Contemporary crop research programs involve the evaluation of numerous micro-plots spread across extensive experimental fields. As a result, there is a growing need to depart from labor-intensive manual measurements when assessing phenological data. The growing significance of high throughput phenotyping platforms (HTTP), including unmanned aerial vehicles (UAVs), has rendered these technologies essential in crop research. The overall objective of this study is to explore and validate the use of HTTP methodologies, specifically the potential of vegetation indices (VIs) derived from conventional RGB images, to forecast the date of heading (DH) and maturity (DM) for various cereal crops under different irrigation conditions. To pinpoint DH and DM prediction, a total of nine UAV surveys were conducted throughout the entire crop cycle. Prediction models for DH and DM using VIs were successfully developed for various crop species, explaining 65 % of the variance in bread wheat and 75 % in oats. The highest percentages of variance explained were achieved when models were developed separately for the two irrigation conditions (well-irrigated and rainfed). However, the percentage of variance explained by these models decreased when applied to barley (R²<0.5 for DH). Notably, including final plant height as a predictor increased the percentage of variance explained by the models only for irrigated bread wheat. Furthermore, the utilization of multi-temporal equations, which amalgamated data from diverse UAV surveys, notably enhanced the percentage of variance explained by the model (+160.71 % improvement in DH predictions), particularly those tailored to each specific crop species and irrigation condition. The investigation additionally established a thorough protocol for modeling the phenological aspects of cereal crops utilizing data acquired from UAVs, thereby enhancing the accessibility of this technology for measurements of phenology in large crop research programs.

中文翻译：

---

利用植被指数预测抽穗期和成熟期：以面包小麦、大麦和燕麦作物为例的案例研究


当代作物研究项目涉及对分布在广泛试验田中的大量微地块进行评估。因此，在评估物候数据时，越来越需要摆脱劳动密集型的手动测量。包括无人机 (UAV) 在内的高通量表型平台 (HTTP) 的重要性与日俱增，使得这些技术在作物研究中变得至关重要。本研究的总体目标是探索和验证 HTTP 方法的使用，特别是源自传统 RGB 图像的植被指数 (VI) 的潜力，以预测各种谷类作物的抽穗日期 (DH) 和成熟度 (DM)不同灌溉条件下。为了精确预测DH和DM，在整个作物周期中总共进行了九次无人机调查。使用 VI 成功开发了适用于各种作物品种的 DH 和 DM 预测模型，解释了面包小麦中 65% 的方差和燕麦中 75% 的方差。当针对两种灌溉条件（良好灌溉和雨养）单独开发模型时，可以实现最高的方差解释百分比。然而，当应用于大麦时，这些模型解释的方差百分比有所下降（DH 的 R²<0.5）。值得注意的是，将最终株高作为预测因子增加了仅针对灌溉面包小麦的模型所解释的方差百分比。此外，利用多时态方程合并了来自不同无人机调查的数据，显着提高了模型解释的方差百分比（DH 预测提高了 160.71%），特别是针对每种特定作物品种和灌溉条件定制的方程。 该调查还建立了一个全面的协议，利用从无人机获取的数据对谷类作物的物候方面进行建模，从而提高了该技术在大型作物研究项目中测量物候的可及性。