Repository logo
 

Geospatial Process Modelling for Land Use Cover Change

Files

Whole thesis(13.32 MB)

Date

2013

Authors

Supervisor

Sallis, Philip
Shanmuganathan, Subana

Item type

Thesis

Degree name

Doctor of Philosophy

Journal Title

Journal ISSN

Volume Title

Publisher

Auckland University of Technology

Abstract

Human activities and effects of global warming are increasingly changing the physical landscape. In view of this researchers have developed models to investigate the cause and effect of such variations. Most of these models were developed for specific locations with spatial variables causing change for that location. Also the application areas of these models are mainly binary transitions, not complex models which involve multiple transitions, for example deforestation models which deal with the transition from forest lands to non-forest areas and urban growth transition from non-urban areas to urban. Moreover these land simulation models are closed models because spatial variables cannot be introduced or removed, rather modellers can only modify the coefficients of the fixed variables. Closed models have significant limitations largely because geospatial variables that cause change in a locality may differ from one another. Thus with closed models the modellers are unable to measure and test the significance of variables before their inclusion. This work investigated existing land use cover change (LUCC) models and aimed to find a geospatial workflow process modelling approach for LUCC so that the influence of geospatial variables in LUCC could be measured and tested before inclusion. The derived geospatial workflow process was implemented in DINAMICA EGO, an open generic LUCC modelling environment. For the initial calibration phase of the process the Weight of Evidence (WoE) method was used to measure the influence of spatial variables in LUCC and also to determine the variables significance. A Genetic Algorithm was used to enhance the WoE coefficients and give the best fitness of the coefficients for the model. The model process was then validated using kappa and fuzzy similarity map comparison methods, in order to quantify the similarity between the observed and simulated spatial pattern of LUCC. The performance of the workflow process was successfully evaluated using the Auckland Region of New Zealand and Rondônia State of Brazil as the study areas. The Auckland LUCC model was extended to demonstrate vegetative carbon sequestration scenario. Ten transitions were modelled involving seven Land Use Cover (LUC) classes and a complex dynamic LUCC for Auckland was generated. LUC maps for 1990 and 2000 were used to calibrate the model and 2008 was used to validate the model. The static spatial variables tested were road networks, river networks, slope, elevation, hillshade, reserved lands and soil. The hillshade and soil variables were found to have no significant impact in the LUCC for the Auckland area, therefore they were excluded from the model. If a closed model had been used these insignificant variables would have been included. The calibration phase revealed that wetland and cropland LUC areas in Auckland have not changed between 1990 and 2000. The validated LUCC model of Auckland, served as a foundation for simulating annual LUC maps for advance modelling of Carbon Sequestration by vegetation cover. In order to test the generic nature of the workflow process model a second case study was introduced that had a different data resolution, area extent and fewer LUC transitions. Compared to Auckland, the new Rondônia case study was a simple LUCC model with only one transition, with coarse data resolution (250m) and large area extent. The evaluation of the Rondônia LUCC model also gave good result. It was then concluded that the derived workflow process model is generic and could be applied to any location.

Description

Keywords

Land use cover change, Cellular Automata, Weight of evidence, Genetic algorithm, DINAMICA, Environmental modelling

Source

DOI

Publisher's version

Rights statement

Permanent link

https://hdl.handle.net/10292/6026

Collections

Doctoral Theses