Regional and global vegetation simulations can be problematic when analysis units to which parameters are assigned do not align with plant productivity and phenology. Having a suite of predefined biophysical regions at a variety of scales that correspond to differences in plant productivity and phenology would allow analysts to select a set of analysis units at the scale needed. In other cases, environmental or social responses may be hypothesized to be related to differences in plant dynamics. One may compare the discrimination in such data that biophysical regions at different scales provide to determine which best distinguishes the responses in question, such that like responses fall within the same regions to the degree possible. If those relationships are significant, the responses may then be extrapolated based on the biophysical regions. I defined hierarchical biophysical regions based on plant productivity and phenology by clustering global 0.083 degree resolution normalized difference vegetation indices (NDVI) over a 10 year period. Agglomerative average-linkage distances based on squared error between clusters were conducted using an iterative sampling approach to merge more than 2 million clusters into fewer and fewer clusters based on NDVI greenness profiles comprised of 240 values over 10 years, until all cells were in a single cluster. Greater and greater differences in greenness profiles were ignored at higher levels of the hierarchy. Using a difference increment of 0.1, 253 non-duplicative sets of clusters were created, and 107 of those were included in animations that may be used to explore differences in global plant dynamics. Differences in clusters were quantified based on comparing the focal set of cluster results with 10 other cluster sets. Analysts may use the hierarchical clusters to improve the alignment of their parameter sets that inform plant growth and other dynamics with real-world plant dynamics.

中文翻译：

---

分层的全球植物生物物理区域作为潜在分析单位。

当分配了参数的分析单元与植物的生产力和物候特性不一致时，区域和全球植被模拟可能会出现问题。在与植物生产力和物候学方面的差异相对应的各种尺度上拥有一套预定的生物物理区域，将使分析人员可以选择所需规模的一组分析单位。在其他情况下，可以假设环境或社会反应与植物动力学的差异有关。可以比较在不同规模的生物物理区域提供的这种数据中的辨别力，以确定哪种方法最好地区分了所讨论的响应，从而使相同的响应尽可能落入相同的区域内。如果这些关系是重要的，则可以基于生物物理区域推断响应。我通过将全球0.083度分辨率归一化差异植被指数（NDVI）聚类在一起，在10年的时间段内，基于植物的生产力和物候来定义分层的生物物理区域。使用迭代采样方法，基于簇之间平方误差的聚集平均链接距离，根据NDVI绿色度分布图（包含10年中的240个值），将超过200万个簇合并为越来越少的簇，直到所有单元都在一个单元中为止簇。在更高层次的层次上，绿色配置文件中越来越大的差异被忽略了。使用0.1的差异增量，创建了253个非重复的群集集，其中的107个包含在动画中，可用于探索全球植物动态的差异。通过将聚类结果的焦点集与其他10个聚类集进行比较，可以量化聚类中的差异。分析师可以使用层次聚类来改善其参数集的对齐方式，这些参数集可以利用实际的植物动态信息为植物生长和其他动态提供信息。