A group at the University of Illinois at Urbana-Champaign offers planners, policymakers, interest groups and citizens a glimpse of the future, using a computational model to simulate land-use change.
If a new road is built, how will the flow of traffic in the city change? Will businesses and people be drawn to the new thoroughfare, and if so will this represent economic development or simply migration from other areas of the city? If the road sparks residential growth, how will city services meet the challenge? Will schools and fire stations need to be built in the area? How will growth affect water quality, wildlife habitat, and other environmental factors?
These questions and many others like them whirl in the brains of government officials and urban planners as they try to make the best decisions for their communities. How can they possibly foresee the impact their choices will have a year, five years, 10 years, 20 years, and even 50 years into the future? It's a challenge that would stump a crystal ball.
A group at the University of Illinois at Urbana-Champaign is able to offer decision makers just such a glimpse of the future, however, not with a crystal ball but with a computational model that simulates land-use change across space and over time. Planners, policymakers, interest groups and laypeople can use Land-use Evolution and Impact Assessment Model (LEAM) to visualize and test the impact of policy decisions.
The computationally intensive simulations have leveraged the power of NCSA's high-performance computing systems, and the LEAM group relies on NCSA's 3-petabyte mass storage system to store the vast quantities of data generated by the simulations.
Developing the model
The LEAM project had its rather unlikely origins in the late 1990s with a paper on the spread of rabies in foxes that was co-authored by Brian Deal, a research professor of urban and regional planning, and Bruce Hannon, a professor of geography. As the two researchers collaborated on visualizing this complex, multi-variate process, it occurred to them that a similar model could be created to simulate how environmental, social, and economic systems interact to drive the evolution of urban areas. As their work expanded, they were joined by experts from the University's landscape architecture, urban planning, geography, economic, and natural resources and environmental sciences departments, including co-principal investigator Varkki George Pallathucheril, an associate professor of urban and regional planning.
The model developed by the LEAM team assesses an area's “growth potential.” The area of interest – a city, a county, a region – is divided into smaller cells. The model draws on population, geography, and land-use data; for each cell, information on economics, transportation, utilities, and neighboring land uses is factored in. All of the factors are weighted to determine the probability that a particular cell will change, and what type of change is most probable.
The model allows decision makers to test scenarios, Deal says, helping them consider what their preferred outcome is and how they can get there.
Assessing policy impacts
The LEAM model got its first real-world test with Kane County, Illinois, a historically agricultural area that has begun to expand rapidly due to expansion from the Chicago metropolitan area. The county's leaders are struggling to manage the pressure for rapid development that threatens to consume the area's agricultural land and open spaces.
First the LEAM group worked with Kane County's leaders and members of the community to understand what issues were important to them and what factors were driving growth. Data were gathered from sources such as state agencies, water and soil conservation districts, and even building permits. Then several simulations were run, analyzing how various policies (a fast-growth policy vs. a slow-growth policy, various open-space set-asides, etc.) would affect the area's development.
A scenario that included a fast-growth policy with a low open-space set-aside, for example, resulted in predictably explosive growth, shown on color-coded maps so the data can be readily understood and interpreted. The LEAM model further described what impacts this growth could have on diverse factors such as energy consumption (with gas and electric consumption soaring to a level that might outstrip what local utilities could provide) and raccoon and bird habitat (the Eastern Meadowlark population would fall with a loss of habitat, while the adaptable, scavenging raccoons would be abundant).
A boost from NCSA
The Kane County simulations encompassed approximately 1 million cells, and the dynamic model carries out multiple calculations for each cell, and the calculations are carried out over many time steps. Each simulation took hours to run, and if an error occurred in the simulation, it took hours to discover and correct course.
That's when Jeff Terstriep, formerly associate director for computing and communications at NCSA, stepped in to give the project a big boost. Terstriep parallelized the LEAM code and has benchmarked it on as many as 1,024 processors, enabling the team to take advantage of the high-performance computing clusters at NCSA. Simulations that had taken hours could now be completed in under 30 minutes using NCSA's IBM p690 system.
“The big boost for us was being able to get at the supercomputers and parallel process our work. We really wouldn't have made any advances without NCSA,” Deal says. “That was a real necessity because it provided computing that we couldn't do here.”
Recently the LEAM lab acquired a small AMD Opteron cluster on which to crunch the number for many of its simulations, but as they experiment with calibrating the model with genetic algorithms, the team will again need to use a high-end system.
“It's just not feasible to do on our own servers,” Terstriep says. He has done some preliminary work on NCSA's Tungsten cluster.
The importance of storage
Of even greater significance for the LEAM team, however, is NCSA's 3-petabyte mass storage system, DiskXtender, which is available to all the center's users for permanent data storage. DiskXtender is a boon for the data-intensive LEAM simulations, which require massive data inputs and generate 300-500 megabytes of data from each computational run.
Terstriep says NCSA's mass storage system offers three key benefits to the LEAM team: permanence (all of their data from all of their projects is stored; there's no need to weed out older files to make room for new ones); a high access rate (meaning it takes just a short time to access the data for a new simulation, or to call up historic data); and bandwidth.
“The mass storage system is one of the most important services we use at NCSA,” Terstriep says. “It's a vital component that is impossible for research projects to duplicate.”
This project is supported by the National Science Foundation, the U.S. Department of Defense, and the State of Illinois.
For more information:
http://www.leam.uiuc.edu/