SBIR Phase I: Radar Snow Retrieval

小型企业创新研究(SBIR)第一阶段项目的更广泛影响/商业潜力将开发出变革性的机器学习算法，将改善水管理。水管理对美国西部当地人口的社会福祉、食物供应和气候适应能力至关重要，但水管理人员缺乏足够的积雪深度和水分信息，这是管理、储存和转移灌溉和消费用水所必需的。当积雪深度错误时，缺水情况会变得更加严重，可能会对农业经济和脆弱人口造成毁灭性的损害。拟议的技术将能够从风暴事件到季节性积雪估计，并为需要水管理的分水岭地区提供准确的积雪深度和水当量映射。这个小型企业创新研究(SBIR)第一阶段项目开发了确定积雪深度和水分的算法。积雪反演算法的发展没有跟上短波长的部署。C波段和X波段雷达在山区山谷中被用作“填补空白”的雷达。对于这些短波长雷达，需要开发有效的雪水当量检测算法。人工智能/机器学习(AI/ML)和优化算法预计将提高与点尺度(传感器)观测和与水管理有关的跨流域地区的估计精度。物理制导的神经网络(PGNN)可以产生物理上一致的结果，并推广到样本外的场景。将PGNN应用于SNOW检索有望比纯粹的数据驱动或确定性算法执行得更好。预期的技术成果将为水资源管理者提供实时更新的基于云的订阅服务，使用历史和当前的雷达数据来改进运营决策。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势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 will develop transformative, machine-learning algorithms that will improve water management. Water management is critically important to the social well-being, food supply, and climate resiliency of the local population in the Western United States, yet water managers lack adequate snowpack depth and water content information necessary for the management, storage, and transfer of water for irrigation and consumption. Water deficits are made worse when snowpack depths are in error, potentially resulting in devastating damage to agricultural economies and vulnerable populations. The proposed technology will be able to scale from storm events to seasonal snowpack estimations and provide accurate mappings of snowpack depths and water equivalents for watershed areas needing water management.This Small Business Innovation Research (SBIR) Phase I project develops algorithms for determining snowpack depth and water content. Snow retrieval algorithm development has not kept pace with the deployment of short wavelengths. C- and X-band radars are used as ‘gap-filling’ radars in mountainous valleys. Developing effective algorithms for detection of snow water equivalent is needed for these short wavelength radars. Artificial Intelligence/Machine Learning (AI/ML) and optimization algorithms are expected to improve estimation accuracy compared with point-scale (sensor) observations and across watershed areas relevant to water management. Physics-guided neural networks (PGNNs) can produce physically consistent results and generalize to out of sample scenarios. Application of a PGNN to snow retrievals is expected to perform better than purely data-driven or deterministic algorithms. Anticipated technical results will provide water managers with a cloud-based subscription service updated in real-time, using historical and current radar data to improve operational decision-making.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.