关键词: 粮食安全, 粮食自给率, 最小人均耕地面积, 谱系聚类(层次聚类)

Abstract: Grain security is fundamental to the sustainable development of our society and national economy. As research regions with small area are vulnerable to the impacts of policy changes, indexes related to grain security of these areas often change in the form of stages, which means that the mathematical regularity of long-term datasets is not significant. As a result, it is difficult to implement grain security programming for the future.We put forward a new method of combining hierarchical cluster analysis with traditional mathematical models, and established a quantification standard for the validity judgment of the clustering results. Meanwhile, a criterion for the using of hierarchical cluster analysis was also proposed, but we strongly recommended that mass data from other research areas are needed to calibrate and perfect it.Kunshan (1985-2007) was chosen as a study region to prove the new method, because it is small in area but with rapid economic development. The results of analysis showed that: the indexes related to grain security did change in the form of stages, which were stage 1 from 1985 to 2001 and stage 2 from 2002 to 2006; new models based on hierarchical cluster analysis showed better results in simulating the grain security situation from 1985 to 2006 and smaller deviations in predicting it in 2007 than traditional models—the prediction deviations of total population,farmland acreage and grain productivity in 2007 were 3.4%, -10.9% and 5.6% with clustering and -25.7%, 96.1% and 122.5% without clustering, while the determination coefficients of three different models were 0.977, 0.981 and 0.914 with clustering and 0.801, 0.518 and 0.426 without it; in 2015, Kunshan would have a decreased grain self-sufficient ratio of 6% and a decreased minimum farmland acreage per capita of 0.022 hm². After a comparison with works of other researchers(on average, the minimum farmland acreage per capita of Zhejiang Province and East China were 0.044 hm² and 0.075 hm² respectively) and a subsequent analysis, we concluded that the new method of combining hierarchical cluster analysis with traditional mathematical models was both feasible and scientific.In the discussion, we thought that although such models with hierarchical cluster analysis can be applied to predict the grain security in the future, the predicting precision of the model is dependent on the duration of the recent stage to a great degree. However, the deficiency does not harm its function as a decision support implement for future grain security programming as well as town planning.

Key words: grain security, grain self-sufficient ratio, minimum farmland acreage per capita, hierarchical cluster analysis