Knowledge Discovery from Sensor Data For Scientific Applications

Wireless sensor networks, in-situ sensor infrastructures and remote sensors offer new opportunities for pervasive monitoring of the built and natural environments. The ability to generate actionable predictive insights from raw sensor data is critical for many scientific applications of high societal priority. One example is sensor-based early warning systems for geophysical extremes like tsunamis or hurricanes, which can help preempt disaster damage through tactical decisions. Indeed, the impacts of the 2004 Indian Ocean tsunami may have been almost entirely be preventable if early warning systems were well developed. One other example is high-resolution risk-mapping for natural hazards, based on predictive insights obtained through a combination of historical and real-time sensor-based data, with physics-based computer simulations. High-resolution risk maps for geophysical extremes, in combination with information about consequences and resiliency, can lead to strategic policy tools, which in turn can be utilized to reduce the impacts of anticipated natural hazards. Sensor-based knowledge discovery requirements need to be geared towards scalable and efficient implementations of offline predictive insights and fast real-time analysis of incremental information. Predictive insights regarding weather, climate and geophysical hazards require models of rare, anomalous and extreme events, nonlinear phenomena, and change analysis, in particular from massive volumes of geographic data. On the other hand, historical data may also be noisy and incomplete, thus robust tools need to be developed for these situations. This chapter describes the problems and research challenges in the interdisciplinary area of hazards mitigation, specifically focusing on new opportunities for knowledge discovery from sensor data and model simulations.