The nature of dark energy is one of the fundamental astrophysical questions of our time. This project will address the computing challenges of current and upcoming cosmological surveys which will seek to answer this question. The research activities will focus on weak gravitational lensing, in which the signal is the distortion of images of distant galaxies, caused by gravitational perturbations of photon paths as they travel through space. The interpretation of these data is a computationally intensive process: the shapes of the raw galaxy images must be accurately determined, taking into account correlated systematic effects such as distortions due to the atmosphere and telescope optics. For the upcoming Large Synoptic Survey Telescope (LSST) project, this process must be performed on-line on around one billion separate objects per night. Once these results are in hand, there is the further challenge of processing them: comparing theory to various statistics derived from the observations can give quantitative insight into fundamental aspects of the universe: its expansion history, the growth of structure, and the nature of the dark matter and dark energy which together make up 95% of the energy-content of the universe.
Great strides have been made in addressing the computational aspects of weak lensing measurement and analysis. Unfortunately, many current methods will have difficulty scaling to the size of future surveys. LSST will image the entire southern sky every three nights over the course of ten years, producing an astronomical database of unprecedented size. Analysis of these data will require the use of new database architectures with new, purpose-built parallel algorithms. The current project will address those issues by fostering a close link between computer science and the "domain sciences" like astronomy. The work will take place within the Database Research Group in the Computer Science& Engineering department at University of Washington (UW), with close ties to the Survey Science Group in the Astronomy Department at UW. This project will contribute fundamental results both in computational and astronomical science, mainly through the development of techniques within the new SciDB parallel data processing engine to address specific computational hurdles for cosmological research with LSST. This research is vital to the science goals of LSST: current methods will not scale to the size of the survey.
The impact of this research will reach beyond astronomy to other domain sciences. Many domain sciences are currently producing data at an increasing rate. Scientists in a variety of fields are beginning to see the need for close collaboration with computer scientists who have experience in the methods required to store, access, and process their data. At the same time, computer scientists who are pushing the limits of computation often lack access to datasets on which to test their methods. The PI will foster interdisciplinary connections by developing a seminar in data-intensive computing, geared toward graduate students in a broad array of disciplines. In addition the PI will build on his volunteer experience with Seattle's Pacific Science Center Science Communication Fellow program, which seeks forums for scientists to share their research with the broader public. Through ongoing involvement in this program, he will continue to share the exciting results of data-intensive astronomy with members of the community.
