Toward Spatio-temporal Models of Biogeophysical Fields for Ecological Forecasting (BDEI).

We are now in an era of intensive earth observation. Orbital platforms generate myriad remote sensing datastreams across a range of spatial, temporal, spectral, and radiometric resolutions. The number and variety of "eyes in the skies" are scheduled to increase significantly over the next few years. This veritable data deluge necessitates new ways of thinking about transforming remote sensing data into information about ecological patterns and processes. These datastreams hold the promise for environmental decision support. Effective use of remote sensing datastreams to characterize and monitor landscape dynamics requires analysis of the temporal variations in spatial patterns. We can distinguish four main phases in the analysis of image time series (Henebry and Goodin 2002): (1) change detection---perceiving the differences; (2) change quantification---measuring the magnitudes of the differences; (3) change assessment---determining whether the differences are significant; and (4) change attribution---identifying or inferring the proximate cause of the change. There is a critical need for theories and tools that will enable efficient and reliable characterization of spatio-temporal patterns contained in image time series. Such tools ought to be based on ecological expectations of land surface dynamics, analogous to climatological expectations. Ecological expectations would summarize across specific regions the typical temporal development of spatial pattern in biogeophysical fields. To make ecological forecasting an operational possibility (Clark et al. 2001; Maier et al. 2001), there is need for computational strategies to establish and to update complex spatio-temporal baselines that will enable prediction of the usual and the identification, quantification, and assessment of the unusual. Beyond computational considerations, there is an urgent need for environmental scientists to dialogue with computer scientists to develop effective and robust spatio-temporal models of biogeophysical fields for database and datamining applications aimed at the investigation of these baselines and their associated anomalies in datastreams. One objective in our Biodiversity and Ecosystem Informatics (BDEI) project is to sponsor an international cross-disciplinary workshop to examine the challenges facing the development and implementation of the next generation of spatio-temporal data models. This workshop is to held 8-10 April 2002 at the San Diego Supercomputer Center. We shall report the results at the dg.o conference. Here we offer some background on principal issues and a list of questions to be examined at the workshop.