From Yahoo News via the Christian Science Monitor
A new team is trying a new approach to climate modeling using AI and machine learning. Time will tell if there is a positive effort or an extremely complicated exercise in curve fitting. Your goal is to create regional-level predictive models that are useful for planning. Few publicly admit that despite thousands of “studies” with scaled-down GCMs, these do not exist today.
“There are some things that have very robust results and other things that those results are not that robust,” says Gavin Schmidt, director of NASA’s prestigious climate modeling program at the Goddard Institute for Space Studies. But the deviations urge skeptics to dismiss the whole field.
“There are enough things out there that people can choose to support their prejudices,” says Dr. Housefather. “Climate skeptics … argued that climate models always predict too much warming.” After studying the models of the past 50 years, Dr. Hausfather: “It turns out they did remarkably well.”
However, climate modellers recognize that accuracy needs to be improved in order to find a way through the climate crisis. Now, a team of climatologists, oceanographers and computer scientists on the east and west coasts of the United States have started a brave race to do just that.
They have brought together some of the best experts from around the world to start building a new, modern climate model. They hope to correct the enormous flow of data from sensors in space, on land and in the ocean and use “machine learning”, a type of artificial intelligence, to bring their model to life and provide new insights into what many believe is it most pressing threat to the planet.
Their goal is to provide accurate climate predictions that can tell local policymakers, builders and planners what changes can be expected by when, with the numerical probability that weather forecasts now describe a rain probability of 70%, for example.
Tapio Schneider, a Germany-born climatologist at the California Institute of Technology and Jet Propulsion Laboratory in Pasadena, California, is leading the effort.
“We don’t have good information for planning,” said Dr. Tailors a gathering of scientists in 2019. Models can’t tell New York City how tall it is to build walls, or California how much it will spend protecting its vast water infrastructure.
They just vary too much. For example, in Paris in 2015, 196 countries agreed that it will have alarming consequences if the planet, measured against the industrial age, warms up by 2 degrees Celsius. But when do we get there? From 29 leading climate models the answer ranges from 20 to 40 more years – almost the difference of a human generation – below the current emission values. This area is too large to schedule actions that require extensive new infrastructure, from replacing fossil fuels to moving to electric vehicles to lifting houses.
“It is important to make better predictions and make them quickly,” says Dr. Cutter.
This is funny
And it threatens to ripple feathers in the world of climate science, especially in established modeling centers like Dr. Schmidt in Goddard. “I think they oversold what they can do,” says Dr. Schmidt. Do you need a new model? “You would say yes. I would probably say no. “
Apparently a fairly modest group.
The other differentiator, notes Dr. Marshall, are the ones who are working on it. “The model is actually less important than the team of scientists you have around you,” he says. In fact, the 60 to 70 researchers and programmers in the CliMA group represent a true United Nations.
Somebody hung a map on the wall of the CliMA house, a converted provost house in Caltech, and asked everyone to identify their houses. “There were a lot of needles,” says Dr. Cutter.
Here’s the AI part
A climate model that “learns”
CliMA decided to take an innovative approach to use machine learning. Satellite and sensor information is freely available – much of it for weather forecasts. Dr. Schneider intends to “train” his model on data from the past three decades and then routinely provide it with the latest updates. The model itself could “learn” from the data and calibrate its performance using formulas refined by AI, even if the climate changes.
Other topics discussed include the reasons for choosing to program in Julia. To read the rest of the way, check out the full article here.
HT / Clyde Spencer
Like this:
Loading…
