Summary
Our research into Intelligent Ecosystem Prediction combines continental scale biospheric models and new computer science techniques for autonomous data understanding to develop new methods for rapid access, analysis, and utilization of large, heterogeneous data sets. The primary goal of this research is to develop, implement, and apply an adaptable architecture for automated conversion of large amounts of data from multiple sources into usable products, including forecast maps of biospheric conditions and predictions of episodic events based on causal relationships identified in data.

To accomplish this, large heterogeneous data sets are first retrived and assimilated by TOPS and the IMAGEbot Planner, and mined by the Tetrad Causal Analysis system and other discovery algorithms. Novel models discovered are returned to TOPS via a Planner API for incorporation into the Ecocast architecture for evaluation and verification. The entire system runs in a distributed, modular architecture with system components physically located in California, Montana, and Pennsylvania.

Work to date has demonstrated the feasibility of using this architecture to forecast fire using MODIS data. By bringing together domain experts from the computer science and Earth science communities, this project is leveraging diverse knowledge and resources to build the next generation of biospheric nowcasts and forecasts.