Edinburgh Research Archive >
Geosciences, School of >
Geography and the Lived Environment Research Institute >
MSc Geographical Information Science thesis collection >
Please use this identifier to cite or link to this item:
Files in This Item:
||Size||Format||Philipp_Graeser_Diss_2008.pdf||File only available to GIS staff and students||1.32 MB||Adobe PDF|
|Title: ||Predicting plant species distributions on a local scale for conservation planning: A comparison of different modelling techniques|
|Authors: ||Graeser, Philipp|
|Supervisor(s): ||Legg, Colin|
|Issue Date: ||5-Dec-2008|
|Abstract: ||In nature conservation planning, knowledge about the geographical distribution of species is critical for the development of adequate protection measures. In that context, spatially explicit models of species distributions can be useful tools to identify sites where species of conservation concern are likely to be present. Thus, species distribution models can help to improve the efficiency of recording efforts for rare and endangered species or can be used to predict the possible spread of invasive plants making it easier to develop adequate counteractive measures such as concentrating efforts for pest control at important sites.
In the present study, predictive models for plant species distributions are created for the Lothian region in the southeast of Scotland. The main objective is to evaluate if floristic presence-only data available for this region can be used to produce species distribution models of sufficient quality for use in conservation planning. For that purpose, different modelling techniques are applied to a set of 26 model species comprising of rare and endangered species as well as alien plants, which have the potential for invasion of native plant communities.
In order to identify the best modelling method, logistic regression, autologistic regression and maximum entropy models are applied to the same datasets. Species distributions are modelled as a response to topography, climate, soil conditions, and landcover of the study area. Model performance is assessed using standard measures of model accuracy and precision. Further-more, possible influences of multicollinearity in model predictors, the number of available records, and species ecology on model performance are investigated.|
|Keywords: ||vascular plants|
Generalised Linear Model
|Appears in Collections:||MSc Geographical Information Science thesis collection|
Items in ERA are protected by copyright, with all rights reserved, unless otherwise indicated.