Combined method of image and cluster analysis to estimate the structural diversity of fungal communities

Abstract

A method was designed to simultaneously estimate the diversity of fungal communities and growth-related indices. It is based on cluster analysis of image data of fungal colonies growing in agar plates. A total of 20 features (colony area, mean colours of the upper and lower colony sides as well as features describing the colour distribution) were recorded. The relative differentiating powers of these features were determined. Since the intensity of colours and their ratio to one another were dependent on the camera distance, colours were recorded at a constant distance of 36 cm. In contrast, growth rates were taken at a distance of 68 cm. Identification of a reproducible clustering level was favoured by the fact that, when similarity was reduced in dendrograms, the evenness in type distribution decreased at a higher rate than the number of image types and, in addition, exhibited distinct minima and intermittent maxima. In all data sets examined, abundant and rare image types, including singletons, always occurred at the first evenness minimum. Comparing fungal samples from a metal-contaminated arable soil and the relevant control showed the richness in image types and three growth-related indices (the average and the variation coefficient of growth rates as well as the portion of isolates with growth rates >3 mm day−1) to be greater in the control. Of these four indices, the former was the least sensitive and the latter the most sensitive indicator of metal stress.