SBIR Phase I: A Physics Guided Machine Learning Framework for Monitoring Rivers using Satellite Imagery

小企业创新研究（SBIR）第一阶段项目的更广泛影响将是为保险，能源和农业利益相关者提供空间范围（洪水测绘）和河流流量（水资源管理）变化的近实时信息。有效管理水资源和相关风险已成为社会面临的一项重大挑战。洪水是世界各地常见的灾害，干旱则导致经济严重中断和生命损失。该项目将利用人工智能和千兆字节的卫星图像来实施物理指导的数据密集型方法，以推进全球水文建模。该项目将在一个方便用户的网络环境中，以相对较低的计算成本（与地面传感器相比），提供对河流水动力的有效和准确的图像观测。这将是改善全球水资源建模和预测的重要一步。这一小企业创新研究（SBIR）第一阶段项目旨在开发先进的人工智能技术，利用大量卫星图像跟踪地球仪中河流的地表水变化。虽然传统的人工智能技术纯粹是由数据驱动的，但所提出的技术将已知的物理定律融入到这些算法中。这种物理学引导的方法使得这些技术对大气扰动（云、阴影、霾等）更加鲁棒，并且能够协同使用不同分辨率的图像数据集，这是卫星图像分析的两个主要问题。此外，拟议的不确定性量化技术将使领域专家将其当地的河流流量的知识纳入框架，以完善results.This奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

The broader impact of this Small Business Innovation Research (SBIR) Phase I project will be to provide near-real time information of changes in the spatial extent (flood mapping) and flow of rivers (water resource management) to insurance, energy, and agricultural stakeholders. Effective management of water resources and associated risks has become a major challenge for society. Floods are common disasters around the world and droughts lead to major disruptions to economies and a loss of life. This project will leverage artificial intelligence, and peta-bytes of satellite imagery to implement a physics guided data-intensive approach for advancing global hydrological modeling. The project will provide efficient and accurate imagery-derived observations of water dynamics in rivers at relatively low computational cost (compared with ground sensors) in a user-friendly web environment. This will be a significant step towards improving the modelling and forecasting of water resources around the world.This Small Business Innovation Research (SBIR) Phase I project aims to develop advanced artificial intelligence techniques to track surface water changes in rivers across the globe using vast amounts of satellite imagery. While conventional artificial intelligence techniques are purely driven by data, the proposed technology incorporates known physical laws into these algorithms. This physics guided approach makes these techniques much more robust to atmospheric disturbances (clouds, shadows, haze, etc.), and enables synergistic use of imagery datasets at different resolutions which are two major issues with satellite imagery analysis. Furthermore, the proposed uncertainty quantification techniques will enable domain experts to incorporate their local knowledge about river flows into the framework to refine results.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.