Visual and Machine Analytics for Environmental Monitoring

被污染的陆地场所，如废弃油田或化学品泄漏是一个主要的环境问题。加拿大20多万个受污染的地点通过土壤和水污染造成健康问题，并代表着数十亿美元的环境和经济责任。受污染的场地必须进行修复以去除污染物，而合作伙伴环境材料科学（EMS）开发了一种新的生物修复方法，比现有方法有效得多，成本也低得多。他们的方法的一个关键要素是比其他方法更频繁地监测当前的污染水平，以便让利益相关者确定补救措施是否正确工作，并预测一个地点何时将符合规定。从污染现场的传感器收集的信息用于污染现场的模型和可视化，以帮助利益相关者做出决策。然而，目前的建模和可视化技术在很大程度上没有在环境背景下进行测试——大量的变量，生成准确模型的困难，每个状态存在的多维不确定性，以及许多不同的利益相关者为分析带来了新的和困难的挑战。该项目将开发分析技术，在污染场地监测和补救方面取得成功。我们的主要成果将是实现新的建模、分析和可视化技术的新颖软件工具，这将大大提高EMS在实际操作中的修复解决方案的有效性，从而更快地修复加拿大受污染的场地。我们有三个研究目标：传感器分析，以改善现场传感器数据的解释、聚合和表征；在高度不确定的领域更准确的预测模型；可视化技术表示数据、模型和不确定性，使多方利益相关者能够对受污染地点做出更好的决策。

Contaminated terrestrial sites such as abandoned oilfields or chemical spills are a major environmental problem. Canada's 200,000+ contaminated sites cause health problems through soil and water pollution, and represent a multi-billion dollar environmental and economic liability. Contaminated sites must be remediated to remove pollutants, and partner Environmental Material Science (EMS) has developed a novel bio-remediation approach that is far more effective and far less expensive than current methods. A key element to their approach is monitoring current levels of contamination much more frequently than other methods, in order to allow stakeholders to determine whether remediation is working correctly, and to predict when a site will come into compliance. The information gathered from sensors at the contaminated site is used in models of the contaminated site, and in visualizations that assist stakeholders in making decisions. However, current modeling and visualization techniques are largely untested in the environmental context - and the large number of variables, the difficulties in generating accurate models, the multi-dimensional uncertainty present at each state, and the many different stakeholders present new and difficult challenges for analytics. This project will develop analytics techniques that can succeed in the context of contaminated-site monitoring and remediation. Our main outcome will be novel software tools that implement new modeling, analytics, and visualization techniques, that will substantially improve the effectiveness of EMS's remediation solutions in real-world operation, leading to much faster remediation of Canada's contaminated sites. We have three research objectives: sensor analytics to improve interpretation, aggregation, and characterization of site sensor data; predictive models that are more accurate in highly-uncertain domains; and visualization techniques that represent data, models, and uncertainty to enable multiple stakeholders to make better decisions about contaminated sites.