances and 79 associations to be defined (Wiser & De Cáceres 2013; the names 'alliance' and 'association' follow usage in Europe and North America; Peet & Roberts 2013).…”

“…Traditionally, quantitative vegetation classifications have been static; when new data are incorporated the original classification data must be reanalysed, which destroys the original hierarchy and classes themselves (De Cáceres et al 2010;). As such, one would be reluctant to develop a quantitatively based national classification until there was comprehensive coverage of vegetation plot data (the term 'plot' is used here in a broad sense to incorporate both quadrat and transect data) collected in a consistent way.…”

Application and comparative performance of network modularity algorithms to ecological communities classification

“…These include the character species, whose occurrence is typical to specific plant communities, and the differential species, whose occurrence can be used to distinguish related plant communities, but are not limited to a single community. Computer based methods including network applications have been used in the classification of plant communities [7][8][9][10][11]. De Cáceres et al [7,8] tested the performance of fuzzy clustering for the classification of communities in mesophytic and xerophytic pastures of Spanish highlands.…”