Premium
A comparison of environmental and biological site classifications for the prediction of macroinvertebrate communities of lakes in Northern Ireland
Author(s) -
Neale Martin W.,
Rippey Brian
Publication year - 2007
Publication title -
aquatic conservation: marine and freshwater ecosystems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.95
H-Index - 77
eISSN - 1099-0755
pISSN - 1052-7613
DOI - 10.1002/aqc.858
Subject(s) - water framework directive , multivariate statistics , variation (astronomy) , environmental science , biological data , ecology , geography , statistics , mathematics , water quality , biology , physics , genetics , astrophysics
1. The Water Framework Directive (WFD) requires member states to establish ‘type‐specific biological reference conditions’ for the assessment of surface waters and describes a method for establishing such biological targets. The methodology described in the Directive is based on geographic and physical characteristics and similar to the ‘multimetric’ approach used widely in North America, but contrasts with the biological classification of the ‘multivariate’ approach which is commonly used in Europe. 2. An investigation was made of the relative efficiency of the multimetric and multivariate classification approaches in partitioning the observed biological variation of the macroinvertebrate communities of 22 minimally disturbed lakes in Northern Ireland. Furthermore, the accuracy of predictive models based on the two approaches was compared. 3. The WFD environmental typologies partitioned the observed biological variation poorly with a maximum ANOSIM R ‐value of 0.216 compared with 0.609 for a multivariate biological site classification. The WFD System B approach partitioned more variation than the System A approach. 4. The use of a predictive model, based on any of the site classification approaches, resulted in more accurate faunal predictions than a null model. However, the biological model produced the most accurate predictions. 5. Canonical correspondence analysis showed that the environmental variables used in the biological classification model explained more variation (28%) than the best‐performing environmental classification model (24%). The most important variable was pH, followed by surface area and altitude. The variables used in all approaches explained more variation when used as continuous rather than categorized parameters. 6. The results of this work, the first such comparative study of predictive models in lakes, concur with similar studies in streams. It may be concluded that large‐scale environmental classifications are unreliable as a method for setting type‐specific reference targets for the ecological assessment of fresh waters. Copyright © 2007 John Wiley & Sons, Ltd.