自然资源学报 ›› 2018, Vol. 33 ›› Issue (3): 515-525.doi: 10.11849/zrzyxb.20170040

• 资源研究方法 • 上一篇    下一篇

基于多智能体和土地转换模型的耕地撂荒模拟研究——以陕西省米脂县为例

宋世雄, 梁小英*, 陈海, 毛南赵   

  1. 西北大学,西安 710069
  • 收稿日期:2017-01-15 修回日期:2017-06-20 出版日期:2018-03-20 发布日期:2018-03-20
  • 通讯作者: *梁小英(1974- ),女,陕西泾阳人,副教授,主要从事地理信息系统、自然方面研究。E-mail: liangxy@nuw.edu.cn
  • 作者简介:宋世雄(1990- ),男,陕西乾县人,硕士研究生,中国自然资源学会会员(S00002058M),主要研究方向为土地集约利用和GIS应用。E-mail: ssx1990@126.com
  • 基金资助:
    国家自然科学基金项目(41671086,41271103)

The Simulation of Cropland Abandonment Based on Multi-agent System and Land Transformation Model: A Case Study of Mizhi County, Shaanxi Province

SONG Shi-xiong, LIANG Xiao-ying, CHEN Hai, MAO Nan-zhao   

  1. Northwest University, Xi’an 710069, China
  • Received:2017-01-15 Revised:2017-06-20 Online:2018-03-20 Published:2018-03-20
  • Supported by:
    National Natural Science Foundation of China, No. 41671086 and 41271103.

摘要: 多模型耦合已经成为国内外模拟LUCC有效途径之一。论文通过如下步骤阐明宏观耕地撂荒格局和微观主体行为间的互动机理。首先,将微观主体间的相互作用纳入其决策,构建有限理性多智能体决策模型(Multi-Agent System,MAS);其次,通过SNNS平台的训练学习以及与历史数据对比分析,验证土地转换模型(Land Transformation Model,LTM)模拟研究区宏观撂荒格局的有效性;最后,依据多模型耦合机理,耦合MAS模型与LTM模型,形成耕地撂荒模拟模型(Cropland Abandonment Simulation Model,CASM),并基于研究区耕地撂荒的实际数据,探讨模型的合理性和可行性。结果表明:与2013年历史土地利用数据对比,CASM模型的PCM(Percent Correct Metric)系数为71%,比单独利用LTM模型的模拟精度提高3%,不仅表明CASM能够较好地模拟分析米脂县耕地撂荒空间格局分布,而且可有效揭示宏观耕地撂荒格局的微观驱动机理;同时,文章指出未来研究中要考虑政策和市场的影响,进一步完善不同层次主体决策对撂荒的影响,以此来提高模型对现实耕地撂荒的解释力。

关键词: 多智能体模型, 耕地撂荒模型, 人工神经网络模型, 土地转换模型

Abstract: The integration of different models has been one of the effective ways to simulate LUCC. In this paper, several steps were designed to reveal the interactive mechanism between macroscopic pattern of cropland abandonment and microcosmic behaviors of agents. Firstly, a multi-agent decision-making model under bounded rationality was constructed which took the interactions among different agents into consideration. Secondly, the effectiveness of Land Transformation Model (LTM) in simulating the macroscopic pattern of cropland abandonment in the study area was tested through the training on SNNS platform and the comparison with historical data. Thirdly, Cropland Abandonment Simulation Model (CASM) was formed by integrating multi-agent system (MAS) and LTM, and the rationality and feasibility of the CASM was explored. The result showed that the percent correct metric (PCM) of CASM was 71%, which was 3% higher than the PCM of LTM. This result indicated that CASM can not only simulate and analyze the distribution of spatial pattern of cropland abandonment in Mizhi County, but also effectively reveal the micro driving mechanism of cropland abandonment. Finally, several advices of future research were given to improve the explanatory ability of CASM, e.g., the effect of policy, market and the decisions of different levels of agents (such as the rural cooperation and government) on the cropland abandonment should be considered.

Key words: artificial neural network, cropland abandonment simulation model, land transformation model, multi-agent system

中图分类号: 

  • F301.21