USE OF SATELLITE IMAGERY FOR SPATIAL PATTERN ANALYSIS ON LANDUSE CHANGE
Q. Zhou1, B. Li2, A. Kurban3
Land cover change is one of the most sensitive indicators that echo the interactions between human activities and natural environment. In arid environment the land cover change often reflects the most significant impact on the environment due to human activities or natural forces. Remotely sensed data have been used for environment change study in decades and large collections of remote sensing imagery have made it possible to analyze changes of environmental elements and impact of human activities.
This study seeks an efficient and practical methodology to quantify spatial pattern of land cover change that can be related to both human activities and natural factors. The basic approach is to derive and interpret spatial pattern metrics of multi-epoch trajectories of land cover change. This method integrates multi-temporal and multi-scale remotely sensed data from various sources with a monitoring time frame of 30 years, including historical and state-of-the-art high-resolution satellite imagery. The history of land cover change for every location in the study area is traced, and nature, area extent and spatial pattern of such changes are also analyzed.
The generic approach of this study is based on post-classification comparison method, which is commonly employed in land cover change detection studies. First, multi-temporal land cover types are derived based on a unified land cover classification scheme and from classification of multi-temporal remotely sensed imagery. Categorical land cover change trajectories are then established and reclassified according to the nature and driving forces of the change. Finally, spatial pattern metrics of the land cover change trajectory classes are computed and their relationships to human activities and environmental factors are analyzed.
A case study has been conducted in the aridzone of Xinjiang Uygur Autonomous Region of China. To establish landuse change trajectories, all classified images were integrated in GIS with a raster format using ArcGIS software. Based on the classification scheme, all possible landuse change trajectories were established and classified according to whether the change was induced by human activity or merely by a natural process.
The results show that from 1994 to 2000, cropland increased from 3.7% to 12.3% of the total area of the study site, reflecting a three-fold increase in 6 years. The trajectory analysis result also shows that increasing cropland largely came from salty grass (5.2% in 1986 ® 2.9% in 2000) and water body (16.1% in 1986 ® 8.0% in 2000), indicating that most new cultivations were located in the river flat. Another significant human impact was the construction of a dam in early 1980s, resulting in flooding of vast area of grass/woodlands, which decreased from 74.4% (1976) to 69.7% (1986) in area. In contrast, water body area increased from 8.4% to 16.1%.
Four landscape metric measures have been
selected to describe spatial pattern of land cover change trajectory classes in
the study area of aridzone in western
The percentage of landscape (PLAND) of change trajectory classes is a good indicator to show the dominant process of the environmental change. In the study area, the dominant process of environmental change was still due to natural forces, but the impact of human activities has increased significantly in recent years.
Aggregation of patches of land cover change trajectory classes shows the nature of change processes. Larger aggregation often suggests more aggressive progress from a dominant driving force. In this study, the human induced changes generally show larger aggregation, indicated by lower Normalized Landscape Shape Index (NLSI), suggesting that human impact is more concentrated than the changes caused by natural forces.
Among human-induced changes, the reservoirs/ponds class was closely associated with only few of the other change trajectory classes (e.g. flooding zones), indicated by its low Interspersion and Juxtaposition Index (IJI). The cultivation classes also show low IJI with the tendency of associating with each other.
The whole human-induced trajectory class shows a lower Area Weighted Fractal Dimension Index (FRAC_AM) than those of natural changes, indicating less complexity in shape. This suggests that large, relatively regularly shaped patches are the general spatial pattern when lands were converted from natural cover types to cultivated lands or reservoirs. The comparison between the FRAC_AM of old and new cultivation also suggests that expansion of old cultivation was likely the representative process of farmland growth in the study area.
Numerous challenges, however, are also raised from this study. First, similar to all other remote sensing studies, the uncertainty in classification and change detection based on multi-temporal and multi-resolution remotely sensed images will always produce significant impact on the analytical results. To what extent that the impact may affect the final analytical results is certainly subject to further study. Secondly, the interactions and associations among different land cover change trajectories need to be further explored by introducing other landscape metrics or other spatial pattern measures. The development of more comprehensive and representative parameters to describe spatio-temporal patterns and process shown by remotely sensed data is also recommended.