The Effects of Spatial Heterogeneity on Land-Use Change Model: A Case Study in Hsin-Chu City
Date Issued
2008
Date
2008
Author(s)
Yang, Shu-Ting
Abstract
The processes and driving forces of landscape evolution are the main research interests of land-use and land-cover change. The linear regression model has been widely used in previous researches. However, land-use change is occasionally non-linear and heterogeneous in time and space. The linear model is supposed to estimate overall conditions without considering the uniqueness/outlier in a concerned area. This study focuses on the effects of spatial heterogeneous on the land-use change modeling. Two major effects, averaging effects and spatial autocorrelation of the residuals are explored from modeling results which correlates to spatial heterogeneity. Specifically, we used a linear model ( logistic regression model ) and a non-linear model ( decision tree model, C5.0 ) to simulate the expansion of building areas ( 1982-1994 ) in Hsin-Chu City, Taiwan. As shown in the result, the decision model performs better than the logit model. Because of the averaging effect and spatial autocorrelation of the residuals, the logit model has higher uncertainties and lower predictabilities within the discrete built hotspots. In summary, this research suggests the spatial heterogeneous effects can be effectively embedded in the non-linear models ( e.g. the decision tree model ) that lead better potential and applicability for the land-use change modeling.
Subjects
land-use change
spatial heterogeneity
Decision tree C5.0
logistic regression
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-97-R94228014-1.pdf
Size
23.53 KB
Format
Adobe PDF
Checksum
(MD5):5f8fa75e304748ba944b787781ba6702