This paper presents an empirical study comparing different uncertainty analysis (UA) and sensitivity analysis (SA) methods, focussing their usefulness for the output analysis of land use/land cover change (LUCC) agent-based models (ABMs). As a result, a workflow to integrate UA and SA is presented to evaluate ABMs outputs. We developed a baseline scenario and performed a comprehensive investigation of the impacts that differences in sample sizes, sample techniques, and SA methods may have on the model output. The analysis is done in the context of a particular agent-based simulator with a LUCC model in a Brazilian Cerrado case study. The experiments indicate that there are known challenges to be overcome by the use of statistical methods. Even though the presented analysis was done over a particular simulator, we intend to contribute to the community that understands the importance of statistical validation techniques to improve the level of confidence in agent-based simulation outputs.

本文提供了一项实证研究，比较了不同的不确定性分析（UA）和敏感性分析（SA）方法，重点介绍了它们在基于土地使用/土地覆被变化（LUCC）主体模型（ABMs）的输出分析中的有用性。结果，提出了整合UA和SA的工作流程以评估ABM的输出。我们制定了基线方案，并对样本量，样本技术和SA方法的差异可能对模型输出产生的影响进行了全面调查。在巴西Cerrado案例研究中，在具有LUCC模型的特定基于代理的模拟器的背景下进行了分析。实验表明，使用统计方法可以克服已知的挑战。即使提出的分析是在特定的模拟器上完成的，