Analysing spatio-temporal pattern of changing farmland in China's arid zone

Multi-temporal imagery has been used for landuse and land cover change detection since the very early stage of remote sensing technology. As large amount of remotely sensed data have been collected, historical land cover changes and change patterns can be reconstructed by a time series recorded by images. This paper reports a study on the methodology for quantifying spatial pattern of land cover changes in an arid zone during a 13-year period and the attempts to identify the key factors for these changes. The approach is based on the post-classification method. Multi-temporal images were independently classified to establish change trajectories for the farmland land cover type. A set of class-level metrics is then calculated on the trajectory classes, including Percentage of Landscape (PLAND), Normalized Landscape Shape Index (NLSI), Interspersion and Juxtaposition Index (IJI) and Area Weighted Fractal Dimension Index (FRAC_AM). These metrics and their relationship were shown as good indicators on the environmental impact in the fragile ecosystem due to the rapid expansion of farmland accompanied with the limited water resources. The results show that spatial pattern metrics of land cover change trajectories provide an effective measurement on landscape changes, which can further be interpreted for agriculture planning and management.