The dynamics of grassland ecosystems are highly complex due to multifaceted interactions among their soil, water, and vegetation components. Precise simulations of grassland productivity therefore rely on accurately estimating a variety of parameters that characterize different processes of these systems. This study applied three calibration schemes – a Single-Objective (SO-SUFI2), a Multi-Objective Pareto (MO-Pareto), and, a novel Uncertainty-Based Multi-Objective (MO-SUFI2) – to estimate the parameters of MONICA (Model for Nitrogen and Carbon Simulation) agro-ecosystem model in grassland ecosystems across Germany. The MO-Pareto model is based on a traditional Pareto optimality concept, while the MO-SUFI2 optimizes multiple target variables considering their level of prediction uncertainty. We used measurements of leaf area index, aboveground biomass, and soil moisture from experimental data at five sites with different intensities of cutting regimes (from two to five cutting events per season) to evaluate model performance. Both MO-Pareto and MO-SUFI2 outperformed SO-SUFI2 during calibration and validation. The comparison of the two MO approaches shows that they do not necessarily conflict with each other, but MO-SUFI2 provides complementary information for better estimations of model parameter uncertainty. We used the obtained parameter ranges to simulate grassland productivity across Germany under different cutting regimes and quantified the uncertainty associated with estimated productivity across regions. The results showed higher uncertainty in intensively managed grasslands compared to extensively managed grasslands, partially due to a lack of high-resolution input information concerning cutting dates. Furthermore, the additional information on the quantified uncertainty provided by our proposed MO-SUFI2 method adds deeper insights on confidence levels of estimated productivity. Benefiting from additional management data collected at high resolution and ground measurements on the composition of grassland species mixtures appear to be promising solutions to reduce uncertainty and increase model reliability.

由于土壤、水和植被成分之间的多方面相互作用，草地生态系统的动态非常复杂。因此，草地生产力的精确模拟依赖于准确估计表征这些系统不同过程的各种参数。本研究应用了三种校准方案——单目标 (SO-SUFI2)、多目标帕累托 (MO-Pareto) 和新颖的基于不确定性的多目标 (MO-SUFI2)——来估计 MONICA 的参数（氮和碳模拟模型）德国草原生态系统的农业生态系统模型。MO-Pareto 模型基于传统的 Pareto 最优性概念，而 MO-SUFI2 考虑多个目标变量的预测不确定性水平来优化它们。我们使用叶面积指数的测量值，地上部生物量和土壤水分来自五个具有不同切割强度（每个季节从 2 到 5 次切割事件）的地点的实验数据，以评估模型性能。MO-Pareto 和 MO-SUFI2 在校准和验证期间均优于 SO-SUFI2。两种 MO 方法的比较表明它们不一定相互冲突，但 MO-SUFI2 为更好地估计模型参数不确定性提供了补充信息。我们使用获得的参数范围来模拟不同切割制度下德国的草地生产力，并量化与跨地区估计生产力相关的不确定性。结果表明，与粗放管理草地相比，集约管理草地的不确定性更高，部分原因是缺乏有关切割日期的高分辨率输入信息。此外，我们提出的 MO-SUFI2 方法提供的有关量化不确定性的附加信息增加了对估计生产力的置信水平的更深入了解。受益于高分辨率收集的额外管理数据和草地物种混合物组成的地面测量，似乎是减少不确定性和提高模型可靠性的有希望的解决方案。