Lack of Warning on Drought Reflects Forecasting Flaws

By Andrew Freedman

In May, the U.S. Agriculture Department predicted a record corn yield after farmers planted the largest area of corn and soybeans since 1937. Three months later, after a searing drought engulfed a wide swath of the continental U.S., those crops lie in ruin.

Despite all of the resources at forecasters’ disposal, the worst drought to strike the U.S. in nearly 50 years came on largely without warning across the fields of the Midwest and High Plains during late spring and early summer. Between May 1 and July 24, the drought footprint in the lower 48 states expanded from an already high 38 percent to a devastating 64 percent, engulfing more than a dozen states in the process, including nearly the entire corn and soybean growing region. Judging by past droughts, the drought of 2012 will likely cost the U.S. somewhere on the order of tens of billions of dollars.

The lack of advanced warning made the drought an especially harsh blow to farmers and ranchers who had been expecting a bumper year. Instead of a season of plenty, what farmers got was a “flash drought,” brought on by a stubborn “heat dome” of High Pressure and its brutally hot days and record warm nights. The 1-2 punch of the heat and lack of rainfall shriveled crops, baked fish in shrunken ponds and rivers, and forced ranchers to sell their cattle.

Despite major advances in seasonal climate forecasting during the past several decades, the failure to predict the scope and severity of the 2012 drought demonstrates that such forecasts are still rife with uncertainty, as forecasters must wrestle with the significant blind spots in their own understanding of the climate system and computer model simulations of that system.

Researchers and forecasters say that in addition to the uncertainties involved with their current tools and techniques, a lack of investment is also hobbling efforts to improve high-impact seasonal-to-yearly climate forecasts, such as the seasonal drought outlooks on which farmers and many others rely.

The three-month seasonal drought outlook, which is revised monthly, is the main drought forecasting tool produced by the federal government. It wasn’t until June 21 that an outlook showed drought conditions were likely to persist and expand in the Midwest and High Plains, and by that time, the country was rapidly heating up and drying out, destined to record its hottest month on record in July.

Prior to that — during the critical planting time when farmers were making key decisions about how to utilize their land during the growing season — the drought outlooks showed no hint of an impending widespread drought, let alone one of the top 10 worst droughts on record.

David Miskus, who prepares the drought outlooks for NOAA’s Climate Prediction Center, said that once drought conditions have set in, forecasters have a solid track record of predicting how such conditions will evolve. Accurately anticipating drought onset, though, is a much trickier problem.

“We don’t have much skill in forecasting drought development,” he said. One reason for this, scientists say, is that the computer models forecasters use don’t accurately capture the ways that land surface conditions interact with the atmosphere. The models tend to have more skill in predicting drought development or tendency out to a few weeks in advance, but beyond that, they have major limitations.

Temperature departures from average during July 2012.
Credit: NOAA/NCDC

According to Tony Busalacchi, the director of the University of Maryland’s Earth System Science Interdisciplinary Center, the severity and magnitude of the drought “caught everybody by surprise.”

“Our ability to predict drought on seasonal timescales is not very good,” Busalacchi said.

Climate forecasters and researchers said that there were signs that the summer of 2012 could feature more widespread drought conditions, since water temperatures in a large part of the Pacific Ocean were cooler than average, and Atlantic water temperatures were warmer than average. Such an arrangement has been linked to major droughts in the past because of the way it influences the formation and movement of weather systems across the U.S. Such a sea surface temperature pattern was in place during a severe drought in 1988, for example.

However, just because this sea surface temperature pattern is in place doesn’t guarantee drought development.

Busalacchi said the sea surface temperature pattern “would suggest drought,” but that forecasters completely missed the scale and scope of the disaster that has been unfolding during the past few months.

