ITR: Statistical Pattern Recognition in Environmental Observation and Forecasting Systems

Oregon Health & Science University, Portland OR

Investigators

Abstract

Project Summary Environmental observation and forecasting systems (EOFS) are emerging new technologies with unparalleled potential to impact sustainable development. EOFS are expected to foster and support new paradigms for generation, transfer and social application of knowledge in domains that range from the global earth to its regional and local sub-systems. At the core of EOFS is the timely and customized acquisition, generation, processing and delivery of reliable, relevant information to many and very diverse audiences. Multiple challenges need to be met to implement this concept. A critical challenge is the development of automated procedures to verify the quality of the huge amounts of observational and simulation data that are generated by EOFS both in real-time and off-line. Process- based strategies for quality control of scientific data, while effective for moderate-size archival data are too labor-intensive to map well into EOFS-scale data sets. Strategies based on pattern recognition and machine learning hold significant promise as an alternative or complement. Under the proposed project, we will develop approaches based in statistical pattern recognition and signal processing, on-line adaptive systems, datamining, and advanced search to address critical quality control issues including: 1) Detecting sensor corruption in non-stationary, spatial-temporal systems, 2) Estimating true signals from corrupted sensor data, and 3) Detecting and characterizing regimes where model anomalies are likely. These quality control techniques will be developed and exercised on CORIE, a pilot EOFS for the COlumbia RIver Estuary and adjacent coastal waters (http://www.ccalmr.ogi.edu/CORIE) The project will have strong social impact, through the role of EOFS (and, specifically, CORIE) on regional and national sustainable development issues. The project will also include cross-disciplinary educational opportunities at multiple levels.