Analysis of Spatiotemporal Aggregation of Land Use Change Processes Considering the Shape of Land Units

Abstract

The processes of land use and cover change (LUCC) are highly diverse and complex, being heavily influenced by natural factors, economic factors, and other related factors. These changes have a significant impact on ecological environments and landscapes, and serve as a reflection of human activity, limited by natural factors. As a result, LUCC has been widely studied across multiple scientific disciplines. In particular, considerable progress has been made with regard to traditional methods of analyzing land use structures, which focus on the overall differences in the land use structure in each spatiotemporal snapshot. However, these methods have overlooked the continuity in the evolution of each land use unit between different snapshots, impeding the development of a comprehensive model for the spatiotemporal evolution of land use processes. In this work, land use change process (LUCP)—constructed using multiple land use data points from different points in time—was employed as the basis to develop a method to measure the spatiotemporal distance between irregular land patches in evolution sequences based on LUCP. Furthermore, the spatiotemporal distribution model was analyzed using Monte Carlo simulation and measurements of the shortest spatiotemporal distance of LUCP. This work employs land use data for Huainan in China, a typical coal resource city, from 2008 to 2017 for an empirical study. A typical kind of spatiotemporal evolution of LUCP (evolution from farmland to grassland within any two years) is evaluated. Taking into account the shape of land use units, the spatiotemporal distances between irregular evolutionary sequences are measured using buffer-based superposition. The results show that the expected mean nearest neighbor distance for the irregularly evolving sequence of land use units is 0.085 in the completely random CSR model, whereas the mean nearest neighbor distance is 0.037 in the real observation model. These results indicate that such LUCPs have generally shown a spatiotemporal aggregation pattern over the past 10 years. However, since the z-score is 1.03, which is in the range of −1.65 to 1.65, this aggregation pattern is not statistically significant. These experiments demonstrate the validity of using the method proposed herein to study similar problems. The results of this work provide valuable insight into the spatiotemporal evolution process of land use units, which could be instrumental in exploring the potential spatiotemporal model of LUCP evolution.