Arizona State University (ASU) will develop a modeling environment that couples the outcome of sophisticated simulations of weather conditions for hurricanes with an agent-based model of social actors who pursue, process, and transmit information. The coupled modeling laboratory will be developed using Open Source software to ensure that its operation is transparent and accessible to the scientific community. The project will simulate hurricanes and associated hazardous weather conditions using the Weather Research and Forecasting (WRF) model developed by the National Center for Atmospheric Research. The ASU Team will be responsible for developing the agent-based modeling (ABM) component. The ASU team will initially use the NetLogo ABM platform developed at Northwestern University for model prototyping and development. For more realistically parameterized experiments, we will move to the Repast ABM platform, developed at Argonne National Laboratory. While Repast is a more challenging development and prototyping environment, new versions can be run in the NCAR high performance computing (HPC) environment that will permit larger-scale simulations. Modeling even regional information networks coupled with storm simulations in this way is a challenging and complex task, and has the potential to become rapidly computationally intensive with large numbers of independent agents. To make this research more tractable, we plan to develop the modeling laboratory and carry out experiments as a two-phase activity. The first phase will be the development of a more abstract modeling environment with a limited number of cells and agents (hundreds to a few thousands). This environment would couple climate modeling to an ABM of sociallymediated information transmission at relatively limited scale to focus on understanding the underlying dynamics of information transfer over space and time for hazards warnings. It will also help to identify particularly more and less sensitive parameters. The second phase will build on results of the first phase. The cells of the virtual world will be parameterized with more detailed geographic information relevant to hurricane hazards, including coastal configuration, elevation, vegetation, and the built landscape. We will use a variety of social science data collected during the first phases of this project to help refine the parameterization of agent characteristics, including the ways in which they pursue and process information. We will then carry out systematic experiments in these more realistic modeling environments. This will serve two important goals. First, it will give us a way to evaluate model behavior and refine our model design to produce results that better match real-world information dynamics. Second, this will result in a prototype tool for testing ways to enhance the effectiveness of hazards warnings in particular realistic settings in ways that would not be possible without an actual storm event.
|Effective start/end date||11/10/14 → 8/31/18|
- NSF-GEO: Division of Atmospheric and Geospace Sciences (AGS): $414,437.00