Management zones are an important aspect of precision agriculture as they help to define spatiotemporal areas that share homogenous attributes for site-specific management such as irrigation. In this study, we evaluated the potential of delineating irrigation management zones in almond (Prunus dulcis) orchards using three different variables: evapotranspiration (ET), crop water stress index (CWSI), and normalized difference vegetation index (NDVI). Multispectral and thermal images were collected on 11 days in June and July of 2020 using an instrumented aircraft flown over an almond orchard in the Central Valley of California. Obtained images were used to compute ET, CWSI, and NDVI. An unsupervised k-means clustering algorithm was used to delineate the field into management zones, and silhouette width was used to determine the optimum number of zones. Regardless of the input variable and collection date, the optimum number of irrigation management zones was identified as two. ET- and CWSI-based management zones addressed higher spatial variability in the field (up to 73.6 %) than NDVI (up to 68 %). Similarly, the level of agreement between management zones delineated using ET and CWSI was strong (kappa coefficient: 0.84 to 1.00). ET- and CWSI-based management zones also showed a trend (p < 0.1) in distinguishing the difference in the irrigation application and almond yield between the two delineated zones. This study shows that CWSI can be as effective as ET in delineating irrigation management zones, and both inputs sensitive to capture the vegetation responses to irrigation management during the summer growing season. Results from this study can be useful for growers and decision makers to practice precision irrigation management in woody perennial cropping systems.

中文翻译：

---

蒸散量 （ET）、作物水分胁迫指数 （CWSI） 和归一化差异植被指数 （NDVI） 的比较分析，以划定杏仁园特定地点的灌溉管理区


管理区是精准农业的一个重要方面，因为它们有助于定义共享同质属性的时空区域，以便进行特定于地点的管理，例如灌溉。在这项研究中，我们使用三个不同的变量评估了在杏仁 （Prunus dulcis） 果园划定灌溉管理区的潜力：蒸散量 （ET）、作物水分胁迫指数 （CWSI） 和归一化差异植被指数 （NDVI）。2020 年 6 月和 7 月的 11 天，使用一架飞越加利福尼亚中央山谷杏仁园的仪表飞机收集了多光谱和热图像。获得的图像用于计算 ET 、 CWSI 和 NDVI。使用无监督 k-means 聚类算法将字段划分为管理区域，并使用轮廓宽度来确定最佳区域数。无论输入变量和收集日期如何，灌溉管理区的最佳数量都被确定为两个。基于 ET 和 CWSI 的管理区解决了比 NDVI（高达 73.6%）更高的现场空间变异性（高达 68%）。同样，使用 ET 和 CWSI 划定的管理区之间的一致性水平很强（kappa 系数：0.84 至 1.00）。基于 ET 和 CWSI 的管理区在区分两个划定区域之间灌溉应用和杏仁产量的差异方面也显示出趋势 （p < 0.1）。这项研究表明，CWSI 在划定灌溉管理区方面与 ET 一样有效，并且这两种输入都对捕捉夏季生长季节植被对灌溉管理的反应敏感。这项研究的结果有助于种植者和决策者在木本多年生种植系统中实施精确灌溉管理。