Cross-Site Comparison of Land-Use Decision-Making and Its Consequences across Land Systems with a Generalized Agent-Based Model

Local changes in land use result from the decisions and actions of land-users within land systems, which are structured by local and global environmental, economic, political, and cultural contexts. Such cross-scale causation presents a major challenge for developing a general understanding of how local decision-making shapes land-use changes at the global scale. This paper implements a generalized agent-based model (ABM) as a virtual laboratory to explore how global and local processes influence the land-use and livelihood decisions of local land-users, operationalized as settlement-level agents, across the landscapes of six real-world test sites. Test sites were chosen in USA, Laos, and China to capture globally-significant variation in population density, market influence, and environmental conditions, with land systems ranging from swidden to commercial agriculture. Publicly available global data were integrated into the ABM to model cross-scale effects of economic globalization on local land-use decisions. A suite of statistics was developed to assess the accuracy of model-predicted land-use outcomes relative to observed and random (i.e. null model) landscapes. At four of six sites, where environmental and demographic forces were important constraints on land-use choices, modeled land-use outcomes were more similar to those observed across sites than the null model. At the two sites in which market forces significantly influenced land-use and livelihood decisions, the model was a poorer predictor of land-use outcomes than the null model. Model successes and failures in simulating real-world land-use patterns enabled the testing of hypotheses on land-use decision-making and yielded insights on the importance of missing mechanisms. The virtual laboratory approach provides a practical framework for systematic improvement of both theory and predictive skill in land change science based on a continual process of experimentation and model enhancement.

土地利用的局部变化源于土地系统内土地使用者的决策与行为，而土地系统本身由本地与全球层面的环境、经济、政治及文化语境所塑造。这种跨尺度的因果关系，为构建关于局部决策如何在全球尺度上影响土地利用变化的通用认知带来了重大挑战。本研究构建了广义基于代理的模型（agent-based model, ABM）作为虚拟实验室，旨在探究全球与局地过程如何影响本地土地使用者的土地利用与生计决策——研究将土地使用者具象为聚落层面的智能体，并在六个真实研究样区的景观中开展模拟实验。研究选取美国、老挝与中国作为样区，以覆盖人口密度、市场影响力与环境条件等全球尺度上的显著差异，研究区的土地系统涵盖从刀耕火种农业到商业化农业的多种类型。本研究将公开获取的全球数据集集成至该基于代理的模型中，以模拟经济全球化对本地土地利用决策的跨尺度影响。本研究构建了一套统计指标，用于评估模型预测的土地利用结果与观测景观以及随机（即零模型）景观之间的吻合度。在六个样区中的四个——这些样区中环境与人口因素是土地利用选择的重要约束条件——模型预测的土地利用结果相较于零模型，与实际观测的景观更为吻合。而在另外两个样区中，市场因素对土地利用与生计决策具有显著影响，此时模型对土地利用结果的预测效果反而不如零模型。本模型在模拟真实土地利用格局时的成功与不足，为检验土地利用决策相关假说提供了支撑，并帮助我们识别出现有模型中缺失机制的重要性。这种虚拟实验室方法为土地变化科学领域的理论体系与预测能力的系统性提升提供了实用框架，该框架基于持续的实验与模型迭代优化过程。