Successful predictions of river characteristics across England and Wales based on ordination

• Published in: Geomorphology (Amsterdam, Netherlands) Netherlands), 2013-07, Vol.194, p.121-131
• Main Authors: Vaughan, Ian P., Merrix-Jones, Faye L., Constantine, José Antonio
• Format: Article
• Language: English
• Subjects: biotopes; Channels; Earth sciences; Earth, ocean, space; England and Wales; environmental management; Exact sciences and technology; Feature based; Freshwater; Geomorphology, landform evolution; groundwater; Hydromorphology; Marine and continental quaternary; Mathematical models; Morphology; Prediction; Principal component analysis; regression analysis; River classification; River Habitat Survey; Rivers; streams; Surficial geology; surveys; Vegetation
• ISSN: 0169-555X; 1872-695X
• DOI: 10.1016/j.geomorph.2013.03.036

Predictions of river channel form under current conditions or in response to environmental or management changes and the rapid comparison of different channel reaches are important tasks in river management. River classification is a common and valuable framework to address these aims but may suffer from the necessity to force a continuum of channel morphology into discrete groups. More generally, the scope to test the ability of predictive tools has been limited because of a shortage of field data. In this study, we used principal component analysis (PCA) to identify the main sources of variation among river reaches in England and Wales based on a set of 20 variables expected to correlate with channel morphology. The PCA scores were then used to predict the distributions of a wide range of hydromorphic features based on >4000 reaches surveyed in the River Habitat Survey baseline. For comparison, the predictive ability of three pairs of variables (channel slope–discharge, slope–catchment area, and specific power–catchment area) was also tested. The PCA identified specific stream power, channel size and groundwater input as the main sources of variation among reaches. Regression models using PCA scores or paired variables were effective predictors of a range of channel characteristics, including predominant substrate, flow biotopes, and channel vegetation. Channel cross sections and anthropogenic modifications were less predictable. All of the approaches permitted simple plots of river reaches and quantitative comparisons of the (dis)similarity among individual reaches, whilst the paired variables also minimised the data requirements. Our work reiterates the value of simple, paired variables as a basis for rapidly comparing river reaches and, for the first time, quantifies the predictive ability of these approaches across a wide range of channel characteristics at a national scale. Principal component analysis provides a valuable exploratory tool for identifying the main sources of variation in complex, multivariate data from which a simplified version (e.g., specific power and area) could be adopted.