Ecological Forecasting: Advanced Technologies for Discovery in Earth Science Data

With NASA sensors onboard satellites, aircraft, and UAVs currently producing over two terabytes of data per day, and considering the wealth of ground-based observation networks, there is a clear need for architectures and systems capable of autonomous analysis and utilization of sensor web data streams. Our research has combined biospheric models with remotely sensed data and new computer science techniques to develop a biospheric monitoring and forecasting system. The Terrestrial Observation and Prediction System (TOPS) is an operational system and has capabilities for rapid access, integration, and utilization of multiple large, heterogeneous data sets. TOPS incorporates cutting edge computer science algorithms for causal discovery and automated planning to provide a robust capability for on-demand data processing. TOPS also provides an operational environment for data-driven modeling and discovery using multi-terabyte Earth observation data archives. Automated data fusion capabilities provided by TOPS have been used in data driven modeling experiments. These experiments have employed machine-learning algorithms for learning causal structures to search terabytes of Earth observation data and develop novel models of Earth science processes such as wildfire risk. Using TOPS, we are also implementing models from multiple domains to develop a range of applications including mapping of wildland fire risk, UAV deployment for wildfire monitoring, irrigation forecasting, tracking anomalies in global net primary productivity, and mapping vector abundance and disease transmission risk. TOPS is currently being used to produce nowcasts and forecasts of biospheric conditions from local to global scales. Products and images from TOPS are distributed via the web and available for use by scientists, educators, and decision makers.