A.S. Mahiny (Iran)
Vegetation Change, Neural Networks, Simulation, Boorowa, Australia
Past changes in vegetation were detected using post classification comparison of the classified Landsat MSS 1973 and TM 2000 images. Variables including bio physical factors, image derivatives and patch and landscape metrics of the remnant vegetation were used as predictors against the detected change. Simulating change was carried out using neural networks method. Performance of the neural networks modeling and the effects of the predictor variables, were assessed using ROC, multi-resolution goodness of fit and visual evaluation. A series of reduced variable models were used to appraise the effect of each individual predictor variable which showed that agricultural activity surrounding the patches, the ratio of MSS band 4 to NDVI and slope are the three most important factors related to vegetation decrease in the area of study. Proper management options can be devised for the vegetation patches using knowledge about the set of important predictor variables. With this knowledge available, modifications can be applied to the important factors with the aim of circumventing their negative effects on the vegetation.
Important Links:
Go Back
