导　　师： 白红英;陈海

学科专业： 070501

授予学位： 硕士

作　　者： ;

机构地区： 西北大学

摘　　要： 土地利用变化研究作为全球环境变化的核心计划之一,受到不同领域学者的广泛关注。利用MAS模型模拟微观个体的决策过程,探讨其对自然与社会经济环境变化的适应机制,是农户土地利用行为模拟研究的一个重要趋势。 本文从宏观尺度总结了银州镇土地利用变化的时空分异规律,分析了土地利用变化的政策影响。在微观尺度上,借鉴多智能体模型,分别运用BDI/(Belief-Desire-Intension/)模型和强化学习模型构建孟岔村农户的土地利用行为模型,从单个农户行为和农户整体行为两个层次,探讨土地利用变化的决策机制。并检验两种模型的准确度,得出以下结论: /(1/)1986-1996年,银州镇耕地面积呈增加的态势,而林地和草地处于减少的态势;1996-2006年,耕地面积较大幅度下降,林、草地面积处于增加态势。 以上两个阶段土地利用变化转折源于退耕还林还草政策的影响。退耕还林还草政策主要影响大于25度的坡耕地,而无定河两岸川地/(水浇地/)更多的受农户土地利用行为的影响。 /(2/)基于单个农户行为的BDI模型。定量探讨了市场对农户的作用,以及同类农户间的相互作用,主要利用连续的函数,而非离散的0、1值反映土地利用变化所受到的影响,比较适合实际情况。 /(3/)基于农户整体行为的强化学习模型。研究了政策、市场、农户自身及农户间的相互作用、土壤肥沃程度等因素对农户土地利用行为决策的影响。通过构建连续函数,表现出农户某种作物的种植概率,利用概率值与农户拥有地块之间的关系,清楚的表现了农户决策向空间转化的机理。 /(4/)BDI模型和强化学习模型虽然在研究对象、考虑因素、因素对农户影响的权重、模型结果向空间转化等方面不同,但模型模拟结果都表明孟岔村年间作物面积整体变化不大,而每个农户地块上的作物种类变化较大。 As one of the core of global environmental change research, land use change was being concerned by scholars in different fields. Use of MAS model to simulate the process of micro-individual's decision-making and to explore the mechanisms to adoption of individual to the natural and socio-economic environmental change, MAS model will be one of the important trend to simulate the households land use behavior. In macro-scale, summary the land use change in spatial and temporal of YinZhou in MiZhi County, and analysis of land use change under the impacts of policies and the mechanism of the overall households' responding to land use change. In micro-scale, use multi-agent model, the use of BDI /(Belief-Desire-Intension/) model and the reinforcement learning model construction of the model of households land use behavior, from the individual households and the overall household's level of behavior separately, to explore the changes in land use decision-making mechanism. Moreover, the models pass the inspection and verification. The conclusions of this paper were as follows: /(1/) During 1986 to 1996, arable land had the increasing trend, while the woodland and grassland in the decrease trend; with the returning farmland to forests and grasslands of policy. During 1996 to 2006, the larger decrease in arable land, woodland and grassland area increased. The policy of returning farmland to forest can be used to explain the land use change occurred in the beam, which slop is greater than 25 degree. However, the arable land beside WuDing river were affected on the household's land use behavior. /(2/) BDI model based on individual household's behavior. Explore the affect of market to households, and the interaction between same type households. It mainly used continue function other than 0 or 1 value to reflect the land use change under the affected of two factor in above, and it suitable for the reality. /(3/) Reinforcement learning model based on the overall household's behavior. Study on the affected of decision-making of household's land use by the factors of policy, market, households themselves, the interaction in households and soil fertility degree. Reinforcement learning model express the probability of household planted certain crop with probability of quantitative methods by the continue function. Moreover, it explained the mechanism of decision-making of household's transformation to space clearly by using the correlation between household possessing land and probability number. /(4/) BDI model and Reinforcement learning model had the different in the factors of research on subject, consideration about factors, the weight of factors impact on households, the result of household's decision-making to the space. The result showed that there had little changes in the overall crop area, but each farmer's crop type had lager change.

关 键 词： 土地利用变化 多智能体模型 模型 强化学习模型 孟岔村

分 类 号： [F321.1 F224]

领　　域： [经济管理] [经济管理]