Real-time Bayesian Anomaly Detection for Environmental Sensor Data

Recent advances in sensor technology are facilitating the deployment of sensors into the environment that can produce measurements at high spatial and/or temporal resolutions. Not only can these data be used to better characterize systems for improved modeling, but they can also be used to produce better understandings of the mechanisms of environmental processes. One such use of these data is anomaly detection to identify data that deviate from historical patterns. These anomalous data can be caused by sensor or data transmission errors or by infrequent system behaviors that are often of interest to the scientific or public safety communities. Thus, anomaly detection has many practical applications, such as data quality assurance and control (QA/QC), where anomalous data are treated as data errors; focused data collection, where anomalous data indicate segments of data that are of interest to researchers; and event detection, where anomalous data signal system behaviors that could result in a natural disaster. This study develops two automated anomaly detection methods that employ Dynamic Bayesian Networks (DBNs). These machine learning methods can operate on a single sensor data stream, or they can consider several data streams at once, using all of the streams concurrently to perform coupled anomaly detection. This study investigates these methods’ abilities, using both coupled and uncoupled detection, to perform QA/QC on two windspeed data streams from Corpus Christi, Texas; false positive and false negative rates serve as the basis for comparison of the methods. The results indicate that a coupled DBN anomaly detector, tracking the actual windspeeds, their measurements, and the status of these measurements, performs well at identifying erroneous data in these data streams.