Klaus Wolter, a researcher at NOAA’s Earth Systems Research Laboratory in Boulder, Colo., said he was one of several scientists who warned of drought problems this summer, although he too was surprised by the “sharpness” and expansiveness of the drought conditions. “The general idea that we would be looking at drought problems this year in a greater acreage, I certainly put that out and I wasn’t the only one,” Wolter said.

“Some of this to me was clearly predictable, but that it would get this bad in the Midwest? That was something that wasn’t on the radar until very late.”

President Obama and Agriculture Secretary Tom Vilsack visiting a farm in Iowa to look at the damaged crops.
Credit: USDA.

Wolter said it was appropriate for the Climate Prediction Center to be cautious about predicting a major drought, since it was likely that such a forecast would be inaccurate, considering the odds of such extreme events. The Midwest had not seen a major drought in recent years and flooding has been a more frequent concern.

One of the reasons why forecasters missed this event may be due to flaws in the computer models they use to simulate short-to-medium term weather and climate conditions. Right now, the models don’t clearly capture the details of how land surface features interact with the atmosphere, and such feedbacks may play a key role in the development of drought conditions.

For example, drier-than-average conditions during the spring can foreshadow drought development during the summer, in part because there is less moisture available to enter the atmosphere in such regions.

According to Eric Wood, a professor at Princeton University who is researching ways to improve drought predictions, the state-of-the-art seasonal forecast models tend to show drought conditions out to a couple of weeks in advance, but then revert back to showing normal precipitation patterns after that. This tendency limits forecasters’ skill in predicting drought conditions a month or more ahead of time.

“What happens is that the models are not able to have very much skill out past a few weeks when it comes to precipitation. They’re better on temperature, but not on precipitation,” Wood said. His research group at Princeton is working to understand why the models perform so poorly with long-range precipitation outlooks, and he said the primary suspect at this point is the failure to accurately account for interactions between the land and atmosphere.

A Question of Resources

In today’s budget-constrained environment, funds for seasonal climate forecasting are squaring off against other priorities within agencies such as NOAA, which also supports weather research and long-range climate change studies.

Improving seasonal-to-yearly climate forecasts could have significant societal benefits. “There are huge impacts on the economy when looking at the seasonal [time]scales,” Wood said.

For example, if farmers and ranchers had advanced notice of the current drought event, perhaps they could have minimized their economic losses by planting less during the spring, and purchasing crop insurance.

However, the main office within NOAA that funds climate research and applications, known as the Climate Program Office, had its budget slashed by 30 percent last year. It is focusing the majority of its cuts on research programs, including seasonal climate forecasting research, while trying to maintain weather and climate observation networks and existing programs, including the National Integrated Drought Information System.

“More resources could be going to the seasonal to interannual [prediction] problem,” said Lisa Goddard, the director of the International Research Institute for Climate and Society at Columbia University. Goddard said it is often difficult to show forecasting improvements as well as the need for further refinement of seasonal predictions, in part because of the way that outlooks are presented to the public and government officials. Another problem is the perception that seasonal-to-yearly forecasts are already as accurate as they are going to get, and that short-term weather and long-term global warming should get the lion’s share of funds.

“It is true that there are a number of individuals in the government that think the seasonal prediction problem is done, and I think that’s unfortunate,” she said.

Busalacchi said it’s important that forecasters perform a post-mortem assessment of what went wrong with this year’s drought outlook to find the important factors they failed to pick up on, and learn how to improve forecasts of the next drought event. Unfortunately, he said, it’s not clear that the funding is available for such studies, nor is there a clear and consistent push for such work from NOAA’s leadership.

For farmers and ranchers across the heartland of the U.S., that post-mortem will come far too late, and at far too high of a cost. As the summer nears an end, their hopes now rest with the prospect for drought-busting rains during the cooler months ahead. And the newest seasonal drought outlook, issued on Thursday, does show improving conditions in parts of the Ohio Valley and Southwest, but persistent drought through the end of November in the High Plains and Mississippi River Valley.

Then again, based on its recent track record, that drought outlook should be taken with a grain of salt.