Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Development of species distribution models and their application to birds in river habitats

Vaughan, Ian Phillip ORCID: 2004. Development of species distribution models and their application to birds in river habitats. PhD Thesis, Cardiff University.

[thumbnail of U584653 (1) dec page removed.pdf]
PDF - Accepted Post-Print Version
Download (8MB) | Preview


1. Distribution models are used as management tools to predict species' distributions and quantify their habitat preferences. Numerous methodological issues require further development, which are explored using the distribution of birds along rivers as a model system in which there is a need to i) develop the quantification and analysis of variation in river habitat features and ii) better quantify species' habitat preferences for conservation and prediction. These themes were linked using a combination of the British Trust for Ornithology's Waterways Breeding Bird Survey (WBBS) and the Environment Agency's River Habitat Survey (RHS), along with similar data from upland Wales and the Himalayan mountains. 2. Training data are the starting point for distribution models and their properties can have profound effects. These issues were investigated via a literature review, which identified key factors including the type of predictors chosen, the approach to environmental sampling and spatial autocorrelation in distribution patterns. Recommendations are made that should optimise model quality, whilst making the most efficient use of available resources. 3. Testing the performance of distribution models is vital. Using a review of the available literature, highlighting weaknesses in current practice, and a case study using a Himalayan river bird, recommendations are made for improved practice. Bootstrapping and independent data should be used to assess overfitting and transportability, respectively. Accuracy statistics should facilitate inter-model comparisons, examining both discrimination and calibration. Nominal presence/absence predictions are problematic: information-theoretic methods may be the most useful approach. 4. Complex habitat data, such as RHS, may create a range of problems during statistical analyses unless the sample size is large. Data reduction, using methods such as principal components analysis (PCA), is an effective solution, but the resulting axes may be difficult to interpret. Using models built with Welsh river bird-RHS data, I compared the interpretability and predictive efficacy of PCA used in its conventional form against PCA preceded by the clustering of RHS variables that referred to the same ecological factors. The two approaches produced similar predictive performance but habitat indices produced by the latter were easier to interpret. A variant of PCA devised for qualitative data was also examined, and benefited RHS analyses involving ordinal variables. 5. Predictive models for 28 river birds, built with the WBBS, represent the first quantitative study linking detailed river habitat data with river bird distributions across the UK. Accuracy varied widely, with better performance for species associated with the river channel, rather than floodplain habitats, reflecting the relative coverage of these river features in RHS. By using variable clustering, the likelihood of species occurrences could be easily related to the observed habitat. 6. By utilising important methodological developments, this project provides important evidence that RHS forms an effective basis for relating many river birds to their habitats, and that when used in conjunction with the WBBS, could bring valuable benefits to river bird conservation. More generally, the work illustrates how RHS can describe variations in river structure and anthropogenic modification in a manner that is relevant to riverine organisms, along with transferable methods for describing and modelling the resulting relationships.

Item Type: Thesis (PhD)
Status: Unpublished
Schools: Biosciences
ISBN: 9781303200236
Date of First Compliant Deposit: 30 March 2016
Last Modified: 11 Dec 2023 15:52

Actions (repository staff only)

Edit Item Edit Item


Downloads per month over past year

View more statistics