Geospatial Computing Collaborations

Geospatial algorithm development is a complex collaborative process involving remote sensing scientists with different expertise needing to cooperate on massive and diverse datasets. These datasets can range from satellite imagery to citizen scientist observations, and include both publicly available and proprietary resources. The traditional mode of work is for each participant to have their own desktop machine with large attached storage and to essentially work in a silo. This approach, however, inhibits effective collaboration because individuals must navigate a complex system setup and use ad-hoc techniques such as sharing a portable hard drive to co-produce data products. The limitation of running on a desktop computer also restricts the scale of questions that can be asked due to excessively long processing times, sometimes spanning several weeks. Another consequence of this process is the lack of data conformance standards, that tend to emerge from shared needs when communities collaborate to advance application frameworks.In this paper, we investigate the social and collaborative aspects of cloud architectures to support geospatial computing. These new architectures provide emergent modes of collaboration. We provide a concrete case study of a typical geospatial collaboration, and detail the build of a collaborative platform as we evolve it from a silo, to a research cluster, and finally to a public cloud. The architectures are evaluated in terms of costs and benefits from systems and software engineering perspectives.