Abstract
ABSTRACT Land use/cover changes (LULC) are complicated and regionally diverse. When mining regional patterns, the use of a spatial relationship that is determined without considering the spatial correlation among geographical objects can lead to problematic results, e.g. mistakenly treating unrelated objects as adjacent. Additionally, traditional prevalence measures are unstable for uneven datasets such as LULC, wherein some land-use change types show small numbers and uneven quantities, and valuable rules for some land-use categories may be ignored. Therefore, we proposed a regional pattern mining method. First, we developed adaptive adjacent criteria, which can be automatically generated for each specific zone to define adjacency for better spatial-temporal mining. Then, a combinational decision model was built to improve the stability of the prevalence measure, which was used to filter out the insignificant spatial-temporal rules. Furthermore, we proposed two levels of land-use pattern mining, i.e. cluster-level mining and polygon-level mining, to first discover hot-spot areas where similar land-use change has occurred frequently and then to determine the location, frequency, and change time of rules related to different land-use activities. The proposed method was used for mining the dependence of land use and regional patterns on land-use changes. Results show that the proposed method can determine the spatial dependence between the land-use categories, as well as regional patterns of land-use changes. According to our research, the study area, Xinbei District, China, is undergoing land-use change involving rapid urbanization, extensive transportation construction, and losses of farmland.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call