FIRST THE PAST, THEN THE FUTURE
There’s little question that modeling is, for Liang, a work of passion. Much like a musician tunes in, Liang’s focus is absolute. “So you can see, see, look, beautifully done,” he says, pointing to where the lines generated by his model almost perfectly coincide with the lines representing storms and stream flow in the past.
Liang’s been working on his own model CWRF – pronounced “sea wharf” – for five years now. Toying with the model’s results, his excitement is clear, but the model’s not quite where he wants it to be yet, which is why he's not ready to release it.
CWRF is not alone in North America; there are at least six other regional models. But not enough has been done to vet their accuracy as guides to the future climate.
“It’s kind of a space that’s open for development, but we need to do more research,” said Dave Bader, the lead author on the U.S. Climate Change Science Program’s recent report about models. Bader is also the director of the program for climate model diagnostics and inter-comparison at Lawrence Livermore National Laboratory in California.
The U.S. government began its first major comparison project – the North American Regional Climate Change Assessment Program – in late 2006. The idea was to line up six regional models at the same start line and run them all using four sets of global conditions established by the larger climate models. Then, the race would be on.
At the end, it would have been possible to declare winners, based on the models’ ability to forecast heatwaves, hurricanes and other climate events from the past accurately.
But the program was shorted on funding, receiving only half of the $5 million requested, so most of the models are run through two of the four possible simulations, according to Linda Mearns, the program’s director and a senior scientist at the The National Center for Atmospheric Research in Boulder, Colo. The results – which will still provide more information then is currently available about the models’ forecasting abilities – are being made available in a public database, so that scientists the world over can start the work of analyzing how they are performing.
As the results are studied, one model might emerge as the best at predicting rainfall, while another excels at forecasting fluke floods.
Across the Atlantic, the European Union completed a comparison study of regional models, and has already begun a second project that involves high-resolution modeling called Ensembles.
The United States’ tardiness in beginning comparison work has meant little information is available about the strengths — and failures — of the regional climate models that center on North America. “We don’t really have the evaluations and comparisons for [regional models] that we have for the global models,” said Gutowski, who is involved in the new comparison project.
Unlike the regional models, the global models have been run from the same start line, in exactly the same conditions. So scientists know which ones are good at predicting long-distance events – things such as the temperature of the ocean, which require long-distance calculations corresponding to their size. And they know which ones are better at sprinting to forecast a tropical storm that’s closer on the horizon.
Mearns got the idea for the comparison project after spending time in Italy at the International Centre for Theoretical Physics with Filippo Giorgi, one of the first regional modelers. The lack of information about regional models also frustrated Mearns while she was writing about regional forecasts for the latest report released by the Intergovernmental Panel on Climate Change.
“We couldn’t make any grand statements, because none of the results were really comparable,” Mearns said.
Like race results where a runner skips out and takes a shortcut, there was no real way to know which model was the best forecaster for the future because they had all been run under different conditions for North America.
On his own, Liang has been comparing his model’s ability to, in effect, forecast the past against that of other models. It’s a standard test for a model; before it can be trusted to make reliable projections about the future, it first has to demonstrate its success with the past.
Climatologists call this “hindcasting.”
“Really the only basis we have for believing these models can be used as a window into the future is the extent to which they can accurately do the past,” said Kenneth Kunkel, formerly with the Illinois State Water Survey and now at the Desert Research Institute in Reno, Nev.
So Liang set two benchmarks for CWRF.
It must accurately recreate the climate between the years of 1982 and 2007, a period chosen because of the rich data set available about temperatures, rainfall, stream flows from ice melt in the mountains – and other climate information.
Liang is also asking CWRF to predict two extreme weather events from the past: the 1988 drought that hit the Midwest and the floods that followed there five years later. For the second goal, he has received a grant from the National Oceanic and Atmospheric Administration.
CWRF is not Liang’s first regional model. Nor is it entirely his model; like most scientific discoveries, models build on existing knowledge and so they have many makers.
CWRF is an effort to improve upon an existing model, in this case a well-known model for weather research and forecasting that’s known as WRF. That model, developed by the The National Center for Atmospheric Research, along with other U.S. science centers, is used for both real-time forecasting and long-term research.
But if Liang succeeds, he expects his model, once fully functional, to be able to say more about the future climate in Illinois and other areas in North America.
Previous Page | 1 | 2 | 3 | 4 | 5 | Next
Print
Email
Bookmark
