This award facilitates scientific research using the large, new, computational resource named Blue Waters being deployed at the University of Illinois. It provides travel funds to support technical coordination between the principal investigators, the Blue Waters project team and the vendor technical team.
This collaborative award to Pennsylvania State University and Princeton University formally links high-resolution astrodynamics design and coordination of space assets with Earth science impacts. This is accomplished through the project's petascale many-objective global optimization framework. The multidisclinary research team will exploit Petascale global sensitivity-based model diagnostics to evaluate the predictive fidelity of the astrodynamics and Earth system models, and allow visualization of design tradeoffs. With Blue Waters, the project's research aims to transform the scale and scope of systems-of-systems engineering design challenges that can be explored. From the design cognition perspective, this use of Blue Waters will radically advance ability to discover and visualize optimal many-objective satellite constellation design tradeoffs that include Earth science applications. In this case the research will demonstrate the planning framework on Blue Waters using Global Precipitation Measurement (GPM). GPM provides a highly challenging design benchmark that has strong links to Earth science through its (1) 4-D characterization of freshwater availability, (2) improved characterization of the microphysical properties of precipitation events, (3) improved climate predictions based on enhanced estimation of surface water fluxes, (4) high value operational data for numerical weather prediction, and (5) global enhancements in hydrologic prediction and management of floods, droughts, landslides, and hurricanes.
Beyond the US and Europe, there is a critical deficiency in the in-situ Earth observations needed to monitor floods and droughts. In the developing world, water managers are reliant on future satellite observations and in particular on satellite-based precipitation information. The proposed framework will strengthen our ability to serve this critical need by providing a specific evaluation of how candidate GPM designs can potentially improve our ability to use satellite data for flood assessment, drought monitoring, and assessing water availability in large river basins. The project will also be used in the training of 15-25 demographically and geographically diverse graduate PhD students from engineering, computer science, mathematics, statistics, operations research, and the Earth sciences.
Providing reliable and timely hydrologic information to policy makers and stakeholders can help mitigate the impact of hydrologic extremes (i.e. droughts and floods). This is a challenging task over data sparse regions due to a lack of robust ground measurement networks. A promising yet elusive solution is to rely on satellite remote sensing to provide an integrated global water cycle observatory. In this project we have addressed this concern by analyzing the global coverage frequency of the current satellite constellation that measures precipitation; this information has allowed us to propose solutions to increase satellite precipitation fidelity. Our team at Princeton, Cornell, and Aerospace has accomplished these goals by (1) formally linking high-resolution astrodynamics design and coordination of space assets with their Earth science impacts within a Petascale global optimization framework, (2) successfully completing the largest Monte Carlo simulation experiment for evaluating the required satellite frequencies and coverage to maintain acceptable global monitoring of hydrologic extremes, and (3) evaluating the limitations and vulnerabilities of the current suite of satellites. Our team has exploited access to the Blue Waters machine to radically improve our ability to discover and visualize optimal tradeoffs. Our design of satellite-based precipitation systems has explored the use of perturbing astrodynamics forces for passive control, the sensitivity of global water cycle simulations on attainable satellite data frequencies, and advancing new technologies for highly scalable many-objective design optimization. This project has been the first to develop a 10,000 member Monte Carlo global hydrologic simulation at a one degree resolution that characterizes the uncertain effects of changing the available frequencies of satellite precipitation on drought and flood forecasts. The simulation—optimization components of the work have set a theoretical baseline for the best possible frequencies and coverages for global precipitation given unlimited investment, broad international coordination in reconfiguring existing assets, and new satellite design objectives informed directly by key global hydrologic monitoring requirements. To provide a robust analysis of the global satellite coverage and propose alternative solutions required millions of computing hours and petabytes a storage – a task that Blue Waters made possible. In simple terms, the scale and ambition of our computational experiments required the ability to compress years of computational work into minutes of wall-clock time. Our applications were extremely data intensive, so Blue Waters’ high core count and high memory were fundamental to realize the discoveries from this project. The global hydrologic ensemble required approximately 30 million core hours yielding up to 2 Petabytes of model output. This output represents a new benchmark dataset that will be of broad interest in a variety of Earth science and engineering applications. Our satellite design tradeoff analysis expended approximately 120 million core hours to discover how quickly we deviate from the optimal observation frequencies with limits on spending, limits in international coordination, neglect of hydrologic objectives, and the simplified astrodynamics simulations currently employed in practice.
