Abstract. Grasslands provide many important ecosystem services globally and forecasting grassland productivity in the coming decades will provide valuable information to land managers. Productivity models can be well-calibrated at local scales, but generally have some maximum spatial extent in which they perform well. Here we evaluate a grassland productivity model to find the optimal spatial extent for parameterization, and thus for subsequently applying it in future forecasts for North America. We also evaluated the model on new vegetation types to ascertain its potential generality. We find the model most suitable when incorporating only grasslands, as opposed to also including agriculture and shrublands, and only in the Great Plains and Eastern Temperate Forest ecoregions of North America. The model was not well suited to grasslands in North American Deserts or Northwest Forest ecoregions. It also performed poorly in agriculture vegetation, likely due to management activities, and shrubland vegetation, likely because the model lacks representation of deep water pools. This work allows us to perform long-term forecasts in areas where model performance has been verified, with gaps filled in by future modelling efforts.

中文翻译：

---

草地生产力模型的多尺度评估

摘要。草原在全球范围内提供许多重要的生态系统服务，对未来几十年的草原生产力进行预测将为土地管理者提供有价值的信息。生产力模型可以在局部范围内进行很好的校准，但通常在某些最大的空间范围内表现良好。在这里，我们评估了草地生产力模型，以找到用于参数化的最佳空间范围，从而将其随后应用于北美的未来预测中。我们还评估了关于新植被类型的模型，以确定其潜在的普遍性。我们发现该模型最适合仅包含草原，而不包括农业和灌木地，并且仅适用于北美的大平原和东部温带森林生态区。该模型不适用于北美沙漠或西北森林生态区的草地。由于管理活动，该植被在农业植被和灌木林植被中的表现也很差，可能是因为该模型缺乏深水池的表示。这项工作使我们能够在已验证模型性能的领域中进行长期预测，并通过将来的建模工作来填补空白。