EAGER-DynamicData: Transforming Wildfire Detection and Prediction using New and Underused Sensor and Data Sources Integrated with Modeling

EAGER-DynamicData：使用新的和未充分利用的传感器以及与建模集成的数据源来改变野火检测和预测

• 批准号: 1462247
• 负责人: Janice Coen
• 金额: $ 10万
• 依托单位: University Corporation For Atmospheric Res
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-01 至 2017-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1462247&HistoricalAwards=false
• 关键词: EAGER; DynamicData; Transforming; Wildfire; Detection

Wildland fires are a costly natural hazard. Newer modeling systems have combined numerical weather prediction models with the traditional tools used to model fire behavior, making them more capable of realistically modeling how fires unfold, however, applying them to accurately anticipate a fire?s growth is a difficult forecasting challenge. The principal challenges are that errors accumulate as the accuracy of weather forecasts decreases with time and that some processes cannot be anticipated by the model such as the lofting of burning embers ahead of the fire (potentially starting new fires) and firefighting. The team?s recent work has combined theCoupled Atmosphere-Wildland Fire Environment (CAWFETM) weather?fire behavior modeling system with satellite-based fire detection data from the Visible and Infrared Imaging Radiometer Suite (VIIRS) instrument to ignite fires already in progress, allowing an accurate forecast of fire growth for the next 12-24 hours; sequences of these simulations can maintain a reasonable forecast of fire growth from the time the satellite detects it until it is extinguished. The remaining challenges limiting the forecast skill are common to traditional approaches to modeling complex, nonlinear natural systems and include accumulating error and optimally exploiting all available data sources. The team will investigate how more tightly integrating new and underused sensor and data sources with the modeling could potentially transform both wildfire detection and prediction. Advances will be integrated into the team?s work transitioning the system into operations, benefiting society with earlier wildfire detection, faster response, and better fire forecasts.The goal is to develop innovative Dynamic Data System techniques that improve wildfire detection and growth forecasting. The work will address three objectives, 1) develop and apply algorithms (steered by other data) to distill new and existing (but underutilized) sources of data on wildfire detection and mapping, 2) develop and apply algorithms to integrate asynchronousdata on wildfire detection and monitoring with coupled weather?wildland fire models, and 3) measure the improvement in wildfire detection time and forecasted fire growth. The methods include creating an adaptive control system for initiating forecasts based on the arrival of new data; allowing sensors to inform algorithms where to look in other underutilized datasets; creatingapproaches for intelligent, iterative processing of large datasets; and using model forecasts to drive these intelligent searches. The techniques could have broad application across other nonlinear systems that are currently done in a traditional manner with rigorous scheduling of routine, repeated modeling relying on fixed detection algorithms and regular, periodic input dataarrival.

荒地火灾是一种代价高昂的自然灾害。较新的建模系统将数值天气预报模型与用于模拟火灾行为的传统工具相结合，使它们更能够真实地模拟火灾如何展开，但是，应用它们来准确预测火灾的发展是一项艰巨的预测挑战。主要的挑战是，随着天气预报的准确性随着时间的推移而降低，误差会累积，并且模型无法预测某些过程，例如在火灾前燃烧余烬的升起（可能引发新的火灾）和消防。该团队最近的工作将大气-荒地火灾环境耦合 (CAWFETM) 天气火灾行为建模系统与来自可见光和红外成像辐射计套件 (VIIRS) 仪器的卫星火灾探测数据相结合，以点燃已经发生的火灾，从而允许准确预测未来 12-24 小时的火势蔓延情况；这些模拟的序列可以对从卫星检测到火灾直至火灾被扑灭的火灾蔓延进行合理的预测。限制预测技能的其余挑战对于复杂、非线性自然系统建模的传统方法来说是常见的，包括累积误差和优化利用所有可用数据源。该团队将研究如何将新的和未充分利用的传感器和数据源与建模更紧密地结合起来，从而可能改变野火检测和预测。进步将被整合到团队的工作中，将系统转变为运营，通过更早的野火检测、更快的响应和更好的火灾预测来造福社会。目标是开发创新的动态数据系统技术，以改进野火检测和增长预测。这项工作将实现三个目标，1）开发和应用算法（由其他数据引导）来提取有关野火检测和绘图的新的和现有的（但未充分利用的）数据源，2）开发和应用算法来集成野火检测和绘图的异步数据。结合天气与野火模型进行监测，3) 衡量野火检测时间和预测火势增长的改进情况。这些方法包括创建一个自适应控制系统，用于根据新数据的到来启动预测；允许传感器通知算法在其他未充分利用的数据集中查找的位置；创建对大型数据集进行智能、迭代处理的方法；并使用模型预测来驱动这些智能搜索。这些技术可以在其他非线性系统中广泛应用，这些系统目前以传统方式完成，通过严格的例程调度、依赖于固定检测算法和定期、周期性输入数据到达的重复建模。