This thesis considers various aspects of the use of remote sensing, geographical information systems and Bayesian knowledge-based expert system technologies for broad-scale monitoring of land condition in the Western Australian wheat belt.The use of remote sensing technologies for land condition monitoring in Western Australia had previously been established by other researchers, although significant limitations in the accuracy of the results remain. From a monitoring perspective, this thesis considers approaches for improving the accuracy of land condition monitoring by incorporating other data into the interpretation process.Digital elevation data provide one potentially useful source of information. The use of digital elevation data are extensively considered here. In particular, various methods for deriving variables relating to landform from digital elevation data and remotely sensed data are reviewed and new techniques derived.Given that data from a number of sources may need to be combined in order to produce accurate interpretations of land use/condition, methods for combining data are reviewed. Of the many different approaches available, a Bayesian approach is adopted.The approach adopted is based on relatively new developments in probabilistic expert systems. This thesis demonstrates how these new developments provide a unified framework for uniting traditional classification methods and methods for integrating information from other spatial data sets, including data derived from digital elevation models, remotely sensed imagery and human experts.Two applications of the techniques are primarily considered. Firstly, the techniques are applied to the task of salinity mapping/ monitoring and compared to existing techniques. Large improvements are apparent. Secondly, the techniques are applied to salinity prediction, an application not previously considered by ++ / other researchers in this domain. The results are encouraging. Finally limitations of the approach are discussed.
| Identifer | oai:union.ndltd.org:ADTP/222458 |
| Date | January 1997 |
| Creators | Caccetta, Peter A. |
| Publisher | Curtin University of Technology, School of Computing. |
| Source Sets | Australiasian Digital Theses Program |
| Language | English |
| Detected Language | English |
| Rights | unrestricted |
Page generated in 0.0022 seconds