Modelling spatio-temporal pattern of landuse change using multitemporal remotely sensed imagery

Remotely sensed data is the most important data source for environmental change study over the past 40 years. Since large collections of remote sensing imagery have been acquired in a time frame of successive years, it is now possible to study long-term spatio-temporal pattern of environmental change and impacts of human activities. This study seeks an efficient and practical methodology for landuse monitoring and spatio-temporal pattern analysis by integrating multitemporal remotely sensed data in a monitoring time frame of 13 years at the middle reach of Tarim River in the aridzone of China. Multi-source and multi-scale remotely sensed images are used, including multispectral images acquired by Landsat 5 and 7, China-Brazil Earth Resources Satellite (CBERS) and Beijing-1 (BJ-1). The temporal trajectories of landuse change have been established for analysing its spatial pattern for a better understanding of the human impact on the fragile ecosystem of China's arid environment. This study analyzed spatial pattern of landuse change trajectories based on the post-classification comparison method. All images were classified into 5 to 6 classes, which were then combined into two main classes, namely, farmland and the others. Area statistics and temporal trajectories of changed farmland were then derived using the classification results. The result shows that in the study period of 13 years, the farmland has increased over two times with an annual growth rate of over 10%. It is also shown that farmland abandon was significant in some areas due to some environmental issues such as shortage of water resource and salinity. Using the method, one can re-establish the history of landuse change and related such change with other environmental and socio-economic data, so as to gain better understanding on the response of natural environment to the human impact that may be introduced as the consequence of economic development and government polices.