How does in-row seed spacing and spatial pattern affect canola yield?

加拿大油菜籽产业价值267亿加元，为加拿大人提供了25万个工作岗位。 油菜产量受到低于最佳种植群体的限制，这是由于作物出苗不良、种子分布不规则和故意降低播种率造成的。卡诺拉植物可以通过增加分枝和豆荚数量来至少部分地补偿较低的植物密度。这些对植物结构的改变，称为表型可塑性，增加了所占据的空间，太阳辐射捕获和每株植物的种子产量。然而，高度分枝的植物需要更长的时间才能达到完全的冠层关闭，这会减少太阳辐射的捕获和作物整体的种子产量。独立于植物密度，修改油菜植物之间的行间和行内间距也会影响表型可塑性，这使我们假设，最佳地安排植物将降低产量最大化所需的植物密度。然而，相邻的油菜植物之间的空间关系和管理它们的机制知之甚少，我们目前缺乏空间明确的种植建议，以利用这种可塑性。拟议的研究将通过田间试验研究油菜冠层如何在不同种植安排下发展，以优化太阳辐射捕获和作物生产力。将开发基于卷积神经网络（CNN）的深度学习方法，使用无人机（无人机）、地面摄像机和3D扫描仪收集的图像来量化单个油菜植物随时间推移所占据的空间。这些方法将被应用于研究相邻油菜植物之间的空间关系。此外，我们将研究不同种植安排导致的冠层光环境，并将其应用于生长室中，以了解油菜表型可塑性如何随着光环境的变化而产生。这项研究的结果将是空间上明确的种植建议，考虑到植物分布，而不仅仅是每个区域的播种密度。我们相信，这些建议将以更低的种子成本获得最大产量，使油菜生产更有利可图，并提高全球粮食安全。

The Canadian canola industry is worth $26.7 billion and sustains 250,000 jobs for Canadians. Canola yield is limited by lower than optimum plant populations, which result from poor crop emergence, irregular seed distribution, and intentionally low seeding rates. Canola plants may at least partially compensate for lower plant densities through increased branching and number of pods. These changes to plant architecture, termed phenotypic plasticity, increase the space occupied, solar radiation captured and seed production per plant. However, highly branched plants take longer to reach full canopy closure, which reduces solar radiation capture and seed yield for the crop overall. Independent of plant density, modifying the inter- and intra-row spacing between canola plants also affects phenotypic plasticity, leading us to hypothesize that arranging plants optimally will lower the plant density required to maximize yield. However the spatial relationships between neighbouring canola plants and the mechanisms that govern them are poorly understood, and we currently lack spatially explicit planting recommendations to exploit this plasticity. The proposed study will examine how the canola canopy develops with varied planting arrangements through field experimentation, with the goal of optimizing solar radiation capture and crop productivity. Deep learning approaches based on convolutional neural networks (CNN) will be developed to quantify the space occupied by individual canola plants over time using imagery collected by a UAV (drone), ground cameras, and 3D scanner. These methods will be applied to study spatial relationships between neighbouring canola plants. Furthermore we will study the canopy light environments that result from different planting arrangements, and apply them in a growth chamber to understand how canola phenotypic plasticity arises with the changing light environment. The outcome of this study will be spatially explicit planting recommendations that take plant distribution into account rather than only seeding density per area. We believe that these recommendations will result in maximum yield with lower seed costs, making canola production more profitable and resulting in greater global food security.