Use this URL to cite or link to this record in EThOS:
Title: Factors associated with abundance and extinction risk in birds and hoverflies at local and global scales : an analysis using 'random forests'
Author: Mansour, Ali
ISNI:       0000 0004 2718 476X
Awarding Body: Manchester Metropolitan University
Current Institution: Manchester Metropolitan University
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
The loss of biodiversity is one of the most urgent crises which the world faces. There have been many known extinctions and probably many more unnoticed extinctions in recent decades and the number of species at risk of extinction is increasing. Extinction is the final consequence of serious environmental and/or genetic stress for a population and the species which are most likely to become extinct in the near future are those which, according to the mCN Red List, are the most highly endangered ones at the present time. It is still inadequately clear why some species become endangered while others seem to be safe from extinction. It is, however, clear that extinction does not happen randomly and there are some characteristics e.g. body size and living on islands, which seem to make species more prone to population decline and extinction. Factors which promote species extinction may differ from one species to another, and some factors may have more impact in initially threatening the species e.g. habitat loss and others causing their final extinction e.g. introduced predators. Also, population decline of species is probably due to the interaction of many intrinsic and extrinsic human impact factors. Although invertebrates make up the vast majority of species on Earth, most studies of extinction risk have concentrated on vertebrates. This study concentrated on a number of bird groups and also one invertebrate group, the hoverflies [Syrphidae]. Most studies of extinction risk have attempted to understand the extinction process but few have extended their analyses to address the more practical question for predicting which species might become threatened in the future. Previous studies have also concentrated on global extinction risk and have not addressed the issue of which factors may be related to decline at local levels. In this study I examined the factors which might be associated with current extinction risk in a number of bird and hoverfly groups and attempted to predict which species could become threatened in the future. My aim was to see if the same factors are associated with risk (as measured by their IUCN category or local abundance) in a number of animal groups and also to see if the factors which are associated with extinction risk at the global scale are the same as those related to low population abundance at a local scale. A number of different techniques of analysis have been used in extinction-risk studies and these techniques differ in their suitability and ease of use. I used a relatively new classification method called Random Forests (RF) in addition to other traditional methods. RF is becoming popular in many fields of science because it is a non-parametric classification which performs very well in comparison to other methods; it can cope with mixed variable types, non-linear and complex interactions and even highly correlated predictors. In most analyses of this study, RF allocated more than 70% of species correctly to their current IUCN threat category and predicted which species might become threatened or more threatened in the future (because they share the same biological characteristics as those which are in a particular threat category). I suggest these latter species which should be carefully scrutinized in the future. In general, the factors which are associated with threat in birds and hoverflies are body size (usually large but sometimes small species are more threatened), geographical range and living on islands and association with particular habitat types. These factors are correlated with extinction risk at local scales as well as at the global scale. Although groups of biological traits can successfully classify and predict threat status, there are actually rather few relationships which hold across all of the bird groups. The RF appears to be a very useful tool for both accurately classifying species and predicting species which could be threatened in the future. It will always be a complex task to understand and predict risk but this method is the first which can handle large numbers of interacting extrinsic factors. The interaction of biological and geographical factors and extrinsic human impact factors such as climate change will be a particularly important area of research in the future.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available