Agent农业土地变化模型研究进展

农业土地变化是全球变化与可持续研究的热点,当前研究虽取得了长足进展,但仍存在诸多不足,集中表现在对农业土地系统复杂性与动态性的认识不够.近年来,基于Agent的农业土地变化研究(农业ABM/LUCC,Agent-based agricultural land change modeling)逐渐兴起,极大的丰富了传统研究的理论与方法,具体表现在:(1)农业ABM/LUCC将微观层面的人类个体行为整合进土地变化研究框架,有助于更加清楚的认识农业土地系统的“人类-自然”综合复杂性问题.(2)农业ABM/LUCC能够动态表达土地系统变化的内生反馈机制,有助于弥补传统的静态土地变化驱动机制分析的不足.(3)基于ABM/LUCC的农业土地利用格局动态研究是整合“人类-自然”综合研究的关键桥梁,农业ABM/LUCC能够与其他生物地球物理模型或经济模型动态嵌套,使多尺度、多维度综合模型研究成为可能.然而,农业ABM/LUCC研究也存在诸多挑战,如理论研究滞后于应用研究,大尺度应用难以开展,以及农户行为的模拟结果很难得到校验等.

Progress of agent-based agricultural land change modeling: a review

Agricultural land use and land cover change (Agri-LUCC) is one of the key issues among global change and sustainability studies. Year-on-year progress makes "agricultural land change" to be an emerging interdisciplinary science. As an effective tool for understanding the driver, process and consequence of Agri-LUCC, spatially-explicit land change models have successfully applied in representing agricultural landscapes and its possible developments across scales. Although several breakthroughs have been achieved by traditional land change modeling, there are still many crucial issues remain unsolved, especially the insufficient cognition on the complexity and dynamics of agricultural land systems. Recently, some researchers begin to combine agent-based models (ABM, one of the key tools for complex system studies) with land change models, bringing a new emergence of model series in the agricultural land change modeling community, which are called as Agri-ABM/LUCCs. Progress in this field can be summarized as: (1) Based on the complexity system theory, most of these models bring theoretical and methodological innovations in analyzing the complexity of agricultural land systems. (2) These models innovatively take land use decisions at individual level into consideration, based on which to recognize the role of decision makers bringing about changes, through their choices, on regional level landscapes. Such "modeling with stakeholders" underlines the role of farmers in agricultural transformation, facilitating the expression of diversified decisions on agricultural land use from heterogeneous farmers. (3) Agri-ABM/LUCC links "land change driving forces" with "land use consequences" as an endogenous feedback loop in agricultural land change processes. This tightly coupled method describes a better feature of agricultural land dynamics, which is essential for analyzing the vulnerabilities, impacts, and adaptation in agricultural land change context. (4) From the recent literature, a wild range of issues related to farmer's decisions on their land were discussed, including deforestation, agricultural expansion, crop allocation, resource management, and settlement and livelihood decisions. In these studies, various methods and approaches were used in representing farmer's decisions. Methods include linear programming model, optimization model, heuristic imitative and innovative decision-making algorithms, utility function, decision tree, evolutionary programming, probabilistic method, participatory modeling, role playing game, bounded-rational approach, spatial multi-nominal logistic functions, among others. (5) This new perspective provides a way to dynamically link agricultural land change assessments for integrated human-natural studies. On one hand, consequence of agricultural land change can be used to forecast crop production then to develop food security scenarios; on the other hand, the same land change result is valuable for predicting carbon-nitrogen cycling processes, consequently for projecting carbon sequestration within large scale agricultural landscapes. Scenarios of food and ecological security provide feedbacks to individual farmers to alter their decisions of land use in turn. Beside the progress, however, problems of current Agri-ABM/LUCCs still exist, such as "theory divorced from practice", deficiency in cross-site comparison, and difficulties in carrying out large-scale modeling. The most critical problem is that other than the common characteristics of complex adaptive systems, some of the special features of agricultural land systems exit in their spatial-temporal dynamics, scaling effects, coupled human and natural issues, and multi-dimension feedbacks. These features are still not well examined in the current studies, which require further in-depth discussions in the future.