Multi-agent model-based historical cropland spatial pattern reconstruction for 1661–1952, Shandong Province, China

• Published in: Global and planetary change Vol. 143; pp. 175 - 188
• Main Authors: Yang, Xuhong, Jin, Xiaobin, Du, Xindong, Xiang, Xiaomin, Han, Juan, Shan, Wei, Fan, Yeting, Zhou, Yinkang
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-08-2016
• Subjects: Accuracy; Computer simulation; Decision making; Farming; Historical cropland; Human behavior; Mathematical models; Multi-agent; Multiagent systems; Reconstructing spatial pattern; Reconstruction; Shandong Province; INW, China, People's Rep., Shandong Prov; Multi-agent; Reconstructing spatial pattern; Shandong Province; Historical cropland
• ISSN: 0921-8181; 1872-6364
• DOI: 10.1016/j.gloplacha.2016.06.010

To advance the research of global land use/cover change (LUCC), biodiversity, global carbon cycle, and other aspects of the earth system, it is essential to reconstruct changes in historical cropland cover with long time series and high-resolution grid. Currently, it is a general approach which is based on the view of combining the overall control of cropland area, selecting grid of high land suitability, and ‘top-down’ decision-making behaviors to reconstruct the historical cropland. Considering various factors that influenced cropland distribution, including behavioral agent's selection by itself and the limitation of nature and human factors, a spatiotemporal dynamical reconstruction model of historical cropland based on the multi-agent systems has been developed from the perspective of ‘bottom-up’, which combine macroscopic and microscopic decision-making behaviors of agents to simulate the government and farmer autonomously implementing the selection behaviors of farming area. Taking Shandong Province as the study area, this model was used to imitate its cropland spatiotemporal pattern with 1km grid-resolution from 1661 combining the contemporary pattern and reconstructed amount of historical cropland as a maximum potential scope and control variable of reconstruction model, respectively, furthermore, followed the accuracy valuation and comparative analysis. The reconstructed results show that: 1) It is properly suitable for Multi-Agent to simulate and reconstruct the spatial distribution of historical cropland; 2) compared with historical map data sets (1930s) from the view of point to point, the correctly classified producer accuracy, user accuracy and overall accuracy of reconstructed result totally up to 59.09%, 80.62% and 62.31%, respectively, and shows our reconstruction map achieved a better agreement with the historical maps; 3) from the view of grid-level or county-level, our reconstruction approach can effectively keep away from the grid with mountain-hilly and easily flooding probability, it showed a good similarity and higher consistency when compared with the research result based on archives and historical records in overall pattern and tendency of cultivation, and difference error decrease gradually, moreover, the cultivation ratios in county-level is more close to the historical situation.
•Apply a spatiotemporal dynamical reconstruction model based on the multi-agent systems to reconstruct the spatial-temporal distribution of historical cropland.•Select two types of agents to characterize the decision-making of different behavioral subjects, i.e., a macro-agent and micro-agents.•Methodology combining of ‘top-down' and 'bottom-up' was proposed to reconstruct historical cropland•The reconstruction results showed strong agreement with the historical map by means of point-by-point verification.