Site-specific agricultural management (SSM) relies on identifying within-field spatial variability and is used for variable rate input of resources. Precision agricultural management commonly attempts to integrate multiple datasets to determine management zones (MZs), homogenous units within the field, based on spatial characteristics of environmental and crop properties (i.e., terrain, soil, vegetation conditions). This study compared several multivariate spatial clustering methods to determine MZs for precision nitrogen fertilization in a citrus orchard. Six variables, namely normalized difference vegetation index, crop water stress index, digital surface model, slope, elevation and aspect, were used to characterize spatial variability within four plots. Six clustering model composites were compared, each including some or all of the following components: (1) spatial representation of the data (e.g., Getis Ord Gi*); (2) variable weights based on their relative contribution; and (3) clustering methods, including different extensions of K -means and hierarchical clustering algorithms. The fuzzy K -means algorithm applied to the weighted spatial representation was found to generate MZs with similar numbers of trees, while the K -means algorithm applied over the spatial representation generated MZs that were more continuous over space, with minimum fragmentation. Spatial variability was not constant across the orchard and among the different variables. Management of the sub-units, or plots, using spatial representation rather than the measured values, is proposed as a more suitable platform for agricultural practices. SSM is dependent upon available variable rate application technologies. Future development of fertilizer application for individual trees will require adjusting the statistical approach to support tree-specific management. The suggested model composite is flexible and may be composed of different models for delineating plot-specific MZs.

中文翻译：

---

果园施氮管理区空间聚类模型比较

特定地点农业管理 (SSM) 依赖于识别田间空间变异性，并用于资源的可变速率输入。精准农业管理通常试图根据环境和作物特性（即地形、土壤、植被条件）的空间特征，整合多个数据集来确定管理区 (MZ)，即田间内的同质单元。本研究比较了几种多元空间聚类方法，以确定柑橘园精确施氮的 MZ。六个变量，即归一化差异植被指数、作物水分胁迫指数、数字表面模型、坡度、高程和坡向，用于表征四个样地内的空间变异性。比较了六个聚类模型组合，每个都包括以下部分或全部组件： (1) 数据的空间表示（例如 Getis Ord Gi*）；(2) 基于其相对贡献的可变权重；(3)聚类方法，包括K-means的不同扩展和层次聚类算法。发现应用于加权空间表示的模糊 K 均值算法生成具有相似树数的 MZ，而应用于空间表示的 K 均值算法生成的 MZ 在空间上更连续，碎片最少。整个果园和不同变量之间的空间变异性并不是恒定的。建议使用空间表示而不是测量值来管理子单元或地块，作为更适合农业实践的平台。SSM 取决于可用的可变速率应用技术。个别树木施肥的未来发展将需要调整统计方法以支持特定树木的管理。建议的模型组合是灵活的，可以由不同的模型组成，用于描绘特定地块的 MZ。