A new generation of satellites is imaging the earth's surface with unprecedented spatial and spectral resolution. With the ability to identify local features related to environmental exposures, this high-resolution imagery is gong to revolutionize health risk assessment. The realization of this potential depends critically on our ability to recognize spatial patterns on these large images. This project will develop fast spatial null models for use in statistical pattern recognition, and will accomplish 4 aims. (1) Implement fast simulation algorithms conditioned on properties of the data, and on spatial functions; (2) Assess project feasibility by evaluating the performance of these algorithms on existing high-resolution, hyperspectral imagery; (3) Implement the simulation algorithms in 2 commercial spatial analysis software packages; (4) Apply the software and methods to demonstrate the approach and unique benefits for risk assessment. The phase 1 research will address the first two aims;
aims three and four will be accomplished in phase 2 once feasibility is demonstrated. The technologic and scientific innovations from this project are expected to greatly enhance our ability to extract knowledge from high resolution imagery.
The imminent launch of over a dozen satellites capable of high-resolution imagery is giving health researchers powerful new data for relating environmental features to health outcomes, but existing software packages cannot undertake spatial analysis of these extraordinarly large data sets. The fast simulation algorithms from this research will be incorporated into 2 commercial software packages, providing advanced spatial analysis for large imagery.
