Development of

(1)在本研究项目中，我完成了全国范围内的土地利用、自然和社会经济条件两级数据库。使用这个数据库，我应用了单级（传统）多元回归模型和两级多元回归模型（第一级：直辖市，第二级：县）。结果表明，地形因素、农场经营条件因素和城市化因素是决定耕地份额及其变化的核心因素。我在地级市层面进行回归分析，其中标准变量为地级市的随机项（多水平分析的输出），解释变量为同级社会经济因素。农业合作社的领导水平、劳动力市场的发展水平、农业基础设施的投资水平、农村社区的社会纽带强度等因素对农地份额的影响显著。与传统的回归方法相比，多层次建模方法能更好地揭示土地利用变化的机理。(2)编制了京都市另外两层数据库（第一层为时间序列数据，第二层为农村社区数据）。运用多层次模型提取了京都市1600个社区近30年（1970-2000年）的耕地变化格局。(3)另一方面，采用基于agent的方法对社区层面的农业用地进行建模，建立了原型模型。通过该模型，我们可以描述各种土地利用政策的影响，如大米价格、大米生产调整、土地改良工程等。多层模型的因子分析与多智能体系统的模拟模型之间的联系是构建“区域敏感全球土地利用模型”最有前途的任务。

(1) In this research project, I finalized the nation-wide, two-level database on land use, natural-and socioeconomic conditions. Using this database, I applied both a single-level (conventional) multivariate regression model and a two-level multivariate regression model (the level 1 : municipalities, the level 2 : prefectures). The results shows that the topographic factors, the farm management conditions and the urbanization factors were the core factors that decided the share of farmland and its change. I carried out regression analysis at prefecture level, in which the criterion variables are the random terms at prefectural level (the outputs form the multilevel analysis) and the explanatory variables are socioeconomic factors at the same level. It became clear that the factors such as the leadership of agricultural cooperatives, the development level of labor market, the investment level for agricultural infrastructure and the strength of social bonds of rural community would significantly influence the farmland share. In this way, the multi-level modeling approach would bring us more appropriate insight on the mechanism of land-use change than the conventional regression approach.(2) Iprepared the other two-level database of Kyoto prefecture (time series data at level 1, rural community at level 2). I could extract the patters of farmland change of 1600 communities in Kyoto Prefecture for 30 years (1970-2000) by applying the multilevel model.(3) On the other hand, I applied Agent-based approach for modeling agricultural land use at community level to develop a prototype model. By the model, we can describe the effects of various land use policies such as rice price, adjustment of rice production, land improvement project and so on. The linkage between the factor analysis by the multilevel model and the simulation model by the multi-agent system is the most promising task to develop "the Regionally Sensitive Global Land-use Model".