Computation of earth science products on spaceborne platforms

The next generation of Earth-observing spacecraft are likely to generate enormous volumes of data. A major challenge lies in the conversion of these mountains of data into information useful to researchers and other users. Hierarchical segmentation is one way to detect relationships among regions in a hyperspectral image. We implemented this algorithm on a next-generation space-capable hardware platform, and studied its performance before and after adapting it to use the platform's unique computational resources. We found that these adaptations enable an order-of-magnitude increase in performance over our initial implementation, and our detailed analysis points to areas for additional improvement.