Alumnus, researcher help develop new Climate Prediction Center forecasting tool

Predicting the weather several weeks in advance would be a major asset to businesses, governments and even individuals or families. If businesses had three weeks’ notice that the start of winter would most likely be warmer than average, they could adjust their inventory accordingly. Governments with money set aside for snow removal could reallocate funds to other public services. Families nailing down vacation logistics would have a better idea of what clothes to bring.

But chaos theory has put limitations on what meteorologists can reliably forecast in this time range.

“Historically, forecasting three and four weeks into the future has been regarded as particularly challenging, which is paradoxical because we are able to make forecasts with a longer lead time, one month or more into the future. The temperature of the ocean surface, which evolves slowly from month to month, has a large impact on the atmosphere, so it’s possible to make forecasts a month or more into the future. However, data from the ocean’s surface hasn’t allowed us to make reliable forecasts in the three-to-four-week range,” said Steven Feldstein, Penn State meteorology professor and senior scientist.

Feldstein and Penn State alumnus Nat Johnson have made the first step in addressing this difficulty.

When Johnson was a Ph.D. student in Penn State’s meteorology program, he collaborated with Feldstein on basic research into two large-scale weather processes occurring in the tropics, El Niño Southern Oscillation (ENSO) and the Madden-Julian Oscillation (MJO). During their investigation, the researchers noticed that weather in the tropics could impact weather thousands of miles away.

They applied their findings to help develop what’s known as a probabilistic forecasting tool. The tool gives a skillful approximation of how temperature and precipitation three to four weeks into the future will compare to historical averages.

“No matter how big our computer and how good our science, there will always be uncertainty in weather predictions, and that’s okay. The challenge is not to make uncertainty go away, but to figure out how we can use it,” said  David Titley, professor of practice in the Department of Meteorology and director of Penn State’s Center for Solutions to Weather and Climate Risk.

Probabilistic models differ from the more common short-term forecasts for temperature, rain or snow. When you look at tomorrow’s predicted temperature, you’re seeing a result from a “deterministic” computer model that predicts a determined value.

“Even though we have uncertainty, we’re able to convey some useful information about a predicted state in a probabilistic form. We couldn’t say the temperature will be a specific amount, but we could say that, three weeks from now, it probably will be wetter than normal, or there’s a 65 percent chance it will be warmer than usual,” said Johnson, who is now an associate research scholar with Princeton University’s Program in Atmospheric and Oceanic Sciences.

3-4 week experimental forecast screenshot

Example of the 3-4 week experimental model used by the Climate Prediction Center. Penn State researcher Steven Feldstein and alumnus Nat Johnson conducted basic research and collaborated with the CPC to develop this tool.

Image: Climate Prediction Center

After Johnson graduated from Penn State, he and Feldstein, along with researchers at the U.S. Climate Prediction Center (CPC), continued investigating the impact of the MJO and ENSO in closer detail.

The MJO is a slow-moving weather system that changes temperature, precipitation, pressure and other meteorological variables where it travels. It moves around the world eastward through the tropics (the area between the Tropic of Cancer and Tropic of Capricorn) and cycles through eight distinct phases on a regular basis, completing a full cycle in 30-60 days. It has regular impacts on tropical cyclones, monsoons and precipitation in the tropics.

Johnson and Feldstein noticed that it also affects weather thousands of miles away from its location.

“We were researching the large-scale circulation impacts of ENSO and the MJO, which occur in the tropics and are associated with clusters of thunderstorms about the size of the U.S. We noticed a relationship between what’s going on with ENSO and the MJO in the Indian and Pacific Oceans and the weather over North America several weeks later,” said Johnson.

“Immediately, we saw the potential for making forecasts into that range, which hadn’t been done before,” said Feldstein.

After documenting their findings in the journal Weather and Forecasting, in 2014, Feldstein and Johnson began working with the CPC to develop a forecasting tool that could be put into operation and made available to the public.

As an integral part of the National Weather Service, the CPC provides climate prediction and real-time weather monitoring services to individuals, businesses and governments across the U.S. It offers outlooks ranging from three days to an entire season in the future, but, until the fall of 2015, they were missing a forecast for the three-to-four-week range.

That changed on September 18, 2015, when the CPC made new experimental three-to-four week probabilistic forecasts available on the CPC website along with their suite of other forecasts.

After collaborating on this research for more than seven years, Feldstein and Johnson said that they were pleased to be able to apply principles of basic science in a way that will benefit people throughout the United States.

“This project has been very satisfying, and I enjoy undertaking both basic and applied sciences, and this is a great example of how we’ve been able to apply science in a way that’s useful for a lot of people,” said Johnson. “I owe a great deal to Penn State and the Department of Meteorology. There was a real focus on developing a solid foundation of the physical principles while also being connected to operational forecasts.”

“We went from doing basic research on the MJO, ENSO and teleconnections and used our findings to develop a forecasting tool. I feel fortunate to have played a role in this project, especially because, for years, week 3 and 4 forecasts have been regarded as particularly challenging,” said Feldstein. “Now we are trying to improve these forecasts by taking into account other processes, such as the impact of changes in Arctic sea ice.”

Contacts: 
Last Updated February 25, 2016