Data Quality in High Performance Distributed Sensing Systems

We explore a class of distributed sensing systems that make use of multiple points of high bandwidth sensor readings that can be time synchronized and analyzed to make scientific inferences. Such high performance wireless embedded systems are needed to collect data for seismic, structural, and acoustic monitoring in challenging environmental conditions. Building and deploying these high performance systems pushes the limit of existing wireless systems components to the point that data can be lost and data quality can suffer. Through several iterations of deployments we have explored techniques, and developed and evaluated methods to prevent data loss and repair data quality.;We have developed a system and a set of supporting tools called Disruption Tolerant Shell (DTS). This system is a redesign of existing Internet tools for data delivery and system management to work without end-to-end connections. The system provides efficient, near real time data delivery as well as status and performance reports of the entire system. DTS has been validated on two real long term deployments, the MesoAmerica Subduction Experiment (MASE) and the PERU Seismic Experiment (PERUSE) which consisted of 50 wirelessly connected seismic stations covering 300 Km.;Preventing data quality loss in high performance embedded sensing systems is entirely a problem of maintaining time synchronization. We have developed a methodology called Data Driven Time Synchronization (DDTS) which provides post-facto time synchronization. DDTS works by using characteristics about the data to provide time synchronization. We have used DDTS to recover incorrectly time synchronize data form our MASE deployment using microseisms. In the process we have made discoveries about how ocean effects such as wave height are directly correlated to microseism. We have also shown that there is potential to use DDTS for acoustic networks by applying similar data processing techniques we used for seismic DDTS.