Jarkovský, JiříDušek, LadislavPavliš, PetrKubošová, KláraHodovský, JanHřebíček, JiříHřebíček, J.Ráček, J.2019-09-162019-09-162005https://dl.gi.de/handle/20.500.12116/27313Monitoring of water organisms communities become a standard approach in surface water monitoring and the analytic approach is commonly based on comparison with reference sites. There are several problems with this analysis including selection of appropriate methods with respect to nature of the data. We tested two different “nonparametric” methods for multivariate analysis of reference and standard sites association (i.e. method suitable for classification of unknown sites) consisting of i) robust true distances of sites based on several data views – biotic, static and dynamic abiotic properties and ii) non-parametric comparison of differences among sites and/or their groups (i.e. classification). The first classification algorithm is based on nearest neighbor method, the second is newly developed “centroids distance” algorithm based on percentiles of multivariate homogeneous clusters of localities. The relevance of algorithms were first tested on the set of well known reference data and the results were compared to “classical” methods in this field like discriminant analysis or neural networks. The evaluated nonparametrics methods revealed the same or better results than classical approach probably due their non-sensitiveness to problems with outliers and multivariate distribution of data and are proposed to become standard methodology for analysis of biomonitoring data in the Czech Republic. The methods were then applied on real data of biomonitoring network of the Czech Republic (macrozoobenthos communities and sites abiotic description data); marked differences among reference sites and sites under environmental stress were found and these differences are in relation to state of sites evaluated by more conventional methods or expert opinion. The presented methodology is under implementation into the multicentric expert system for analysis and management of biomonitoring data – Triton.Text/Conference Paper