SBIR Phase I: A highly-scalable, rapid, in-season approach to tune a nitrogen model for accurate prediction of a corn crop’s remaining nitrogen need

这项小企业创新研究（SBIR）第一阶段项目的更广泛影响/商业潜力是开发一种新型人工智能技术，使美国玉米种植者能够根据田地特征、降雨和生长条件优化氮肥施用。更有效地使用氮肥将减少美国农业的碳足迹，因为美国农业部门消耗的10%以上的能源用于玉米氮肥的生产。由于氮肥是投入成本和产量的关键驱动因素，这项技术将通过降低投入成本和最大化玉米产量来提高美国农民的盈利能力。该项目可能使玉米生产产生更少的氮肥污染，氮肥污染威胁人类健康，退化水生生态系统并排放导致气候变化的温室气体。这个项目的一个关键的社会元素是一个面向中学理科学生的教学模块，该模块将土壤科学、农学和作物管理的内容与美国玉米种植者遵循最佳管理实践所面临的挑战相结合。这个小型企业创新研究（SBIR）第一阶段项目旨在证明使用机器学习来评估从安装在地面机器人上的低成本相机拍摄的图像中可见的最低玉米叶片黄度的技术可行性。玉米叶片的特征性发黄是氮素（一种关键营养物质）不足引起的胁迫的强烈指示。原型神经网络模型将被迭代改进，部分是通过在这个第一阶段项目的过程中显著增加可用的训练图像。图像将在中西部几个州的实地试验中收集。初步数据表明，特征变黄的程度是累积氮胁迫的一个指标，只能在冠层下观察到，而不能通过机载传感器观察到。当玉米长到1-2英尺高时，一个商业上可用的氮模型将被用来估计通过控制氮肥添加量而产生的小块土地上累积的氮胁迫。调整将通过模拟调整关键模型参数，直到在田地的小块上观察到的和模拟的累积氮应力之间的差异最小化。并行软件开发将改进运行模拟领先氮气模型的原型代码，使模型快速调整成为可能。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is the development of a novel artificial intelligence technology that enables U.S. corn farmers to optimize nitrogen fertilizer applications based on field characteristics, rainfall and growing conditions. More efficient use of nitrogen fertilizer will reduce the carbon footprint of U.S. agriculture because more than 10% of the energy consumed in the U.S. agricultural sector goes toward the production of nitrogen fertilizer for corn. Because nitrogen fertilizer is a critical driver of both input costs and yield, this technology will improve the profitability of U.S. farmers by reducing input costs and maximizing corn yields. This project may enable corn production that creates less nitrogen fertilizer pollution, which threatens human health, degrades aquatic ecosystems and emits greenhouse gases that contribute to climate change. A key social element of this project is a teaching module for middle-school science students that will blend content on soil science, agronomy, and crop management with the challenges faced by U.S. corn farmers to follow best management practices.This Small Business Innovation Research (SBIR) Phase I project seeks to demonstrate the technical feasibility of using machine learning to evaluate the amount of yellowness on the lowest corn leaves visible in images taken from low-cost cameras mounted on ground robots. Characteristic yellowness on corn leaves is a strong indicator of stress caused by insufficient nitrogen, a key nutrient. A prototype neural network model will be iteratively improved, in part by dramatically increasing the available training imagery over the course of this Phase I project. Imagery will be collected on field trials set up across several Mid-Western states. Preliminary data suggest that the extent of characteristic yellowness is an indicator of accumulated nitrogen stress that is observable only under the canopy and not via airborne sensors. A commercially available nitrogen model will be used to estimate accumulated nitrogen stress across small plots created by manipulating the amount of added nitrogen fertilizer when corn is about 1-2 feet high. Tuning will occur by adjusting key model parameters through simulation until the differences are minimized between observed and modeled accumulated nitrogen stress across a field’s small plots. Parallel software development will improve prototype codes running simulations of a leading nitrogen model, enabling rapid model tuning.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.