University of Twente Student Theses

This website will be unavailable due to maintenance December 1st between 8:00 and 12:00 CET.

Predictive Distribution Modelling of Timon lepida in Spain

Kgosiesele, Ednah (2010) Predictive Distribution Modelling of Timon lepida in Spain.

[img] PDF
Abstract:Ecological factors determining the geographic distribution of Timon lepida are poorly known. This work modelled the potential geographic distribution of Timon lepida at two spatial scales: 1. Landscape (Andalucía) 2. Regional (Spain) and analyzed the degree to which this distribution is associated with different predictor variables (e.g. temperature, solar radiation, topography, vegetation etc). The objectives of this study are to: (1) determine the most important predictor variables influencing the spatial distribution of Timon lepida; (2) generate potential geographic distribution maps for this species and (3) compare the predictive powers of environmental variables and hyper temporal NDVI to predict this distribution. Maxent, a presence-only distribution modelling technique was used to model predicted ranges for Timon lepida, using a large dataset of 10*10 km UTM presence only records collected between 1998-2002 period over Europe and a set of biophysical variable of 1*1 km resolution. To test the average behavior of the algorithm, 10 iterative models were produced by dividing all the presence records into 70% for training and 30% for testing. Three sets of model scenarios were generated: (i) models including environmental variables, (ii) models including environmental variables and vegetation and (iii) models including vegetation indices. Model accuracy was measured with binomial tests of omission rates and the area under the curve (AUC). All models were significantly better than random by the binomial test and AUC measure. The AUC score for models built using environmental variables was always higher indicating better discrimination of suitable and unsuitable areas for the species. For the two spatial scales, environmental variables models had a superior predictive ability than vegetation models. These findings did not support our hypothesis. The results indicate that at a landscape level, topographic variables (aspect and slope) are the most important whereas at a regional scale, climatic variables (temperature seasonality, solar radiation, altitude) and hyper temporal NDVI appear to have a significant effect on this distribution pattern. The results of the present study can be an important contribution to a better understanding of the ecological requirements of the species. The conclusions would be more precise if the adequate precise high resolution environmental data is included in the future application and reliable datasets of current conditions are identified to improve results. Keywords: Maxent, AUC, Timon lepida, NDVI, environmental variables
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page