Modelling of habitat conditions by self-organizing feature maps using relations between soil, plant chemical properties and type of basaltoides

Abstract

The paper shows the use of Kohonen's network for classification of basaltoides on the base of chemical properties of soils and Polypodium vulgare L. The study area was Lower Silesia (Poland). The archival data were: chemical composition of types of basaltoides from 89 sites (Al₂O₃, CaO, FeO, Fe₂O₃, K2O, MgO, MnO, Na₂O, P₂O₅, SiO₂ and TiO₂), elements contents in soils (Cd, Co, Cu, Fe, Mn, Mo, Ni, Pb, S, Ti and Zn) and leaves of P. vulgare (Ca, Cd, Co, Cu, Fe, K, Mg, Mn, Mo, N, Ni, P, Pb, S, Ti and Zn) from 20 sites. Descriptive statistical parameters of soils and leaves chemical properties have been shown, statistical analyses using ANOVA and relationships between chemical elements were carried out, and SOFM models have been constructed. The study revealed that the ordination of individuals and groups of neurons in topological maps of plant and soil chemical properties are similar. The constructed models are related with significantly different contents of elements in plants and soils. These models represent different chemical types of soils and are connected with ordination of types of basaltoides worked out by SOFM model of TAS division. The SOFM appeared to be a useful technique for ordination of ecological data and provides a novel framework for the discovery and forecasting of ecosystem properties.